Model Reference Paper (rev B) Guidelines for the LRIC bottom-up and top-down models

Transcription

1 DATE 12 September 2007 Dnr /23 Model Reference Paper (rev B) Guidelines for the LRIC bottom-up and top-down models

2 1 Introduction and background Policy and objectives Policy Objectives Methodology Process Document structure... 4 Part A: Common guidelines Long Run Incremental Cost Definition of LRIC Long Run Incremental Forward-looking costs Definition of increment Size of increment Access and core increment Other increments Cost Causality Treatment of common costs Allocation of common costs Services modelled Core and access services PSTN (PSTN/ISDN) and Broadband services Leased lines Other services Co-location services The nature of co-location services List of co-location services covered by LRIC Demand and growth Demand Margins for growth Model outputs Cost types Level of detail Cost categories Cost of co-location Costs of shared access to unbundled local loop Cost of Bitstream access General costing issues Annualisation methodologies Post- och telestyrelsen i

3 5.1.1 Annualisation criteria Capital charge Depreciation charge Annuities Guidelines Cost of capital Geographical differentiation Geographical differentiation Geo-types Other costing issues Base year Working capital Set-up and duration related charges Routing factors Part B: Specific guidelines for top-down model Overview of top-down modelling Determine homogenous cost categories (Step 1) Group cost category by activity and network elements (Step 2) Re-value assets, calculate GRC, NRC and CCA depreciation (Step 3) Developing cost-volume relationships (Step 4) Cost access and interconnection services (Step 5) Top-down asset valuation and capital costs Gross asset valuation Current cost accounting Revaluation methods Asset prices New technology Land and buildings Utilisation rates Valuation of major asset categories Access network Core network Trenching costs, incl. poles Poles Indirect network costs (buildings, IT, motor vehicles, etc.) Co-location Annualisation Asset lives Annualisation of relatively new assets Fully depreciated assets Capital maintenance Net asset valuation NBV/GBV methodology Rolling forward methodology Assets under construction Post- och telestyrelsen ii

4 8 Operating costs Operating costs Cost categories Efficiency Activity based allocation of operating costs Documentation Costing services Homogeneous cost categories Service usage Trenching and duct Copper and fibre cable Local Exchanges Tandem exchanges Transmission equipment Indirect network costs and overhead costs Outsourced costs Assigning costs to services Calculating incremental costs Calculating costs of services within an increment Treatment of common costs Model functionality and documentation Model requirements Sensitivity analysis Model documentation Justification Audit of model Part C: Specific guidelines for bottom-up model Overview of bottom-up modelling Measuring demand and establishing unit costs (Step 1&2) Building hypothetical network (Step 3) Determining the cost of network elements (Step 4) Costing services (Step 5) Optimisation Scorched node assumption Technology Switching technology Transmission technology Access technology Requirements of the optimised network Quality of service Service equivalence Correct dimensioning of the network Post- och telestyrelsen iii

5 13 Demand Demand in the access network Demand in the core network Estimation of end-user demand Estimation of dimensioned demand Bottom-up modelling Equipment prices and cost data Modelling exchanges The hierarchy of the exchanges Choosing between different nodes Modelling transmission Transmission hierarchy Network configuration Dimensioning the network Modelling access Collecting information by geo-types Estimation of equipment requirements Modelling infrastructure Trench in the core network Costing of trench Duct Poles Cable requirements in the core network Cable modularity and length Modelling co-location Bottom-up costing issues Indirect network costs Overheads Operating costs Cost allocation Model functionality and documentation Model requirements Sensitivity analysis Model documentation Appendices Appendix 1 Summary of criteria Appendix 2 Abbreviations used Appendix 3 Glossary of terms Post- och telestyrelsen iv

6 Appendix 4 Detailed list of services modelled or covered by LRIC A.4.1 Interconnection services A.4.2 (Wholesale) copper access services A.4.3 Bitstream Access A.4.4 Co-location services Post- och telestyrelsen v

7 1 Introduction and background This Model Reference Paper (MRP) outlines the features and key principles applying to the bottom-up and top-down models to be developed with the purpose of producing a hybrid model and determining LRIC based cost for certain (wholesale) access and interconnection services in Sweden. The 'LRIC method' refers to a method for the calculation of cost-orientated pricing that is - based on the long run incremental costs for an efficient operator who makes use of modern technology, and - includes a mark-up for common costs for an efficient operator under competitive conditions. 1 The hybrid model shall be used, together with a pricing methodology approved by PTS, when PTS assesses whether the prices that TeliaSonera applies for interconnection and LLU, including co-location products, satisfy the requirement regarding cost orientation. During the years , PTS, together with operators, has produced an LRIC model, a 'hybrid model', for calculating the costs of interconnection in the fixed network and local loop unbundling (LLU). The first version of the hybrid model was approved on 19 December PTS has thereafter updated the hybrid model on an annual basis, resulting in the current version (v4.1) released in December According to the PTS Regulations on the LRIC method for the calculation of cost-oriented pricing, PTS shall, at least every three years, review the need to revise the respective hybrid model. PTS shall then take into account, among other things, economic life, required return and the application of new technology. A review of the need to revise the hybrid model for the fixed network means that PTS needs to consult operators regarding a multitude of technical and financial issues. The consultation with the operators confirms that there is a need for revising the hybrid model. PTS has decided to revise the Model Reference Paper as a prelude to preparing a revised bottom-up model and requiring the SMP operator to prepare a revised top-down model. The revision of the hybrid model for the fixed network comprises: - products and services in the access network that are covered by PTS's obligation decision on local loop unbundling: shared access, full access and co-location, - interconnection services in the fixed networks that are covered by PTS's obligation decision on interconnection. 1 PTSFS 2005:5 PTS Regulation on the LRIC method for the calculation of cost-orientated pricing. 2 LRIC, The final hybrid model, 19 December Post- och telestyrelsen 1

8 The revision also includes the development of the hybrid model to produce cost calculations under the LRIC method for: - products and services covered by PTS's obligation decision on bitstream access. 1.1 Policy and objectives Policy One of the main principles in PTS policy for the access regulation of last mile networks is that infrastructure-based competition should be promoted when replication of infrastructure is feasible. 3 To give the right investment signals and promote effective competition, the prices for interconnection and LLU should be based on the long-run incremental cost (LRIC) for an efficient operator. When the prices are determined on the basis of the LRIC-methodology infrastructurebased competition is promoted in those areas where it is efficient to have replication of infrastructures, while service-based competition is promoted in areas where replication of infrastructures is not efficient. The hybrid model puts the policy into practice by creating neutrality in the choice between building your own infrastructure and paying for access to the SMPoperator s infrastructure. Conditions for neutral incentives are created since the assumptions underpinning the hybrid model represent a balanced view on what it would cost for an efficient operator of TeliaSonera s size to build and run a network today. Thus the assumptions take account both of those within the bottom up model, which are by necessity somewhat theoretical, and those in the top down model, that encompass some of the realities of TeliaSonera s actual network Objectives The objective of using LRIC are to: Encourage the use of existing facilities of the SMP operator where this is economically desirable, avoiding inefficient duplication of infrastructure costs by new entrants (incentive to buy); Encourage investment in new facilities where this is economically justified by 1. new entrants investing in competing infrastructure 2. the SMP operator upgrading and expanding its networks (incentive to build); Increase the transparency of the cost calculations underlying the access and interconnection charges; and 3 Policy för tillträdesreglering i accessnät - PTS-ER-2006:26, 10 July Post- och telestyrelsen 2

9 Increase predictability for both the SMP operator and the other operators with regards to future determination of access and interconnection charges. When access and interconnection charges are based on LRIC they do not distort the build/buy decision of new entrants they will be encouraged to use existing facilities if, and only if, it is economically desirable to do so. Just as important, LRIC-based access and interconnection charges also mean retaining investment incentive for incumbents to upgrade or extend the existing network when new technology is available. When charges are set on the basis of LRIC, infrastructure competition is encouraged in those areas where it is efficient to have competing infrastructure, whereas service competition is encouraged in those areas where the investment in competing infrastructure is not efficient. 1.2 Methodology To send the right investment signals and promote efficient competition, prices should reflect the LRIC of an efficient operator facing the demand of the existing SMP operator (currently this means TeliaSonera) 4. The efficient operator is defined as the theoretical operator that would exist if it were in a fully competitive market in Sweden, but with the same scope and demand of the existing SMP operator. This approach ensures that the economies of scale, scope and density are divided equally between the SMP operator and the interconnecting operators allowing the interconnecting operators to compete with the SMP operator on equal terms. Ultimately, the cost results derived from the hybrid model will be based upon a fair comparison of the costs calculated in a top-down and a bottom-up model respectively. The purpose of the top-down model is to calculate the LRIC on the basis of the existing network and cost structure of the SMP operator, eliminating inefficiencies and replacing outdated equipment with new, more cost-effective technology. The purpose of the bottom-up model is to calculate the LRIC on the basis of an efficient network using the newest technology actually employed in large-scale networks. In principle, the bottom-up model should model the network that an efficient operator would build today to meet the forward-looking demand of the SMP operator. The costs (if any) for migrating to the efficient operator from today s operations are not included. 1.3 Process The review of the hybrid model will observe the principles prescribed by PTS's Regulations. This revised model reference paper sets out the guidelines and criteria for the modelling work. All guidelines set out as criteria as well as other guidelines should 4 These costs are likely to be different from the costs of an actual new entrant, entering the current market, as the new entrant will not be able to achieve the same economies of scale, scope and density as the SMP operator when the SMP operator is already in the market. Post- och telestyrelsen 3

10 be followed during the modelling work unless it explicitly appears that the text is of a guiding nature that the responsible modellers may choose to depart from. Other deviations require prior clarification with PTS. Upon finalisation of the revised models, PTS will undertake a thorough review of the two models to ensure that they meet the guidelines set out in this MRP. TeliaSonera is responsible for developing a revised top-down model while PTS is in charge of developing a revised bottom-up model in co-operation with interested parties from the industry, including TeliaSonera. A reconciliation of the two models will again be undertaken by PTS and used as the basis for PTS's development of a revised hybrid model on which the final price setting will be based. A reconciliation report will be submitted to public consultation in order to seek opinions from the industry. The report will identify, quantify and where possible explain the differences between the two revised models. Also the revised hybrid model and PTS s final proposed pricing methodology will be subjected to public consultations. 1.4 Document structure The guidelines provided in the MRP are structured in three parts plus appendices: Part A includes common guidelines for the two models. Part B includes specific guidelines for the top-down model. Part C includes specific guidelines for the bottom-up model. In part A: Chapter 2 discusses concepts involved in the definition of long run incremental costs. Chapter 3 describes the services that need to be modelled and the services that should be priced according to LRIC. Chapter 4 discusses the outputs of the two models including the level of detail. Chapter 5 discusses a number of general costing issues such as depreciation methodologies, cost of capital, working capital, and the use of routing factors. In part B: Chapter 6 provides an overview of the main steps involved in building a topdown model. Chapter 7 discusses the gross and net asset valuation of assets in the top-down model. Chapter 8 describes the calculation of working capital and allocation of operating costs in the top-down model. Chapter 9 discusses a number of the issues involved in the costing the access and interconnection services including the allocation of costs from cost categories to network elements (using cost volume relationships) and on to services (using routing factors). Post- och telestyrelsen 4

11 Chapter 10 sets out the requirements for the functionality, documentation and audit of the model. In part C: Chapter 11 provides an overview of the main steps involved in building a bottom-up model. Chapter 12 discusses the level of optimisation in the bottom-up model including constraints such as the scorched node assumption. Chapter 13 describes how demand in the core and access network should be estimated and applied in the bottom-up model. Chapter 14 covers issues related to the estimation of equipment prices and specific guidelines for the modelling of exchanges, transmission and access network and infrastructure. Chapter 15 discusses a number of other costing issues such as the calculation of indirect costs, overheads, operating costs and working capital, and allocation of costs to network elements and on to services. Chapter 16 sets out the requirements for the functionality and documentation of the bottom-up model. In the appendices: Appendix 1 includes a summary of criteria. Appendix 2 lists the abbreviations used in the MRP. Appendix 3 presents a glossary of terms. Appendix 4 details the list of services included within the various Reference Interconnect Offers that should where practical be modelled. Post- och telestyrelsen 5

12 PART A: COMMON GUIDELINES Post- och telestyrelsen 6

13 2 Long Run Incremental Cost 2.1 Definition of LRIC In this section, the concepts involved in the definition of (forward-looking) Long Run Incremental Costs (LRIC) are discussed Long Run Using a long-run measure of costs, such as LRIC, implies a time horizon where all inputs, including capital equipment, may vary in response to a change in demand. This means that the cost models should adapt all input factors to the forecasted demand for services respecting practicalities like minimum size of input and quality of service Incremental Incremental costs are the costs caused by the provision of a defined increment of output given that some level of output (which may be zero) is already being produced. Equivalently, incremental costs can be defined as the costs avoided (i.e. saved) by not providing the increment of output. For the purpose of interconnection charges, the increments have usually been defined as the entire group of services using the core (or access) network. These services (PSTN, 5 Broadband, leased lines, etc.) include those provided by the SMP operator as well as those provided by interconnecting operators using the SMP operator s network. The costs of the network providing this wider group of services are then divided by the total volume of demand (for example, number of subscribers, calls or traffic minutes, Gigabytes) in the increment to produce the average incremental cost or per unit LRIC 6,7. Figure 1 illustrates the concepts of incremental and average incremental costs: 5 For the purpose of this Model Reference Paper, PSTN should be considered to include both PSTN and ISDN. 6 The terms LRIC and LRAIC (long run average incremental costs) are often used interchangeably. 7 This averaging does not imply that all the services included in the increment will be attributed the same costs. It simply means that the costs of using a given network element will be the same for the included services. Services will use the network differently. As discussed in section 5.4.4, this will be taken into account through the use of so-called routing factors. Services may also have different cost drivers, such as subscribers, calls and minutes, or packets. Nor does the averaging imply that prices will need to be the same throughout the day. When setting interconnection charges, costs may also be "de-averaged" to a peak and an off-peak charge. Post- och telestyrelsen 7

14 Figure 1: Long Run (Average) Incremental costs Costs Incremental costs 1 Average incremental costs Fixed common costs Increment Volume Forward-looking costs Costing systems can be backward-looking, forward-looking, or a mixture of the two. Backward-looking systems are based on the historic cost basis. They may contain outdated technologies and inefficiently incurred costs reflecting, for example, the costs of manpower required to maintain outdated technologies. Forward-looking costs are not the costs in the future but reflect the costs that a network operator, building a network today, looking forward, would incur. Costing measures should be forward-looking to reflect the true economic costs of producing an increment of output. In practice, however, there is likely to be considerable debate about the precise definition of forward-looking. Networks evolve over time with the result that the network of even an efficient SMP operator may look very different from the network design that would be used if starting from scratch (often referred to as a scorched earth assumption). To make no allowance for the starting position of the operator could result in cost estimates, and therefore interconnection charges, that are too low. Such an outcome would not provide the right incentives for SMP operators to invest in, and maintain, their networks. Moreover, other operators would have no incentive to invest in their own infrastructure, since they could immediately purchase the interconnection services from the SMP operator when an investment is successful, while not paying for unsuccessful investments. Therefore, the scorched node assumption discussed in section 12.1 should be applied. How forward-looking the models should be in their choice of technology is described in more detail in sections 7 and As mentioned, the forward-looking costs are the costs of building a network today, looking forward. "Looking forward" implies that the expected development in prices, first of all asset prices, and expected development in demand will need to be taken into account. Finally, it should be noted that the models should model this optimised network as if it were already in place. No migration costs (additional costs associated with moving from the existing network to the optimised network) should be included. Post- och telestyrelsen 8

15 Criterion CG 1 The models should be based on forward-looking long run incremental costs. No migration costs should be included. 2.2 Definition of increment Size of increment In principle, there are an infinite number of different sized increments that could be measured. However, these increments can effectively be grouped into three different categories: 1. a small change in the volume of a particular service; 2. the addition of a whole service; or 3. the addition of an entire group of services. The first definition of the increment is equivalent to a measurable version of marginal cost, that is the cost associated with a one-unit change in output. The second definition may apply to services of very different sizes, such as interconnection, local calls and national calls. In telecommunications regulation, the third definition of the increment is generally used to set interconnection charges. The use of large increments minimises the amount of common costs and ensures an equal treatment of the SMP operator's internal (on-net) and external (interconnection) traffic Access and core increment Two main increments are defined 9 : The access increment defined as all (regulated as well as unregulated) services using the access network The core increment defined as all (regulated as well as unregulated) services using the core network. The incremental costs of the core are those costs incurred when adding a core network when an access network is already in place. Similarly, the incremental cost of the access network is the cost incurred when adding an access network when a core network is already in place. The LRIC of co-location is the cost incurred when providing co-location services. 8 When both the internal and external traffic is included, the cost of using a given network element will be the same for internal and external traffic. This may not be the case if a separate increment is defined for interconnection services. 9 These definitions are consistent with the original definitions offered by the Commission in Commission Recommendation on Interconnection in a Liberalised Telecommunications Market Part 1 Interconnection Pricing, 15 October Post- och telestyrelsen 9

16 Criterion CG 2 For the core network, the increment should include all services using the core network. For the access network, the increment should include all services using the access network. The LRIC of co-location is the cost incurred in providing co-location services. These definitions should include the services provided by the SMP operator s network division to its own retail division as well as the services provided to other operators. Costs in the core increment are driven by, for example, the volume of traffic, packets or calls, whereas costs in the access network are mainly driven by the number of subscribers. Costs are also driven by the quality of service (QoS), particularly in the core network. Until QoS is specifically charged for separately, it will need to be treated as a minimum requirement. The access network is typically defined from the first connection point at the customer s premises up to and including the line card (hereafter referred to as the access-core line card). Having said this, current developments in technology are now blurring the borderline between the access and core networks. For PSTN services, the access-core line card can either be placed as part of the local exchange (often referred to as Host Subscriber Stage - HSS) or away from the local exchange (often referred to as Remote Subscriber Stage - RSS). For leased line services, the access-core line card will be placed as part of the transmission equipment at the first node (ie the serving node), and for broadband services, the access-core line card placed as part of the DSLAM, which might even be placed in a street cabinet. For Remote Multiplexers (RSMs) locations the situation is not so clear cut. In some cases, the RSM was installed to replace an old analogue exchange and in others has been used merely as a work-around for a lack of copper in the access network. In either case, there will also be a line card at the RSS/HSS and so there is an option for the access/core demarcation to be deemed to occur at either the line card of the RSM (in which case the RSM would have the access-core line card) or the line card of the RSS/HSS (in which case the RSS/HSS would have the access-core line card). Potentially, therefore, each RSM location needs to be treated on its own merits, with factors taken into consideration including: the historical reason for the RSM deployment the number of connected customers the existence (or not) of other equipment at that location (particularly DSLAMs) the physical distance from the RSM to the next equipment location The access-core line card should be included in the access network, as the number (and thus costs) of such line cards are related to the number of lines rather than the amount of traffic (whether minutes, calls, or packets). The cost of the line card should therefore be included in the cost of access services. However, the Post- och telestyrelsen 10

17 access-core line card(s) should be excluded from the costs of unbundled local loops, as the line card is not used in the provision of this service. Criterion CG 3 The access-core line card (typically, though not necessarily, for the PSTN located in the concentrator) should be included in the access network, whereas other equipment related costs should be included in the core network, except where costs are common between the two networks. The access-core line card(s) should be excluded from the costs of unbundled local loops but included in the costs of the access service it relates to (such as telephone line rental for a PSTN line card, broadband access for a DSLAM line card etc.) Other increments Other potential increments include a retail increment for the access and core networks; an international increment; an increment for premium rate services; an increment for the mobile network; and an increment for other services: Retail Increment. The costs referred to for the access network and core network are wholesale costs. They exclude any costs incurred in packaging and selling services delivered over these networks to final customers, as opposed to wholesale customers. Such costs include marketing and customer billing costs, customer service costs and retail elements of the finance and human resources departments, land and buildings. These costs belong in the retail increment. International Increment. This increment covers the costs of the transmission links between tandem and international exchanges as well as the cost of international exchanges and assets at these exchanges. International calls will typically use both the core increment and the international increment. Premium Rate Services Increment. Premium rate services include toll free (0800) phone calls and various information services which can be accessed at premium rate charges. These services may require a separate network of switches (in which additional transmission assets will be required) or may be delivered over existing switches. Mobile Increment. Assets included in this increment include the base stations, basestation controllers, mobile-switching centres and transmission links required by the mobile network. Other Increment. The SMP operator may provide a range of other services, such as the provision of customer premises equipment, and may have investments in other companies either at home or abroad. These increments will not need to be modelled separately. However, the associated traffic, using the core and access network, needs to be included in the core and access increment. Also the core and access increment may share costs with these other increments. For example, the international switches and mobile switches will often be co-located with the PSTN switches. Where this is the case, a fair proportion of the building costs (and other common costs) should be allocated to the mobile and international increment respectively. Similarly, the cost Post- och telestyrelsen 11

18 of the head office function should have a portion allocated to overseeing any investments in other companies. 2.3 Cost Causality The definition of the access and core increments set out in section implies that fixed costs 10 that are specific to either the core or access networks are included in LRIC. To the extent practical, costs (both capital costs and operating costs) should be allocated to services on the basis of cost causality. This might be on either a direct or indirect basis. A good example of how this might apply in practice is to consider the issue of trench, duct, cable and fibre where, as a starting point: Trench should be allocated on the basis of duct Duct should be allocated on the basis of cables Cables should be allocated on the basis of system lines (fibre pairs connecting distant items of transmission equipment) System lines should be allocated on the basis of usage, where necessary taking account of usage translation techniques (for example, translating voice minutes into Mbps) The above example assumes that the network is constructed in an efficient manner and does not, without good, justifiable reasons, separate services such that not all services attract a fair proportion of cost. We distinguish here between three types of costs: Directly attributable costs, shared costs and common costs. Directly Attributable Costs are the costs incurred as a direct result of the provision of a particular service in a particular increment. These costs fall into two types. Firstly, the costs of some inputs vary with the level of output, so that even if the output of more than one service requires this input, the extent to which a single service causes the costs can be calculated. Secondly, there are assets and operating costs which are fixed with respect to the level of output but which are service specific. Shared Costs are the costs of those inputs necessary to produce two or more services within the same increment, where it is not possible to identify the extent to which a specific service causes the cost. Examples of shared costs in the core network include optical fibre, transmission equipment and related overheads, all used by PSTN, leased line and other services Common Costs are the costs of those inputs necessary to produce one or more services in two or more increments, where it is not possible to identify the extent to which a specific increment causes the cost. Trenching costs provide a good example of the difference between shared and common costs. The costs of trenching specific to the access network (or the core network) will generally be shared costs since the trenching is likely to be used by two or more services. 10 Fixed costs are defined here as costs which do not change with the level of output Post- och telestyrelsen 12

19 However, some trenching will be used by both the access and the core network. In these instances, the costs will be common costs. Another example of common costs is corporate overheads. Figure 2: Cost concepts Access Network Costs Core Network Costs Attributable costs in the access network (e.g. line card) Shared costs in the access network (e.g. trench in the access network) Attributable costs in the core network (e.g. tandem exchange) Shared costs in the core network (e.g. trench in the core network, ADMs etc.) Common costs (e.g. trench shared by access and core) Figure 2 illustrates the relationship between directly attributable, shared and common costs. The first definition of the increment discussed in section would only include some of the directly attributable costs in the core and access networks. The second definition would include all directly attributable costs. But the variant of the third definition taken to be the increment for LRIC purposes would include all directly attributable and shared costs in the core and access networks. Only common costs would be excluded. Criterion CG 4 To the extent practical, costs (both capital costs and operating costs) should be allocated to services on the basis of cost causality. This assumes that the network is constructed in an efficient manner and does not, without good, justifiable reasons, separate services such that not all services attract a fair proportion of cost. 2.4 Treatment of common costs The large proportion of fixed common costs in telecommunications means that setting interconnection charges equal to incremental costs does not allow the SMP operator to recover the costs that span increments, even when those costs are efficiently incurred. However, setting interconnection charges based on (but not set at ) LRIC permits recovery of efficiently incurred common costs. This can be achieved via the use of mark-ups, where, for example, the LRIC of each increment is marked Post- och telestyrelsen 13

20 up by an equal proportion so as to recover (but not over-recover) the common costs. Criterion CG 5 The models should allow recovery of common costs. These costs should be shown separately. Criterion CG 6 The models should identify the costs that are common between the other increments and the core and access networks Allocation of common costs As defined above, common costs are the costs of those inputs necessary to produce one or more services in two or more increments, where it is not possible to identify the extent to which a specific increment causes the cost. The allocation of such common costs will therefore always be somewhat arbitrary. Otherwise the common costs would not really be common and should instead be allocated directly to the increment. Common costs are therefore typically allocated using some sort of mark-up 11. Before turning to the description of mark-ups, it should, however, be underlined that broad mark-ups should only be used where it is not possible to establish a clear causal relationship between costs and services/increments. In many instances it will be possible to establish such a relationship by carefully examining the direct and indirect costs drivers 12. Exchange buildings, for example, might at first sight be considered a common cost, as they are used to provide both access and core services. It would however, be inappropriate to allocate the total costs of exchange buildings on the basis of simple mark-ups. The costs of exchange buildings are to a large degree driven by the number of square meters required by the equipment installed in the buildings. Most of the equipment in the exchange building is core related. Only the distribution frame and part of the subscriber stage are access related. It will therefore be possible to allocate most of the building costs to the core network, e.g. on the basis of the occupied square meters. Mark-ups can be either additive or multiplicative and similarly be either differentiated or uniform. A (uniform) additive mark-up implies that common costs are divided by the number of increments and the resulting total is added to each increment (thus, if common costs were 2,000, 1,000 would be added to both 11 It may be noted that a price based on LRIC + mark-up for common costs will lie between LRIC and the so-called stand-alone costs (SAC), which are the costs of producing the increment, assuming no other increments were produced. A price set at SAC would correspond to allocating all the relevant common costs to the increment, whereas a price set at LRIC would correspond to not allocating any of the common costs to the increment. 12 Some underline this point by using the term residual common costs instead of common costs (the common costs that can not be satisfactorily allocated using direct or indirect cost drivers). Post- och telestyrelsen 14

21 the core and access increments). An additive mark-up implies that the allocation of common costs is independent of the costs of the various increments. A multiplicative mark-up implies that common costs are split in relation to the relative level of incremental costs of each of the increments. For example, if the incremental cost of access is 75% of total incremental costs and the incremental cost of core is 25% of these costs, then the access increment would be allocated 75% of common costs and the core increment 25% of these common costs. This means that if LRIC for the access network is 15,000 and LRIC for the core network is 5,000 and commons costs are 2,000, the mark-up will be 10% (2,000/ (15, )). Thereby 1,500 (=0.1x15,000) of the common costs are allocated to the access network and 500 to the core increment. Multiplicative mark-ups are sometimes referred to as equi-proportionate mark-ups. An alternative approach would be to calculate differentiated mark-ups. That is to say the mark-up as a proportion of incremental cost would vary between services. There are at least two potential rationales for using differentiated mark-ups. Firstly, differentiated mark-ups could be used to encourage entry or to ensure that common costs are not assigned to increments that are not prone to competition. 13 Having said this, differentiated mark-ups could well be not allowed under established regulatory practice. Secondly, differentiated mark-ups may be used to reflect the fact that services have different demand elasticities. So-called Ramsey mark-ups are determined on the basis of demand elasticities. Where a service has a high elasticity, the markup should be low, because changes in price have a significant impact on purchasing patterns thereby distorting market signals. The converse is the case where elasticities are low. From an economic point of view, a Ramsey mark-up is the theoretically correct way to efficiently recover the common costs, ensuring that resources are put to their best possible use. Indeed, it can be shown that a profit maximising company operating in a competitive market would also set its prices in this way. Ramsey pricing, however, has a number of weaknesses when implemented in practice. First of all, price elasticities are very difficult to estimate and verify. This is of special concern since an operator operating in both competitive and regulated markets will have a strong incentive to attribute a disproportionate amount of the common costs to the regulated products. Price elasticities would also be likely to vary over time, with price, and be dependent on the level of competition in various segments of the market. Also multiple price elasticities could occur depending on the intended use of the product. The method therefore faces a number of operational difficulties See footnote 11 in the Commission's recommendation on interconnection in a liberalised telecommunications market (98/C84/03) and point 696 in FCC's "First Report & Order in the Matter of Implementation of the Local Competition Provisions in the Telecommunications Act of 1996". 14 Proper Ramsey mark-ups would also take into account the impact of cross-elasticities and externalities. This could significantly improve the allocative properties of the resulting prices, but would also add to the complexity of the analysis and the associated data requirements Post- och telestyrelsen 15

22 Secondly, it may seem unfair that consumers should bear a larger burden of the costs just because they are so dependent on provision of the services or have so few alternatives that their demand is not very sensitive to the price. Finally, it is not always clear how to estimate demand elasticities for access and interconnection services, since these services are sold to other operators reselling and re-packaging the services to end-users with very different demand elasticities. Criterion CG 7 As far as possible, common costs should be allocated to increments and services using appropriate (direct or indirect) cost drivers. Only common costs, for which it is not possible to identify the extent to which a specific increment or service causes the costs, should be allocated via mark-ups. The starting point should be equi-proportionate mark-ups. The models should allow for equi-proportionate mark-ups to be used for all cost categories. It is possible that there could be instances where there might be good reasons for departing from equi-proportionate mark-ups. However, if this is the case, it must be properly justified in the model documentation. Post- och telestyrelsen 16

23 3 Services modelled Telecommunications operators typically carry a wide range of services across their networks. In addition to the traditional voice services, operators provide leased lines, broadband and other data services, and other services such as cable TV. The models need to account for all of these services. To exclude some would result in an under-dimensioned network and increased costs for the remaining services as costs, such as ducts, would be allocated to fewer services. Therefore more services need to be modelled than the actual number of services that should be priced on the basis of LRIC. Appendix 4 includes a detailed list of the RIO (Reference Interconnect Offer) services. The models should include and categorise services under the following headings: Core and access services PSTN/ISDN services Broadband services (including Bitstream access) Leased lines; and Other services Co-location services 3.1 Core and access services PSTN (PSTN/ISDN) and Broadband services PSTN services include standard call services that originate and terminate on exchange lines, whereas broadband services include both retail and wholesale (Bitstream) services. The services comprise a broad variety; below is listed the most important of these. Post- och telestyrelsen 17

24 Table 1: List of standard PSTN/ISDN and Broadband services Core Access National calls PSTN Line Rental - FO 15 -internal calls ISDN 2 Line Rental - FO-external calls ISDN 30 Line Rental International calls Inbound Wholesale Line Rental** International calls Outbound Shared Access** International calls Transit Full Access** Fixed to mobile calls Other Access (including fibre and wireless technologies)** Mobile to fixed calls IN services Broadband Mass call services Internet dial-up (network product) Retail broadband Interconnection - double segment (double Bitstream access tandem) Interconnection - region segment (single tandem) Interconnection - metro segment Interconnection - local segment Interconnection - single transit Interconnection - double transit Interconnection - international call Operator Services Other calls ** Where the demand for these services has not been included in other categories, such as PSTN or ISDN line rentals. The purpose is to estimate demand for all access lines, avoiding double counting. Criterion CG 8 The models should include all standard PSTN/ISDN and Broadband services Leased lines To ensure that the volume information in the top-down and bottom-up models is consistent, the leased lines should be defined in a uniform way ensuring compatibility between the models. Leased lines may thus be classified in the three following groups: 15 Förmedlingsområde, or transit area. An FO internal call covers calls up to and including those using a single transit switch, whereas an FO external call covers calls using more than one transit switch. Post- och telestyrelsen 18

25 retail customers, usually requiring leased lines to provide a permanent connection between customer premises; other operators, usually requiring leased lines to provide a permanent connection between networks; network operators, requiring leased lines for internal use. In addition to this, SMP operators may carry data or other services over leased lines. Such services should be modelled and shown separately. Criterion CG 9 When dimensioning the network, the leased lines traffic volume should include leased lines provided to retail customers, to other operators and to the network operator itself. The models will not need to calculate the costs of leased lines explicitly. Leased lines should only be included for dimensioning the network and for ensuring that a fair amount of the costs shared with PSTN and Bitstream services are allocated to leased lines services Other services Demand for other services using the core and access networks should also be included to ensure that the core and access increments are dimensioned properly. Inclusion of this demand will allow a fair distribution of shared and common costs. Cable television, Virtual Private Networks (VPN) and packet-switching technologies such as frame relay are examples of these services. Criterion CG 10 Where possible, the models should categorise the other services into two major groups: one category comprising for example cable TV services, IP TV services and other services using their own nontelecommunications electronic equipment. one category comprising non broadband data services (by type) using the core network. It is not the purpose of the models to calculate the LRIC of these services, but only to ensure that a fair proportion of costs is attributed to these services. 3.2 Co-location services The nature of co-location services Co-location enables the placement and operation of telecom equipment in buildings housing technical plant. Co-location services provide access to the infrastructure of these buildings such as power supply, cooling, ventilation, security, and common amenities. Post- och telestyrelsen 19

26 From a modelling point of view, no network modelling is required in order to model co-location services. What has to be modelled is the cost incurred by producing the co-location services. Co-location may be relevant in relation to switched interconnection, access to unbundled local loop, or for other purposes. However, only the co-location services related to access to the unbundled local loop are to be priced on the basis of LRIC. Despite this, the co-location services to be modelled in the top-down and bottomup models may include more services than only the services which are to be costed according to LRIC principles. The reason behind the inclusion of more services is that some of the costs that make up the LRIC-specific co-location services are shared with other co-location services (i.e. co-location in relation to switched interconnection or "tele hotels"), services in the access network, and services in the core network. Examples of such shared costs may be the cost of accommodation, power supply, cooling/ventilation, and also the cost of administrative and technical staff. On the other hand, some co-location services include only co-location specific cost categories, such as racks and cables. In this case, only the services related to access to the local loop (that are to be costed according to LRIC principles) have to be modelled. Last, the costs of power consumption used by electronic equipment and cooling/ventilation can be attributed directly to relevant co-location services on the basis of the unit price paid to the power supplier List of co-location services covered by LRIC According to Skanova s co-location agreement, co-location includes the following services: Fee for tenders Location of equipment Transportation of equipment Installation and mounting of equipment Station wiring Placing Power, cooling and ventilation Demonstration of co-location node In Skanova s co-location agreement, a number of service variations are listed for each of these services. The variations are associated with different prices. The reason may be that different costs are incurred for the different service variations. The models should model the cost of all service variations set out in Appendix 4. Post- och telestyrelsen 20

27 3.3 Demand and growth Demand Although costs should be allocated to the total amount of traffic using the network, the network should be dimensioned to carry the traffic in the "busy hour" subject to the required QoS. The busy hour may vary between the different parts of the network. Criterion CG 11 The models should identify busy-hour information for traffic. The models should be flexible enough to allow for changes in these figures Margins for growth The modelled network should be able to meet the demand not only in the base year but also in the foreseeable future. It will therefore be necessary to develop forecasts for the development in demand. The network dimensioning should correspond to what an efficient operator facing these forecasts would do. Whereas margins for growth will typically be implicitly incorporated in the topdown model via the existing network, margins for growth will need to be modelled explicitly in the bottom-up model. The models may use different planning periods for different parts of the network. Forecasts for growth should be specified for each set of services. Criterion CG 12 The network dimensioning should correspond to what an efficient operator facing the forecasted demand would do. The models should show the anticipated Cumulative Annual Growth Rate (CAGR) for each service for a five year period, following the base year, The models should allow for a change in the margins for growth. The models should use the following planning horizons as a starting point: 5 years for the access network and infrastructure in the core network; 3 years for exchange equipment; 3 years for transmission equipment; 3 years for backhaul broadband equipment (core routers etc); and 1 year for DSLAMs. For line cards a 1-year planning horizon should be used as the starting point. If different planning horizons are used, this will need to be justified. Post- och telestyrelsen 21

28 4 Model outputs This chapter outlines the types of outputs that both models should provide and sets out the level of detail for which costs should be shown. The treatment of colocation, shared access and Bitstream access receives particular attention. It is important that the top-down and bottom-up models produce comparable results at different levels. The reason for this is twofold. First, in order to compare the cost of network elements and ultimately the cost of interconnection, access and co-location services; and second, in order to go beyond the network elements to compare variables such as utilisation rates, volumes, direct network costs and operating costs. If this objective is realised, it will lead to a more efficient reconciliation process. 4.1 Cost types Costs should be broken down into three types in terms of the way they relate to the network: direct network costs, such as processors, ports, multiplexers, duct, and fibre; indirect network costs, such as power, accommodation, network management and maintenance; and overheads, such as the personnel department. Network costs measure the costs of those inputs necessary for the network to run. They can be divided between direct and indirect network costs. A direct network cost is defined to be one where the level of inputs, and therefore the cost, depends on factors exogenous to the network, such as the level of demand. For example, the number of line cards, and therefore their total cost, will depend on the number of subscribers. In contrast, an indirect network cost is one where the level of inputs and hence cost depends on choices made concerning other inputs, and therefore only indirectly on external factors such as the level of demand. An example is racks, since the number and size of necessary racks will depend on the choices made concerning ports and line cards. The types of network costs included in the models will depend on the technology and configuration modelled. Therefore, it is not possible to provide a complete list. A top-down model can determine these costs explicitly as can the bottom-up model, although the bottom-up model may in some cases need to estimate them by using mark-ups. Overheads (also referred to as common business costs) cover those costs that are not necessary for running a network, but must nevertheless be incurred by the company running the network in order to function. 4.2 Level of detail Some aggregation of costs is desirable to make the models manageable, but this aggregation should be limited to ensure that the models provide a detailed Post- och telestyrelsen 22

29 breakdown of costs. This is important, since the reconciliation process requires the models to be transparent Cost categories The cost categories that fall under the heading of direct network costs should be sufficiently disaggregated so that each cost category has only one cost driver. For example, a local switch consists of both ports and processors, and thus, its costs depend on call minutes and call attempts. Therefore, there should be two cost categories, the costs of ports and the costs of processors, instead of a single cost category measuring the cost of local switches. Criterion CG 13 Cost categories should, as far as possible, be identified to obtain only one exogenous cost driver for each category. The models should identify operating costs and asset costs separately. Only those operating costs necessary to bring an asset into working for its intended use, such as transport, installation and commissioning should be capitalised. Other operating costs should be included in separate cost categories. Criterion CG 14 Costs related to assets can include capitalised operating costs when there is a rationale for it. These costs should be shown separately in the documentation. As stated in the previous section, it is difficult to provide a complete list of cost categories to be modelled without knowing exactly the technology and configuration modelled. 4.3 Cost of co-location Experience shows that the costs of co-location services produced in top-down and bottom-up models respectively may be difficult to reconcile. There may be three main reasons for this difficulty. First, cost categories may be different. Second, the models may apply different units of measurements, which are not easily convertible for reconciliation purposes. Finally, costs may be expensed in one model, whereas the same costs are annualised in the other model and vice versa. Reconciliation of co-location costs may be eased if modelling guidelines are more precise on these main issues the cost categories, units of measurement and whether costs should be expensed or annualised. Hence, this MRP provides detailed guidance on these issues. The cost categories which are regarded common to both co-location and other services in the core and access network are: Land and buildings (annual costs); Site preparation and fit-out of buildings (one-off and/or annual costs); Post- och telestyrelsen 23

30 Security systems, fire surveillance, etc. (one-off and/or annual costs); Power supply (annual costs); and Cooling/ventilation (annual costs). The above cost categories, which are common to both co-location and other services, may be distributed among the relevant services once the costs are estimated. Costs should be distributed according to an appropriate cost driver. For the cost categories that relate to space, an appropriate driver will be the number of square meters occupied by the equipment associated with particular services. For the cost categories that relate to power, an appropriate key may be the power and cooling/ventilation requirement measured in Watts. The power and cooling/ventilation requirement may be too difficult to estimate, however. A proxy could be the number of square meters occupied by the equipment associated with particular services. Where appropriate, costs in the common cost categories should be allocated between services according to square meters occupied. The SMP operator should co-ordinate with PTS to ensure that the common cost categories are allocated using the same principle in both models. The cost categories which are regarded as being specific to the co-location services related to unbundled local loop - and therefore should be costed according to LRIC principles - are: Administrative staff (one-off costs and annual costs); Technical staff (one-off costs); Racks ( ETSI-skåp ) (one-off and annual costs); Co-location specific power supply inclusive power consumption (one off and annual costs); Co-location specific cooling/ventilation (one-off and annual costs); and Cables (one-off and annual costs). Note that power supply and cooling ventilation are included in the co-location specific cost categories as well as the common cost categories. The reason is that some of the costs associated with power supply and co-location are only incurred for specific co-location agreements. Criterion CG 15 The modelled co-location services should include the following cost categories common to both co-location and other services in the core and access network: Land and buildings (annual costs); Site preparation and fit-out of buildings (one-off and/or annual costs); Security systems, fire surveillance, etc. (one-off and/or annual costs); Power supply (annual costs); and Post- och telestyrelsen 24

31 Cooling/ventilation (annual costs). The specific co-location services to be costed include: Administrative staff (one-off costs and annual costs); Technical staff (one-off costs); Racks ( ETSI-skåp ) (one-off and annual costs); Co-location specific power supply inclusive power consumption (one off and annual costs); Co-location specific cooling/ventilation (one-off and annual costs); and Cables (one-off and annual costs). Each cost category should include one-off and annual costs as shown above. 4.4 Costs of shared access to unbundled local loop Shared access to the unbundled local loop implies that the other operator gains access to the high frequency band of the copper pair whereas the SMP operator continues to use the lower frequency band for providing PSTN services 16. A large amount of the costs, in particular the costs of the copper pair, will be shared between the PSTN and shared access service. It is important that the models distinguish between these shared costs and those which are directly attributable to the two services. Criterion CG 16 The models should distinguish between the costs that are specific to PSTN services, costs that are specific to shared access, and the costs that are shared between the PSTN services and the shared access service. The additional cost of shared access (compared to PSTN) should be shown as a separate output of the models. Shared access, full access, PSTN subscription and Bitstream access (discussed below), are naturally closely linked together. It is important to ensure consistency between the included network element costs. The total annualised cost of the actual copper pair might be considered to be the same whether it is used for providing PSTN services, full access, shared access or PSTN services plus Bitstream access. On the other hand, a case might be made for separating out copper pairs that are too long for broadband or those on nodes that do not contain a DSLAM from the remainder of the copper pairs. How the shared costs are allocated between the services for the final price setting is a different issue. Costs that are specifically caused by full access, shared access 16 In principle, the reverse situation could also occur. Post- och telestyrelsen 25

32 or Bitstream access, such as frequency planning, should be allocated to these services. Raw copper used for xdsl and other services may need to be of a better quality than copper used for PSTN services. Where this is explicitly considered in the models, other differences such as a shorter average loop length would also need to be taken into account Criterion CG 17. The total annualised cost of the actual (raw) copper pair should as a starting point be the same whether it is used for providing PSTN services, full access, shared access or Bitstream access plus PSTN services. If it is decided to cost Bitstream capable copper pairs separately from non-bitstream capable copper pairs then this must be justified and documented. 4.5 Cost of Bitstream access The models should be capable of calculating the cost of Bitstream access. Bitstream access allows other operators to provide Internet access to end-users by using ADSL transmission on the existing copper pair used for PSTN services. Instead of installing their own equipment, operators use the existing DSLAM equipment of the incumbent (SMP) operator. The Bitstream access service will normally include: Lease of capacity on the copper (similar to shared access); Lease of capacity in the SMP operator's DSLAM; Transport of traffic from the DSLAM to the nearest point in the SMP operator's network available for Bitstream interconnect. The service might also include installation of the ADSL filter at the customer's premises. In order to use the service, the other operator might also need to have a separate agreement regarding connectivity between the first available interconnect point and the other operator s required interconnect point (e.g. using a Virtual Private Circuit or VLAN). Figure 3 provides a schematic illustration of Bitstream access. The figure is only included for illustrative purposes and relates to ATMbased DSLAMs rather than those with Ethernet-based backhaul. Post- och telestyrelsen 26

33 Figure 3 Bitstream access Customer premises Other operator premises ADSL modem ADSL filter NTP Line Leased line ATM VPC ATM network POI PSTN DSLAM Data Exchange building Splitter RSS/HSS MDF PSTN Criterion CG 18 When modelling the cost of both shared and full Bitstream access, the models should be able to distinguish between the costs of: Capacity on the copper; capacity in the DSLAM; and transport of traffic from the DSLAM to the nearest point in the SMP operator's network available for Bitstream interconnect. The costs of the modem at the customer's premises, and any costs related to installation of a filter at the customer s premises, should not be included in the cost of Bitstream access and, if included within the modelling, should be shown separately. The additional cost of Bitstream access (compared to PSTN) should be shown as a separate output of the models. Post- och telestyrelsen 27

34 5 General costing issues This chapter discusses a number of general costing issues relevant for both models, such as annualisation, cost of capital, geographical differentiation of costs, working capital, base year, distinction between set-up and duration related costs and the use of routing factors. 5.1 Annualisation methodologies Annualisation criteria Three criteria should guide decisions on the appropriate approach to modelling the annualisation charges for the various assets: Accuracy; Consistency; and Tractability. An accurate annualisation charge should have a profile which reflects the expected levels and changes in replacement costs, obsolescence, operating costs, output levels, asset productivity, the cost of capital and the asset life. Consistency requires that annualisation charges should be set in such a way that there are no arbitrage opportunities available for purchasing assets at certain stages of their lives. Where the output produced by the asset is constant, consistency requires that the sum of the annualisation charge and operating costs of an asset purchased in Year t will be the same in Year t+1 as if the asset had been purchased in Year t+1. Tractability means that there is sufficient information to calculate the annualisation charge using the chosen approach. The annualisation charge consists either of a capital charge and a depreciation charge or alternatively a combined annuity charge Capital charge The capital charge is simply the cost of capital (the required rate of return on capital) multiplied by the average value of the asset for the year being reviewed. It reflects the costs of having capital tied up in the fixed assets, thereby not being able to use it for alternative purposes. Assuming the cost of capital is 10 per cent, the capital charge for an asset worth 1000 at the start of the year and 900 at the end of the year would be Depreciation charge The depreciation charge should reflect the change in the asset value during the year being reviewed. The depreciation can be determined using a number of different methodologies: Economic depreciation; Straight-line depreciation; or Post- och telestyrelsen 28

35 Sum of year's digits. These are discussed in more detail below: Economic depreciation Economic depreciation measures the change in an asset s economic value. It can be calculated as the estimated Net Present Value (NPV) of future cash flows at the start of a given year less the estimated NPV of future cash flows at the end of the year. The depreciation profile will depend on a number of factors such as, in particular, the expected development in annual operating costs, the revenue generated by the asset and the asset s price (acquisition cost). Economic depreciation is in practice very difficult to calculate. First of all the calculation of economic depreciation is information intensive. For example, it will be necessary to form a view on what technological advances might take place, and how these advances might affect an asset s value. There is also likely to be uncertainty concerning future demand for the output of the asset and the future operating costs of the asset. It may also be difficult to assign revenues to particular assets, as they often have complementary roles. Finally, changes in operating costs and prices over time will change the economic lifetime of the asset over time. In a regulatory context, there is also a risk of making a circular argument, as the calculation of economic depreciation depends on the expected development in revenue which in turn depends on the calculated depreciation charge included in the regulated charges Straight-line depreciation Straight-line depreciation divides the asset s price by the asset s life to produce an annual depreciation charge. To calculate the annualisation charge, a capital charge is added. The straight-line depreciation charge will typically be higher than the economic depreciation charge in the early years of an asset s lifetime except where operating costs are rising very rapidly or output levels produced by an asset decrease rapidly as the asset becomes older. A further related limitation with this approach is that the annualisation charge depends on the vintage of the assets being considered Sum of year digits The sum of year digits (SOYD) is a simple method for generating a front-loaded depreciation schedule. It may be a useful approximation if the asset s operating costs are expected to rise or its price or the revenue it generates is expected to fall. For each year, the depreciation charge s share of the gross asset value is determined as (remaining lifetime+1) divided by the sum of lifetime years. It is sometimes argued that sum of digits depreciation provides a reasonable proxy to economic depreciation. However, this is only likely to be the case where the operating costs for an asset are rising rapidly or the output produced by the asset Post- och telestyrelsen 29

36 is falling rapidly. Estimates suggest that a combination of these factors would be needed to justify this methodology Annuities The annuity approach calculates a single charge that replaces the depreciation charge and the capital charge. It should achieve the criteria of consistency and tractability outlined in the introduction to this chapter. A standard annuity calculates the charge that, after discounting, recovers the asset s purchase price and financing costs in equal annual sums. If the price of the asset is expected to change over time, a tilted annuity would be more appropriate. A tilted annuity calculates an annuity charge that changes between years at the same rate as the price of the asset is expected to change. This results in declining annualisation charges if prices are expected to fall over time. As with a standard annuity, the tilted annuity should still result in charges that, after discounting, recover the asset s purchase price and financing costs 17. The (tilted) annuity charge is estimated according to the following formula: r p I t 1 + p r Where: r = cost of capital p = rate of price change ("tilt") t = asset lifetime I = investment Guidelines Since economic depreciation is in practice very difficult to calculate, the two models should not use that approach. Instead, the models should apply one of the other approaches. These focus on recovering the replacement cost, rather than the economic value of the asset. The most important factors that will inform the decision on the appropriate approach are: expected movements in the asset s price; the expected revenue the asset will generate; the asset s expected annual operating costs. 17 As regards the top-down model, the introduction of a tilt, where asset prices are changing, means that the annuity method is consistent with the FCM approach discussed in section 7.5. Post- och telestyrelsen 30

37 If the asset s expected price or the revenues it generates is expected to fall, or its operating costs expected to rise, any depreciation schedule should be front loaded with larger charges in the early years. It should be noted that top-down models are based on multiple vintages of assets, whereas bottom-up models are based on a single, new vintage of assets. The existence of multiple vintages of assets can dampen out the differences between different approaches. Using straight-line depreciation in the top-down model has the advantage that it is in line with the depreciation principles used for accounting purposes. This increases the objectivity and reduces the problem of windfall gains and losses arising from a change in the depreciation principle. The starting point for the topdown model should therefore be straight-line depreciation. For the bottom-up model, the starting point should be to use (tilted) annuities. The annuity approach has the advantages that the annualisation charge is independent of the age of the asset. The fact that the bottom-up model is (artificially) modelling new assets therefore becomes less of an issue 18. Criterion CG 19 The starting point for the top-down model should be straight-line depreciation, whereas the starting point for the bottom-up model should be (tilted) annuities. 5.2 Cost of capital The cost of capital measures the opportunity costs of the sources of capital (debt and equity) invested in the company (the SMP operator). PTS believes that an estimated nominal pre-tax cost of capital of 10.8% currently represents the best available estimate of the cost of capital of a Swedish SMP operator. Therefore, this estimate should be used as an interim cost of capital in the two models. However, it should be possible to change this value in the models. Prior to the final price setting, PTS will evaluate whether the estimated parameter values, and hence cost of capital, are still appropriate. Criterion CG 20 The models should use an interim nominal pre-tax cost of capital of 10.8%. The models should allow the cost of capital to be altered. 5.3 Geographical differentiation Geographical differentiation The average subscriber related costs are likely to differ significantly between urban and rural areas. Factors driving these differences are: 18 Otherwise, it could be argued that the bottom-up model should not model the costs in year 1 but rather in year 3 or 5 for example. Post- och telestyrelsen 31

38 Trenching and ducting costs, which tend to be higher in urban areas; Distances between the exchange and the customer which tend to be shorter in urban areas; and Cable size (number of pairs), which tends to be larger in urban areas. The overall impact of these factors can be determined by calculating access costs separately for different types of areas, so-called geo-types. Whereas costs may vary substantially in the access network, cost differences are likely to be much smaller in the core network. Moreover, it would be difficult to demarcate the areas, as the core network as opposed to the access network will pass through multiple geotypes. The model should therefore only distinguish between different geo-types in the access network Geo-types For access services, the model should provide separate costs by geo-type along with the average national costs. Criterion CG 21 For access services, the models should provide separate costs by geotype along with the average national costs. 5.4 Other costing issues Base year The base year, the year to which all data should be related, is If data for one reason or another is not available for that specific year, an extrapolation should be made from relevant historic data to calculate the proper reference data for the base year 20. Criterion CG 22 The models should model the costs for Working capital There is typically a delay between paying out cash for inputs and receiving cash for outputs. For that reason, working capital is required at the beginning of trading to be able to cope with the delay that arises from normal business activities. Once the investment in working capital is made, this cash is tied up in the running of the business until trading ceases. Therefore, there is an opportunity cost as this cash could be invested elsewhere. Working capital is calculated as the current assets less current liabilities: 19 For core, the models should naturally distinguish between local and national calls. However, that has nothing to do with geographical differentiation of costs. 20 This does not necessarily imply that the regulated prices will be based on the costs of the base year. Post- och telestyrelsen 32

39 Working capital = Inventories + Receivables + Cash Payables The cost of working capital is then calculated by multiplying the working capital by the cost of capital. Working capital may also include stock of network spare parts, capital work in construction, i.e. assets not yet being activated, and other current assets and liabilities. Criterion CG 23 The models should include a calculation for the cost of the working capital of an efficient operator, unless zero is used in which case it can be shown as an input Set-up and duration related charges The PSTN interconnection charges are made up of set-up and duration related costs. This reflects the fact that certain costs are caused by a call attempt whereas others are caused by the duration of the call. For instance, whereas the costs related to processing and signalling are driven by the number of call attempts, the costs of exchange ports and transmission costs are driven by the number of minutes (in the busy hour). The models should be able to reflect this structure and should therefore, where relevant, distinguish between per call network elements and per minute network elements. For example, the models will need to distinguish between the costs of using the network element local exchange for a call and for a minute. It will therefore also be necessary to develop routing factors for both calls and minutes. IN costs should be separately identified from other signalling related costs. This will imply defining separate IN network element(s) and corresponding routing factors. Criterion CG 24 The models should distinguish between set-up related and duration related costs. This requires the calculation of both set-up and duration related costs, network elements and routing factors. IN costs should be separately identified from other signalling related costs. This will imply defining separate IN network element(s) and corresponding routing factors Routing factors Routing factors specify, for each type of service, the average use made of each type of network element. Each service therefore has a routing profile indicating Post- och telestyrelsen 33

40 how the service uses the network elements (distinguishing between the different types of exchange and the different parts of the transport network 21 ). In bottom-up models, routing factors are used both to dimension the network and to cost the services, whereas in a top-down model, routing factors are normally only used to cost the services. Having calculated the annual cost of each network element, service costs may be calculated using routing factors. These have to be applied twice: First to calculate the total cost of using a network element for one minute or one call (the costs per element stage) Second to calculate the service cost by multiplying the cost per element stage by the routing factor (the number of element stages used). Network elements may be specific transmission links or different types of exchanges. In the case of transmission links, element stages can be measured in minutes; in the case of exchanges, these are measured in minutes and busy hour call attempts. Network elements will form the building blocks from which services will be costed. The cost per element stage is simply the total annual cost of the element divided by the number of element stages. The number of element stages can be calculated by multiplying the number of call minutes (busy hour call attempts) by type of service by the average number of routing stages for that type of service and then adding up results per element stage. For example, if there are 1,000 call minutes of a particular type and on average these pass through 2 local exchanges (routing factor = 2), these minutes will give rise to 2,000 local exchange element stage minutes. Assuming that adding up the results for local exchange gives 10,000 local exchange element stage minutes and the annual cost of a local exchange is 40,000 the cost per element stage is (40,000 divided by 10,000) 4 per minute. The cost of services can be calculated by multiplying routing factors of the service for each of the elements used by the cost per element stage and adding up the results. For example, a double transit call might use the following mix of elements as indicated by the routing factors in the table below, where the cost per element stage are also shown. Table 2: Example: calculating the service costs of double transit. Network element RSS/HSS RSS-LE LE LE-TE TE TE-TE Routing factors (double transit) 2 1.2* Cost per element stage Service cost per element Service cost = 30.2 (per minute or call) * this is less than two since some customers are connected to host switching systems 21 The examples in this section are PSTN based, although the principles apply equally to other services such as Bitstream. Post- och telestyrelsen 34

41 By multiplying the routing factors for the double transit service with the relevant element stage cost, the service cost per element is calculated. To arrive at the service cost, each of the individual element costs are added, equalling 30.2 in the example above. It will be necessary to show that routing factors are efficient. One way of investigating this is to collect data on routing profiles (the percentage of calls following particular routings). Criterion CG 25 The models should show routing factors for (at least) each of the following PSTN related network elements Concentrator (RSS or HSS); Local Exchange (LE); Tandem Exchange (TE); RSS-LE transmission; LE-LE transmission; LE-TE transmission; and TE-TE transmission. Information should be provided separately for all the major call types. In addition, the documentation should include information (for all calls) showing the percentage of calls following a particular routing pattern (e.g. 2 RSSs, 1 local exchange, 2 RSS-LE transmission links). Routing factors for Bitstream related network elements should also be provided as necessary. The list of network elements above should not be considered a complete list. Separate routing factors will e.g. be required for calls and minutes. For some calls it will also be necessary to include an IN network element. As with interconnection services, the costs of the different access services may be derived using access routing factors that specify the intensity with which each service uses each network element. The access network elements should be chosen in order to facilitate an accurate allocation of costs to services, whilst keeping the number of network elements manageable. However, as a minimum access network elements should be split into geo-types where appropriate. Examples of network elements that typically make up the access network are: Copper pair; Fibre pair; PSTN electronics; ISDN electronics; Other electronics; and Post- och telestyrelsen 35

42 Line card. Again, this should not be considered a detailed list. The copper pair will e.g. be made up of several sub-elements such as MDF, jumpering, cable in the primary access network, distribution point(s), cable in the secondary access network, and final drop cable. The models should separately identify the costs related to these network components. Whether they are modelled as specific network elements or treated as cost categories is left to the modelling. However, the models should also show the aggregate costs of the copper line to facilitate comparison between the models. Criterion CG 26 Access network elements should be split into geo-types where appropriate and "routing factors" (usage factors) specified for each service to be costed. Post- och telestyrelsen 36

43 PART B: SPECIFIC GUIDELINES FOR TOP-DOWN MODEL Post- och telestyrelsen 37

44 6 Overview of top-down modelling The procedure to build a top-down model can be summarised in five steps illustrated in Figure 4 Figure 4: Simplified overview of the steps in building a LRIC top-down model Step 1: Group into homogenous cost categories Fixed Assets Net Current Assets Operating Costs Depreciation Cost Categories (1,2,...N) Step 2: Group Cost Category by Activity and Network Elements Core Elements Access Elements Retail and Other Activities Common Costs Step 3: Re-value Fixed Assets and Calculate CCA Depreciation Develop Gross Replacement Costs Calculate NRC and CCA Depreciation Annualised capital costs and operating costs per cost category for each increment Apply the cost of capital Annualised capital costs and operating costs Step 4: Develop Cost-Volume Relationships (CVRs) Cost categories allocated to each service (PSTN, leased lines and other services) Apply CVRs PSTN s share of cost categories allocated into network elements Step 5: Cost interconnection and access services Deriving of unit costs (e.g. per minute) for each network element Apply routing factors Mark-up Aggregation of network elements to derive LRIC costs for I/C and access products I/C and access charges 6.1 Determine homogenous cost categories (Step 1) As illustrated in Figure 4, the first step is to group costs that have similar characteristics into individual cost categories, also called homogenous cost Post- och telestyrelsen 38

45 categories. The level of homogeneity is determined by the need to identify individual cost drivers and to account for changes in costs over time. The cost driver should explain the costs of a particular activity and should be quantifiable call minutes and number of calls are examples of readily quantified cost drivers. A further consideration is that the cost driver should be measurable in a way that enables it to be identified with individual products or services. For example at the exchanges, the two cost drivers in the core network are call duration (e.g. exchange ports) and call attempts (e.g. processing capacity). These components the cost of an exchange port and the cost of processing capacity need to be separately costed. Since network costs have to be re-valued using Current Cost Accounting (CCA) discussed below - when developing a LRIC model, costs also have to be disaggregated in a way that allows different cost trends to be tracked. 6.2 Group cost category by activity and network elements (Step 2) Once the homogeneous cost categories have been identified, the next step is to determine the activities using the cost categories and to attribute them to different network elements. It is important to define network elements and determine, at an operational level, mappings of network elements to interconnection services. This is not a straightforward process and should be done with as much detail as the accounting system will allow. Whereas the quantity of network equipment has a direct relationship with network elements, many classes of asset and operating costs cannot be directly attributed to either services or network elements. The fundamental requirement is to develop a model which reasonably reflects the complex set of relationships between some of these classes of cost and ultimate outputs. 6.3 Re-value assets, calculate GRC, NRC and CCA depreciation (Step 3) Since LRIC is a forward-looking concept, current cost accounting (CCA) principles have to be used to determine the appropriate net asset value of assets and associated depreciation charges. This involves re-valuing assets on the basis of the replacement cost of the modern equivalent asset (MEA). The MEA may be defined as one with the required capacity and functionality that, summing for all future years, has the lowest (discounted) cost. If there are differences in operating costs between the MEA and the existing asset, the MEA valuation of the existing asset must also be adjusted to reflect these, or a specific operating cost adjustment made. The differences may arise, for example, due to differences in maintenance costs, network management costs and associated indirect costs. Finally, when selecting the MEA, differences in asset lives should also be taken into account. Having established the gross asset value, the net asset value can be found as the gross asset value net of accumulated depreciation. The net asset value represents Post- och telestyrelsen 39

46 the capital tied up in network assets, valued on a current cost basis. If gross asset values are estimated on the basis of replacement costs, this net asset value is referred to as the Net Replacement Cost (NRC). Capital costs can then be found by multiplying the average net asset valuation for the year with the cost of capital. 6.4 Developing cost-volume relationships (Step 4) The penultimate step in the development of a top-down model is to derive costvolume relationships (CVR). In particular, CVRs: trace how individual costs vary with underlying cost drivers; and identify all variable, fixed, common and joint costs. In simple terms, a CVR is a curve which describes the relationship between the volume of a cost driver and its related costs (see illustration). The two key characteristics of cost volume relationships are the gradient of the curve describing the marginal cost for each value of the cost driver volume and the intercept with cost axis describing the fixed (common) costs. Figure 5: Illustration of different cost-volume relationships Cost Cost Cost Volume Volume Volume (a) (b) (c) The cost-volume relationships will depend on the cost under consideration. In figure a, there are no fixed or common costs and the straight CVR curve reflects constant returns to scale. In the telecom sector, a cost category will often display economies of scale and/or economies of scope. In figure b, there are no fixed or common costs, but the CVR curve now reflects economies of scale. Finally, figure c illustrates the CVR curve for a cost category with fixed or common costs and economies of scale. In some cases, the relationships may be based on existing engineering models or based on simulations undertaken by engineers. In other cases, the estimates may be based on regression analysis or examinations of the processes underlying particular activities by conducting interviews and field research. 6.5 Cost access and interconnection services (Step 5) The final step in the process is to develop a model that incorporates the various steps outlined above. The model is likely to be large and complex given the large number of cost categories that may need examining. This is because a cost Post- och telestyrelsen 40

47 category may have a number of cost drivers and because the potentially complex relationships between cost categories and end services. All costs within the LRIC model will be directly or indirectly related to the volume of output of the increments. Certain costs are directly related to those volumes whereas others will only have an indirect relationship mediated through other intermediate cost drivers. The method for calculating the incremental cost category, however, is always to: identify the cost driver volume associated with the increment; derive the cost driver volume of the particular cost category; and finally calculate the associated cost of adding the cost category to the increment. The output of the LRIC model is long run incremental costs for each cost category and for each main increment and per unit costs for the access and interconnection services. As set out in section 5.4.4, routing factors are applied to calculate the costs of using the individual network elements. The costs of services are then found by multiplying the routing factors of the individual services by the costs of the individual network elements. Finally, a mark-up may be included to allow the operator to also recover common costs. Post- och telestyrelsen 41

48 7 Top-down asset valuation and capital costs 7.1 Gross asset valuation Current cost accounting When valuing assets using CCA, there are three alternative valuation methodologies: Replacement Cost (RC) measures the costs of replacing the existing asset with another asset of similar performance characteristics; Net Realisable Value (NRV) is the amount which would be obtained by selling an asset when sale costs are deducted; and Economic Value (EV) is a measure of the value of an asset based on the sum of discounted cash flows which an asset is expected to generate during its remaining lifetime. One widely applied criterion for selecting between valuation methodologies is known as the Value to the Owner Convention. Under this convention, the methodology for measuring current cost is as follows: Current Cost = min [replacement cost, max (NRV, EV)] If the EV were greater than the NRV, the operator would keep the asset in its current use. If the NRV on the other hand were greater than the EV, the operator would sell the asset now, as the proceeds from the sale would exceed the economic value that it would be expected to generate from its continued use. Therefore, the deprival value or recoverable amount of the asset is the higher of EV and NRV. The current cost is therefore the lower of its deprival value and the net replacement cost. That is, the lower of the amounts that the operator could recover from the asset and the costs of replacing the asset with an identical one. The formula above suggests that the choice of valuation method for current cost accounting should be determined on a case-by-case basis. However, traditionally, top-down cost models have tended to use the replacement cost methodology to calculate the current cost of most assets as the replacement costs. Replacement costs are easier to estimate and easier for other parties to verify. To calculate the net asset value, it is necessary to first estimate the gross asset value. The gross asset value is the value of a new (non-depreciated) asset. If asset values are estimated on the basis of replacement costs, the gross asset value is referred to as the Gross Replacement Costs (GRC). Criterion TD 1 Asset valuation should reflect the replacement costs of the modern equivalent asset (MEA) The MEA is discussed in more detail below. Post- och telestyrelsen 42

49 Criterion TD 2 The MEA is the asset that produces the same outputs produced by the existing asset at lowest costs Revaluation methods In many cases, the modern equivalent asset will be the same as the existing assets. For example, the modern equivalent asset for trenching is likely to be trenching 22. There are several different methods for the valuation of assets on a current cost basis: Historic cost of an asset could be used as a proxy for the current cost of an asset where it is unlikely that the use of historic cost would give a materially different value compared to the value obtained with current costs. This is the case where the asset is either of low value, where the asset life is relatively short or for additions during the year. Indexation (where the historic cost/book value is multiplied by a relevant price index) could be used for assets where there has been very little technological change and all direct costs that have been incurred and capitalised to date would have to be incurred if the asset were replaced today. Absolute valuation multiplies the physical quantities of assets with the current unit prices. This methodology should always be used where there has been technological change. In this case, the replacement cost is based on the cost of a modern equivalent asset (MEA) with similar service potential. Absolute valuation is preferable to the alternative of indexing historic cost valuations, since: Assets may consist of a number of different cost categories each requiring separate indices. The importance of the various cost categories is likely to change over time; Indexation can result in duplicated costs, for example, where a trench is redug to install more cable. Further, some assets in the asset register may no longer be required; A true replacement cost valuation requires an inventory of equipment actually provided whereas an indexed valuation does not. This means that it is much easier to reconcile top-down asset values based on replacement cost than those based on indexation. Criterion TD 3 The top-down model should value assets on the basis of an absolute valuation. Any use of indexation will need to be justified by supporting 22 The MEA may not be the same type of trenching, though. Trenching laid today might for example use different digging depths and different qualities of surface reinstatement than was used when the existing trenching were laid. Post- och telestyrelsen 43

50 documentation and should only be used where there has been no technological change. Where the difference between the current and historic cost of the asset is likely to be small relative to the overall gross asset valuation, or the asset life is short (3 years or less), the top-down model may use historic costs. No more than 5% of the total value of the assets may be valued according to historic costs. The model documentation should justify the decision to use historic costs Asset prices Asset prices may be sensitive to the quantities purchased. Asset prices should generally be based on recent contracts. However, the SMP operator should include all reasonable volume and trade discounts it would expect to receive for an average purchase. For example, recent additions to the trench network may involve the addition of small amounts of trench on different routes. The associated per kilometre cost may be significantly higher than if much larger quantities were added. It would therefore be misleading to use such digging prices for the entire network. Criterion TD 4 The SMP operator should have available for inspection documentary evidence of asset prices used within the model New technology In many cases, new technologies may have been developed since the existing asset was installed. It may be that the existing assets can or would no longer be purchased. Provided the new technologies can perform the functions carried out by the existing asset (with the same or better quality), the modern equivalent asset (MEA) may therefore be an asset using the new technology. In principle, this should be independent of whether or not the SMP operator has plans to replace the existing technology. Examples of the above include PDH equipment, which would normally be valued on the basis of SDH equipment, and co-axial cable, which would normally be valued on the basis of optical fibre. If the existing asset has excess capacity, there is no requirement for the MEA to also have excess capacity; conversely, in cases where the cheapest replacement provides additional functionality or capacity, this should still be the basis for the MEA MEA adjustment If the MEA implies material differences in operating costs, quality, asset lives or space requirements, the asset value should be adjusted to account for the effect these differences will have on the net revenues generated by the asset for each year of the asset s lifetime. For material differences in operating costs, the Post- och telestyrelsen 44

51 adjustment can either be applied to the asset value or as a separate specific adjustment to the operating costs for those assets. Where the MEA is superior (inferior) to the existing asset, these differences should be discounted by the cost of capital and subtracted from (added to) the asset valuation. This is necessary to ensure that other operators do not have to pay interconnection costs for MEA assets that have a different level of functionality than those of the SMP operator. Differences in operating costs may arise from differences in maintenance cost, network management costs and associated indirect costs. If the differences are to be applied to the asset value (as opposed to being included as an operating cost adjustment), then such differences should, in principle, be estimated for each year of the asset s life. The differences should then be discounted by the relevant cost of capital and summed. Where the MEA is more expensive to operate than the existing asset, the resultant sum should be added to the valuation; where, as is more likely, the MEA is cheaper to operate than the existing asset, the resultant sum should be subtracted from the valuation. Space requirements may differ between technologies. Any such differences should be quantified and multiplied by an appropriate cost of space per unit. If the MEA requires less (more) space, the cost associated with the space difference should be subtracted from (added to) the MEA s valuation. Finally, when selecting the MEA, the model should take account of differences in asset lives. For example, the MEA should be the one with the required capacity and functionality that, summing for all future years, has the lowest (discounted) net replacement cost. Later chapters provide insights on how to estimate depreciation charges and calculate an asset s net replacement cost. Criterion TD 5 If the MEA implies differences in operating costs, quality, asset lives or space requirements, the asset value should be adjusted to account for the effect these differences will have on the net revenues generated by the asset for each year of the asset s lifetime. For material differences in operating costs, the adjustment can either be applied to the asset value or shown as a separate specific adjustment to the operating costs for those assets. Where the replacement cost of the new asset is higher than the existing asset, the operator should document (through the MEA adjustment) that the total current cost associated with the new asset is lower than the total current cost of the existing asset. Following the MEA adjustment, the replacement cost of the modern equivalent assets should always be equal to or lower than the replacement cost of the existing asset (provided that the existing asset can still be purchased) This has nothing to do with the relationship between historic costs and current costs. The current cost of an asset may well be higher than the historic cost. Post- och telestyrelsen 45

52 7.1.5 Land and buildings Buildings may be treated in two ways in the top-down model, as capital costs or operating costs Land and buildings treated as capital costs If the SMP operator owns the building, buildings should be treated as a capital costs like any other asset. In principle, buildings should be valued at their market value. To increase objectivity and transparency, however, the SMP operator should, where possible, value buildings on the basis of an objective measure such as the taxation value, adjusted as appropriate to consider any systematic difference between the market value and this objective measure. Land and buildings assets fall into two classes: Specialised and general-purpose buildings. Specialised land and buildings house equipment specific to telecommunications, such as exchanges and transmission equipment. Generalpurpose buildings include office buildings, stores etc. Some buildings will be used for both specialised and general purposes. For specialised land and buildings, the market value may need to be adjusted. For example, there will be vacant space in many exchange buildings, often reflecting the fact that they were built to accommodate analogue switching equipment, which has a larger footprint than the equivalent digital equipment, now used throughout the network. In this case, the surplus space could be used by other parts of the company s business or, in some cases, be rented to another business. There is no justification for including the cost of such excess space as a cost of access or interconnection. However, there may be a case for some spare capacity in terms of space, where the provision of additional space represents an economically sensible contingency, e.g. due to future demand for co-location space, or to provide adequate space for technology upgrades where there will be an element of parallel running of the legacy and new technology platforms. In order to produce the valuation of buildings consistent with that required by an efficient operator, the value should be scaled down by the ratio of inefficient vacant to total space: vacant space CCA building value = market value of building 1 - total space Inefficient vacant space should be defined as the difference between actual space and the space that would be required for a 2-3 year planning horizon. Criterion TD 6 No value should be attached to vacant space, except where it may be shown that it is economically rational to maintain this vacant space using a 2-3 year planning horizon. The amount of vacant space that has been excluded should be made clear either within the model itself or within the associated documentation. Another adjustment that may be necessary relates to the specialised costs that may have to be incurred, but for which the market valuation makes no allowance. For Post- och telestyrelsen 46

53 example, the cost of specially protected floors for exchange equipment should be included in the calculation of incremental costs (provided that the costs were necessary and efficiently incurred). Such costs could either show up as a separate cost category or as an adjustment to the asset value of the building Land and buildings treated as operating costs Buildings may also be owned by a separate company or subsidiary and leased back to the network division. In this case, building costs may also be treated as an operating cost (rent). The rent should then reflect the market value for rental space. The costs of the sub-company, running the exchange buildings, should obviously be excluded, as these costs are already included in the rent. Where the buildings are owned by a company with different owners than the SMP operator, and rented to the SMP operator on market terms, these rents may be used as accommodation costs in the top-down model. If the buildings are owned by a company with the same owners as the SMP operators' network division, the (internal) transfer prices may be used as a proxy for the market based rent, only if the SMP operator can justify that these transfer prices reflect the market based rent. The justification would typically include a comparison between the transfer prices and estimated costs calculated on the basis of the building value. This would need to be done for a sample of buildings. Again, the costs of inefficient vacant space should not be included. Criterion TD 7 Where land and building costs are treated as operating costs, it should be justified that the rent corresponds to a market-based rent. Again, the costs of inefficient vacant space should not be included. 7.2 Utilisation rates In some cases, the SMP operator s utilisation level of equipment may be too low. 24 It is not possible to specify efficient utilisation levels in advance and indeed, the level of utilisation is likely to vary by part of the network. 25 For example, erlangs per circuit utilisation levels might be justified on the basis of a variety of factors. Modularity the fact that equipment may not always be purchased in the required size but has to be purchased in fixed pre-defined sizes. This factor is likely to be a significant constraint on achieving high utilisation rates, such as for example high erlangs per circuit utilisation on thin routes, such as between the RSS and the local exchange. 24 Utilisation level measures will vary by part of the network but include erlangs per circuit, utilisation of transmission equipment, line cards, copper cable, fibre, distribution points, etc. 25 For example, erlangs per circuit utilisation is likely to be much higher on busy routes, such as between local exchanges and tandem exchanges than on less heavily used routes, such as between RSSs and local exchanges. Likewise, the utilisation level for copper cable in the access network (pairs in use as a percentage of pairs in the ground) will drop the further one moves away from the MDF towards the subscriber. This is due to modularity as well as the need to allow for growth. Post- och telestyrelsen 47

54 Growth requirements it is reasonable to install sufficient capacity to allow for anticipated growth over a certain period. The SMP operator will need to provide evidence that the period used for dimensioning purposes is reasonable and that estimated growth during this period is reasonable. Also, it should show how the network takes account of the frequency of customers moving, changes in demographics and changes in interconnection traffic. Acceptable blocking levels the SMP operator needs to ensure that the number of calls that are blocked during busy hour is at an acceptable level. Traffic surges the amount of traffic carried in busy hour is likely to vary by month of year and by day of week. The operator will need to provide evidence of traffic patterns over the course of the week and by time of year. Network resilience local exchanges, for example, will often be linked to two or more tandem exchanges, implying that if a particular tandem exchange goes down, part or all of the traffic from the local exchange may be carried by another tandem exchange. This will have implications for exchange dimensioning. The operator should provide: details on the maximum utilisation reached before additional capacity is added the rationale for the chosen maximum utilisation rate the actual utilisation levels on a selection of items of equipment to show the relationship between the average and maximum utilisation levels; and justification for excess capacity based on rational economic considerations taking into account modularity and margins for growth. Criterion TD 8 An SMP operator should show the utilisation level for exchanges, transmission equipment, optical fibre, and copper cable and justify why this is efficient. The top-down model should be sufficiently flexible to permit adjustments to the utilisation level. In a few cases, the use of the modern equivalent asset may imply a lower utilisation rate than the existing network. Where this is the case, it should be documented that this does not lead to higher overall costs than the existing asset 26. Criterion TD 9 The top-down model documentation should show both the utilisation rates in the SMP operator's actual network as well as the utilisation rates used in the model. 26 If, for example, the lower utilisation rates imply increased flexibility and reduced operating costs, the value of the associated benefits should be subtracted. Post- och telestyrelsen 48

55 Where the model applies utilisation rates that are lower than the actual utilisation rates, it should be justified that this does not lead to higher overall costs. 7.3 Valuation of major asset categories Access network Important components in the access network are: copper cable; optical fibre; radio; line cards; trenching, which may be ducted in some cases; and indirect network costs and overhead costs, including land and buildings. This section focuses on the first three components Copper cable Copper cable costs consist of a number of cost categories including the cost of the cable itself, jointing costs, installation costs and various indirect costs. Cable should be valued on the basis of an absolute valuation of the cable required for the existing level of demand, allowing for a reasonable planning margin. The first step involves selecting a statistically valid sample of routes for the SMP operator s network as a whole and for particular geo-types within that network. The next step is to estimate each of the individual items of cost on the selected sample of routes. The estimates are then applied to the entire network to provide a gross asset valuation for all copper in the access network. Operators typically provide for a larger number of pairs than necessary for the current level of demand. The SMP operator will need to show the ratio of its actual number of pairs to the number of lines in use in all parts of the access network (primary, secondary and final drop). Criterion TD 10 SMP operators should justify their ratio of actual number of pairs to subscriber lines in different parts of their access network and in different geo-types. The statistical validity of the sample of routes chosen should be explained within the model documentation Optical fibre The valuation principles for optical fibre are similar to those for copper cable. However, the SMP operator should demonstrate that optical fibre is the appropriate technology in particular situations. Where, for example, optical fibre is used to deliver some sub-2mbit/s over a single fibre leased lines, it could be Post- och telestyrelsen 49

56 argued that the appropriate MEA is copper (using xdsl technology). The same argument applies to 2 Mbit/s leased lines although in this case, the appropriate technology will depend on distance from the exchange and expectations of future requirements from the customer. Criterion TD 11 SMP operators should justify their use of fibre in the access network. If copper would be cheaper and there is no other justification for using fibre then the asset should be valued with copper as the MEA for fibre Radio In some cases, access may be provided over radio. This is likely to be the most cost-effective technology in areas with low population density (rural areas). As with optical fibre in the access network, the SMP operator should show that radio usage in the network is optimal given the costs and other considerations stated above. The gross asset valuation should reflect any differences between actual and optimal radio usage. Criterion TD 12 SMP operators should justify their use of radio in the access network. If copper would be cheaper and there is no other justification for using radio then the asset should be valued with copper as the MEA for radio Core network Important components in the core network are: exchanges; transmission equipment; DSLAMs; Ethernet switches; IP routers; optical fibre; trenching; and indirect network costs and overheads, land and buildings Exchanges This section covers remote and host subscriber stages, local exchanges, tandem exchanges and ATM switches for data traffic. The valuation should identify separately the cost of each of these items, distinguishing, where appropriate, between the values of different makes of switches. Post- och telestyrelsen 50

57 The top-down model requires less optimisation than the bottom-up model, a point that is particularly relevant when considering how to value exchanges. When optimising in the top-down model, it is not necessary to consider, for example, whether tandem exchanges should be replaced by local exchanges or local exchanges by remote subscriber stages. However, some optimisation in the topdown model is still necessary. At the exchanges, the two cost drivers in the core network are call duration (e.g. exchange ports) and call attempts (e.g. processing capacity). These components the cost of an exchange port and the cost of processing capacity need to be separately costed. If current equipment order prices do not provide sufficient information, it may be necessary to obtain additional information from switch manufacturers. A significant part of exchange costs consists of software. Since software costs are regularly updated and extended with new functionality, this suggests that it may be appropriate to derive separate asset lifetimes for software and hardware. This will also mean that separate cost-volume relationships are required for software and hardware. Criterion TD 13 SMP operators should quantify separately the costs of call duration and call attempts. SMP operators should also provide information to support their policy on exchange lives and, where appropriate, justify attaching the same asset lifetime to hardware and software. Finally, where software update contracts relate to exchanges in different levels of the network hierarchy, costs should be split in proportion to determine exchange costs for each class of exchange Transmission equipment Transmission equipment includes Add Drop Multiplexers (ADMs), line termination equipment, cross-connects, Wavelength Division Multiplexing (WDM) equipment, Dense Wavelength Division Multiplexing (DWDM) equipment and regenerators. The model should be able to show the values for each of these classes of equipment separately. PDH equipment can encompass both classical PDH equipment (now largely replaced by SDH equipment) and also PCM equipment (often used at the edge of the core network for low capacity routes). For the former, the top-down model should value PDH equipment using SDH as the MEA. For the latter low capacity PCM routes, the SMP operator should adopt an MEA that is both efficient and cost effective and should justify its choice. The use of SDH assumptions should reflect efficient deployment of SDH. This could involve SDH rings in some parts of the network, SDH point-to-point links for other routes, and access SDH technology for serving more remote (or lower traffic intensity) routes of the network. Post- och telestyrelsen 51

58 The use of digital cross-connects in relation to leased lines and data services traffic can be justified in many circumstances. However, it is more difficult to justify their use for PSTN. Criterion TD 14 PDH equipment can encompass both classical PDH equipment (now largely replaced by SDH equipment) and also PCM equipment (often used at the edge of the core network for low capacity routes). If circuit switching is retained for the PSTN then, for the former, the top-down model should, as a starting point, value PDH equipment using SDH as the MEA. For the latter low capacity PCM routes, the SMP operator should adopt an MEA that is both efficient and cost effective and should justify any alternative to the equipment currently used. The model documentation should justify the value of cross-connects attributed to PSTN DSLAMs, Ethernet Switches and IP Routers The main optimisation issue with regard to broadband/bitstream technology is likely to revolve around whether the DSLAM backhaul should be based on ATM or Ethernet/IP technology. The SMP operator should provide justification if it decides to retain ATM-based backhaul as the appropriate technology within the model. Criterion TD 15 The SMP operator should provide justification if it decides to retain ATM-based backhaul as the appropriate technology within the model Optical fibre A major issue in valuing optical fibre is to determine how much optical fibre an efficient operator would provide in its network. Typically, operators provide more optical fibre than currently required. For example, many operators install one or more 96-fibre cables on routes where much smaller cables would satisfy existing and prospective requirements in the near future. While this practice is understandable given the costs of installing additional fibre cable, the SMP operator will need to justify its installation practices. Similar remarks apply where insufficient fibre has been installed in the past requiring either additional fibre or the use of DWDM equipment. The assessment of fibre provision should take account of the likely trade-offs between additional fibre provision versus more costly higher-order multiplexing equipment such as DWDM. The model documentation should provide: The relevant costs inclusive of installation costs of different size cables; Data showing how the demand for fibre has grown on a range of routes; and Post- och telestyrelsen 52

59 The planning period for fibre, why this period has been chosen and what demand growth is expected to be over this period, in terms of overall traffic and for fibre. Criterion TD 16 SMP operators should justify current installation practices for optical fibre in the core network. If no justification can be provided for excess optical fibre, it should have no valuation Trenching costs, incl. poles Trenching costs are the costs of digging and reinstatement, costs which have not changed much as a result of technological developments. Both the access and core networks incur trenching costs, so unless otherwise indicated, the remarks in this section apply to both parts of the network. Trenching costs will especially be a significant part of access costs. Three important issues should be addressed. First, trenching costs are likely to vary between areas. For example, it is likely to be more expensive to dig in urban areas than it is to dig in non-urban areas. The model should be sufficiently disaggregated to permit estimation of the costs of trenching in different areas. Secondly, trenching costs should only reflect the cost of a modern network. If, for example, larger trenches were required in the past because of the need to install multiple co-axial cables in the core network, the trenching valuation should not reflect this. Finally, per kilometre costs will vary according to the number of kilometres of trenching required, as well as the possibility for co-digging with other telecommunications operators or utilities. Trenching costs actually incurred may relate to relatively minor extensions and modifications of existing routes, resulting in higher per kilometre charges than if the network were constructed in its entirety in a single period. Consequently: costs for the access network should be based on recent contracts since lengths in this part of the network tend to be short; and costs for the core network should be based on the costs that would accrue if longer lengths of trenching occurred, since an SMP operator could have planned for construction of these longer lengths. This approach means that per kilometre costs may differ between the core and the access networks even in the same terrain. Where information regarding the amount and type of trench/duct in the SMP operator's network is not available, the SMP operator may use sampling techniques. If a sampling technique is used, it is important to ensure that the sample(s) is (are) representative, both for the overall network as well as within the defined geo-types. Post- och telestyrelsen 53

60 Criterion TD 17 The top-down model (documentation) should identify trenching costs for different terrain types. The model documentation should explain the rationale for differences in these costs in different parts of the network. In many cases, telecom networks and other utilities share trench and duct. Where this is the case, costs should be split pro rata to the number of used ducts or used cables depending on the trench sharing agreement adopted. This will ensure that all users of the trench take a fair share of the common costs. Criterion TD 18 Where the telecommunications network and other utilities share duct and trench, costs of shared stretches of trench should be split pro rata to the number of used ducts or used cables depending on the trench sharing agreement adopted unless a more appropriate cost driver exists. The chosen cost driver should be documented and justified Poles In rural areas, copper, or indeed fibre, cable may be strung on poles rather than buried in trenches, as the low number of lines may not justify the costs of trenches. The model documentation should distinguish between copper/fibre cable that is trenched and copper/fibre cable strung on poles. Criterion TD 19 The model documentation should distinguish between copper/fibre cable that is trenched and copper/fibre cable strung on poles Indirect network costs (buildings, IT, motor vehicles, etc.) The top-down model should also include the indirect network costs of support functions such as buildings, IT systems, computers, motor vehicles, etc. Buildings have already been addressed separately. Within each of these rather broad classes of indirect network costs, a number of sub classes (cost categories) may be identified. For example, computers may be divided into PCs, equipment for network planning, network management and billing systems. The model should divide assets into homogenous cost categories before valuation, cf. section 6.1. For example, all relevant IT systems should be identified. IT systems may either be an asset owned by the network operator or outsourced to another company (or another internal division). The cost of outsourced IT systems should reflect the cost of outsourced IT systems for an efficient network operator. Outsourcing cost may be calculated on the basis of costs of actual outsourcing agreements with appropriate adjustments due to efficiency. The cost of a specific IT system should be allocated to network elements (or activities) according to appropriate cost drivers. Post- och telestyrelsen 54

61 7.3.6 Co-location The modelling of co-location services is discussed in more detail in sections 3.2 and 4.3. A potential problem area, which is particularly relevant for the top-down model, is double counting, i.e. that costs in cost categories that are common between colocation and the core/access network are included in both co-location and core/access. Cost categories that are common to both co-location and access/core should be included only once, and these costs should be apportioned between co-location and access/core according to appropriate allocation keys. Another problem is that the components used may be too detailed to be defined by top-down cost allocations. The model should therefore define each component separately identifying the ones having a realistic driver and value (such as floor space and rack space) and those that do not (such as cable trays). Criterion TD 20 The apportionment of cost in cost categories that are common to colocation and access/core should be appropriately documented to ensure that there is no double-counting of costs. 7.4 Annualisation Asset lives In theory, the asset lives used in the top-down model should correspond to the economic lifetime of the assets. In practice, however, top-down models often use the book asset lives. Although this approach is likely to underestimate the economic asset lives, due to conservative accounting practices, it has the advantage of being more objective. It is also more in line with the rationale underlying the top-down model: to build a model that relies on the actual (and audited) accounts of the SMP operator. Finally, the asset lives used in Swedish accounts should in principle reflect the economic lifetime of the assets. If assets lives were adjusted, it would be necessary to also adjust the net asset values, as the assets would then be less (or more) depreciated than appears from the accounts. Longer asset lifetime would lead to a lower depreciation charge but higher capital costs. The overall effect on annual capital costs might therefore be limited 27. Further, such adjustments will reduce the transparency and objectivity of the calculations. The top-down should therefore generally use book lives. Corrections should only be made in extreme cases where book lifetimes deviate substantially from economic lifetimes. Criterion TD 21 When determining asset lives in the top-down model, the starting point should be book lives. If asset lives are set on the basis of economic lifetime, the model should also be able to show the costs calculated on 27 This is further complicated by the existence of fully depreciated assets, discussed below. Post- och telestyrelsen 55

62 the basis of book lives. When the asset lives used in the model differ from the book lives, this will need to be justified for each category and the impact on costs should be documented. Any related adjustments to the capital base will also need to be justified and documented. When using a MEA that is different from the existing asset, the model should apply the asset life of the existing asset. As discussed in section , the valuation should be adjusted for any differences in asset lives Annualisation of relatively new assets As set out in section 5.1.5, the starting point for the top-down model should be straight-line depreciation Fully depreciated assets In many cases, the operational lifetime of assets may differ from their book life. This gives rise to a phenomenon known as fully depreciated assets whereby assets are still operational but have been fully depreciated. 28 Such assets will have a positive Gross Book Value (GBV) and Gross Replacement Cost (GRC) but will have a zero Net Book Value (NBV) and Net Replacement Cost (NRC). Fully depreciated assets will give rise to differences in top-down and bottom-up models. In a top-down model, such assets have no value and hence no cost. In a bottom-up model all assets are new by definition. Therefore assets that are fully depreciated in the top-down model, if required by an efficient operator, will have a value in the bottom up model. Hence in this respect, a bottom-up model will give a higher level of costs than a top-down model. Intuitively, the conventional treatment of fully depreciated assets in top-down models is unappealing, since it attributes zero value to assets which clearly have a value. On the other hand, there is no convincing alternative. If a bottom-up approach, based on the assumption of new assets were used in a top-down model, this would imply that the top-down operator would be able to charge depreciation and capital cost on assets for which the cost has already been recovered. Such `double-dipping can be rejected on fairness grounds. Not only would the operator be able to recover past depreciation once again, more importantly, if the fully depreciated assets are a result of having too short asset lives, the operator would also be allowed to recover the replacement costs more than once in the future. If the economic lifetime of an asset is 15 years, but the asset lives used in the model are 10 years, this means that the operator after 10 years would be able to recover the depreciation charge once again if the asset were attributed a positive value after year 10. Alternatively, asset lives in the top-down model could be set on the basis of the assets' operational economic lifetimes given the assumption of no radical technological change. However, this could give rise to problems of stranded assets 28 In some cases, the asset s life may be less than its book life either due to technological obsolescence or quality problems. Again, the likelihood of this happening should be reflected in assets book lives. Post- och telestyrelsen 56

63 in the future and, in any case, might also give rise to double-dipping. Finally, the SMP operator could introduce more sophisticated asset life policies involving, for example, different asset lives for different vintages of equipment. This would make the production of results more complex, however, and might only have a limited impact on results. What is really important is to ensure that the treatment of fully depreciated assets is consistent with the assumptions made regarding asset lives. The top-down model should therefore include a calculation of costs using book asset lives (as prescribed in Criterion TD 21) where no value is attributed to fully depreciated assets 29. In order to appreciate the importance of fully depreciated assets and analyse the interrelationship between asset lives and fully depreciated assets in the SMP operator s network, the top-down model may in addition include a separate calculation of costs where the value of fully depreciated assets still in use is included. The methodology and assumptions used to revalue such assets should be carefully documented and should be consistent with the assumptions made regarding economic asset lives. Such assumptions should be applied consistently to both fully depreciated assets and other assets. The SMP operator should identify and quantify fully depreciated assets by asset class and, where possible, vintage. Criterion TD 22 The top-down model should ensure consistency between asset lives and the treatment of fully depreciated assets. The model should show a calculation where FDA is attributed no value (using book lives). In addition the model may show a calculation where the economic value of fully depreciated assets is included, provided this is consistent with the assumptions made regarding asset lives. Asset lives should be applied consistently to both fully depreciated assets and other assets. The SMP operator should show the extent of fully depreciated assets by asset class and, where possible, vintage. 7.5 Capital maintenance When re-valuing assets on a current cost basis, a number of additional adjustments are required when calculating the annualised costs. There are two alternative approaches for making these adjustments. The two approaches differ in their approach to "capital maintenance" - the manner in which capital of an operator is viewed when determining profit. Capital can either be viewed in operational terms or in financial terms. The two approaches are therefore known as Operating Capital Maintenance (OCM) and Financial Capital Maintenance (FCM): 29 The issue of fully depreciated assets does not arise in the context of bottom-up modelling, where asset valuations should reflect all the assets required to support the relevant service volumes. Post- och telestyrelsen 57

64 Under the OCM approach, capital maintenance requires the company to have as much operating capability at the end of the period as at the beginning i.e. to maintain a set of inputs that can produce the existing level of output. Under the FCM approach, capital maintenance requires that shareholders funds at the end of the period are maintained at the same level in real terms as at the beginning of the period. Under both OCM and FCM, the depreciation charge of an asset is calculated on the basis of the new asset valuations the Gross Replacement Cost (GRC). If the price of an asset is falling (rising), the current cost depreciation charge will be lower (higher) than if calculated on the basis of historic costs 30. Under the FCM approach, additional adjustments are made to include the socalled "holding gains" and "holding losses" related to the individual asset categories and the effect of general inflation on shareholders funds. A holding gain (loss) arises when the price of an asset rises (falls) during the course of the year. Under the FCM approach, holding gains (losses) should be deducted from (added to) the depreciation charge 31,32. The effect of general inflation on shareholders funds is taken into account through an adjustment to shareholder s funds, determined by multiplying the opening value of shareholders funds by the change in the index of general inflation in the period. This amount should be debited to the profit and loss account and credited to a financial capital maintenance reserve. There are at least three reasons for preferring FCM to OCM. First, the concept of physical capital maintenance (OCM) will be of limited value in a world where the mix of assets and the mix of outputs are rapidly changing, as is the case for telecommunications. Secondly, accounting data can provide essential information about whether a firm should continue or discontinue an activity and whether, from a regulator s perspective, the firm is making acceptable, excessive or insufficient profits. However, one of the conditions for accounting information to perform this role is that it includes holding gains and losses. In other words, any inferences drawn about the firm s performance from OCM measures of profitability, either from a shareholder s perspective or a regulator s perspective, 30 The difference between the historical cost depreciation and current cost depreciation is often referred to as supplementary depreciation. Under CCA, this supplementary depreciation should be charged against profits in the profit and loss account. 31 If holding gains are taken account of and asset prices are rising, the FCM depreciation charge (current cost depreciation + holding gains) may be less than the historic cost charge. In fact, it may even be negative in some cases. 32 The idea is that the operator in addition to the standard depreciation charge should be compensated for holding an asset for which the acquisition cost is falling. If this were not the case, it would bias the investment decision of interconnection operators, encouraging them to postpone their investments until the equipment had fallen in price. In addition, the interconnecting operators would be able to push down prices later on, when equipment prices had fallen. In a competitive market, the SMP operator would have to match these prices, set on the basis of lower equipment costs. Consequently, the SMP operator would not be able to recover its forwardlooking costs. Therefore, its investment incentives would be distorted as well. A similar (simply reversed) argument applies when asset prices are increasing. Post- och telestyrelsen 58

65 may be incorrect. 33 Thirdly, OCM depreciation implies that the firm will not recover the cost of its investment when asset prices are falling and will overrecover its costs when asset prices are rising. 34 As the Commission notes: The use of the OCM concept may systematically incorporate insufficient or excess returns into the level of allowed revenue (depending, respectively, on whether asset-specific inflation was expected to be lower than or higher than general inflation). This is not a desirable feature of any regulatory regime. 35 Criterion TD 23 The top-down model should measure annualised costs using the FCM approach. Costs should include holding gains or losses, provided the model documentation can justify their inclusion. 7.6 Net asset valuation Net asset value is the gross asset value net of accumulated depreciation. Multiplying the resulting average net asset valuation for the year by the cost of capital, and adding a depreciation charge for that year will give the annualised cost of the SMP operator s asset base. There are three principal methods for calculating a net asset valuation: The NBV/GBV methodology The rolling forward methodology The NPV methodology. The last approach, which implies the use of economic depreciation, is theoretically preferable. However, for the reasons outlined in section of the common guidelines, this approach is not practical. This section focuses on the two alternatives to the NPV methodology 36 : 33 Essentially, the evidence that accounting information may provide useful information about the firm s performance when based on the Value to the Owner Rules further requires that holding gains be incorporated into the profit and loss statement. 34 In the former case, the sum of depreciation will be less than the cost of the investment; in the latter case it will be greater. The FCM methodology does not suffer from this problem. 35 See Commission Recommendation 8 April 1998, op cit. 36 As opposed to the bottom-up model, the purpose of the top-down model is to calculate the costs of the existing network (incl. the existing mix of asset vintages) of the SMP operator using current cost accounting (CCA). Assets are valued on the basis of the replacement costs of the modern equivalent asset not new assets. The historic depreciation of assets is used as a proxy for the deterioration of the assets. Such use of existing accounting information is an integral part of CCA and has nothing to do with historic costs. Post- och telestyrelsen 59

66 7.6.1 NBV/GBV methodology The simplest approach to calculating the net asset valuation is to multiply the gross asset valuation by the historic cost ratio of net book value (NBV) to gross book value (GBV) 37 : NBV NRC = * GRC GBV This should be done asset category by asset category 38. However, the approach will not provide accurate results when asset prices are changing. Where asset prices are rising, the methodology places too much weight on recent observations. This is because the asset price increases will result in a higher GBV per unit of output for more recent observations whereas the gross asset valuation per unit of output should be the same for all observations. The impact of this bias will lead to overestimation of net asset valuations, and hence capital costs. The converse holds when asset prices are falling. There are other factors that might in practice affect the bias. For example, the investment pattern is unlikely to be even. The actual investment pattern will affect the NBV to GBV ratio, which may result in biases, either positive or negative, if using this ratio to calculate net asset values Rolling forward methodology The rolling forward methodology calculates the net asset valuation as the gross asset valuation less the current cost accumulated depreciation. The rolling forward approach produces the correct net asset valuation if two assumptions hold. First, it requires that current cost depreciation plus holding gains and losses are equal to economic depreciation in each and every year. Secondly, the starting net replacement cost must be correct. This may be difficult to do in practice, since it requires details on the installation dates of each of the assets included in the gross replacement cost. Such information may not be available, particularly not for asset categories that include a large number of items or where individual items have been modified at various stages during the asset s lifetime. In such circumstances, an initial net asset valuation could be calculated using the NBV/GBV methodology. Clearly, the longer the period for which the application of the NBV/GBV is used, the greater is the potential error in the calculation of net replacement cost. Although the rolling-forward methodology is the theoretically correct methodology, it is associated with a number of practical difficulties. The SMP operator may therefore choose between either of the two methodologies. As the NBV/GBV methodology will lead to higher (lower) annualised costs than the rolling forward methodology where asset prices are rising (falling), it must therefore be ensured that the methodologies are used in a consistent manner. If 37 Gross Book Value is the historic cost valuation of assets before depreciation. Net Book Value is the historic cost valuation of assets after accumulated depreciation. 38 For each main RSS type, for example, the GRC of the RSSs should be multiplied by the ratio between NBV and GBV of these RSSs. Post- och telestyrelsen 60

67 different methodologies are used for different assets, this will need to be documented and justified in the documentation. Criterion TD 24 The top-down model should estimate net asset values on the basis of either the NBV/GBV methodology (for each asset category) or the rolling forward methodology. Preference should be given to the rollingforward methodology subject to data availability. If different approaches are used for different asset categories, this should be documented and justified Assets under construction Capitalised interest arising from assets in the course of construction should be included in the gross replacement cost of the assets. Depreciation should only begin when the assets are in service. Post- och telestyrelsen 61

68 8 Operating costs 8.1 Operating costs Cost categories The top-down model should examine operating costs at a disaggregated level to ensure that they are assigned to the correct part of the network. Only network (or wholesale) costs should be included in the access and core increments. Any costs related to retail activities such as marketing 39 as well as the retail cost categories related to both wholesale and retail activities should be excluded from these increments. Operating costs for activities closely related to the network include provision, maintenance and network planning and installation. The difficulty here is in identifying whether the activity relates to the access network, the core network or both. For other operating cost activities, an additional problem is that they are only indirectly linked to the network; they represent indirect network costs or overhead costs. Such operating costs comprise among other things: Transport. 40 Although the link to the network will sometimes be straightforward, e.g. transport for personnel maintaining the network, in other cases the link may be less direct, for example company cars. Accommodation. Here again the link to network activities may be straightforward in some cases, such as accommodation used to house exchange equipment, and much less so in other cases, such as accommodation costs related to office buildings. Finance. Activities include pay-roll financing, maintaining asset registers (primarily network related), preparing company reports and accounts and management accounts. Network planning and network optimisation. Computing. Again there will be direct costs, such as costs of network management and indirect network operating costs such as PC s. Human Resources. General Management. Costs falling into this category include the costs of legal and regulatory departments, insurance, royalty costs and the costs of senior management. Interconnection. Specific operating costs, such as the costs of staff involved in interconnection billing. 39 If the SMP operator believes that marketing and other retail activities include network elements, it should provide justification for its belief. 40 Transport costs consist of both capital items (the vehicles themselves) and operating cost items such as petrol and maintenance. The discussion here relates to the latter class of costs, for transport and for the other categories of costs discussed in this section. Post- och telestyrelsen 62

69 Many of these costs affect both the wholesale (network) and retail business. Activity based allocation methods should be used to determine their magnitude. Where relevant, the top-down model should distinguish between "pay" costs (cost related to salaries) and "non-pay" costs (costs that are not related to salaries) Efficiency The top-down model should only include efficiently incurred costs. An SMP operator may include operating costs incurred to satisfy legal and regulatory requirements, such as the provision of accounts and information, even when it would not be efficient to incur these costs if the legal obligation was not in place. The SMP operator's accounts are unlikely to distinguish between efficiently and inefficiently incurred operating costs. Reasons for inefficient operating costs are: using an asset which is not the MEA; inefficient processes; and excessive labour and other inputs, even given efficient technologies and processes. With respect to the first class of costs, any adjustment arising from, for example, higher operating costs for PDH than for SDH equipment can be made either to the asset value or to the operating costs. Section explains how to make such an adjustment to the asset value. To be correct, this adjustment requires determination of the efficient operating costs for the existing asset and its MEA; care must be taken not to include the costs of inefficient processes or excessive use of inputs in this process. There are a number of techniques for identifying efficient operating practices. Benchmark comparisons may compare the ratio of operating costs to capital costs, ideally at a disaggregated level, for the SMP operator and a range of other operators. 41 Typically, the benchmark is with US operators, since that is where data are most widely available, although comparisons with other operators in Sweden or in Europe may be preferable. An alternative benchmarking analysis would explicitly compare how the SMP operator performs activities with how other operators perform the same activity. The SMP operator s operating costs would then be adjusted to reflect the costs of the most efficient of the benchmark operators, including the SMP operator (the efficient benchmark operator may differ by activity). However, this approach is subject to two potential problems, namely that operators may classify costs in different ways and that there may be a trade-off between different classes of costs. Therefore, it should be used with caution. In addition to simple unit cost (ratio) analysis and basic (OLS) regression analysis, there are two main techniques for estimating relative efficiency across firms in practice: Stochastic Frontier Analysis (SFA) and Data Envelopment Analysis (DEA). Whereas SFA uses econometric techniques, DEA uses mathematical 41 When using benchmarking, appropriate corrections should be made for relevant differences between an SMP operator in Sweden and operators in other countries. Post- och telestyrelsen 63

70 programming. Neither of these techniques are free from criticism and therefore efficiency studies should ideally use both DEA and SFA, with the results of each method being used as a cross-check on the results of the other 42. The SMP operator may for historic reasons be burdened by redundancy charges or pension liabilities for retired employees (typically civil servants) according to a financing scheme that is not in tune with the requirements of the new competitive environment. Such costs would not be incurred by a new (forward-looking) operator and should therefore not be recovered through the access and interconnection charges. Criterion TD 25 The top-down model should only include efficiently incurred costs related to the wholesale activities. No redundancy costs should be allocated to access and interconnection services. The model documentation could be accompanied by an independent efficiency study of the SMP operator. Where the efficiency study identifies inefficiently incurred costs, operating costs should be adjusted to reflect an efficient operation. An independent efficiency study combined with any appropriate efficiency adjustments would increase the reliability of the cost data used in the top-down model Activity based allocation of operating costs Having identified the classes of operating costs and made adjustments to ensure that they only reflect efficiently incurred costs, the top-down model must then allocate these costs to the different services provided. This may be a difficult and time consuming exercise for operating costs that are common to more than one service. Here, an activity based costing (ABC) approach should be used to achieve a more satisfactory allocation of operating costs. Figure 6 illustrates how the ABC approach allocates operating costs See e.g. the recent study conducted for Oftel, "The Profitability and Efficiency of the UK Network Operators", 26 September 2001, available at 43 The ABC approach is explicitly recommended in the Commission Recommendation, 19 September 2005 on accounting separation and cost accounting systems under the regulatory framework for electronic communications. Post- och telestyrelsen 64

71 Figure 6: Activity Based Costing (ABC) Direct Costs Indirect Costs Activity 1 Activity 2 Activity 3 Product A Product B ABC is a two-stage costing system. It assumes that activities cause costs and that products (or services) create demand for activities. In the first stage, costs are allocated to activities; in the second stage, the costs of activities are allocated to products. ABC focuses on why expenditure is incurred, i.e. what activities did the expenditure support? Having identified the reason for incurring the expense, it is possible to trace the costs through to the particular services that cause these costs to be incurred. If an expenditure cannot be related to an activity of the firm, or if an activity cannot causally be related to a specific product or service supplied by a firm, then the expenditure should not be accounted for in the costs related to that product or service, because there can be no argument for recovering such expenditure from the user of that service. To assign the costs of each activity to products or services will require identification of a cost driver for each activity. The cost driver should explain the costs of that activity and should be quantifiable call hours and number of calls are examples of readily quantified cost drivers. A further consideration is that the cost driver should be measurable in a way that enables it to be identified with individual products or services. Criterion TD 26 The top-down model should allocate operating costs to the various services on the basis of cost causality, such as ABC. Supporting documentation should describe the cost drivers and how the model assumes they affect operating costs for each activity. The documentation should also describe which activities the different services consume Documentation Allocation of operating costs is complex. It is therefore important that the SMP operator carefully document the different steps involved in the allocation procedure. The documentation should describe how the different cost categories have been allocated from the general accounts of the SMP operator to the network division, Post- och telestyrelsen 65

72 distinguishing between PSTN network services, broadband/bitstream services and other services such as leased lines and data services. Within the PSTN network and broadband/bitstream network, it should be documented how costs have been split between the access network, the transport network, exchanges and other. The documentation should include a description of the allocation keys used (percentages as well as cost driver). In particular, it is also required to ensure that there is no double counting of operating costs. This is particularly important where costs of a given service is estimated separately. This could e.g. be the case for co-location or interconnection specific services. Operating costs allocated to these services will need to be deducted elsewhere. The documentation should include a table showing all the operating costs by category included in the LRIC cost base summing up to the total. Post- och telestyrelsen 66

73 9 Costing services 9.1 Homogeneous cost categories As discussed in section 6.1, the first step in the top-down modelling is to group costs with similar characteristics into individual cost categories, also called homogenous cost categories. The cost driver should explain the costs of a particular activity and should be quantifiable. It is important to ensure that the model is sufficiently disaggregated to accurately allocate costs to the various network services. Criterion TD 27 The model should separate costs into homogeneous cost categories. The SMP operator will need to provide justification for its selection of cost categories. 9.2 Service usage An important stage in the development of a top-down model is to quantify the extent to which increments and, where relevant, services use a particular asset or operating cost category. This means Identifying an appropriate basis for measuring volumes; Determining the extent of volume usage by different increments; and Where appropriate, determining the extent of volume usage of services within these increments. In some cases, the measurement of volumes is relatively straightforward, in others it is more complex. We briefly discuss the measurement of volumes in relation to a number of asset and operating cost categories. The focus is on the contribution to volumes of different increments and the determination of the relative contributions of PSTN, broadband/bitstream, and leased line services within the transport network Trenching and duct The volume of trenching and duct can be measured in terms of kilometres with separate measures required according to whether a duct is in the access or core network. In addition, where trenching and duct is required for other increments, this should be shown in a separate cost-volume relationship. However, in most cases, trenching and duct will be shared with the core, and perhaps the access network. Criterion TD 28 Separate volume measures are required for access and core trench/duct and potentially for other increments. Post- och telestyrelsen 67

74 9.2.2 Copper and fibre cable The relevant volume measures here are subscriber lines, where cable is in the access network, and system lines, where cable is in the core network. Clearly, in the core network, the network should be designed assuming a modern efficient operator with respect to the number of system lines required and their usage Local Exchanges A range of separate volume drivers is required for local exchanges. First, it is important to have separate drivers for the subscriber stage/concentrator and the local exchange processors because these form different core network components. Secondly, different makes of exchanges may have different cost structures; hence different sets of cost-volume relationships and volume measures are required. Thirdly, both the subscriber stages and the local exchange processor have more than one cost driver. Different volume measures are required to account for the respective contributions of access and core on the one hand and call attempts and duration, within core, on the other. Finally, a further split may be required between hardware and software if it is found that these exchange components have different asset lifetimes. Before these different drivers are treated in greater detail, a comment should be made on the relationship between the number of cost categories and volume drivers. If asset lifetimes are different for software and hardware they should have different cost-volume relationships. In such cases, there should be a one-to-one relationship between cost category and cost-volume relationship Subscriber stages (concentrators) There are two main volume drivers for remote subscriber stages: subscriber lines and the volume of call minutes. In addition, some costs of remote subscriber stages may be related to busy hour call attempts, i.e. the switches may have some processing functionality. Where this is the case, a third volume measure - busy hour call attempts - is required Local Exchange Processors In addition to subscriber stages/concentrators, the local exchanges consist of a range of sub-systems that process calls, including the switch block and the actual processor. While all costs relate to the core network, some are driven by call minutes (such as ports) while others, such as signalling support and processing, are driven by call attempts. Separate identification of these costs is necessary to determine separate duration and set-up charges. Criterion TD 29 Separate volume measures are required for subscriber stages and processors, by manufacturer of exchange, for lines, call duration and call attempts. In addition, where software and hardware lifetimes differ significantly, separate cost-volume relationships are required. Post- och telestyrelsen 68

75 9.2.4 Tandem exchanges The issues that need to be addressed, namely separate volume measures for duration and set-up related costs and potentially different relationships for hardware and software are the same as for local exchanges Transmission equipment While a single volume measure can be derived for all transmission equipment, problems arise in measuring the respective volume usage by different services and increments. The most significant of these problems is to account for the respective usage of PSTN traffic, leased lines and other services running over a specific system line (where a system line is defined as a fibre connection between two distant items of transmission equipment), since the former is measured in call minutes whereas leased lines and data services are measured in Kbit/s or Mbit/s. This problem can be overcome by converting PSTN traffic volumes into Mbit/s equivalents. This in turn involves the following steps: Adjusting for non-paid time (such as holding time); Applying routing factors, which indicate the extent to which different calls use the network; Converting into Busy Hour Call Minutes; Conversion from 64 Kbit/s into 2 Mbit/s; and Adjusting for utilisation. However, converting on this basis will not necessarily reflect the intensity with which PSTN services and leased lines use transmission equipment. Adjustments should also be incorporated to take account of resilience requirements for leased lines and differences in the way the services use the network (routing factors). Criterion TD 30 In order to measure respective volume usage over a specific system line containing both PSTN and non-pstn services, PSTN minutes should be converted into Mbit equivalents. Further adjustments are required to take account of diversity and differentials in the intensity of usage DSLAMs, Ethernet Switches and IP Routers For DSLAMs, the line cards should have subscriber lines as the volume driver. For the common parts, there are arguments to use both subscriber lines and Gigabytes as the volume driver, largely depending on the extent to which broadband/bitstream services are priced (that is, the extent to which the prices reflect usage). The SMP operator should justify the volume driver selected. For Ethernet switches and IP routers, Gigabytes should be used as a starting point for the volume driver, although for IP routers there is also an argument for the use of packets. Where certain routes through the network, or indeed types of traffic flowing over the network use either reserved capacity or enjoy Post- och telestyrelsen 69

76 prioritisation, then adjustments could be made to reflect this. If the SMP operator does include such adjustments then these should be justified. Criterion TD 31 The starting point should be to use subscriber lines as the volume driver for DLSAM line cards and Gigabytes as the volume driver for DSLAM common parts and also for Ethernet switches and IP routers. Departures from this, for example to take account of Quality of Service requirements, should be justified and documented Indirect network costs and overhead costs Many capital assets and operating costs are indirectly related to particular end services or increments. The fundamental requirement is to develop a model which reasonably reflects the complex set of relationships between some of these classes of costs and ultimately outputs. The costs of land and buildings are examples of indirect network costs, as these costs are not driven directly by traffic, calls or the number of lines. Rather, these costs are driven by the floor space required, which in turn is determined by the quantity of equipment housed in the buildings. Costs will therefore need to be allocated on this basis. In many cases, the volume of particular activities is linked to the number of people within the organisation. For example, the number of human resources staff required is primarily driven by the number of people in the organisation and the complexity of personnel policies (itself partially related to the number of people in the organisation). In this case, it is important to establish a causal chain between the volume of other activities and the volume of human resources staff required. To provide a couple of examples: The volume of local exchange assets is a key driver of the volume of local exchange maintenance. This in turn will give rise to a demand for human resources staff. The volume of assets will be a key driver of maintenance and installation personnel requirements. These in turn will be one of the drivers of transport personnel requirements. In each case, personnel will give rise to the need for human resources staff. In the case of some cost categories, the intensity of usage will vary according to people s function in the organisation. For example, motor vehicles may be particularly heavily used by maintenance, provision and installation staff; while office accommodation usage will depend on a person s position within the organisational hierarchy. In such cases, weighted cost-volume relationships are required. Post- och telestyrelsen 70

77 Criterion TD 32 For staff related volume measures, the SMP operator should provide documentation justifying the way in which usage intensity has been measured by functional area within its organisation Outsourced costs A substantial amount of the SMP operator s operating costs may be outsourced to another company (or subsidiary). In this case it may be difficult for the SMP operator to establish a link from the operating costs paid to the production company and the underlying cost driver. As far as possible, the SMP operator will therefore need to undertake the cost-modelling in close co-operation with these production companies. Criterion TD 33 Where the construction and/or operation of the network, or any associated activities, have been outsourced to another company or subsidiary, the SMP operator should, where possible, undertake the cost modelling in co-operation with these companies to ensure that it is possible to establish a clear link between the underlying cost-drivers and the costs incurred by the SMP operator for those activities. 9.3 Assigning costs to services Calculating incremental costs The incremental costs should be calculated for each of the cost categories separately. Where a cost category is not directly related to a final increment, the model will need to reflect the causal chains between the cost category under consideration and the final increment, e.g. one of the many causal chains for human resources is through maintenance personnel and hence final increments Calculating costs of services within an increment. Having determined the incremental costs, the next step is to calculate the costs of each of the services within an increment. For example, what is the cost of PSTN in the core network? In some cases, the cost should be allocated entirely to one of the services since only one service requires that particular cost category. Sometimes the appropriate response will be to look at the volume of costs that each of the services are responsible for, and share the costs between the services on this basis. But this will not work for shared costs. For example, if PSTN, leased lines and other services all use the trenching in the core network, using kilometres of trenching to determine the allocation of the costs of the service may not be sensible. Instead, the model should allocate the costs of trenching according to some other weighting factor, such as Mbit/s. Post- och telestyrelsen 71

78 Criterion TD 34 In the core and access increments, the top-down model should assign costs to the various services within those increments. The total of the costs assigned to the various services should correspond to the total cost of that increment. For each cost category, the documentation should provide a rationale for the assignment. There should be consistency in the assignment rules used for the different cost categories. 9.4 Treatment of common costs As mentioned in the common guidelines, a mark-up should be added to cover the common costs. It should be ensured that the total amount of annual costs allocated to services corresponds to the total amount of annual costs related to the individual cost categories. Post- och telestyrelsen 72

79 10 Model functionality and documentation 10.1 Model requirements Top-down models tend to be extensive, complex and not very transparent. However, the complexity (and associated lack of transparency) is often unwarranted. When building the top-down model, a separate aim should be to make the model as transparent as possible. In many cases it may for example be possible to build the model using standard-software such a MS Excel, Visual Basic and Access. Unless standard-software such as MS Excel is used, PTS should be provided with the facility to run the model. Criterion TD 35 The model should show how the costs of services are derived by multiplying the routing factors for each service to the costs of network elements. The total LRIC of a service should show a breakdown between the LRIC and the mark-up for common costs. Unless standard-software such as MS Excel is used, PTS should be provided with the facility to run the model Sensitivity analysis In the reconciliation process, it will be necessary to examine and quantify the impact of the differences between the two models. Ideally, estimates of these impacts should be made with both models. However, this will not always be easy to do in the top-down model. For example, since the top-down model is not required to consider the mix of exchanges, such functionality will not need to be included in the model. Where possible, the model should have the flexibility to examine the impact of a change in: Equipment prices; Utilisation Rates; Cost of capital; Volumes; Annualisation methodologies; The inclusion/exclusion of fully depreciated assets; Asset lives; and Price trends. To assist the reconciliation of the top-down and bottom-up model, it should be possible to carry out sensitivity analyses. In some cases, these sensitivities should be formally tested within the model, for example, alternative costs of capital. In other cases, it may not be possible to run formal sensitivities within the model. Post- och telestyrelsen 73

80 However, it should be possible to produce off-line estimates of the impact of alternative assumptions in many cases Model documentation The model documentation should include but not be limited to: Cost of access and interconnection services; Cost of individual network element stages; Routing factors; Volumes; Number of sites and exchanges by type; Description of methodology, assumptions, samples, etc. List of cost categories and network elements; GRC and NRC for all cost categories; Quantity and unit price information underlying the GRC for cost categories; CVR-curves with an indication of the cost driver and explanation of methodology used for estimating the CVR; Annualisation assumptions (depreciation methodology, assets lives, price trends for all assets); Operating costs and allocation keys; Working capital; Description of the network structure, indicating changes compared to the existing network; Average cable lengths in the access network by geo-type distinguishing between the primary and secondary network and the final drop; Trench length in core and access; Utilisation rates (existing as well as modelled); Documentation of efficiency and efficiency adjustments; User manual; and A list indicating how general and specific top-down criteria have been met. The model as well as the model documentation should be in English. Supporting documentation, such as contracts, or analyses carried out in Swedish may be provided in Swedish. Post- och telestyrelsen 74

81 Justification In several places of this MRP, it is noted that the approach taken in the top-down model should be "justified". In many cases it is not possible (or would not be appropriate) for PTS to specify the nature of such a justification. Unless otherwise stated, it is therefore left to the SMP operator to decide on the most appropriate kind of justification. The general principle is that the justification should allow an independent party (such as PTS) to evaluate whether the approach taken is reasonable or not, given the overall objectives and purposes of the model Audit of model The model should be subject to an independent audit 44. Based on an assessment of significance and risk, the audit should include, but not necessarily be limited to: Spot checks on the relationship between the dimensioned network and underlying registers, statistics, etc. not included in the annual report Spot checks of the GRCs/current cost valuation of assets (comparison with price lists/contracts, asset register, etc.) Balancing book values applied against book values recorded in the fixed asset register Spot check of annualisation calculations Balancing working capital and operating costs against annual reports Spot checks on ABC methodology used to allocate operating costs Spot checks on the cost allocation methodology and allocation keys Spot checks of volumes, traffic statistics, etc. not covered by the annual report. The purpose of the audit is to ensure that the information provided in the model and in the model documentation is correct. It is not the purpose of the audit to verify that the model meets the criteria set out in this MRP. Ideally, the model should be accompanied by the audit report. However, it will be sufficient if the audit report is provided within 6 weeks of the delivery of the model. Criterion TD 36 Ideally, the model should be accompanied by the audit report. However, it will be sufficient if the audit report is provided within 6 weeks of the delivery of the model. 44 The audit may well be undertaken by Telia's external auditor. Post- och telestyrelsen 75

82 PART C: SPECIFIC GUIDELINES FOR BOTTOM-UP MODEL Post- och telestyrelsen 76

83 11 Overview of bottom-up modelling The procedure to build a bottom-up model can be summarised in the five steps illustrated in the figure below: Figure 7: Simplified overview of the steps in building a LRIC bottom-up model Step 1 & 2: Measure demand and establish unit costs End-user demand (internal and external) Dimensioned demand Equipment prices and cost data & Technology/configuration choice Step 3: Build hypothetical network both in terms of assets and operating activities Apply the cost of capital Annualised capital cost per cost category for each increment Design/Equipping of hypothetical network Apply routing factors Common capital costs Operating cost per cost category for each increment Common operating costs Annualised cost per cost category for each increment Allocation of cost categories to each service (PSTN, leased lines and other services) Step 4: Determine costs of different network elements Allocation of PSTN s share of cost categories into network elements Step 5: Cost interconnection and access services Deriving of unit cost (e.g. per minute) for each network element Apply routing factors Mark-up Aggregation of network elements to derive LRIC cost for I/C and access products I/C and access charges Post- och telestyrelsen 77

84 11.1 Measuring demand and establishing unit costs (Step 1&2) Figure 7 illustrates that the first two steps can be carried out simultaneously. Demand data is collected from the SMP operator and transformed to dimensioned demand (demand used for dimensioning purposes). Also information is collected on equipment prices and other cost data and decision is taken on the overall choice of technology and network topology. The information generated by these steps will be necessary to optimally equip the network. The costing of the individual network elements and services can only occur after the network has been equipped Building hypothetical network (Step 3) Engineering principles will inform the dimensioning process for the network. For example, the number of ports in an exchange might be estimated by reference to the busy hour erlangs in the exchanges. The equipment prices and other cost data such as the unit cost of trench collected during step 2 will form the basis for determining the capital costs of the dimensioned network. However, the bottomup model will also need to include operating costs and overhead costs Determining the cost of network elements (Step 4) The bottom-up model then needs to estimate the capital cost of each network element. However, the model needs to calculate annual costs, so the investment costs will need to be annualised to generate an annual figure for the capital expenditure involved in using each asset. Depending on the choice of depreciation profile, this requires an estimate of the purchase price of the asset, asset lives, the price trend of each asset, the scrap value of the asset at the end of its economic life, and the cost of capital. Operating costs typically make up a significant share of the total annual costs in a network. These need to be added to the annualised capital costs to produce the total annualised costs of network elements. Operating costs will typically be estimated as a mark-up on capital costs Costing services (Step 5) The final step in the bottom-up process will be to cost the various services to be provided. First, the sum of the capital costs and operating cost needs to be transformed into a per-unit cost for each network element. This could e.g. be the cost of using a local switch for one minute. In order to do this, it is necessary to calculate the average traffic handled by the average network element of this type. The total cost of the network element is then divided by this average traffic to come up with a cost per minute. In order to calculate the LRIC of a service, the per unit costs of network elements then need to be aggregated. The linkage between the cost of network elements and service costs is provided by routing factors specifying the average number of each network element used by a particular type of service. Finally, a mark-up is added to include a share of the common costs in the cost of the service. Post- och telestyrelsen 78

85 12 Optimisation 12.1 Scorched node assumption Different optimisation constraints may be put on the modelled network. At the extreme is the scorched earth assumption, where the bottom-up model is not constrained in any way by the existing network design of the SMP operator. Optimally sized switches may be employed at locations optimal to the overall transmission design, as if the network was to be redesigned from scratch. However, for LRIC modelling purposes, the scorched node assumption is often used. The scorched node assumption implies that the SMP operator s existing number and location of its nodes are taken as given. However, to ensure that the SMP operator has incentives to migrate to a more efficient architecture, the scorched node assumption can be modified to allow for certain optimisations. International experience shows that there are many hybrid forms of the scorched node assumption depending on the definition of a node and degree of optimisation. In this MRP, a node is defined as a technical house location (which might contain PSTN switches, concentrators, DSLAMs and potentially multiplexers). This implies that there may be more than one node at a site since different types of equipment are often co-located. However, the number of sites are fixed. The degree of optimisation refers to changes in the nature of nodes. The mix of equipment, therefore, may be changed. For example, a local exchange may be replaced by a remote subscriber stage. Multiplexers or similar equipment with no switching capability may also be placed at a node. Criterion BU 1 The bottom-up model should comply with the modified scorched node assumption where nodes are defined as technical house locations (which might contain PSTN switches, concentrators, DSLAMs and potentially multiplexers), that is to say the existing number and locations of sites are fixed, but no empty sites are allowed although it is possible to change the number and mix of equipment at a site. As regards the modelling of the access network, the main purpose is to estimate the cost of unbundled local loops. As a general rule, the location of the line card (typically located in the concentrator) in the SMP operator's network and the network termination point (located at the subscriber premises) would therefore need to be taken as given. This means that the boundary between the core and access network will be the same in the modelled network as in the actual network. When replacing an RSS with a multiplexer, the line card should remain at the location of the multiplexer. This is also consistent with the requirement for the multiplexer to have concentrating capacity 45. Pair gain systems and the like should clearly not be considered as concentrators but as a part of the access network. Post- och telestyrelsen 79

86 12.2 Technology The bottom-up model should show the costs of a network, implemented with a modern (forward-looking) technology. Modern technology is to be interpreted as the most cost effective technology currently available and actually deployed in large scale fixed networks, or likely to be deployed within the next few years Switching technology This implies that the core network in the bottom-up model could now either be based on circuit-switched technology or packet-switched technology 46 according to accepted industry standards. However, if packet-switched technology is assumed, then the network must still be able to offer standard wholesale products where these are circuit-switched. Criterion BU 2 The core network in the bottom-up model could be based upon either circuit-switched or packet-switched technology according to accepted industry standards. The choice adopted should be justified and documented Transmission technology The implications for the transmission network of the extent of optimisation undertaken within the scorched node assumption will depend, to a large part, on the decisions made for the switching network. When deciding on the choice of technology to be used in the transmission network, it is e.g. necessary to consider: Whether circuit-switched or packet-switched technology has been assumed for switching. SDH technologies and configuration. SDH is considered a well-established technology in circuit-switched networks that is widely used in modern large scale operations. SDH systems are flexible and easy to deploy in different network structures. Ethernet Layer 2 switches and IP Layer 3 routers. These devices are considered a well-established technology in packet-switched networks that are also widely used in modern large scale operations. As with SDH systems, they are flexible and easy to deploy in different network structures. the extent of microwave in the network. Microwave transmission can be cost effective when the costs of trenches are relatively high. In mountainous areas on the border to Norway, for example, it may be more cost effective to send signals using microwave provided that a line of sight can be established; and 46 It could be noted that the development of other emerging technologies like IP already may have caused price reductions for circuit-switched equipment. Post- och telestyrelsen 80

87 the extent of optical transmission systems such as WDM and DWDM. The starting point for a Bottom Up model, however, should be to assume that sufficient fibres are provided in the cable to satisfy foreseen demand. Criterion BU 3 The predominant transmission technology for circuit-switched networks should be SDH, whereas for packet-switched networks it should be Ethernet switches and IP routers. As a starting point, DWDM should not be included on the basis the Bottom Up model should assume sufficient fibres are available within the cables. Microwave transmission should be used only where fibre is not cost effective Access technology In principle, the definition of modern technology in the access network implies that the bottom-up model can include any technology provided that the technology modelled has the potential to produce services with at least the same functionality and quality to the customers and the interconnecting operators as the existing technology. However, given that the model is constrained by technology that is actually deployed (or likely to be deployed over the next few years) in large scale fixed networks and that service equivalence (cf. section ) must be ensured, the model should not adopt fibre in the local loop for serving copper-connected customers. If some customers are currently connected via radio in the access network, then the SMP operator has accepted that this technology provides a suitable level of service. Therefore, in such a case, radio may be modelled as an alternative to copper where this is cost effective. Criterion BU 4 The access network in the bottom-up model should be modelled according to accepted industry standards. As a starting point it should model a fibre access network for customers connected with fibre today and equivalently model copper in the local loop for customers connected by copper today. However, if some customers are currently connected via radio in the access network, then the SMP operator has accepted that this technology provides a suitable level of service. Therefore, in such a case, radio may be modelled as an alternative to copper where this is cost effective Requirements of the optimised network The optimisation performed in the bottom-up model must meet certain minimum requirements. These include: the (modified) scorched node assumption must be met; services must be provided with a quality of service equal to that provided by the SMP operator to end-users and interconnecting operators; Post- och telestyrelsen 81

88 service equivalence must be ensured; and the network must be dimensioned correctly. Each of these minimum requirements (except scorched node) is discussed below Quality of service The bottom-up model needs to demonstrate that the optimised network provides services at a level of quality and functionality corresponding to the level that the SMP operator offers to interconnecting operators and end-users. This means that the starting point should be the SMP operator s existing quality of service (QoS) levels. However, these may not be appropriate in all circumstances, as they may not represent the QoS levels provided by an efficient operator in a fully competitive market. It will therefore be necessary to review relevant international benchmarks taking careful account of factors specific to Sweden. One of the most important QoS parameters for PSTN services in bottom-up modelling is the blocking margin (also known as grade of service) which measures the proportion of call attempts made in the busy hour period that fail to mature because, for example, equipment is faulty or fully occupied on other calls. It is the view of PTS that the maximum blocking margin (end to end) should be set at 0.5%, as a starting point in the bottom-up model. Criterion BU 5 The bottom-up model should demonstrate that the optimised network provides services at an appropriate level of quality for an efficient SMP operator Service equivalence Although the services modelled in the bottom-up model may be of a level of quality corresponding to the level provided by the SMP operator, there may still be differences between them since the network modelled is not an exact replica of the SMP operator's network. Modelling a different network means that it may not always be possible to fully align costed services and actual services provided by the SMP operator. However, viewed from the perspective of the interconnecting operators and the end-user, the modelled services should be equivalent to services provided by the SMP operator and no external costs should be incurred enabling a similar service to be offered. This means, for example, that in areas where ULL is (or is likely to be) commercially viable the bottom-up model should model copper where a copper service is currently provided so ULL services like raw copper are copper based as opposed to being based on alternative technologies such as PONS Correct dimensioning of the network The optimised network must be able to meet the dimensioned demand. It must be documented that the network can carry the dimensioned traffic. In practice, this means that the model must be able to show: Post- och telestyrelsen 82

89 that exchanges are dimensioned sufficiently to carry all subscriber lines; that exchanges are dimensioned sufficiently to carry all of the relevant traffic taking account of the busy hour (both in terms of number of calls and call duration) and blocking margin; and that the traffic, processed by exchanges using the transport network, can actually be carried over the transport network, and that the network has been dimensioned with sufficient resilience. Criterion BU 6 The bottom-up model must be able to demonstrate that the optimised network can carry the dimensioned demand. Post- och telestyrelsen 83

90 13 Demand This section covers guidelines on estimating demand in the core and access network Demand in the access network The bottom-up model should be based on the existing number and mix of subscribers using the SMP operator's network. This means that information will be required for each of the services listed in section 3.1 including leased lines and other services using the access network not captured in the numbers for leased lines. This data will need to be requested from the SMP operator. Once the demand information has been collected, assumptions will need to be made about the growth rate expected for each of those services when dimensioning the network. Moreover, it may be useful to collect information on the type of dwelling occupied by subscribers and distribute the identified level of demand over a convenient set of dwellings. However, the benefits of collecting information on dwellings may not justify the effort required. The number and the size of business customers will also affect the costs of the access network. This means that information may also be useful on the distribution of business customers Demand in the core network Estimation of end-user demand The main source of information on the current level of demand in Sweden will be the SMP operator. The model should include all the current traffic, including: PSTN traffic; Broadband/Bitstream; leased lines; and other services, including those offered by other operators to end customers via the SMP operator's network. An assumed rate of growth over the assumed planning period needs to be added to the current volume of traffic in order to get to the dimensioned end-user demand. For some services, the growth rate may be negative. With regard to PSTN traffic, the bottom-up model should show both annual minutes and number of calls for all the PSTN products listed in section 3.1. The obvious source of this information is billed minutes and number of calls recorded by the SMP operator. Billed minutes and number of calls, however, do not capture the total demand for the core network. This is because: billed calls do not include unsuccessful calls, i.e. calls for which a connection is established but remain unbilled as no answer is received; and Post- och telestyrelsen 84

91 billed minutes do not include ringing time. Criterion BU 7 When measuring the level of PSTN traffic, the bottom-up model should take unsuccessful calls and ringing time into account. With regard to leased lines, the bottom-up model should show total demand of leased lines circuits in terms of number of circuits by capacity bandwidths. Other services such as broadband/bitstream and data services may have different characteristics and should be grouped into different categories of service and by different capacity bandwidths. Criterion BU 8 The model should show the demand for leased lines by number of circuits by capacity bandwidths. Demand for broadband/bitstream and other services should be shown by different categories of services and, within each category, by different capacity bandwidths Estimation of dimensioned demand Once the end-user demand has been estimated, the model will need to show how this has been adjusted to estimate the dimensioned demand to be satisfied by the network. The adjustments include: the application of routing factors; allowance for resilience; and application of busy hour estimates. Also the model needs to take account of the grade of service as indicated in section The modelled network dimensioning should incorporate the actual demand from end-users and interconnecting operators. In addition, the growth rates should be used. The modelling of the network should thus be based on a total assessment of demand from end-users and interconnecting operators including relevant growth rates (forecasts) as well as the above-mentioned adjustment factors The application of routing factors End-user demand is insufficient when dimensioning the network because traffic will flow through the network in different ways and create knock-on requirements. Two main methodologies may be used in order to spread end-user demand over the network: Demand on a route by route basis. This consists of estimating the traffic flowing over the network on a route by route basis. The amount of traffic originating from each node of the network is computed aggregating the amount of traffic from all the lines directly or indirectly connected to that node. Each kind of Post- och telestyrelsen 85

92 subscriber line is assumed to generate, during the busy hour, an average amount of traffic (minutes and corresponding erlangs for PSTN services). Routing factors. Routing factors are defined as the average frequency at which a particular service uses a given network element. These can be based on the SMP operator's existing set of routing factors or calculated from scratch. If routing factors from the SMP operator are used, these should be adjusted to reflect the underlying network architecture of the bottom-up model. The recommended approach is to use routing factors to distribute the forwardlooking level of demand over the different parts of the network. Criterion BU 9 The routing factors used in the bottom-up model need to be consistent with the underlying network architecture and identified separately for each service Allowance for resilience In the event of switching or transmission failure, alternative routes need to be available in order to avoid loss of traffic. The network needs to allow for redundancy. The ability to provide service under breakdown or overload conditions is known as network resilience. Methods of improving resilience include: Improving reliability of components and systems by design and manufacture; Adopting network structures less sensitive to surges and equipment breakdowns; Diversifying the physical routing of circuits making up specific traffic routes; Using rings in the configuration of the transport network, e.g. ring structured, self healing SDH networks, where circuits routed both ways around the ring preserve capacity should the ring be cut by failure of a transmission system; Controlling automatic alternative routing of traffic to avoid congested or failed nodes/links; and Real-time network management to monitor network performance and take action to overcome congestion by re-routing of traffic, blocking traffic surges at their source exchanges and initiating repair of equipment faults. Moreover, different services require different resilience standards. Adjustments for resilience, therefore, need to take into account how the network has been configured and that different services requiring different reliability standards imply different resilience adjustments. However, care must be taken to ensure that the different means of obtaining network resilience complement rather than duplicate each other. Post- och telestyrelsen 86

93 Criterion BU 10 The bottom-up model should show how service-specific adjustments for resilience have been taken into account in the given network architecture Application of the busy hour estimate In the bottom-up model, the core network needs to be dimensioned to carry the amount of traffic in the busy hour. The calculation of the busy hour traffic needs to be done so that adjustments are made to take into account traffic variations over different weeks, over different days of the week and, finally, over different hours of the day. Also account must be taken of traffic surges caused by special events. Other issues must also be considered when working out the busy hour traffic. The distribution of traffic may vary significantly with specific parts of the network, e.g. some switches may serve a bigger share of business as opposed to residential customers than other switches and therefore traffic using those switches is clearly more concentrated. Relying on average busy hour may under-dimension these switches. The usage of the network may vary over time not least because other services, such as data services come along (e.g. internet traffic may make the usage of the network more uniform over the day). Criterion BU 11 Data on "busy hour" traffic should be requested from the SMP operator, where major differences in terms of traffic distribution over time should be identified between different parts of the network and the impact of other services. Post- och telestyrelsen 87

94 14 Bottom-up modelling This section covers issues related to the estimation of equipment prices and specific guidelines on the modelling of exchanges, transmission, the access network and infrastructure Equipment prices and cost data The bottom-up model should use the equipment prices that an efficient operator with the bargaining power of an SMP operator in Sweden would be able to obtain. However, for many assets used in a telecom network there are no recognised market prices; instead if an operator wishes to acquire an asset it will engage in private bilateral negotiations with one or more suppliers. Consequently, the bottom-up model will have to rely on operators providing information on the prices they have paid to acquire a given type of equipment. These may be documented with reference to price lists or contracts and corrected for SMP bargaining power. Naturally, this will only be the case of prices provided from operators other than the SMP operator and where a price-volume relationship can be established and documented. No correction should be made unless it can be clearly documented. It must be demonstrated that the prices collected are appropriate. For example, that prices represent a number of recent contracts and must not be the results of any extraordinary discounts. Equipment prices may include prices for bundled products as long as all the products are related to the modelled network. There may be difference in timing between when an asset becomes operational and when it is paid. Therefore, account may be taken of this by capitalising the interest payable in the meantime. If prices from other LRIC models are used or are taken from operators not operating in Sweden, it should be ensured that they are relevant for the Swedish environment. If this is not the case, the prices should be adjusted to reflect the Swedish environment for the network. Relevant factors include the demographic and geographical profile in Sweden. The exact nature of the adjustment will depend on the equipment in question, and should therefore be undertaken on a case-by-case basis. For example, the costs of trenching should take into account the density of connections and the terrain in Sweden; while the costs of switching should consider the number of switches and the services the network carries in Sweden. Criterion BU 12 Equipment prices and other cost data used in the bottom-up model should reflect those of an efficient operator with the bargaining power of an SMP operator in Sweden. One potential issue of discussion is the implicit relationship between the assumptions guiding the quantity discounts received and the network build out horizon including the possibility of trench sharing with other utilities. Post- och telestyrelsen 88

95 For the purpose of modelling it should be assumed that the network from a technical perspective is built over night, but all input parameters (trench sharing, equipment prices, etc.) should be verifiable and reflect the costs of actual networks built over time. This means that equipment prices may follow from normal operator purchases and sharing may reflect normal planning and construction activity where co-ordination of trench sharing and co-diggings may be planned some years ahead with other operators and utilities. There may be significant differences between the cost estimates provided by different operators. In such cases, clarification will be needed to ensure that the estimates refer to equipment with equivalent specifications Modelling exchanges Section 12 provided the criteria that will need to be met when determining the extent of optimisation in the bottom-up model. The decisions made with regard to optimisation will have implications at two levels. First, they will determine the most cost effective mix of switching technologies that should be used. Second, they will determine the nature and size of each exchange (node). This section provides guidance and high level criteria on the types of issues that are likely to arise on how to determine the nature and size of each node when developing the bottom-up model The hierarchy of the exchanges The trend in telecommunications network over the past few years has been for a flatter switching structure, and for circuit-switched networks, most EU Member States currently tend to have networks with only three levels - transit switches, local switches, and remote subscriber stages. For packet-switched networks, the switching (or routing) structure tends to be more varied. Criterion BU 13 As a starting point, the bottom-up model should consider a hierarchical exchange structure with three layers where a circuit-switched network is assumed. However, provided that a different hierarchical exchange structure can be justified (for example, as a result of assuming a packetswitched network) this may be modelled. The definition, and purpose, of each layer of the hierarchy should be clearly defined Choosing between different nodes The total number of exchanges (nodes) in the SMP operator's network will be the starting point for the modelling of the exchanges in the bottom-up model. Based on this structure, the size and nature of each node must be decided upon given the modelled hierarchy The highest layer of the exchanges The first step in the modelling process will be to model the top layer of the exchanges in the core network. Post- och telestyrelsen 89

96 There are a number of ways to determine the number of nodes in this layer of the exchanges, including. The need for high level switches (and/or routers in a packet-switched network) can be derived from traffic needs. Under this approach, the need would be determined through existing routing factors. For example, for a circuit switched network, the total traffic using this layer of the exchanges would be divided by the average Busy Hour Erlang (BHE) capacity of these switches (allowing for utilisation ratios, etc.) and the optimal number of switches determined. International experience and the experience of other efficient operators could also be used to determine the appropriate number and size of these high level switches/routers. However, corrections are subject to the scorched node constraint and differences in the terms that apply to an efficient operator in Sweden The other layers of the exchanges Once the number of switches/routers at the highest layer of the hierarchy has been estimated, the bottom-up model will need to allocate the remaining nodes to the other layers of the hierarchy. Extensive data is required to provide information on this mix. For instance, information will be needed by type of service (for example, PSTN, broadband/bitstream, leased line) using the network. The mix of switches/routers in the model will depend on the following factors: cost: the cost of serving a certain threshold of customers. The cost will need to include not only the cost of the equipment, but also other costs such as installation, operating costs, accommodation, power and network management. impact on other parts of the network: the bottom-up model should be able to show the cost implications of the chosen mix of switches/routers on other parts of the network. For example, the impact on transmission equipment, the amount of fibre, and the size of trench. security: the implications on the quality of service should also be taken into account. For example, the bottom-up model should show how any proposed mix of switches/routers will affect the grade of service and the resilience of the network. One important aspect of this will be the ability of the network to cope with the breakdown of one or more items of equipment. technical feasibility: any revised mix compared to the SMP operator must be technically feasible. This means that the equipment must be able to handle the increasing amount of traffic and there must be enough switches/routers to effectively host all of the remote subscriber stages, and/or DSLAMs. consistency with the evolution of telecom networks: the optimised equipment must be consistent with the evolution of network design. Post- och telestyrelsen 90

97 Criterion BU 14 The optimisation in the bottom-up model should consider the following factors: cost, the impact on other parts of the network, security, technical feasibility, and consistency with the evolution of the telecom networks Modelling transmission This section defines the minimum requirements with respect to the network architecture and configuration for the bottom-up modelling of transmission Transmission hierarchy The first stage when modelling transmission is to structure the network into tiers. SMP operators in other EU Member States, for example, are increasingly structuring their network into tiers along the following lines: the highest tier, or Tier 1, is used to connect the major nodes for example, transit switches, ATM switches or MPLS routers and to provide connectivity between regions. a second tier, or Tier 2, connects other nodes to the Tier 1 network. the lowest tier, Tier 3, can be used to connect more remote nodes to either Tier 1 or to Tier 2. The nodes on Tier 3 will be smaller switches, concentrators, or DSLAMs. One of the first decisions to be made when developing the bottom-up model is whether the hierarchy presented above is appropriate and if not, provide justification for an alternative choice Network configuration The bottom-up model should investigate the consequences of different configurations in each layer of the transport network. For example, for a circuit switched network, certain parts of the transmission hierarchy may be configured as SDH rings rather than be configured as a meshed network and within the same configuration, there could be different parenting arrangements. When considering different configurations, the bottom-up model will need to take account of different factors, including: the impact on resilience. Double parenting provides more resilience than single parenting; the impact on network management costs. A ring structure usually implies less management costs than a point-to-point structure; and the use of cross-connects. These provide flexibility to the network. Different technologies may also be more appropriate for different parts of the transport network depending on the distribution of the traffic or on the geographic characteristics of the network. Post- och telestyrelsen 91

98 Criterion BU 15 The bottom-up model should show and justify the technologies used in each part of the transmission network Dimensioning the network Once structure, technologies and configurations 47 have been established for each part of the network, the bottom-up model should optimally dimension the transport network on the basis of traffic distribution and equipment costs. The distribution of the traffic in different parts of the network will allow for a more accurate dimensioning than the use of averages, insofar as the cost minimisation problem is repeated for each set of routes of different size. Dimensioning the transmission equipment for each set of routes is another cost minimisation problem, given the amount of capacity which needs to be served on that route. The cost function to be minimised is a linear combination of the costs of the transmission equipment and the costs of the fibre through which the signal travels. The modularity of equipment should be taken into account. Criterion BU 16 Given network technology and configuration, the optimal size of the transmission equipment has to be the result of a cost minimisation problem that also takes into proper account the associated infrastructure costs Modelling access The bottom-up model should estimate the costs of the access network using a mix of actual data of the SMP operator s network as a starting point, but with some optimisation of the equipment in the network. For example, the model could for each geo-type use existing demographic information on the number and type of customers, the mix of dwellings in which they live, and their average distance from exchanges. The model will need to show the following: direct network costs (such as the number of lines, the thickness of cables, the number of primary and secondary connection points); indirect network costs, which can be defined as the cost of those assets that support direct network costs (such as power, accommodation, manholes, etc.); and overhead costs (such as accounting, human resources, etc.). The steps that are involved in modelling the access network may be summarised as: 47 The configuration of the network will also have important implications on how this is dimensioned. So, for example, SDH equipment laid in a ring configuration is dimensioned differently than SDH equipment laid in a point-to-point configuration. Post- och telestyrelsen 92

99 collecting information by geo-types; and given the chosen technology estimating the costs of the relevant access products Collecting information by geo-types The bottom-up model should use the defined geo-types set out in section All the information will need to be collected on a geo-type-by-geo-type basis. There are two types of basic inputs required - demand information and geographic information. Demand information includes the demand for each of the services provided over the access network (such as PSTN, broadband, leased lines, etc.). Geographic information includes data on the average lengths between customers and exchanges. Demand information includes the number of access lines categorised by type. Information is also required on the average distance of customers from exchanges in each geo-type. The model, however, should be able to show the distance from the exchange, RSS or DSLAM etc to the PDP, and the distance from PDP to the SDP and finally from the SDP to the NTP. Information is needed for each of these distances separately because the dimensioning rules for these different parts of the access network will differ. Finally, information will be required on the length of the final drop to the customer s premises. Once all of the information mentioned above is available, the bottom-up model should be able to show demand for access services by geo-type. For each geotype, therefore, the model should show the following information: Number and distribution of products used by all subscribers; Average cable length in the primary access network; Average cable length in the secondary access network; and Average length of final drop to customer's premise Estimation of equipment requirements The recommended approach involves the following stages: selecting a statistically significant sample of exchange areas from each geotype; setting the boundary for each of these exchange areas on the basis of the boundary in the SMP operator s own network; on the basis of detailed maps the most likely source is GIS maps determining the optimal layout of a network, given the known number of subscribers for that exchange area, the dwelling and street pattern; and repeating the exercise for each of the exchange areas and then aggregating up to estimate costs for the geo-type as a whole. Post- och telestyrelsen 93

100 This approach implies the availability of detailed maps. If such information is not available, strong assumptions would be necessary to estimate equipment quantities by using international benchmarks or information from the SMP operator. Criterion BU 17 The model should estimate equipment quantities for the access network using detailed maps and other information for a statistically valid sample of areas. In the absence of such information, alternative approaches, such as data from the SMP operator and international benchmark data may be used. The optimal design of the network, and the mix of costs incurred, will be different in each geo-type and also vary, although to a smaller degree, within geo-type. The equipment requirements will need to take account of those factors considered by an efficient operator when building and operating a network. These factors are sometimes grouped under the general heading of utilisation rates and include, among other things, an efficient level of excess capacity and an allowance for resilience. It should also be ensured that the bottom-up model provides a sufficient level of flexibility in the access network Modelling infrastructure Modelling infrastructure is a vital part of modelling the core and access networks because the cost of infrastructure comprises a large part of the total cost of the network. The definition of infrastructure used in this paper is all the equipment between nodes or distribution points used to carry traffic over the network. Infrastructure equipment is used to provide a number of different services. This implies that a significant amount of shared and common costs, will need to be allocated. This section focuses on issues involved in modelling the following asset classes: the cable containing copper or the optical fibre (cable dimensions vary with the number of pairs contained in it), the duct containing the cable (duct dimensions vary with the number of bores contained in it); and trenches containing the duct or cable buried directly in the trench. Criterion BU 18 The bottom-up model should show infrastructure costs of cable, duct and trenches separately Trench in the core network Given the scorched node assumption, the amount of trenches to be costed in the core network is a function of: the configuration of the network; and Post- och telestyrelsen 94

101 the actual distances between different nodes. If modelling a different configuration than the existing one, the amount of trench needed would depend on actual distances between nodes belonging to the same layer and, to a lesser extent, between nodes belonging to different layers. The amount of necessary information will be substantial and in case averages are used instead of actual numbers, the results are likely to be very sensitive to this information. It should be noted, for example, that many routes may share a single trench. If this factor is not taken into consideration, trench requirements may be significantly over-estimated. The total amount of existing trench by different network layers may be used as a cross check of whether the modelled configuration bears some resemblance with the actual one. Criterion BU 19 The assumptions regarding distances between nodes belonging to the same layers and also between nodes belonging to different layers, on which the amount of trench modelled rely on, should be clearly identifiable and justified for each part of the network. If a theoretical method is used to calculate distances between nodes, this method should be comprehensible and explainable. This may, for instance, mean that a map is drawn of the network's nodes and lines Costing of trench The cost of trench will depend on: the type of terrain; and the geo-type (with trenches to be dug in metropolitan areas being more expensive than trenches in urban and rural areas). Criterion BU 20 The bottom-up model should take proper account of different terrains and geo-types when costing trench, accepting that the shortest route is not necessarily the most cost effective Duct The amount of duct depends on how much cable, within the trench, needs to be ducted. Here, cost and quality considerations should be taken into account. Ducted cable is more costly than the cable buried directly in a trench but also more resilient and providing increased flexibility. Criterion BU 21 The bottom-up model should show and justify the amount of cable (out of the total amount of trenches modelled) that is put in duct (as opposed to buried cable) on the basis of general cost and quality considerations. Post- och telestyrelsen 95

102 Given the amount of duct to be modelled, its cost mainly depends on its size, usually measured in number of bores. The number of bores mainly depends on the number of cables contained in the same duct, given that a lot of routes may share the same trench/duct. Duct sharing mainly varies with geo-types and with network layers (with the higher layers of the network likely to share more routes than the lower ones). Other considerations, such as planning, for example, also need to be taken into account. Criterion BU 22 The bottom-up model should show and justify, for each part of the network, the size of the modelled duct. Trenches and ducts are the main source of common costs between the core and the access network and they can be common to other utilities as well. Rental costs in connection with laying cable in other utilities' trench/duct should be included in the modelling and documented. As stated in section 14.1, the network is technically considered to be built over night. However, input parameters (trench sharing, equipment prices, etc.) should be verifiable and should reflect costs of actual networks (built over time). This means that opportunities for co-digging should be considered and justified. Criterion BU 23 The bottom-up model should show the amount, or proportion, of duct and trench that is common to the core and the access network and any other utility. Where the telecommunications network and other utilities are deemed to share a proportion of duct and trench, costs of shared stretches of trench should as a starting point be split pro rata to either the number of used ducts or used cables. The chosen cost driver should be documented and justified Poles In some parts of the network, especially in rural areas or in particular types of terrain, it may be more cost effective to use poles rather than bury cable in trenches. The bottom-up model should consider whether copper/fibre cable strung on poles is a cost-effective solution Cable requirements in the core network Given that each transmission system usually needs a pair of fibre dedicated to it, one for each direction the signal travels, the results of the transmission modelling will represent the starting point to estimate cable requirements in the core network. Post- och telestyrelsen 96

103 However, it is not sufficient to establish the length and the size of the fibre-cable, since fibre serving different transmission systems share routes (route sharing) and routes share trenches (trench sharing). This sharing enables multi-pair fibres to be used to economise on fibre usage. Criterion BU 24 The bottom-up model should as a minimum consider, for each layer of the network, route sharing, trench sharing, average route lengths and total amount of modelled duct in order to determine the cable size requirements for each part of the core network in terms of number of fibre pairs per cable Cable modularity and length Cable size requirements may be satisfied by different combinations of cables of different sizes. The needs for different cable sizes should be determined taking into consideration the future demand to mirror the fact that digging of new cable represents a substantial cost. The need for excess capacity should therefore be based on rational economic considerations taking into account modularity and margins for growth Criterion BU 25 Given cable size requirements, for each part of the network, the bottomup model should take into account cost considerations and modularity of cables to work out the optimal combination of cables of different sizes. Trench length, for each part of the network, should be applied to the optimal combinations of cable of different sizes in order to work out the total length of cable. Criterion BU 26 Given optimal combinations of cables of different sizes, the bottom-up model should consider the trench length for each part of the network to work out the total length of cable of different sizes. This should include the cable waste that an efficient operator should expect due to cutting-off and modularity Modelling co-location The bottom-up model should include the co-location services related to unbundled local loop. The cost categories to be modelled are set out in section 4.3. Post- och telestyrelsen 97

104 15 Bottom-up costing issues 15.1 Indirect network costs Indirect network costs are those costs that are required for the network to function, but often depend on other inputs. They include costs such as power and racks. These types of assets are difficult to model directly in a bottom-up model and consequently, they are often estimated on the basis of a mark-up on direct network costs. In the bottom-up model, the appropriate approach to modelling indirect network costs should be decided on a case-by-case basis. The guiding principles should be the availability of information and the materiality of the cost category. For some costs, such as accommodation, an effort should be made to estimate the costs directly. The costs of accommodation could be estimated in a bottom-up model in two ways - the first is to use a mark-up, the second is to estimate the space occupied by different equipment and calculate accommodation costs based on a building space cost per square meter. The latter approach should be adopted to estimate the costs of accommodation where care has been taken to dealing with common accommodation costs. Criterion BU 27 Building space costs should be determined as a cost per square metre. The values should be categorised by geo-type. For other indirect network costs, such as racks and power, an effort should also be made to modelling these costs directly. However, for indirect costs such as network management, a mark-up approach may be more sensible. The mark-up used should reflect the costs that would be incurred by an efficient operator. Criterion BU 28 Indirect costs should be calculated in the most appropriate manner taking into account the availability of information and the materiality of the cost category in question. Where mark-ups are used, they should be justified and should reflect an efficient level of indirect network costs. All indirect network costs should be shown separately Overheads Overhead costs (or indirect non-network costs) are costs that are incurred to operate a telecommunications company but which are not directly incurred to provide a core and access network. Examples include the human resources, legal, and planning departments. Only the indirect non-network costs associated with the wholesale core and access service should be included. Retail costs, for example, should be excluded. Further, only an efficient share of these costs appropriate for an efficient network operator building and operating a core and access network in Sweden should be included. Post- och telestyrelsen 98

105 Criterion BU 29 Overhead costs should only be included if they are efficiently incurred in building and operating a wholesale core and access network in Sweden Operating costs The major operating costs are those concerned with maintaining the network and providing or rearranging services to customers. 48 With regards to calculating access and interconnection charges, only costs related to the wholesale division should be included. Retail costs should not be included. Ideally, the bottom-up model should use theoretical cost estimates as its basis for operating costs. This would imply identification of all the major activities by an SMP operator giving rise to operating costs and estimation of the operating costs of each activity. In practice, this is a difficult exercise and may result in activities being missed without a very detailed understanding of the SMP operator s operations. Alternatively, operating costs may be estimated on a more aggregated level and used to estimate operational costs as a percentage of the equipment capital costs. To estimate the ratio of operating costs to investment costs, the model might use: the SMP operator s actual performance; the performance of SMP operators in other countries, typically the US; or the performance of other operators. Another option could be to base operating costs on an event based system, where costs are driven by the number of times a given event (e.g. repair) occurs. Apart from estimating the frequency of events this would entail estimating e.g. average repair times, the associated capital expenses, hourly wages, etc. All these methodologies have limitations. All of them may measure the operating costs of networks that are very different to the network modelled in the bottomup model. Consequently, adjustments may be necessary to account for any differences. Criterion BU 30 Operating costs should be calculated in the most appropriate manner taking into account the availability of information. Where mark-ups are used, they should be justified and should reflect an efficient level of operating costs. The approach used should be justified. Modern equipment and the introduction of computer support systems have reduced the manpower element associated with running a network. Nevertheless, labour costs remain a substantial element of operating costs, especially in the local loop. They are also likely to be geo-type dependent; the operating costs in a rural area are usually significantly higher because of the absence of scale economies. 48 The definition of operating costs does not include depreciation. Post- och telestyrelsen 99

106 To measure the labour costs that an efficient operator would pay requires an allowance for what might be termed ineffective time. This covers such things as when an employee is sick, on annual leave, training or travelling. How to take account of ineffective time will depend on the methodology used to calculate operating costs. Criterion BU 31 When estimating operating costs, the model should allow for ineffective time when calculating labour costs. The percentage of ineffective time should be justified in the model documentation Cost allocation Once the total annual costs of network elements have been estimated, these have to be allocated to services. As stated in section 5.4.4, the linkage between the cost of network elements and service costs is provided by routing factors that specify the usage of a given network element. This usage should be determined with reference to the total volume of traffic. Criterion BU 32 Final cost allocation, from network elements to core services, should be based on the total volume of traffic. Post- och telestyrelsen 100

107 16 Model functionality and documentation 16.1 Model requirements The bottom-up model will form the basis for the hybrid model and will need to be reconciled with the top-down model. Further, the bottom-up model and the hybrid model will be subject to an extensive hearing process. For these reasons, the model should: be based on a standard software package such as MS-Excel; have a simple and logical structure with many simple steps in the processing rather than a few complex ones; have formulae which draw on inputs using links to the relevant cells rather than having inputs hard-coded as part of the formula and there should be no redundant formulae or inputs; make use of different styles (e.g. colour coding) to ensure that cells in the model are easily identified; and have explanatory notes and a clear audit trail, allowing a third party to follow through the calculations from inputs to results. Criterion BU 33 The bottom-up model should be structured so that the key principles and the most significant algorithms used are clearly shown Sensitivity analysis As stated in the previous section, the inputs of the model should be easily identifiable. It is important, given that most of the inputs will involve assumptions on the structure of an hypothetical optimal network in Sweden, to identify those inputs to which the model outputs are most sensitive. For this reason, the bottom-up model should include a tool that enables separate sensitivity analyses to be carried out. Criterion BU 34 The bottom-up model should be able to identify key inputs as the ones the cost estimates, at least at service level, are most sensitive to and perform a sensitivity analysis on these. These include: Traffic volumes; Equipment prices; Utilisation rates; Quality of service parameters; Sharing parameters; Post- och telestyrelsen 101

108 Key technical input and network design rules; Cost of capital; Asset lives; Price trends; Operating costs; and Indirect costs Model documentation The aim of the requested documentation is to put any model user with a reasonable knowledge of the issues treated in a position to understand both the main features of the model and its technicalities. The topics covered by the model documentation should in general provide the rationale for the choices made when following the modelling guidelines. Many of these already require certain decisions to be justified or documented. Therefore, they form a natural part of the model documentation. The documentation should explain: all algorithms and formulae. For example, how the model derives annual charges out of the equipment investment costs and derives costs for the relevant products. Further, the documentation should clearly show how indirect network costs and overheads have been modelled and the methodology used to model operating costs should be clearly shown; how common and shared costs have been allocated and the functionality in place to allow the model user to use different allocation keys, for both common and shared costs; and how the different cost categories are aggregated into cost estimates of network elements and, finally, into the cost estimates of the relevant services. Apart from the justifications and explanations provided above, the model documentation should aid the user or reader of the model with overviews and lists. Therefore, a list of all numerical inputs and a rationale for their value should be provided. These include: Information on volumes. PSTN minutes and number of calls, broadband usage, leased lines number of circuits by bandwidth, etc; Network structure and configuration. Routing factors, average length of transmission routes, amount of trench modelled, etc. Costs for each asset, information on unit costs, lifetimes, price trends, etc. Finally, a list of how general and specific bottom-up criteria have been met should be provided. The model as well as the model documentation should be in English. Supporting documentation, such as contracts, or analyses carried out in Swedish may be provided in Swedish. Post- och telestyrelsen 102

109 APPENDICES Post- och telestyrelsen 103

110 Appendix 1 Summary of criteria Criterion CG 1 The models should be based on forward-looking long run incremental costs. No migration costs should be included. Criterion CG 2 For the core network, the increment should include all services using the core network. For the access network, the increment should include all services using the access network. The LRIC of co-location is the cost incurred in providing co-location services. These definitions should include the services provided by the SMP operator s network division to its own retail division as well as the services provided to other operators. Criterion CG 3 The access-core line card (typically, though not necessarily, for the PSTN located in the concentrator) should be included in the access network, whereas other equipment related costs should be included in the core network, except where costs are common between the two networks. The access-core line card(s) should be excluded from the costs of unbundled local loops but included in the costs of the access service it relates to (such as telephone line rental for a PSTN line card, broadband access for a DSLAM line card etc.). Criterion CG 4 To the extent practical, costs (both capital costs and operating costs) should be allocated to services on the basis of cost causality. This assumes that the network is constructed in an efficient manner and does not, without good, justifiable reasons, separate services such that not all services attract a fair proportion of cost. Criterion CG 5 The models should allow recovery of common costs. These costs should be shown separately. Criterion CG 6 The models should identify the costs that are common between the other increments and the core and access networks. Criterion CG 7 As far as possible, common costs should be allocated to increments and services using appropriate (direct or indirect) cost drivers. Only common costs, for which it is not possible to identify the extent to which a specific increment or service causes the costs, should be allocated via mark-ups. The starting point should be equi-proportionate mark-ups. The models should allow for equi-proportionate mark-ups to be used for all cost categories. It is possible that there could be instances where there might be good reasons for departing from equi-proportionate mark-ups. However, if this is the case, it must be properly justified in the model documentation. Criterion CG 8 The models should include all standard PSTN/ISDN and Broadband services.

111 Criterion CG 9 When dimensioning the network, the leased lines traffic volume should include leased lines provided to retail customers, to other operators and to the network operator itself. Criterion CG 10 Where possible, the models should categorise the other services into two major groups: one category comprising for example cable TV services, IP TV services and other services using their own non-telecommunications electronic equipment. one category comprising non broadband data services (by type) using the core network. Criterion CG 11 The models should identify busy-hour information for traffic. The models should be flexible enough to allow for changes in these figures. Criterion CG 12 The network dimensioning should correspond to what an efficient operator facing the forecasted demand would do. The models should show the anticipated Cumulative Annual Growth Rate (CAGR) for each service for a five year period, following the base year, The models should allow for a change in the margins for growth. The models should use the following planning horizons as a starting point: 5 years for the access network and infrastructure in the core network; 3 years for exchange equipment; 3 years for transmission equipment; 3 years for backhaul broadband equipment (core routers etc); and 1 year for DSLAMs. For line cards a 1-year planning horizon should be used as the starting point. If different planning horizons are used, this will need to be justified. Criterion CG 13 Cost categories should, as far as possible, be identified to obtain only one exogenous cost driver for each category. Criterion CG 14 Costs related to assets can include capitalised operating costs when there is a rationale for it. These costs should be shown separately in the documentation. Criterion CG 15 The modelled co-location services should include the following cost categories common to both co-location and other services in the core and access network: Land and buildings (annual costs); Site preparation and fit-out of buildings (one-off and/or annual costs); Security systems, fire surveillance, etc. (one-off and/or annual costs); Power supply (annual costs); and Cooling/ventilation (annual costs). The specific co-location services to be costed include:

112 Administrative staff (one-off costs and annual costs); Technical staff (one-off costs); Racks ( ETSI-skåp ) (one-off and annual costs); Co-location specific power supply inclusive power consumption (one off and annual costs); Co-location specific cooling/ventilation (one-off and annual costs); and Cables (one-off and annual costs). Each cost category should include one-off and annual costs as shown above. Criterion CG 16 The models should distinguish between the costs that are specific to PSTN services, costs that are specific to shared access, and the costs that are shared between the PSTN services and the shared access service. The additional cost of shared access (compared to PSTN) should be shown as a separate output of the models. Criterion CG 17 The total annualised cost of the actual (raw) copper pair should as a starting point be the same whether it is used for providing PSTN services, full access, shared access or Bitstream access plus PSTN services. If it is decided to cost Bitstream capable copper pairs separately from non-bitstream capable copper pairs then this must be justified and documented. Criterion CG 18 When modelling the cost of both shared and full Bitstream access, the models should be able to distinguish between the costs of: Capacity on the copper; capacity in the DSLAM; and transport of traffic from the DSLAM to the nearest point in the SMP operator's network available for Bitstream interconnect. The costs of the modem at the customer's premises, and any costs related to installation of a filter at the cusomer s premises, should not be included in the cost of Bitstream access and, if included within the modelling, should be shown separately. The additional cost of Bitstream access (compared to PSTN) should be shown as a separate output of the models. Criterion CG 19 The starting point for the top-down model should be straight-line depreciation, whereas the starting point for the bottom-up model should be (tilted) annuities. Criterion CG 20 The models should use an interim nominal pre-tax cost of capital of 10.8%. The models should allow the cost of capital to be altered. Criterion CG 21 For access services, the models should provide separate costs by geo-type along with the average national costs. Criterion CG 22 The models should model the costs for 2006 Criterion CG 23

113 The models should include a calculation for the cost of the working capital of an efficient operator, unless zero is used in which case it can be shown as an input. Criterion CG 24 The models should distinguish between set-up related and duration related costs. This requires the calculation of both set-up and duration related costs, network elements and routing factors. IN costs should be separately identified from other signalling related costs. This will imply defining separate IN network element(s) and corresponding routing factors. Criterion CG 25 The models should show routing factors for (at least) each of the following PSTN related network elements Concentrator (RSS or HSS); Local Exchange (LE); Tandem Exchange (TE); RSS-LE transmission; LE-LE transmission; LE-TE transmission; and TE-TE transmission. Information should be provided separately for all the major call types. In addition, the documentation should include information (for all calls) showing the percentage of calls following a particular routing pattern (e.g. 2 RSSs, 1 local exchange, 2 RSS-LE transmission links). Routing factors for Bitstream related network elements should also be provided as necessary. Criterion CG 26 Access network elements should be split into geo-types where appropriate and "routing factors" (usage factors) specified for each service to be costed. Criterion TD 1 Asset valuation should reflect the replacement costs of the modern equivalent asset (MEA) Criterion TD 2 The MEA is the asset that produces the same outputs produced by the existing asset at lowest costs Criterion TD 3 The top-down model should value assets on the basis of an absolute valuation. Any use of indexation will need to be justified by supporting documentation and should only be used where there has been no technological change. Where the difference between the current and historic cost of the asset is likely to be small relative to the overall gross asset valuation, or the asset life is short (3 years or less), the topdown model may use historic costs. No more than 5% of the total value of the assets may be valued according to historic costs. The model documentation should justify the decision to use historic costs.

114 Criterion TD 4 The SMP operator should have available for inspection documentary evidence of asset prices used within the model. Criterion TD 5 If the MEA implies differences in operating costs, quality, asset lives or space requirements, the asset value should be adjusted to account for the effect these differences will have on the net revenues generated by the asset for each year of the asset s lifetime. For material differences in operating costs, the adjustment can either be applied to the asset value or shown as a separate specific adjustment to the operating costs for those assets. Where the replacement cost of the new asset is higher than the existing asset, the operator should document (through the MEA adjustment) that the total current cost associated with the new asset is lower than the total current cost of the existing asset. Criterion TD 6 No value should be attached to vacant space, except where it may be shown that it is economically rational to maintain this vacant space using a 2-3 year planning horizon. The amount of vacant space that has been excluded should be made clear either within the model itself or within the associated documentation. Criterion TD 7 Where land and building costs are treated as operating costs, it should be justified that the rent corresponds to a market-based rent. Again, the costs of inefficient vacant space should not be included. Criterion TD 8 An SMP operator should show the utilisation level for exchanges, transmission equipment, optical fibre, and copper cable and justify why this is efficient. Criterion TD 9 The top-down model documentation should show both the utilisation rates in the SMP operator's actual network as well as the utilisation rates used in the model. Where the model applies utilisation rates that are lower than the actual utilisation rates, it should be justified that this does not lead to higher overall costs. Criterion TD 10 SMP operators should justify their ratio of actual number of pairs to subscriber lines in different parts of their access network and in different geo-types. The statistical validity of the sample of routes chosen should be explained within the model documentation. Criterion TD 11 SMP operators should justify their use of fibre in the access network. If copper would be cheaper and there is no other justification for using fibre then the asset should be valued with copper as the MEA for fibre. Criterion TD 12

115 SMP operators should justify their use of radio in the access network. If copper would be cheaper and there is no other justification for using radio then the asset should be valued with copper as the MEA for radio. Criterion TD 13 SMP operators should quantify separately the costs of call duration and call attempts. SMP operators should also provide information to support their policy on exchange lives and, where appropriate, justify attaching the same asset lifetime to hardware and software. Finally, where software update contracts relate to exchanges in different levels of the network hierarchy, costs should be split in proportion to determine exchange costs for each class of exchange. Criterion TD 14 PDH equipment can encompass both classical PDH equipment (now largely replaced by SDH equipment) and also PCM equipment (often used at the edge of the core network for low capacity routes). If circuit switching is retained for the PSTN then, for the former, the top-down model should, as a starting point, value PDH equipment using SDH as the MEA. For the latter low capacity PCM routes, the SMP operator should adopt an MEA that is both efficient and cost effective and should justify any alternative to the equipment currently used. The model documentation should justify the value of cross-connects attributed to PSTN. Criterion TD 15 The SMP operator should provide justification if it decides to retain ATM-based backhaul as the appropriate technology within the model. Criterion TD 16 SMP operators should justify current installation practices for optical fibre in the core network. If no justification can be provided for excess optical fibre, it should have no valuation. Criterion TD 17 The top-down model (documentation) should identify trenching costs for different terrain types. The model documentation should explain the rationale for differences in these costs in different parts of the network. Criterion TD 18 Where the telecommunications network and other utilities share duct and trench, costs of shared stretches of trench should be split pro rata to the number of used ducts or used cables depending on the trench sharing agreement adopted unless a more appropriate cost driver exists. The chosen cost driver should be documented and justified. Criterion TD 19 The model documentation should distinguish between copper/fibre cable that is trenched and copper/fibre cable strung on poles. Criterion TD 20 The apportionment of cost in cost categories that are common to co-location and access/core should be appropriately documented to ensure that there is no double-counting of costs.

116 Criterion TD 21 When determining asset lives in the top-down model, the starting point should be book lives. If asset lives are set on the basis of economic lifetime, the model should also be able to show the costs calculated on the basis of book lives. When the asset lives used in the model differ from the book lives, this will need to be justified for each category and the impact on costs should be documented. Any related adjustments to the capital base will also need to be justified and documented. When using a MEA that is different from the existing asset, the model should apply the asset life of the existing asset. As discussed in section , the valuation should be adjusted for any differences in asset lives. Criterion TD 22 The top-down model should ensure consistency between asset lives and the treatment of fully depreciated assets. The model should show a calculation where FDA is attributed no value (using book lives). In addition the model may show a calculation where the economic value of fully depreciated assets is included, provided this is consistent with the assumptions made regarding asset lives. Asset lives should be applied consistently to both fully depreciated assets and other assets. The SMP operator should show the extent of fully depreciated assets by asset class and, where possible, vintage. Criterion TD 23 The top-down model should measure annualised costs using the FCM approach. Costs should include holding gains or losses, provided the model documentation can justify their inclusion. Criterion TD 24 The top-down model should estimate net asset values on the basis of either the NBV/GBV methodology (for each asset category) or the rolling forward methodology. Preference should be given to the rolling-forward methodology subject to data availability. If different approaches are used for different asset categories, this should be documented and justified. Criterion TD 25 The top-down model should only include efficiently incurred costs related to the wholesale activities. No redundancy costs should be allocated to access and interconnection services. Criterion TD 26 The top-down model should allocate operating costs to the various services on the basis of cost causality, such as ABC. Supporting documentation should describe the cost drivers and how the model assumes they affect operating costs for each activity. The documentation should also describe which activities the different services consume. Criterion TD 27 The model should separate costs into homogeneous cost categories. The SMP operator will need to provide justification for its selection of cost categories. Criterion TD 28

117 Separate volume measures are required for access and core trench/duct and potentially for other increments. Criterion TD 29 Separate volume measures are required for subscriber stages and processors, by manufacturer of exchange, for lines, call duration and call attempts. In addition, where software and hardware lifetimes differ significantly, separate cost-volume relationships are required. Criterion TD 30 In order to measure respective volume usage over a specific system line containing both PSTN and non-pstn services, PSTN minutes should be converted into Mbit equivalents. Further adjustments are required to take account of diversity and differentials in the intensity of usage. Criterion TD 31 The starting point should be to use subscriber lines as the volume driver for DLSAM line cards and Gigabytes as the volume driver for DSLAM common parts and also for Ethernet switches and IP routers. Departures from this, for example to take account of Quality of Service requirements, should be justified and documented. Criterion TD 32 For staff related volume measures, the SMP operator should provide documentation justifying the way in which usage intensity has been measured by functional area within its organisation. Criterion TD 33 Where the construction and/or operation of the network, or any associated activities, have been outsourced to another company or subsidiary, the SMP operator should, where possible, undertake the cost modelling in co-operation with these companies to ensure that it is possible to establish a clear link between the underlying cost-drivers and the costs incurred by the SMP operator for those activities. Criterion TD 34 In the core and access increments, the top-down model should assign costs to the various services within those increments. The total of the costs assigned to the various services should correspond to the total cost of that increment. For each cost category, the documentation should provide a rationale for the assignment. There should be consistency in the assignment rules used for the different cost categories. Criterion TD 35 The model should show how the costs of services are derived by multiplying the routing factors for each services to the costs of network elements. The total LRIC of a service should show a breakdown between the LRIC and the mark-up for common costs. Unless standard-software such as MS Excel is used, PTS should be provided with the facility to run the model. Criterion TD 36

118 Ideally, the model should be accompanied by the audit report. However, it will be sufficient if the audit report is provided within 6 weeks of the delivery of the model. Criterion BU 1 The bottom-up model should comply with the modified scorched node assumption where nodes are defined as technical house locations (which might contain PSTN switches, concentrators, DSLAMs and potentially multiplexers), that is to say the existing number and locations of sites are fixed, but no empty sites are allowed although it is possible to change the number and mix of equipment at a site. Criterion BU 2 The core network in the bottom-up model could be based upon either circuit-switched or packet-switched technology according to accepted industry standards. The choice adopted should be justified and documented. Criterion BU 3 The predominant transmission technology for circuit-switched networks should be SDH, whereas for packet-switched networks it should be Ethernet switches and IP routers. As a starting point, DWDM should not be included on the basis the Bottom Up model should assume sufficient fibres are available within the cables. Microwave transmission should be used only where fibre is not cost effective. Criterion BU 4 The access network in the bottom-up model should be modelled according to accepted industry standards. As a starting point it should model a fibre access network for customers connected with fibre today and equivalently model copper in the local loop for customers connected by copper today. However, if some customers are currently connected via radio in the access network, then the SMP operator has accepted that this technology provides a suitable level of service. Therefore, in such a case, radio may be modelled as an alternative to copper where this is cost effective. Criterion BU 5 The bottom-up model should demonstrate that the optimised network provides services at an appropriate level of quality for an efficient SMP operator. Criterion BU 6 The bottom-up model must be able to demonstrate that the optimised network can carry the dimensioned demand. Criterion BU 7 When measuring the level of PSTN traffic, the bottom-up model should take unsuccessful calls and ringing time into account. Criterion BU 8 The model should show the demand for leased lines by number of circuits by capacity bandwidths. Demand for broadband/bitstream and other services should be shown by different categories of services and, within each category, by different capacity bandwidths.

119 Criterion BU 9 The routing factors used in the bottom-up model need to be consistent with the underlying network architecture and identified separately for each service. Criterion BU 10 The bottom-up model should show how service-specific adjustments for resilience have been taken into account in the given network architecture. Criterion BU 11 Data on "busy hour" traffic should be requested from the SMP operator, where major differences in terms of traffic distribution over time should be identified between different parts of the network and the impact of other services. Criterion BU 12 Equipment prices and other cost data used in the bottom-up model should reflect those of an efficient operator with the bargaining power of an SMP operator in Sweden. Criterion BU 13 As a starting point, the bottom-up model should consider a hierarchical exchange structure with three layers where a circuit-switched network is assumed. However, provided that a different hierarchical exchange structure can be justified (for example, as a result of assuming a packet-switched network) this may be modelled. The definition, and purpose, of each layer of the hierarchy should be clearly defined. Criterion BU 14 The optimisation in the bottom-up model should consider the following factors: cost, the impact on other parts of the network, security, technical feasibility, and consistency with the evolution of the telecom networks. Criterion BU 15 The bottom-up model should show and justify the technologies used in each part of the transmission network. Criterion BU 16 Given network technology and configuration, the optimal size of the transmission equipment has to be the result of a cost minimisation problem that also takes into proper account the associated infrastructure costs. Criterion BU 17 The model should estimate equipment quantities for the access network using detailed maps and other information for a statistically valid sample of areas. In the absence of such information, alternative approaches, such as data from the SMP operator and international benchmark data may be used. Criterion BU 18 The bottom-up model should show infrastructure costs of cable, duct and trenches separately.

120 Criterion BU 19 The assumptions regarding distances between nodes belonging to the same layers and also between nodes belonging to different layers, on which the amount of trench modelled rely on, should be clearly identifiable and justified for each part of the network. If a theoretical method is used to calculate distances between nodes, this method should be comprehensible and explainable. This may, for instance, mean that a map is drawn of the network's nodes and lines. Criterion BU 20 The bottom-up model should take proper account of different terrains and geo-types when costing trench, accepting that the shortest route is not necessarily the most cost effective. Criterion BU 21 The bottom-up model should show and justify the amount of cable (out of the total amount of trenches modelled) that is put in duct (as opposed to buried cable) on the basis of general cost and quality considerations. Criterion BU 22 The bottom-up model should show and justify, for each part of the network, the size of the modelled duct. Criterion BU 23 The bottom-up model should show the amount, or proportion, of duct and trench that is common to the core and the access network and any other utility. Where the telecommunications network and other utilities are deemed to share a proportion of duct and trench, costs of shared stretches of trench should as a starting point be split pro rata to either the number of used ducts or used cables. The chosen cost driver should be documented and justified. Criterion BU 24 The bottom-up model should as a minimum consider, for each layer of the network, route sharing, trench sharing, average route lengths and total amount of modelled duct in order to determine the cable size requirements for each part of the core network in terms of number of fibre pairs per cable. Criterion BU 25 Given cable size requirements, for each part of the network, the bottom-up model should take into account cost considerations and modularity of cables to work out the optimal combination of cables of different sizes. Criterion BU 26 Given optimal combinations of cables of different sizes, the bottom-up model should consider the trench length for each part of the network to work out the total length of cable of different sizes. This should include the cable waste that an efficient operator should expect due to cutting-off and modularity. Criterion BU 27

121 Building space costs should be determined as a cost per square metre. The values should be categorised by geo-type. Criterion BU 28 Indirect costs should be calculated in the most appropriate manner taking into account the availability of information and the materiality of the cost category in question. Where mark-ups are used, they should be justified and should reflect an efficient level of indirect network costs. All indirect network costs should be shown separately. Criterion BU 29 Overhead costs should only be included if they are efficiently incurred in building and operating a wholesale core and access network in Sweden. Criterion BU 30 Operating costs should be calculated in the most appropriate manner taking into account the availability of information. Where mark-ups are used, they should be justified and should reflect an efficient level of operating costs. The approach used should be justified. Criterion BU 31 When estimating operating costs, the model should allow for ineffective time when calculating labour costs. The percentage of ineffective time should be justified in the model documentation. Criterion BU 32 Final cost allocation, from network elements to core services, should be based on the total volume of traffic. Criterion BU 33 The bottom-up model should be structured so that the key principles and the most significant algorithms used are clearly shown. Criterion BU 34 The bottom-up model should be able to identify key inputs as the ones the cost estimates, at least at service level, are most sensitive to and perform a sensitivity analysis on these. These include: Traffic volumes; Equipment prices; Utilisation rates; Quality of service parameters; Sharing parameters; Key technical input and network design rules; Cost of capital; Asset lives; Price trends;

122 Operating costs; and Indirect costs.

123 Appendix 2 Abbreviations used ABC AD ADM ADSL APT ATM BHCA CAPM CAPEX CAGR CC CCA CoC CVR DEA DGM DP DSLAM DWDM EPMU EU EV FCM FDA FL-LRIC FO GBV GIS GoS GRC HCC HSS Activity Based Costing Accumulated Depreciation Add Drop Multiplexer Asymmetrical Digital Subscriber Loop Arbitrage Pricing Theory Asynchronous Transfer Mode Busy Hour Calling Attempts Capital Asset Pricing Method Capital Expenditure Cumulative Annual Growth Rate Current Cost Current Cost Accounting Cost of Capital Cost Volume Relationship Data Envelopment Analysis Dividend Growth Method Distribution Point Digital Subscriber Line Access Multiplexer Dense Wavelength Division Multiplex Equi-Proportionate Mark-Up European Union Economic Value Financial Capital Maintenance Fully Depreciated Assets Forward Looking Long Run Incremental Costs Förmedlingsområde, or Transit Area Gross Book Value Geographic Information System Grade of Service Gross Replacement Cost Homogenous Cost Category Host Subscriber Stage

124 I/C IN IP IT ISDN Kbit LE LLU LRAIC LRIC Mbps Mbit MDF MEA MPLS MRP MS NBV NPV NRC NRV NTP OCM OLS PC PCM PDP PDH POI PONS PSTN PTS QoS RC Interconnect Intelligent Network Internet Protocol Information Technology Integrated Services Digital Network Kilobit Local Exchange Local Loop Unbundling Long-Run Average Incremental Cost Long-Run Incremental Cost Megabits per second Megabit Main Distribution Frame Modern Equivalent Asset Multi-Protocol Layer Switching Model Reference Paper Microsoft Net Book Value Net Present Value Net Replacement Cost Net Realisable Value Network Termination Point Operating Capital Maintenance Ordinary Least Squares Personal Computer Pulse Code Modulation Primary Distribution Point Plesiochronous Digital Hierarchy Point of Interconnect Passive Optical Network System Public Switched Telephony Network Post och Telestyrelsen Quality of Service Replacement Cost

125 RIO RSM RSS SAC SDP SDH SFA SMP SOYD SSP TE ULL USO VLAN VPC VPN WDM xdsl Reference Interconnect Offer Remote Multiplexer Remote Subscriber Stage Stand Alone Costs Secondary Distribution Point Synchronous Digital Hierarchy Stochastic Frontier Analysis Significant Market Power Sum Of Year Digits Service Switching Point Transit (also Tandem) Exchange Unbundled Local Loop Universal Service Obligation Virtual Local Area Network Virtual Private Circuit Virtual Private Network Wavelength Division Multiplexing x Digital Subscriber Loop (includes such technologies as ADSL)

126 Appendix 3 Glossary of terms The glossary provides an overview of the terminology used in this LRIC model reference paper. The terminology has been established for the purpose of the LRIC process in Sweden. Where possible the general terminology of the telecommunication industry has been adopted, although there will be cases where this is not appropriate. The list is not intended to provide a comprehensive overview of the economic and technical terms. Rather the purpose is to define the broad concept of the terms where misunderstandings potentially could arise due to differences in terminology. The definitions provided in this glossary are intended to provide clarification to information and guidance contained in the Reference Model Paper. The network definitions regard physical and not "logical networks". Access Network: A set of assets, and associated operating activities, put in place to connect the customer s premises to the nearest exchange. The access network also carries leased lines and other services. The line card (or its equivalent) is included in the access network and represents the boundary between the access and core networks. The access network can be divided into three sections: the primary access network; the secondary access network; and the final drop cable to the customer. Each of these sections is described briefly in this glossary. A schematic illustration of the access network is found at the bottom of this glossary. Annualised Costs: The cost to be assigned to annual accounts for an asset. Annualised costs are made up of a capital charge and a depreciation charge. Capital Charge/Costs: The cost of capital multiplied by the average value of the asset for the year under review. Common Costs: The costs of those inputs necessary to produce one or more services in two or more increments, where it is not possible to identify the extent to which a specific increment causes the cost. Trenching costs provide a good example of the difference between shared and common costs. The costs of trenching specific to the access network (or the transport network) will generally be shared costs since the trenching is likely to be used by two or more services. However, some trenching will be used by both the access and the core network. In these instances, the costs will be common costs. Another example of common costs is corporate overheads. Common costs can be either fixed common costs or joint costs. Concentrator (Unit): Equipment capable of concentrating the traffic (reducing the number of subscriber lines from the access network to a lower number of transmission links towards the local exchange). May or may not include switching capability.

127 Core Network: A set of assets, and associated operating activities, put in place to switch and transport the traffic over the network. Includes the transport network and the exchanges. An illustration of the asset categories included in the core and access network is included at the bottom-of this glossary. Cost Category: A grouping of costs with an identical cost driver. (This does not mean that all costs with the same cost driver should be included in a single cost category.) Cost Driver: The factor that causes a cost to be incurred. For example, the number of subscribers is the cost driver for the cost of line cards. Cost of Capital: The required rate of return on capital. Cost-Volume Relationship: The relationship between the cost and the volume of the cost driver. Demand Elasticities: The percentage change in quantity divided by the percentage change in price. A variety of elasticity measures can be calculated. Own price demand elasticity measures the percentage change in the quantity demanded of good X divided by the percentage change in the price of good X. The cross-price demand elasticity measures the percentage change in the demand for good Y as a result of a percentage change in the price of good X. It is also important to distinguish between market elasticities (percentage change in demand for a good divided by percentage change in price for a good for the aggregate of all providers of a product) and firm elasticities (percentage change in demand for a good divided by percentage change in price for a good for a particular provider). Depreciation Charge: The charge made to reflect the change in value of equipment between the start and end of the year. Direct Network Cost: The cost of inputs necessary for the network to function for which the volume of the input depends directly on the outputs required, i.e. the cost drivers are exogenous not driven by other costs categories. An example is the cost of ports, where the cost driver is the number of call minutes. Directly Attributable Costs: are those costs that are incurred as a direct result of the provision of a particular service in a particular increment. These costs fall into two types. Firstly, the costs of some inputs vary with the level of output, so that even if the output of more than one service requires this input, the extent to which a single service causes the costs can be calculated. Secondly, there are assets and operating costs which are fixed with respect to the level of output but which are service specific. Duration Related Costs: Costs, which are driven by the duration of the call. Exchange Area: Bounded area, which includes one exchange building to which subscribers in the exchange area are connected.

128 Exchange: Equipment capable of switching or concentrating traffic. Examples are transit exchanges, local exchanges and subscriber stages. Transit Exchanges comprise 4 parts. A processor (with the different services realised through software), a switch block (for switching the calls), a transmission/signalling part (for connecting to other exchanges), and an operation- and maintenance part (to perform changes/correct errors in the tandem exchange, billing). May also referred to as Tandem exchange. Local exchanges are built similar to a tandem exchange, but will also contain equipment for connecting the access network. The connection may be direct or in the form of subscriber stages. A Local Exchange may also be referred to as a Local Switch. Subscriber stages reduce the number of subscriber lines (from the access network) to a lower number of transmission links (towards the local exchange). They can either be placed as part of the local exchange, (also termed a Host Subscriber Stage - HSS), or - connected via an external transmission system - away from the local exchange, (also termed a Remote Subscriber Stage - RSS). Subscriber stages may also be referred to as concentrators or subscriber exchanges (when located outside the local exchange building). Subscriber Stages placed at the local exchange (HSS) will only be included in the definition of an exchange when it appears as a (physically and logically) separate unit in relation to the local exchange. Externalities: Where the provision of a service confers benefits or costs to parties other than paying party gives rise to externalities. These can either be positive benefits (a call from the originating party will generally also provide benefits to the terminating party) or negative (pollution). Access externalities arise because when somebody joins the network existing customers can make calls to an extra person. Call externalities arise because a call can provide benefits to both the originating and terminating parties. Fixed Common Costs: The cost of an input that produces outputs for two or more different increments, that does not change with the volume of output. Fixed common costs and joint costs make up common costs. An example of a fixed common cost is the cost of the installation of a concentrator unit. Fixed Costs: Costs, which do not change with the level of output. Indirect Network Costs: costs of inputs, necessary for the network to function, for which the volume of the inputs depends on other inputs. They depend only indirectly on the volume of outputs required, i.e. the cost drivers are endogenous. An example is the cost of racks. Infrastructure: The full set of assets over which the traffic travels, mainly fibre, copper, duct and trenches.

129 Interconnection Point: A point in an operator's network where traffic can be exchanged between different operators networks. Interconnection points can be established at different layers in the network hierarchy (local, regional, national and international interconnection points). Joint Costs: The cost of an input that produces outputs for two different increments in fixed proportions. Reducing the output of a single activity will not reduce joint costs; reducing the output of all activities will reduce these costs. Fixed common costs and joint costs make up common costs. Final Drop Cable: The section of the access network connecting the last distribution point with the first connection point at the customer's Layers of the Network: Different levels of the network hierarchy. Long Run: The length of time over which all inputs can vary in response to a change in demand. Multiplexor: Electronic communications equipment that combines or multiplexes several signals for transmission over a single medium. A multiplexor may terminate and combine the subscriber lines into a single transmission line. The multiplexor is normally combined with a demultiplexor into a single device capable of processing both outgoing and incoming signals. Network Component: A grouping of cost categories that shows the cost of assets at a more detailed level than network elements. For example, whereas a network element will show the cost of a local exchange, a network component will show the cost of the ports that make up the local exchange. The cost of ports will aggregate a number of different cost categories (such as the unit cost of ports, maintenance, installation, accommodation etc). Network hierarchy: Ordering of the different types of exchanges in the core network. A traditional telecommunications network hierarchy includes three layers: A top-layer, which includes the transit exchanges, a middle-layer, which includes the local exchanges, and the lowest layer, which includes the remote subscriber stages. Network Elements: The aggregation of cost categories that uniquely identifies the LRIC-based products for the purpose of this study. So, for the access network, they correspond to the copper line and the fibre line. For the core network, they correspond to the different parts of the network and to the exchanges defining them. In a conventional network they would correspond to concentrator units, local exchanges and tandem exchanges and the transmission and infrastructure equipment connecting them. Node: See exchange. One-off Costs: Costs incurred to establish an activity or service which can not be recovered (irreversible).

130 Operating Costs: The cost of running a network. Examples include the cost of maintaining capital equipment and the cost of running administrative divisions. Overhead Costs: The costs of inputs, which are not necessary for the network to function, but are necessary for the organisation running the network to function. An example is the cost of a personnel department. Port: Point of access where signals may be inserted or extracted. On exchanges one distinguishes between access ports and trunk ports. Access ports face the subscriber whereas trunk ports face other exchanges. Primary Access Network: The section of the access network between the Main Distribution Frame (MDF) and the first distribution point. The line card will be one of the assets in the primary exchange network. Ramsey prices: Prices for individual products/services set so that the contribution to cover common costs (the relative deviation from LRIC) is inversely related to the price elasticity of the service/product (see demand elasticity). Secondary Access Network: The section of the access network connecting the primary access network with the customer s drop-cable (up to and including the last distribution point). Set-up Specific Costs: Specific costs, which relate to a call attempt, answered or not answered. Shared Costs: Costs of those inputs necessary to produce two or more services within the same increment, where it is not possible to identify the extent to which a specific service causes the cost. Examples of shared costs in the core network include optical fibre, transmission equipment and related overheads, all used by PSTN, leased line and other services Short Run: The length of time over which at least one input is fixed. Typically, the inputs that cannot vary are the capital investment decisions. Site: The location where exchanges are placed. Specific Fixed Costs: Fixed costs, which are caused by a single increment. Switching Network: Those assets that provide the functionality to route traffic over the network. Consist of the exchanges in the network. Also referred to as exchange network or exchanges. Transport Network: Those assets that allow traffic to travel over the network. The assets include line termination equipment and transmission electronic equipment such as cross connects and multiplexers as well as infrastructure such as trenches and fibre. Variable Costs: Costs which change with the level of output.

131 Figure A3.1: Categorisation of network terms Core network Transport network Access network Switching/Exchange Exchanges/ Switches network Transmission electronic equipment Infrastructure Infrastructure Other Subscriber stage, local exchange, tandem exchange, ATM-switch, IP router etc. Cross-connects, Multiplexers etc. Fibre,copper, coax trench, duct, manholes, etc. Fibre,copper, coax trench, duct, manholes, etc. MDF, line card, distribution point, etc. Figure A3.2: Schematic of the local access network NTP SDP PDP DF/DDF DF/DDF DF/DDF Distribution network/ Final drop Secondary Network Primary Network PSTN RSS PSTN LE, TE Business & Private Customers RSM RSS ÄS,KS,RS Business Customers Non PSTN NTP = Network Termination Point SDP = Secondary Distribution point PDP = Primary Distribution Point DF = Distribution Frame DDF = Digital Distribution Frame Non PSTN Exchange building Exchange building Access network Core Network Source: Adapted from COMPIS 2000 documentation Note that the access network is defined as including the line card, typically placed in the concentrator (RSS) as opposed to the remote multiplexer (RSM).

132 Appendix 4 Detailed list of services modelled or covered by LRIC A.4.1 Interconnection services 49 Connection services Traffic services Access Termination Transit Regional point of interconnect Local point of interconnect Connection capacity Double segment Single segment Metro segment Local segment Double segment Single segment Metro segment Local segment Double transit Single transit A.4.2 (Wholesale) copper access services 50 New copper access Copper Access S50 Copper Access S300 Copper Access S600 Copper Access A1100 Copper Access A900 Copper Access ADEL900 Installation fee / Installation fee including work in access network and/or customer premises wiring Variable fee (quarterly fee) Installation fee / Installation fee including work in access network and/or customer premises wiring Variable fee (quarterly fee) Installation fee / Installation fee including work in access network and/or customer premises wiring Variable fee (quarterly fee) Installation fee / Installation fee including work in access network and/or customer premises wiring Variable fee (quarterly fee) Installation fee / Installation fee including work in access network and/or customer premises wiring Variable fee (quarterly fee) Installation fee / Installation fee including work in access network and/or customer premises wiring Variable fee (quarterly fee) 49 The services should be defined as in Skanova's Reference Interconnection Offer "SAMTRAFIKAVTAL mellan Skanova Networks bifirma till Telia AB (publ) och OPERATÖREN AB" of 25 September The services should be defined as in Skanova's standard offer on copper access: "RAMAVTAL OM TILLGÅNG TILL KOPPARACCESS I SKANOVAS ACCESSNÄT I SVERIGE", of 5 July 2007.

133 Installation fee / Installation fee including work in access network Copper Access ADEL1000 and/or customer premises wiring Variable fee (quarterly fee) Installation fee / Installation fee including work in access network Copper Access V12 and/or customer premises wiring Variable fee (quarterly fee) Installation fee / Installation fee including work in access network Copper Access V12SUB and/or customer premises wiring Variable fee (quarterly fee) Installation fee / Installation fee including work in access network Copper Access VDEL12 and/or customer premises wiring Variable fee (quarterly fee) Installation fee / Installation fee including work in access network Copper Access VDEL12SUB and/or customer premises wiring Variable fee (quarterly fee) Info request on delivery of Copper Access S300, S600, A1100,,A900, ADEL900, ADEL 1000 Info request on delivery of Copper Access V12, V12SUB, VDEL12 or VDEL12SUB) incl. network investigations. Excl. visit on location Non-recurrent fees for inquires, network investigations, change of product and change of operator Network data Cabinet data Disconnection/reconnection DACS and line conditioner Network data Cabinet data Info request on delivery of Copper Access V12, V12SUB, VDEL12 or VDEL12SUB incl. network investigations. Incl. visit on location (hourly rate) Change of delivered product not requiring switching in the access network Change of delivered product requiring switching in the access network Change of delivered product requiring switching in the cross connect (shared/full) Change of port (shared/full) Change of operator (shared/full) Change from shared access to full access Info request Info request Disconnection of shared access Reconnection of shared access Request for removal of DACS (bärfrekvens) Removal of DACS Request for removal of line conditioner (pupinspolar) Removal of line conditioner Info request Info request A.4.3 Bitstream Access 51 Variable fee Installation fee Full access level 1: ADSL 250, ADSL 2000, ADSL 8000, ADSL Full access level 2: ADSL 250, ADSL 2000, ADSL 8000, ADSL Shared access level 1: ADSL 250, ADSL 2000, ADSL 8000, ADSL Shared access level 2: ADSL 250, ADSL 2000, ADSL 8000, ADSL Installation of shared access 51 The services should be defined as in Skanova's standard offer on bitstream access RAMAVTAL OM TILLHANDAHÅLLANDE AV BITSTRÖMSACCESS TILL OPERATÖRER of 15 August 2007.

134 Operator access Installation of full access Change from shared access to full access Change of delivered product (capacity) One-time fee level 1 One-time fee level 2 Monthly fee level 1 Monthly fee level2 A.4.4 Co-location services 52 Tender fee; co-location at Skanova s node Fee for tenders Tender fee; connection of operator owned copper cable in Skanova s cabinet (connection point of sub-loops ) per installation Location of operator owned equipment in Skanova s node Installation and mounting Station wiring Placing Tender fee; extension of existing co-location Installation fee Mounting, Skanova ETSI cabinet Mounting, operator owned ETSI cabinet Installation of power supply cable at Skanova s node Installation of operator owned copper cable at Skanova s node Installation of operator owned opto cable at Skanova s node Connection of operator owned copper cable in Skanova s cabinet First cable Installation Additional cable Quarterly fee (cabinet) Placing at Skanova s nodes Quarterly fee (Installed operator owned copper and opto cable) Connection plinth at Skanova s Quarterly fee cabinet Power watt Quarterly fee watt Quarterly fee watt Quarterly fee watt Quarterly fee Demonstration of co-location areas in node 52 The services should be defined as defined in Skanova's standard offer on co-location in connection with copper access: RAMAVTAL om tillgång till samlokalisering i samband med Kopparaccess, of 5 July 2007.