characteristics of the top down and bottom up cost analyses · pdf...

REPORT ON:

CHARACTERISTICS OFTHE TOP-DOWN AND BOTTOM-

UP COST ANALYSES

15 March, 2002

ii

Table of ContentsPREFACE 1

PART A: REVIEW OF BOTTOM-UP MODEL .......................................................................................................2

A.1 OVERVIEW OF MODEL .....................................................................................................................................3A.1.1 Structure..................................................................................................................................................3A.1.2 Key Results...............................................................................................................................................3A.1.3 Summary of the initial findings in the Bottom-up Model .........................................................................4A.1.4 Revising the Bottom-up Model .................................................................................................................5

A.2 REVIEW OF ACCESS NETWORK ........................................................................................................................6A.2.1 Overview of Approach..............................................................................................................................6A.2.2 Justification for Approach Taken and Issues for Further Comments..........................................................8

A.3 REVIEW OF EXCHANGE STRUCTURE............................................................................................................. 11A.3.1 Overview of Approach........................................................................................................................... 11A.3.2 Justification for Approach Taken and Issues for Further Comment ........................................................ 12

A.4 REVIEW OF TRANSMISSION NETWORK AND INFRASTRUCTURE .................................................................. 19A.4.1 Overview of Approach........................................................................................................................... 19A.4.2 Justification for Approach Taken and Issues for Further Comment ........................................................ 20

A.5 REVIEW OF CO-LOCATION MODEL.............................................................................................................. 26A.5.1 Overview of Approach........................................................................................................................... 26A.5.2 Justification for Approach Taken and Issues for Further Comment ........................................................ 27

A.6 COSTS IN THE BOTTOM-UP MODEL............................................................................................................. 30A.6.1 Approach and Assumptions to Estimate Replacement Costs .................................................................... 30A.6.2 Approach and Assumptions to Estimate Annualised Costs ...................................................................... 32A.6.3 Approach and Assumptions to Estimate Operating Costs........................................................................ 34A.6.4 Approach and Assumptions to Estimate Indirect Costs ........................................................................... 35A.6.5 Approach and Assumptions to Estimate Overheads Costs........................................................................ 36A.6.6 Approach and Assumptions to Estimate Working Capital Costs ............................................................. 37

A.7 SENSITIVITIES IN THE BOTTOM-UP MODEL................................................................................................. 38A.7.1 Sensitivities in the Access Network......................................................................................................... 38A.7.2 Sensitivities in the Core Network .......................................................................................................... 39A.7.3 Sensitivities in Co-Location Model........................................................................................................ 40

PART B: REVIEW OF TOP-DOWN MODEL...................................................................................................... 42

B.1 INTRODUCTION ............................................................................................................................................ 43B.1.1 The Top-down Cost Analyses................................................................................................................. 43B.1.2 Key Results............................................................................................................................................ 43

B.2 OVERALL COMMENTS ON MODEL AND DOCUMENTATION ....................................................................... 45B.2.1 Theoretical Modelling Approach to Capital Costs.................................................................................. 45B.2.2 SAS model and GRC calculations ......................................................................................................... 45B.2.3 Application of CVR to allocate fixed and variable costs.......................................................................... 46B.2.4 Utilisation Ratios.................................................................................................................................. 47B.2.5 Building Costs....................................................................................................................................... 47B.2.6 Annualisation ....................................................................................................................................... 47B.2.7 MEA Adjustments................................................................................................................................. 49B.2.8 Operating Costs..................................................................................................................................... 50B.2.9 Co-location and Interconnection Specific Services.................................................................................. 50B.2.10 Common and Shared Costs ................................................................................................................... 50

B.3 GROSS ASSET VALUATION............................................................................................................................. 52B.3.1 General Comments on Dimensioning.................................................................................................... 52B.3.2 Overview of TDC’s Network ................................................................................................................ 52B.3.3 Access Network ..................................................................................................................................... 54B.3.4 Exchanges ............................................................................................................................................. 59B.3.5 Transmission ........................................................................................................................................ 62

B.4 ANNUALISATION............................................................................................................................................ 65

iii

B.4.1 Economic Depreciation Model .............................................................................................................. 65B.4.2 Conclusions........................................................................................................................................... 69

B.5 WORKING CAPITAL ....................................................................................................................................... 70B.5.1 Method................................................................................................................................................. 70B.5.2 Potential Problems ................................................................................................................................ 70B.5.3 Compliance with Criteria ..................................................................................................................... 71

B.6 OPERATING COSTS........................................................................................................................................ 72B.7 ALLOCATION OF COSTS................................................................................................................................. 73

B.7.1 Treatment of other increments - shared and common costs..................................................................... 74B.7.2 Differences between Figures in Documentation and Model.................................................................... 74B.7.3 Access Costs ........................................................................................................................................... 76B.7.4 Exchange Costs...................................................................................................................................... 76B.7.5 Transmission Costs................................................................................................................................ 77B.7.6 Other Costs........................................................................................................................................... 78B.7.7 Summary/main issues............................................................................................................................ 80

B.8 CO-LOCATION AND RELATED SERVICES....................................................................................................... 81B.8.1 Methodology.......................................................................................................................................... 81B.8.2 Potential Problems ................................................................................................................................ 81B.8.3 Compliance with Criteria’s ................................................................................................................... 82

B.9 EFFICIENCY STUDY........................................................................................................................................ 83B.9.1 Total Reliance on DEA......................................................................................................................... 83B.9.2 No adjustments made to place TDC and US LECs on a comparable basis............................................. 85B.9.3 No Sensitivity Analysis on Assumptions made by TDC.......................................................................... 86B.9.4 No Weight Restrictions.......................................................................................................................... 86B.9.5 Excessive Reliance on the Output “Main Switched Minutes” ................................................................. 87B.9.6 Conclusion ............................................................................................................................................ 89

B.10 GENERAL COSTING ISSUES............................................................................................................................ 91B.10.1 Volumes................................................................................................................................................ 91B.10.2 Margins for Growth.............................................................................................................................. 91B.10.3 Routing Factors..................................................................................................................................... 91

Telestyrelsen

1

Preface

This report discusses the characteristics of the bottom-up and top-down cost analyses with theaim of assessing the value of the two analyses undertaken by the LRAIC Forum and TDC, as adecision basis, cf. section 15(2) in the executive order on interconnection etc.

The report has a companion report which reconciles the results in the two analyses.

The parties have also been requested to provide responses focusing on the value of the two costsanalyses as a basis for Telestyrelsen’s decision making during the hybrid modelling. Theseresponses are summarized and discussed in a separate hearing note. This hearing note refers tothe relevant sections in the report.

The views of Telestyrelsen expressed in this report should neither be seen as comprehensive nordefinitive.

In terms of the bottom-up model, the review of the model and the reconciliation exercise haverevealed a number of areas where further investigation is needed. These include the lack ofjustification of some key inputs which have a significant impact of results and as to whether thebottom-up model – both in access and core – offers sufficient quality of service. The bottom-upmodel is discussed in Part A of this report.

In terms of the top-down model, the review of the model and the reconciliation exercise havealso revealed areas of some concern. These include the lack of transparency and the inconsistencyof results in a number of areas. The reconciliation exercise has also thrown up a number ofsignificant differences in a number of areas with the bottom-up model and these will need to beexplored in detail with TDC. In addition, some data are outstanding from TDC as part of thereview of the top-down model. The top-down model is discussed in Part B of this report.

Telestyrelsen is very interested in the views of the two parties on the assumptions andmethodologies behind the top-down and bottom-up model. These views will help Telestyrelsento decide about the hybrid model. Certain key issues have been highlighted. However, the partiesmay provide comments on all aspects of the report.

Telestyrelsen

2

PART A: REVIEW OF BOTTOM-UP MODEL

Telestyrelsen

3

A.1 Overview of ModelThis section provides a very brief overview of the bottom-up models developed by the LRAICForum and summarises some of the main findings

A.1.1 Structure

The structure of the bottom-up models has been described by the LRAIC Forum in theirdocumentation to Telestyrelsen and will not be explained in detail here. In short, the Forum hasdeveloped four models.

Three of the models are directly connected to the services being modelled. These models are thecore model, the access model, and the co-location model. The core model deals with PSTNswitching, transmission services and equipment items that are driven by traffic. The access modeldeals with the calculation of the equipment and costs below the existing switch site (the cablesand systems that connect the customers to the switch site). The co-location model deals with thecosts associated with sharing facilities, such as buildings and exchange equipment.

The forth model is the consolidation model which contains the generic values and sharedoutputs, such as annualisation estimates, common costs, and results.

The model is flexible to a degree. Certain inputs can be changed and their impact on the resultschecked. Some changes — such as large scale changes in the number and mix of nodes — maybecome inconsistent with the overall network design. The complexity in the access model alsomakes some types of changes difficult.

A.1.2 Key Results

The results of the bottom-up model are presented at a number of different levels - cost categories,network components, network elements, and interconnection services. The table below showsthe results of the major services that form part of the LRAIC process in Denmark.Table 1: Core results (Øre)

Interconnection service per minute per call Average per minute

Local interconnection 1.49 0.39 1.56

Regional interconnection 2.08 0.61 2.22

National interconnection 2.37 0.79 2.55

Table 2: Access results (DKK)

Interconnection service Annual cost per year

2 wire raw copper 213.45

Dark fibre 2,617.59

Telestyrelsen

4

A.1.3 Summary of the initial findings in the Bottom-up Model

Telestyrelsen believes that there are a number of areas where the assumptions of the bottom-upmodel needs to be further documented. Telestyrelsen will examine these areas when developingthe hybrid model. Telestyrelsen specially notes that:

• Many key inputs are not estimated in a detailed and systematic way. For example, althoughthe access model is quite detailed, the “conversion factor” which has a crucial bearing on thecosts of the largest single network element in the access network is produced with littlesupport. The same can be said for the trench sharing factors in the core network.

• The model appears to depart from common practice in some areas without providingsufficient justification as to why such a different approach has been taken. One example ofthis is the assumption made about utilisation in some parts of the access network (e.g. 100per cent in some parts).

• The model lacks sufficient justification in many areas. Telestyrelsen has raised some of thesewith the Forum throughout the reconciliation phase. An example is in the area of quality ofservice where the documentation and subsequent papers provided by the Forum inTelestyrelsen’s view not has been sufficient to support that the approach adopted is alwaysrobust.

• Telestyrelsen believes that further investigation in the hybrid model is needed to support thatthe core network modelled by the Forum is technically feasible.

• The local interconnection product, modelled by the LRAIC Forum is not readily comparableto the local interconnection product currently provided by TDC.

• Much of the unit cost information, particularly in the core model, does not appear to alwayscorrespond to the equipment model. The very large and very intelligent tandem exchangesappear to be costed using information that is not appropriate to the equipment in the modeland it is not clear whether the additional functionality has been included in the costsassumed by the Forum. The modelled RCUs are also more intelligent than those used inother bottom-up models and in Telestyrelsen’s view, RCUs have a functionality that inpractice is equivalent to a local exchange (compare with definitions in LRAIC terminologilist). Telestyrelsen therefore believes that the Forums use of the word RCU in some instancesis questionable.

• Some of the price trend information is, in some places, not sufficiently documented and leadsto annualised results that are quite low.

• The co-location services modelled in the bottom-up do not always match up to those offeredin Denmark at present. In addition, some of the modelled service costs in the co-locationmodel are very sensitive to changes in demand. Although Telestyrelsen acknowledge, thatthere are very substantial differences in forecasted demand in the analysed years which willimply differences in service cost, the sensitivity analysis shows that the differences are of sucha scale, that Telestyrelsen believes further documentation is needed to support the co-locationmodel.

• The bottom-up model does not provide sufficient justification of how some one-off costs -such as changes to the numbering plan and the treatment of prefixes - have been modelled.

Telestyrelsen

5

A.1.4 Revising the Bottom-up Model

In the course of the reconciliation phase, a number of changes have been made to the bottom-upmodel. Some of these were in response to errors identified by Telestyrelsen during their initialreview of the model.

It is likely that Telestyrelsen will need to make further revision to the bottom-up model. Theserevisions will fall into two categories. One category relates to errors that continue to be found inthe model. These are, in part, inevitable in complex models of these sorts and will be correctedby Telestyrelsen throughout the remainder of the process.

The other category relates to more significant changes in the modelling approach such as changesin the way that an allowance for growth is made or in the way that routing factors are calculated.These changes will if needed be included in the hybrid model and will be explained byTelestyrelsen in full.

Telestyrelsen

6

A.2 Review of Access NetworkThe purpose of this chapter is to review the methodology followed by the LRAIC Forum toestimate the costs of the access network. The approach that the Forum has used has beenpresented in the model documentation provided to Telestyrelsen and in subsequent, and moredetailed, papers providing further documentation and justification. This chapter will, therefore,not repeat the modelling methodology but rather focus on the main assumptions made, theirimportance, and present some issues for further comment.

The access model works out capital requirements needed to provide access services in Denmarkby network elements.1 The network elements are trench, duct, mini-duct, copper, networkterminating points, cabinet equipment, fibre, and line cards. (Line cards requirements areestimated in the core network and included as access costs in the Consolidation model).

A.2.1 Overview of Approach

This section provides an overview of the approach and assumptions used to model the capitalrequirements for the two most important network elements in access — trench & duct andcopper.

A.2.1.1. Trench and Duct

The main methodological assumption that has been adopted in the bottom-up model in order tomodel trench requirements is that trenches in Denmark have a direct relationship with roadpaths. GIS maps have been used to classify all roads in Denmark in the sixteen road categoriescreated and in the four given geo-types. For each of the sixteen types of road, a factor (rangingfrom 0 to 2) has been assumed to convert road length into trench length. These are referred to as“conversion factors”.

Total trench requirements by different geo-types have been split into different types of terrain byassumption and by distinguishing between trench lying inside or outside the urban areas.

The methodology used to model duct requirements relies on the modelled amount of trench.Total duct length has been obtained by subtracting the amount of cable, which is supposed to beburied, rather than in ducts, from the total trench length. Total duct length is then allocated intodifferent duct sizes in order to obtain duct requirements (kilometres of duct) by duct size. Mostof the duct is assumed to be 4 or 5 bore duct, but a very small proportion is assumed to be “largeduct”.

In addition, the bottom-up model estimates the amount of trench and duct that links thecustomer premises to the trenches and ducts. This is referred to as “mini-duct” and it only comesin one size.

A.2.1.2. Copper

The methodology adopted to work out copper requirements is quite complex and relies on theresults of a dimensioning exercise performed on a sample of twenty MDF areas (hereafter calledthe sample) selected by the Forum among TDC’s 1,183 exchange areas. The methodologyadopted to select the twenty sample zones can be described as follows.

1 The terminology used in the bottom-up model differs, in many respects, to the terminology used in the top-down

model. The differences are clarified in the reconciliation report.

Telestyrelsen

7

First, a number of zones were selected for each geo-type on the basis of the relative importance(as measured by number of lines, area covered etc) of that geo-type to the overall network.Within each geo-type, the particular zones on which the cost analysis was then performed havebeen selected in order to form a sample that is aimed to be representative and balanced in termsof teledensity, i.e. number of lines per square kilometre.

Second, for each of these zones, the model works out the crow flight distances of the connectionsbetween the joints of the tree configuration assumed in the model, i.e. network terminationpoint (NTP), exit from street duct (EFSD), secondary distribution point (SDP), primarydistribution points (PDP) and remote concentrator unit (RCU). The methodology adopted toobtain the crow flight distance is quite complicated and relies on a number of assumptions.Conversion factors are then used to convert crow flight distances into actual distances.

Third, the model works out the size of the cable for each of the segments connecting thecustomer to the concentrator at the demand level first. For each zone in the sample, cable fromNTP to SDP is assumed to contain only one pair cable; the size of the cable from SDPs to PDPsis a model input, the size of the cable from PDPs to RCU is obtained by dividing the number ofcopper pairs served in each zone by the number of PDPs (an input of the model) assumed toserve that zone.

Finally, the capital equipment needed to serve each zone is then multiplied by the weight givento it, in order to obtain access requirements for the whole network. The weight given to each ofthe selected areas is a number that should inform on how many zones, similar to the one that hasbeen chosen, are assumed to be present in the network. Weights have been selected so that thetotal number of modelled lines, total number of modelled customer sites and total area covered -obtained by multiplying the values these variables take for the selected zones by the weightsattached to each zone and then adding the twenty results together - match against the valuesthese variable take for the whole network, taken as given in the model.

A.2.1.3. Cabinet Equipment

The modelled cabinet equipment includes: SDP and PDP cabinets (and if necessary SDP andPDP circuit boards).

The number and the size of modelled SDP and PDP cabinets depends on the number and thesize of the modelled SDPs and PDPs for each of the twenty sample zones (the requirements forthe whole network are worked out through the means of the weights attributed to each zone).The number of modelled SDPs per zone is worked out as the ratio between number of subscriberlines served in each zone and connected copper pairs per SDP, an input of the model. Thenumber of modelled PDP per zone is an input of the model. Collocation of SDP equipment inPDP cabinets and of PDP equipment in the concentrator site is taken into account.

The size of each cabinet, classified as small, medium and large, depends on the number of linesserved by each cabinet. This is an input of the model for SDPs and the ratio between number ofPDPs in the zone and total number of lines served in the zone, both inputs of the model, forPDPs. Margins for spare are also taken into account.

• SDP and PDP circuit boards, placed within the cabinets, are dimensioned to carry theaverage number of lines per cabinet. An optimal combination of different sizes for circuitboards is chosen.

Telestyrelsen

8

A.2.1.4. Other Network Elements

The other network elements modelled in the bottom-up model make up a small proportion ofthe costs of access. The methodology will be summarised below and can be explored in moredetail through the documentation and by reviewing the model.

A.2.2 Justification for Approach Taken and Issues for Further Comments

A.2.2.1. Trench Lengths

The methodology used in the bottom-up model to work out trench requirements, as discussedabove, assumes that trenches in Denmark follow roads, with no additional trench beingmodelled. The Forum has provided some justification of this methodology in the January 2002document, Model Changes and Clarifications, arguing that:

We do not expect there to be a material number of customer sites that need to be connected to the Accessnetwork, that are not also adjacent to the road network. This assumes that we consider, say, a block of flats to be"adjacent" to the road network even though some of those flats are not directly adjacent to the road. Frominspecting the maps we believe that this assumption is reasonable other than in two cases:

- Inhabited islands; the cost of connecting these to the mainland (or to a larger island) is calculated inthe Core model and passed from there to the Consolidation model

- Opportunities to reduce the cost of the access network by using other wayleaves, e.g. following a railroute or power line rather than the road, to shorten the length of some connections and/or reduce thecost per km of such links. We believe that the benefits from this are likely to be modest and it seemsreasonable to simply write them off against any additional non-road routes required in the Accessnetwork, if any should arise.

Telestyrelsen has examined the methodology used by the Forum and believes that — on thewhole — it is a sensible way to determine trench requirements in a bottom-up model. Due to therelative impact that trenching costs have on the price of the final service it is important that theinputs used by the Forum is well documented. Within the methodology adopted in the bottom-up model, there are a number of inputs that may affect trench requirements and trench unit costsin a significant way. For example, the estimates of the relationship between road length andtrench length may be subject to a considerable margin of inaccuracy in some cases. One problemis that the relationship may depend on more than the density of sites as assumed in the model.For example, a link between an MDF and an area of houses may pass through an area in whichthere are no sites.

Some of these inputs have been derived from an analysis of GIS maps while others have beenestimated by the LRAIC Forum.

Bottom-up issue-I Telestyrelsen would welcome views from all parties on whether there is likely to be any significanttrench that does not follow the road network and whether the conversion factors used by the Forum areappropriate. Any evidence from TDC’s own network or from other actual networks would be helpfulto Telestyrelsen when developing the hybrid model.

A.2.2.2. The Approach to Estimating Common Costs, Particularly Common Trench

Common costs in the bottom-up model are driven by two factors - the length of trench (andduct) that is assumed to be used by more than one increment and the way that these commoncosts are allocated to different increments.

Telestyrelsen

9

The LRAIC Forum has provided some supporting arguments for the assumptions that they havemade, but little actual justifications. They have not, for example, provided any clear justificationfor the amount of trench and duct that is assumed to be common with cable TV or otherincrements nor any evidence from other countries. Also the LRAIC Forum has not documentedthat the time period implicitly assumed when estimating the scope for co-digging is consistentwith the time horizon implicitly assumed when estimating equipment prices (the scope fordiscounts).

Telestyrelsen acknowledge that it is very difficult to justify any amount of trench/duct that canbe deemed to be efficiently and feasibly shared with other increments. One approach may be toreview the experience in other countries and make adjustments to account for the fact that thenetwork might not be rolled-out in a manner consistent with that assumed by the LRAICForum.

In terms of allocating common costs to different increments, Telestyrelsen believe that a numberof different options could have been explored. One such option may be to review the space takenby the different cables in the trenches and/or ducts.

Bottom-up issue-II Telestyrelsen would welcome 1) Evidence from parties about international experience in order toestimate the amount of trench/duct that is common between the core/access network and otherincrements. 2) The most appropriate way to allocate common costs in the access network

A.2.2.3. The Copper Requirements and Distributors

The methodology that has been used by the LRAIC Forum to estimate copper requirements inthe bottom-up model is complex. However, there are at least three assumptions that requiredeeper examination.First, the lengths of the local loops and in particular, the length of the SDP to NTP part of theloop. This appears to Telestyrelsen to be relatively long.Second, the assumptions about utilisation will need to be examined. The model accounts forutilisation in two ways:

• through an allowance for growth. An allowance of 20 per cent on cable size is assumed forcables from SDP to RCU. No allowance for growth or spares has been assumed for cablesfrom NTP to SDP.

• through modularity. Cables are provided in modules and therefore more cables are put inplace than the ones required by demand considerations. However, from NTP to SDP themodel puts only in place 1 pair cable. Modularity, for this part of the network, does notallow for any spare.

Thirdly, the model assumes that a separate cable is required for each link between a PDP andSDP. Telestyrelsen believes that there may be scope for sharing of cables in some cases.

• With regard to distribution points, the bottom-up model contains relatively large distributorsand subsequently only as a few distribution points. Also the utilisation levels of thesedistributors are very high.

• While the costs associated with cabinet equipment are not significant compared to the costestimates of other network elements, it is important that the bottom-up model accounts for

Telestyrelsen

10

an appropriate number of primary and secondary cabinets. This is because the bottom-upmodel relies on these estimates to work out the number of links in each sample zone and,therefore, the estimated size and length of copper cables.

Bottom-up issue-III Telestyrelsen would welcome information about the utilisation levels in place in other countries fordifferent parts of the access network.

A.2.2.4. The Assumptions Underlying the Access Routing Table

In contrast to the core model where the routing table has been used to both dimension thenetwork and allocate network costs to final services, the table that allocates access network coststo the final services does not enter the dimensioning phase of the model. This means that thisallocation table is quite important and has a significant impact on the final service cost estimates.

The access allocating table is used to allocate the costs of network elements to access services. Foreach network element, the numbers referring to each service should mirror the way each servicecauses the costs of that network element (cost causation principle). Although the allocation ofsome network elements to services seems obvious, others appear not to have been estimated in asystematic way. These figures are then weighted against the amount of the service that iscurrently provided, when the network costs are finally allocated.

The Forum has responded to Telestyrelsens earlier questions about the routing table used inaccess. However, as many of the estimates used in the routing table are based on assumptions, itis important that Telestyrelsen has confidence in the assumptions that have been made. To date,the assumptions used in the bottom-up model have not been sufficiently documented andTelestyrelsen will need to review them carefully.

Bottom-up issue-IV Telestyrelsen would appreciate comments about the assumptions in the access routing table.

A.2.2.5. Other Issues

There are a number of other issues that arose following Telestyrelsen’s review of the LRAICForum’s access network. These are briefly described below:

• the cost of mini-duct. The unit costs that the bottom-up model uses for trench and duct in themini duct part of the access network are considerably cheaper than the ones used elsewhere inthe access network;

• customer connections. The bottom-up model assumes that every subscriber is directlyconnected to an SDP, with PDPs only serving the function of conveying copper cablescoming from SDPs into bigger copper cables going into the concentrator and thatconnections from NTP to SDP are served solely by 1 pair cable. One issue that needsclarification whether the policy of direct connections allows sufficient flexibility, e.g. in termsof delivery times, in the event of a customer requiring a second connection.

Telestyrelsen

11

A.3 Review of Exchange StructureThe purpose of this chapter is to review the methodology followed by the LRAIC Forum toestimate the costs of the exchanges. The approach that the Forum has used has been presented inthe model documentation provided to Telestyrelsen and in subsequent, and more detailed,papers providing further documentation and justification. This chapter will, therefore, not repeatthe modelling methodology but rather focus on the main assumptions made, their importance,and present some issues for further comment.


This section provides an overview of the approach followed by the LRAIC Forum whenmodelling exchanges, particularly in the following areas:

• the network hierarchy;

• the approach taken to meet the scorched node assumption;

• quality of service; and

• demand and routing factors.

A.3.1.1. Exchange Hierarchy

The main feature of the exchange structure modelled by the LRAIC Forum is that it is a twolayer switch hierarchy. It contains a large number of remote concentrator units (RCUs) whichserve as customer access switches and a small number of tandem or transit switches.

The RCUs connect to the customers over the access network and are described as “fullyfunctional” switches. That is, they can set-up calls and complete calls to other customers if theother customer is on the same switch. Calls from one RCU are routed to other RCUs via one ormore tandem exchanges. RCUs can vary in size from small (less than 500 lines) to large (20,000lines or more). Some RCUs are also used as a point of interconnection (POI). Theinterconnection to the RCU POI enables calls to/from any subscriber who connects to thatRCU. Calls to non-POI RCUs must be routed via a tandem switch.

The tandem switches route incoming calls from RCUs or other TSs and send the calls to othertandems or another RCU for call delivery. There are no customers directly connected to tandemswitches. All TSs also serve as POIs.

A.3.1.2. The Scorched Node Assumption

The Forum claims that the (total) number of RCUs and tandem switches has been determinedby following the scorched node assumption. This means that they have placed at least one switchwithin each switching zone and that no sites have been removed. Technical houses have not beenmodelled; this will be discussed in greater detail below.

The scorched node assumption used for this study allowed the mix of exchanges to be altered.The Forum appears to have chosen their optimal number of tandems and then allocated theremainder of nodes as RCUs. The number of tandem switches was chosen by assuming that atleast one tandem switch was needed for each of the three main land masses in Denmark and that“several” TS sites were needed on each land mass for diversity.

Telestyrelsen

12

A.3.1.3. Quality of Service

The LRAIC Forum claims that the service levels supplied by the network are equal to or greaterthan that expected by a telecoms national operator. There are three main ways that this is said tobe achieved in the modelled network.

First, by a GoS assumption which is used to calculate transmission circuit requirements. TheForum assumes a Grade of Service of 0.5 per cent is used to calculate transmission circuitrequirements. They also suggest that since traffic demand is rounded up to the next 2Mbit/slevel, this GoS is always met or exceeded. The allowance for growth also adds further headroomto ensure that the GoS is met according to the Forum, although this will only be the case in theinitial years before the growth arrives.

Second, by assuming that there are two logical paths from an RCU to the tandem switch. TheForum argue that this additional capacity in the transmission network can be used to cover thedifferent QoS-related requirements in the core network. Exchange and transmission equipmentcapability corresponding to such a network management functionality is not consistentlydocumented by the Forum. The question regarding a failure of a tandem switch is discussedbelow, while related transmission aspects are dealt with in chapter A4.

Third, the SDH rings provide automatic re-routing of traffic via the other ring part in the case ofring disconnection. This will be discussed in more detail in A.4.

A.3.1.4. Demand and Routing Factors

The LRAIC Forum’s core model is driven by a routing table, which attributes volumes of trafficto different network elements. This table is used both for capacity calculations and to attributethe calculated costs of the core network back to each of the call products. A probability has beenattached to each (or most) of the possible routes for each call type in order to determine theactual routing factor for each call type.

The assumptions that have been used by the Forum are discussed in more detail below, but thereare two important points to note at this stage.

• First, the routing table and the probabilities assume a two tiered network rather than TDC’sexisting three-tiered network.

• Second, some of the data used by the Forum has used Forum assumptions even in placeswhere TDC data were available.

A.3.2 Justification for Approach Taken and Issues for Further Comment

There are a number of areas where Telestyrelsen has raised questions regarding the exchangenetwork. The main concern was with the level of justification provided by the LRAIC Forum fortheir choice of the network design. In particular, Telestyrelsen requested further information onwhere such a network had been laid out in the past, how the optimal mix of exchanges wasdetermined, the implications of changes in the mix, and the ability of the network to handleoutages. The Forum responded to these concerns in a note entitled “Network DesignJustification” in January 2002.

This section provides a summary of the Forum’s justification for the approach that they havetaken, and Telestyrelsen’s comments on the chosen approach. The areas that are discussed are:

• compliance with the scorched node assumption;

Telestyrelsen

13

• network design and quality of service issues;

• the lack of a local level interconnection product; and

• general concerns with the modelling approach and assumptions used.

A.3.2.1. Compliance with the Scorched Node Assumption

The LRAIC Forum claims that their model is based on the scorched node approach and furtherthat at least one switch is placed within each switching zone.

In their initial model, the Forum modelled 1,224 RCUs included 41 island RCUs. The 10tandems switches were assumed to be co-located with RCUs.

On this point Telestyrelsen questioned whether the number of sites modelled by the Forum wascorrect or whether “technical houses” was to be captured within the scorched node definition.

TDC have a number of technical houses in their network. Telestyrelsen understand that these aresmall, potentially mobile containers (lightweight constructions) that may in some cases housesmaller concentrator units. Information provided by TDC indicate that technical houses onaverage serve 240 subscribers, less than 10% of the amount served by the average subscriberstage. They have no switching functionality. Information on the number – and indeed location –of these technical houses was provided by TDC to the LRAIC Forum.

As Telestyrelsen has indicated in the hearing note, the agency finds that in these technical housesin principle is to be included as a consequence of the node definition in the MRP. However, theagency has not yet reached a final decision in this issue. To facilitate a more detailed assessment,there is a need for precise clarification on the technical houses, especially regarding theirconstruction, the equipment contained within them and the number of connected subscribers.

Telestyrelsen will need to understand how customers are actually connected to the core networkvia the technical houses. For example, if the customer’s line card is in the technical house then itmay be that it performs the function of a remote concentrator. If, on the other hand, thesetechnical houses are simply “intermediate” stages for customers on route to an exchange, whichhouses the line card, then they would not normally be considered to fall within the definition ofan exchange. It is therefore conceivable that there are cases where it is not readily possible toassess whether a technical house falls under the definition of a node. In its assessment,Telestyrelsen will therefore also need to consider the intention of the law and the internationalexperience in relation to the scorched node assumption.

It is worth noting that the scorched node assumption does not make any judgement aboutwhether the node in question is necessary in an efficient network or not.2 If the node falls withinthe definition of a scorched node as outlined in the model reference paper, then it has to beincluded in the modelled network. However, Telestyrelsen note that the inclusion of all types oftechnical houses may lead to a disproportionate large number of nodes in Denmark, comparedwith other countries and the implications of this will need to be closely examined.

Given the above Telestyrelsen would like to have more documentation, enabling it to carry outan overall assessment, with the view of reaching a final decision on the issue.

2 Although in some countries, for instance, Australia, the scorched node assumption in the bottom-up model was

based on Telstra’s “future mode of operation” estimate of the number of nodes which Telstra itself considered tobe efficient. This number was different to the number of nodes actually in their network at the time.

Telestyrelsen

14

Bottom-up issue-V Telestyrelsen welcomes views from the parties on criteria for defining a minimum size of a technicalhouse, seen in a scorched node context.

A.3.2.2. Network Design and Quality of Service Issues

Another issue of concern to Telestyrelsen is the efficacy of the two tiered exchange network andits ability to deal with quality of service aspects, e.g. availability.

The LRAIC Forum argued that as switching costs were considered to be a major cost element, itwas important to reduce these as much as possible. They believed that larger switches had lowerunit costs due to economies of scale. In particular, they argued that if one large switch replacedseveral smaller ones cost savings would accrue from reduced operational costs, less spacerequirement, and less duplication of common equipment. They also argued that if the costs ofconcentrators and fully functional switches were similar, the ease of operations and reducednetwork capacity could be introduced by using only the fully functional switches.

The Forum recognise that a two tiered network requires longer high capacity links but theyconclude in their January 2002 “Network Design Justification” document that:

“The reduction in switching system costs is clear from having fewer larger switches. The total costs are reduced, ifthe transmission costs are a small fraction of the total switch costs, and because larger capacity transmissionsystems provide economies of scale (double the size does not double the cost). The existing Denmark indirectaccess operators have indeed adopted a similar strategy. A low number of transit switches are in operation, withlong distance SDH technology being used to collect traffic from existing LE interconnect points, or directly fromtheir own RCU located on or near customer premises. These operators are all satisfied, and have no intentions ofchanging the network architecture to a three layer model. Over 30% of Danish telephony traffic is handled inthis manner, indicating the underlying practicality of such a model.”

At this point Telestyrelsen have two comments on the important issue of quality and resilienceissues. First, at this stage and based on current information, Telestyrelsen still questions whetherthe two network elements modelled by the Forum – RCUs and tandem switches – are, in fact,capable of performing the functions which they have been given. For example, the Forum statesin the documentation, that an RCU is an ”fully functional switch” including e.g. route analysis,SS7 signalling, charging and a capacity of connecting up to 20 000 subscribers. It is the view ofTelestyrelsen, that such a unit more consistently could be described as a local exchange (asopposed to the broadly used term ”concentrator” describing a unit without switching, routing,charging and signalling capability. Also compare with the LRAIC Terminology List).

Telestyrelsen note the fact, that the RCU in the BU-model – which is a rather complex unit – istheoretical and does not refer to an existing product. Given that, it is important that the Forumprovide a detailed picture of the functionality of the RCU; in Telestyrelsen’s view this has notbeen provided in the documentation to date. This, in turn, leads to difficulties in the assessmentof the contribution from the RCU (and the transit exchange as well) to the overall networkfunctionality, e.g. quality of service related aspects as network management.

More specifically:

• Tandem exchanges typically do not direct calls to “terminating” RCUs, but rather shunttraffic from one tandem to another or to a local switch. The local switch then performs therelevant call look-up functions to find the customer. If the tandems were required to performthese functions, then they would need to be able to provide additional services. It is not clearwhether this additional functionality has been included in the costs assumed by the Forum.

Telestyrelsen

15

• The RCUs in the Forum’s model are also more intelligent than those used in other bottom-up models including the adaptable bottom-up model developed by Europe Economics forthe European Commission. The RCUs in the Forum’s model appear (compared with thetypical functionality of a concentrator) to need additional routing and subscribermanagement functions and, if they send traffic to other RCUs or own customers, will alsoneed billing and other software functions. As with tandem exchanges above, it is not clear towhich extent this additional functionality has been included in the costs assumed by theForum.

Bottom-up issue-VI In Telestyrelsen’s view it has not to a sufficiently extent been documented that the switches can providethe quality necessary and the functionality stated. Telestyrelsen would welcome any views orinformation from the operators on this area.

Second, Telestyrelsen raised questions to the Forum on whether the network design is secure andwhat would be the impact of a fault or outage on one of the 10 tandem switches. The Forumacknowledge that if one tandem switch were to fail then all of the traffic to and from the tandemswitch would be interrupted. However, they also argue that the network design has a number ofoptions that could be used to minimise the impact of such a fault and so avoid a large percentageof the traffic being interrupted. These options included re-routing the traffic to another tandemswitch.3 The LRAIC Forum also noted that statistically, a failure was not likely during the busyhour and that the practical experience showed that there had been no busy hour failure of atandem switch operated by any member of the BU group in Denmark to date.

Telestyrelsen believes that some of the Forum’s assumptions may be too optimistic. In modernnetworks it is not only fault conditions that reduce the grade of service, but traffic overloads due,for example, to radio and television “phone-in” shows. Moreover, as software complexityincreases with additional services, network integrity is reduced since most outages are softwarerelated. Also high operational costs may generally be expected based on the use of advancednetwork management functionality.

In addition, Telestyrelsen notes that island RCUs are connected to the mainland networkthrough a spur. An RCU with a spur would have no resilience against failure unless duplicate andseparate transmission paths are provided.

Telestyrelsen notes that the network design assumed in the bottom-up model is quite differentfrom that in place in Denmark and that operated by other incumbent operators with networks ofthe size and coverage of TDC. Although the onus was always on the Forum to show that theirnetwork was technically feasible and secure, this onus is particularly greater when a very differenthierarchy to that in place in Denmark has been modelled. The documentation and furtheranalysis by the Forum does not so far suggest that a two-tiered network would necessarilyproduce the same or better quality than TDC’s three-tiered network.

Telestyrelsen therefore believes that it would be helpful to see a higher level of quality of serviceincluded in the model and would welcome views from participants about how best to achievethis. One key dimension of quality is the ability of the network to deal with fault situations. Forinstance, the bottom-up model reference paper referred to the need for the modelled exchanges 3 This would be possible as there are two logical paths from an RCU to the tandem switch (presupposed the

necessary network management capability and the required additional equipment capacity in the transit exchangesto cater for redundant RCU-connections) . There will, however, be some loss of quality of service as neither theback-up tandem nor the inter tandem switch transmission routes are dimensioned to carry the additional burdenduring the peak hour.

Telestyrelsen

16

to have the ability to cope with the breakdown of one or more switches, and more generally theneed to have enough switches to host all of the remote subscriber stages.

One method which might achieve this goal is to place two tandems on one ring, either keepingthe number of tandems constant (and increasing ports etc. on the tandems) or by doubling thenumber of tandems.4 Another might be to create more sideways routing between the largerRCUs (perhaps, as a starting point, those that act of a point of interconnection) in order torelieve some of the pressure on tandems, particularly during periods of stress. This approach mayhave the effect of moving at least, in part, to a three-tiered network.

Bottom-up issue-VII Telestyrelsen would welcome any views from operators about how to improve the level of quality ofservice in the network modelled by the LRAIC Forum.

A.3.2.3. The Lack of a Local Level Interconnection Product

One of the main implications of the Forum’s two tiered network is that there can be no “localexchange interconnection” product. At present, that is, in TDC’s existing network, the localinterconnection product allows access to a number of RCU sites in the local exchange region.(This product does not use any tandem switches and hence is the cheapest type of interconnectcall). In the Forum’s model, an equivalent interconnect product would require all of the RCUsto also be interconnection points. This is likely to be expensive and the Forum have stated thatthey do not expect any operators to consider this.

The justification that the Forum have used to offer two rather than three interconnectionproducts (that is, not local interconnection production) is based on cost efficiency. They arguethat by upgrading the largest 100 (or so) RCU’s to make them interconnection points, anoperator can cover the vast majority of the Danish population. The rest of the country can becovered by using transit interconnection calls. They claim that this is economic because:

• Only a small percentage of traffic needs to use the TS interconnection route;

• The TS interconnection call is cheaper that today’s TS call;

• The TS interconnection call is not much more expensive than today’s LE interconnectioncalls; and

• The LE interconnect (RCU sites) that is used to cover most of the population is also cheaperthan today’s LE call.

The Forum conclude that the overall cost for other operators to cover the entire country istherefore lower than today.

The model reference papers made two comments about the types of interconnection productsthat need to be modelled. In the general model reference paper, Telestyrelsen stated that:

“The bottom-up model’s choice of technology and optimisation is subject to the scorched node assumption and therequirement that the notional network as a minimum should offer the same quality of service required of theSMP operator, and be able to provide a functionality similar to that of the SMP operator’s products.”

In the bottom-up mode reference paper, Telestyrelsen stated that:

4 If the former option is adopted, it will be necessary to show that the tandems have the necessary functionality to

deal with the additional traffic they must be capable of handling.

Telestyrelsen

17

“The bottom-up model is not constrained to replicate a network structure with five regional interconnectionareas. But provided the bottom-up model, developed as part of the LRAIC project, comes up with a switchinghierarchy that includes at least two levels of exchanges, then it will be able to replicate the currentinterconnection products, with a corresponding identification of network elements”.

Telestyrelsen accepts that, in principle, the absence of a local interconnection product in theForum’s optimised model may not actually matter provided that it can be replicated using themix of exchanges in the model. However, the question of whether it is more efficient dependson the actual costs of the network.

Telestyrelsen does not believe that the arguments provided by the Forum are fully documented,but accepts that based on the Forum’s current assumptions and costs, it may be more costeffective to cover the population with a mix of RCU interconnection and some tandem levelinterconnection. The relevant “local interconnection” product in the bottom-up model, musttherefore include some single transit interconnection and some form of weighting will benecessary to create a new product that covers those areas reached by the existing localinterconnection product offered by TDC.

It is equally important to state that this may not be the case under a different set of assumptions,and perhaps for those assumptions used in the hybrid model.

Telestyrelsen, therefore, accepts that the interconnection products offered in the bottom-upmodel do not need to replicate exactly those in TDC’s existing network. However, the hybridmodel may need to make adjustments to the model to ensure that the same functionality isrequired and that the Forum’s assumptions hold when costs and assumptions change. Also, thereshould not be any “external” costs incurred that enable a similar product to be offered.

Bottom-up issue-VIII Telestyrelsen accepts in principle that a different set of interconnection products can be modelled in thebottom-up model and note that the services modelled in the top-down and bottom-up model are notequal. Telestyrelsen therefore invites comments from the two parties on how to make the localinterconnect product in the bottom-up model correspond better to the local interconnect productactually provided by TDC.

A.3.2.4. General views on the Modelling Approach and Assumptions Used

The LRAIC Forum have adopted an approach to the modelling of exchange costs (although notnecessarily all the related costs) which is of more an apportionment exercise than a costingexercise. Essentially, the group starts from a total cost (for each size of switch), subtracts costs forknown bits and then uses percentage apportionments to estimate the costs of the rest. Thus themodel estimates port costs for TS’ as follows: (Unit Cost – Processor) x Port % with a furtheradjustment to take account of BHEs carried by the average switch in a given size category. Attimes such calculations are done with inconsistent sources (the RCU Unit Cost allegedly comesfrom EE; the other RCU costs come from the Forum).

This approach implies that the total cost of switches could be calculated by multiplying thenumber of switches in a given size category (an input) by the total cost of that category (aninput). The only role of the rest of the exchange calculations is to allocate these total costsbetween access and core on the one hand and between set-up and duration within core, on theother. Therefore exchange costs are allocated indirectly in the bottom-up model.

Bottom-up issue-IX

Telestyrelsen

18

Telestyrelsen believes that a more direct approach to modelling exchanges i.e. calculating the costs ofeach exchange component directly is more appropriate and welcomes views on this.

In Telestyrelsen’s view further justification on key assumptions such as routing factors and otheradjustments made in the model are needed. The routing factors have been based on ahypothetical two-tiered network where using probabilities for the routings of each service basedon key assumptions on the distribution of calls and traffic. These figures are difficult to verify. Insome cases, for instance busy hour and duration information, the Forum have used engineering“rules of thumbs” or even replaced TDC data with their own data arguing that their data is moreaccurate. These adjustments have a significant impact, not least in how costs are allocated to finalservices. Some of the adjustments made have the effect of allocating more costs to some servicesand less to others, particularly the final interconnection services. Telestyrelsen will need to reviewthese data carefully and ensure that accurate and consistent data has been used.

Bottom-up issue-X Telestyrelsen would welcome comments any views on the assumptions used to calculate routing factors.

Telestyrelsen

19

A.4 Review of Transmission Network and InfrastructureThe purpose of this chapter is to review the methodology followed by the LRAIC Forum toestimate the costs of transport. The approach that the Forum has used has been presented in themodel documentation provided to Telestyrelsen and in subsequent, and more detailed, papersproviding further documentation and justification. This chapter will, therefore, not repeat themodelling methodology but rather focus on the main assumptions made, their importance, andpresent some issues for further comment.


This section provides an overview of the approach followed by the LRAIC Forum whenmodelling transport, particularly in the following areas:

• ring based network;

• the approach taken to model total capacity;

• trenching requirements and sharing; and

• cables.

A.4.1.1. Ring Based Network

The Forum have modelled a transport network with three tiers of rings – RCU rings, secondaryrings and primary rings. The RCU rings link localised RCU sites and are usually low capacityrings. The traffic from the small RCU rings is then combined over larger secondary rings. Thesecondary rings link the RCU’s traffic to the tandem exchanges, which are themselves linkedtogether on the primary rings. This structure has been modelled because the Forum argues thatit is more flexible and more resilient than the types of transmission systems used in TDCs actualnetwork. SDH rings, they argue, all supply diverse paths with auto re-routing of traffic over theother ring path of the ring in the event of failure.

RCU rings link the neighbouring RCU sites and there are between 5 and 12 RCUs on eachrings. In fact, 50 per cent of rings have eight nodes, 20 per cent of rings have five nodes, and 30per cent of rings have 12 nodes.

Secondary rings link a number of RCU rings together. Each secondary ring links into onetandem exchange site.

The connection of traffic between sites is based on optical fibre technology. SDH multiplexing— at speeds up to STM 64 — are used to combine PSTN traffic with other services (such asleased lines and datacomms). As such, all services share the same SDH transport layer.

A.4.1.2. Modelling Total Capacity

The transmission requirements for PSTN capacity have been modelled in a way that is commonfor bottom-up models. That is, they start from actual call minutes, allow for growth etc and thenapply routing factors that reflect the way that the different PSTN services use the transmissionnetwork.

Non-PSTN services, on the other hand are modelled in a different way to PSTN services. Themain difference is that routing factors have not been applied, even though this was requested inCriterion BU XIX of the bottom-up model reference paper. The Forum have argued that the

Telestyrelsen

20

methods used for routing of leased line data over the core network are uncertain to them andinformation that would have helped them was not made available by TDC. Instead, theyassumed that some leased lines were “short” and that some were “longer”. The long lines wereassumed to require additional TS-TS capacity. The impact of the assumptions made regardingleased line routing factors will be addressed below in more detail.

Additional link capacity is included for resilience and to ensure that there is adequate overheadfor traffic grooming, tests and to ease the engineering implementing. The Forum describe twoways of achieving this:

• first, a 100 per cent additional capacity factor is allowed for that allows all traffic to have aone-for-one additional logical path;

• second, each physical link section of a ring is dimensioned to be able to carry the entirecapacity of the ring (this allows the ring to be broken and the entire traffic to be routed inany direction to the destination).

A.4.1.3. Trenching Requirements and Sharing

Trench requirements for the core network are estimated by reference to the distance betweensites. Adjustments are then made to allow for sharing between different rings and between otherutilities.

The physical distance between two RCU sites is calculated by multiplying the average “crowsflight” distance between RCU sites in TDC network (an input of the model) by the “crows-flightfactor”, that converts crows flight distances into actual distances (another input of the model).Multiplying this distance by the number of nodes on each sized ring gives the total physicaldistance for trenching before adjustments for trench-sharing.

Secondary rings are assumed to share as much trench as possible with trench with RCU rings. Toconnect several RCU rings into a secondary ring, some additional trenches are required to fillgaps. Primary rings, like secondary rings may require some stand-alone trench where sharing isnot possible.

The cost of digging trenches depends on the types of surfaces. The total trench length is allocatedto different surface types in order to improve the accuracy of the trench cost estimate.

A.4.1.4. Cables

All fibre cables that are used to connect the sites have 24 fibres. These cables are placed in ductsand all ducts are assumed to be buried to the depths and under the different surfaces as requiredlocally.


There are four main areas where Telestyrelsen has raised questions about the approach taken tomodel the transport network. These are:

• the modelling of primary rings and other transmission assumptions;

• the size and modelling of non-PSTN capacity;

• the appropriateness of the “crows flight” adjustment to determine trench requirements; and

• the extent of trench sharing.

Telestyrelsen

21

A.4.2.1. The Modelling of Primary Rings and Other Transmission Assumptions

The LRAIC Forum has assumed that there are three primary rings in their network — one foreach land mass. The secondary rings are attached to the primary rings. Further, the Forumassumes that only double tandem PSTN traffic and national non-PSTN traffic is routed in theprimary rings.

In November 2001, Telestyrelsen requested information from the LRAIC Forum about thecalculation of equipment in primary rings. In particular, Telestyrelsen asked:

“Do the calculations used to estimate equipment requirements for rings take into account traffic flowing betweenprimary rings, and if so how?”

The Forum did not respond to the request.

The Forum has estimated the capacity in each primary ring by taking total capacity andapportioning this between rings according to a given rings share of total nodes. This procedureworks reasonably well in some circumstances but where there are a number of parallel flowswithin and between rings it may result in highly misleading results. For example, for a call whichuses a switch in Ring A and a switch in Ring B and on the further assumption that neither switchis at a gateway node, both rings need to be dimensioned to carry such traffic – in this case theformula is clearly under dimensioning capacity. Another example that could lead to a generallyhigher use of transmission capacity than the one actually dimensioned for, is the choice of awidespread use of long traffic routes (connection between two RCU’s) in the bottom-up model,instead of using more switching points. While the traffic between RCU’s is switched only intransit exchanges, the distribution of traffic from RCU’s between different transit exchangesimply several long traffic routes in parallel on one transmission route, and that implies a lowtrunking effect (in principle that is: simultaneous use of a specific number of circuits by trafficbelonging to different traffic routes – instead of the routes being transferred one by one inparallel and separated on the same transmission system - would mean a more effective use oftransmission resources, and thereby a higher trunking effect).

This will cause traffic flowing over the primary rings to be understated. Telestyrelsen will takethis into account when developing the hybrid model.

There are a number of other transmission assumptions that also need to be reviewed carefully byTelestyrelsen. These include:

• The dimensioning of the meshing factor. The model includes 5 links which Telestyrelsenpresumes are to link the 3 main land masses of Denmark together. In the first place linkswill need to be dimensioned to take account of traffic flows between the land masses – thishas not been properly done in the model. Secondly, it is unclear why 5 such links have beenchosen. In principle, it would be possible to create a fully resilient network with as few as 3links. Dimensioning on this basis would mean, for example, that if a link between landmasses A and B were to go down traffic would flow to B via C. Both the links between Aand C and B and C would need to be dimensioned for this eventuality as would the SDHring in C. The same principle applies to all other traffic flows between land masses. Using 5links rather than 3 means that the dimensioning problem arising from traffic between landmasses having to pass, in the event of link failure, via an intermediate land mass, however itdoes not eliminate it entirely. Six links are required for this purpose. Thirdly, as a result ofthe above, there is a trade-off between the number of links and the capacity required on eachring and link. Account of this trade-off should be taken in the modelling. Fourthly, it maybe better to use Line Terminating Systems rather than ADMs for the links between landmasses;

Telestyrelsen

22

• The dimensioning of secondary rings does not appear to have taken proper account of the needfor more than one STM64 on individual sites. The model appears to take this into accountin determining the number of optical line cards but not in terms of the number of ADMs;

• The mix of gateway systems and ADMs. The model appears to use both systems at each nodewhereas one would anticipate the gateway systems to be used at the ring head;

• The modelling of cross connects. Firstly, it is unclear from the modelling (and from thedocumentation) the role actually performed by cross-connects. While in the top-downmodel, DXC 4/1 systems are necessary because of the different rates of tributary cards on linesystems and switch ports this is not generally the case in the bottom-up model, although seecomments on tributary cards below. There may well be a need for cross-connects at thehigher level of the network for flexibility particularly in respect of leased line traffic but thecross-connects in the model do not appear to perform this function. As a result, theinclusion of a very large number of cross-connects, far larger than in the top-down model, issomewhat puzzling. Secondly, the model has selected cross-connects which match thedimensions of the ADMs. While this simplifies the modelling, it appears to Telestyrelsen thatthis might not necessarily be the most cost-effective option. Thirdly, the formula in themodel multiplies the number of intersections with the average price of different size cross-connects where this average price is based on STM capacities of the cross-connects at thehigher part of the network hierarchy. Telestyrelsen believes that this formula is incorrect.Furthermore it is not clear, as to how cross connects are used in relation to networkmanagement (e.g. re-routing of traffic in the case of single point failure).

• The modelling of tributary cards. An initial issue is that the model apparently assumes thatthere are 2x2 Mbit/s on each card, whereas equipment specifications typically refer to cardswith either 16x2Mbit/s or 32x2 Mbit/s. A related issue is that the number of tributary cardsdimensioned may not be consistent with the capability of the ADMs to handle these cards,particularly as most traffic is modelled as being at the 2 Mbit/s level. Thus, an STM16 has 8slots each of which can have 32x2 Mbit/s line cards giving a total of 532 Mbit/s compared tothe STM16 itself which has a capacity of 2,480 Mbit/s. In order to overcome this problemmore ADMs may needed or cards with higher speeds could be used in conjunction withcross-connects.

Telestyrelsen will examine each of these in detail and will make any necessary adjustments in thehybrid model.

A.4.2.2. The Size and Modelling of non-PSTN Capacity

The percentage of non-PSTN capacity is very high in the bottom-up model. The Forum put thisdown to two reasons:

• the large volumes of non-PSTN traffic. The values for this traffic were supplied by TDC andwere not able to be verified by the Forum; and

• PSTN volumes will be lower in the bottom-up model because the routing factors are basedon an efficient network design.

The LRAIC Forum state that the relative importance of non-PSTN services depends on anumber of factors. For example, they note that the use of leased lines by other mobile networkshas a major effect as mobiles can be one of the largest buyers of leased line services. However, in

Telestyrelsen

23

some countries, mobiles can self provide transmission capacity. They also make the point thatmost of the growth in demand in operators’ networks is from data services, and not from PSTN.

The size of PSTN capacity does matters from a costing perspective. This is because all of thetransport equipment is shared by the services that use it, then a way needs to be found to allocatethe costs to each of the services. The LRAIC Forum has chosen to allocate the costs to eachservice on the basis of the Mbit/s used by each service. The result is that almost three quarters ofcosts are taken away from the PSTN costs.

Telestyrelsen acknowledges that the approach taken by the Forum is one that is often used inexercises such as this. However, it also recognises that it may not be the best way to allocate someof the shared costs in the transport network. For example, it may be that leased lines — or atleast some leased lines — use the network in a different way to other services. They may, forinstance, use the network more or less intensively that other services, for example, a leased linebetween two adjacent RCU sites does not need to be directed via a tandem switch. If leased linesuse the network less intensively, then there may be an argument that they should bear less of thecost than other services.

In Telestyrelsen’s view, routing factors for non-PSTN services should be used for allocating costsbetween non-PSTN and PSTN services and proposes to use this allocation methodology in thehybrid modelling phase.

Bottom-up issue-XI

Telestyrelsen would welcome views from parties as to whether there are more appropriate ways toallocate costs that are shared by services over the transport network other than routing factors.

When deciding how to allocate shared costs, Telestyrelsen is aware of the need for more accurateinformation about how services actually use the network. The LRAIC Forum noted that:

“To enable any modelling of the routing to be carried out accurately would require information from TDCabout how many of each leased line type are routed locally (within one RCU zone), regionally (between nearbyRCUs or within one transit switch zone), or nationally (between transit switch zones). This information has notbeen made available.”

The LRAIC Forum also claim that:“The model did not attempt to derive “more accurate” values for the factors by producing a “leased line routingtable” as data for the routing table was not available. Introducing the table would not have fundamentallyincreased the model accuracy — it would have only increased the number of input assumptions required.”

Telestyrelsen do not entirely agree with this statement. The bottom-up model reference paperrequested that the model should identify, for each service, routing factors for the switching andtransport network. Given that leased line and datacoms services make up such a large part of thetransport network — and more importantly, take away so much of the costs of the transportnetwork — it is important that they are estimated accurately. This has two implications:

• first, that the routing factor information be applied to these services in order to ensure thatthe dimensioned demand is estimated accurately;

• secondly, that the appropriate amount of network capacity (an costs) is allocated to theseservices. As discussed above, this may mean that different allocation rules are used to allocateshared costs to these services.

Telestyrelsen

24

A.4.2.3. The appropriateness of the “crows flight” adjustment

In section A.4.1.3 we noted that the physical distance between two RCU sites is calculated bymultiplying the average “crows flight” distance between RCU sites in TDC network by the“crows-flight factor”, that converts crows flight distances into actual distances. Multiplying thisdistance by the number of nodes on each sized ring gives the total physical distance for trenchingbefore adjustments for trench-sharing.

Telestyrelsen has received from the Forum detailed information on the methodology used toestimate the average “crows flight” distance between RCU sites. Telestyrelsen, however, has notreceived yet any more precise information on the source of the “crows-flight factor” than the oneprovided by the Forum in a cell of the core model, 5 which states that this has been derived from“survey of own network link”.

Bottom-up issue-XII Telestyrelsen invite both parties to provide their views on the distance between RCU sites in TDC’snetwork as modelled by the Forum and on the relevant assumption for a “crows flight” factor forDenmark.

A.4.2.4. The Extent of Trench Sharing

Trench sharing reduces costs and given the extent of sharing in the Forum’s model, Telestyrelsenneeds to have confidence that the estimates provided are accurate. In November 2001,Telestyrelsen asked the Forum to provide any support for the trench and duct sharingproportions used and to justify the amount of trench and duct common with the access networkand with other utilities. Telestyrelsen also asked that any evidence from Forum members beprovided.

The Forum responded to this request by producing supporting documentation on the extent ofsharing with other utilities and with the access network.

In terms of sharing with other utilities, the Forum assumed that sharing was possible, desirableand that it reduced costs. Consequently they argued that a forward-looking efficient operatorwould have more sharing than is done today. However, the Forum also added that there was nodefinitive evidence to support the percentage sharing factor that an efficient operator should beable to obtain. The value used in the model is:

“an average of opinions from the modelling team and Forum members. It is unlikely to be much lower, and itcould be higher”.

The LRAIC Forum conclude that:“The original 11% and 29% were averaged views. Since there is no definitive evidence for any one number wehave altered the values in the model to 10% and 30% as the original values suggest a degree of accuracy thatsimply does not exist.”

Telestyrelsen agrees that trench sharing with other utilities is desirable and that an efficientoperator would attempt to share trenches where possible and practical. However, account mustbe taken of the trade off between build out period and discounts in equipment prices. Whenassuming a short time period for network build out it is difficult to achieve large sharingpercentages with other utilities, however, large discounts on equipment are likely to be feasible.

5 Cell H30 of the spreadsheet Network design rules of version 1.3 of the core model.

Telestyrelsen

25

Telestyrelsen does not believe that it has been sufficiently documented that the extent of sharingin the bottom-up model is in accordance with the cost information used in the model.

In Telestyrelsen’s view no firm evidence has been delivered by the Forum to estimate the extentof trench sharing with other utilities including a discussion of the relationship with build outperiod and discounts. However, Telestyrelsen accepts that it is difficult to find definitiveevidence, but did ask the Forum to provide evidence for any estimates provided. However, ratherthan provide any evidence from the networks of Forum Members, it appears that a range of“opinions” were produced and an average taken. It appears to Telestyrelsen that it has not beendocumented that these opinions are supported by actual experience in the Forum Member’snetworks.

The other type of sharing is sharing between the core and the access network. The Forumappears to have been a little more systematic when estimating the extent of sharing between thecore and access networks. In particular, they argue that:

“A core network link between two sites will leave each site in duct/trenches. These exit points will need to beshared in almost all cases to avoid huge numbers of exit ducts from each site. The core link will usually be alonga road or street, and we would expect most of these to have at least some houses or businesses nearby. Thereforealmost all roads that the core network link will use, will also have an access duct/trench. An efficient operatorwould endeavour to share the expensive digging as much as possible, even if this required slightly longer cablelength.”

It is unclear how this theoretical approach was implemented. Indeed, it appears to Telestyrelsenthat rather than doing any calculations, a “best guess” approach was taken with the final numberbased on some consensus,

“From inspection of street maps and a consideration of the ubiquity of the access network, we estimated 95%of the core network linkages could share facilities with the access network. This figure was the consensus viewof the Forum modelling team. There is no empirical supporting evidence for this value as none of the membershave a similar access network to TDC. We took a more pessimistic view on primary rings (80%) as these havemore inter-city links and hence are more likely to pass along roads without houses.”

Telestyrelsen does not believe that the resulting percentages have been supported sufficiently andthat a more systematic approach is needed. Telestyrelsen accepts that the Forum does not operatesimilar net as TDC in Denmark. Telestyrelsen does however believe that some Forum membersoperate national networks — including access networks — in other countries. Supportingevidence from these networks could have been provided in order to cross-check the resultsderived from the Forum’s “best guess” approach.

Bottom-up issue-XIII Telestyrelsen accepts that trench sharing with other utilities and between core & access is importantand invites the parties to submit evidence supported by actual network experience in other networks.

Telestyrelsen

26

A.5 Review of Co-Location ModelThe purpose of this chapter is to review the methodology followed by the LRAIC Forum toestimate the costs of co-location. The approach that the Forum has used has been presented inthe model documentation provided to Telestyrelsen and in subsequent, and more detailed,papers providing further documentation and justification. This chapter will, therefore, not repeatthe modelling methodology but rather focus on the main assumptions made, their importance,and present some issues for further comment.


The approach adopted by the LRAIC Forum when estimating the costs of co-location has beento estimate the costs of specific cost components and then determine the amount of co-locationspace required. These two are then brought together to produce the cost of co-location services.These are shown in Table 3.Table 3: List of co-location services provided by the bottom-up model

Co-location services Unit NotesCo-location space m2 Includes indoor cost per m2, racking and power supply. Space can be

used for exchange of traffic or access lines.Co-location access Line Paired cable related costs.Co-location traffic Link Coaxial cable related costs.Caged co-location m2 Cost of guarded room fit-out.Co-location shared Shared line Paired cable related costs assuming shared usage of raw copper".Security visit m2 Additional costs of carrying out security visits.

The bottom-up model does not estimate the costs of all types of co-location. The Forum arguethat this could be quite operator specific. Instead, LRAIC based costs for co-location will belimited to costs associated with sharing facilities, such as buildings and exchange equipment.

Each of the main components of the co-location model are examined below.

A.5.1.1. Estimating Costs

There are two types of costs included in the co-location model — the costs associated withbuilding and access co-location services and the costs associated with co-location equipment.

The costs of building and accessing co-location services are included in the model either fromTDC data or input assumptions from the Forum. The actual (annualised) building costs comefrom the main core model. Four room sizes are assumed — 10 m2, 30 m2, 80 m2 and 200 m2 —and the cost of fitting out each of these sizes is estimated. The “fit out” costs include powerwiring/fuse systems and fit out of area with smoke detectors.6 Other costs such as the incrementalcost of cooling and guarded room fit out (e.g. cage installation) are also added Guarded room fitout costs also include the costs of doors access control, plan & design. There are also additionalcosts for room fit out (e.g. electric power meters per customer and alarms).

6 The default data includes central-site costs for power relating to co-location space.

Telestyrelsen

27

In addition there are other costs, some of which vary with the number of visits. These costsinclude cost of electric power per kWh, co-location planning and training (which includesbusiness wide costs to set up the business) and security person for guarded access. Some of thesecosts depend on the number of visits per operator per site and their duration.

The second category of costs in the co-location model are those associated with co-locationequipment. These costs include the costs of tie cable from MDF to HDF, tie cable from HDF toDDF coax, DSLAM "splitter", HDF (Cu pairs), HDF (coax DDF), and cable run/channelling.7

A.5.1.2. Estimating Demand

Much of the demand data comes from TDC although the LRAIC Forum has made a number ofimportant assumptions. In short, the model looks at the number of sites where co-location isdemanded and groups these sites into geo-types and ultimately to differently sized co-locationrooms. The average number of operators per site is also estimated and is based on estimatesprovided by TDC.

Co-location demand data is a mix of demands for 2000, 2001, 2003 and 2005. Only one set ofdemand data can be used in the model at a time.


At this stage, Telestyrelsen has four main areas in the co-location model that require furthercomment. These are:

• the assumptions about the growth rates in the model;

• the justification for the costs that have been inputted into the model;

• the technical features of the co-location services modelled; and

• the service modelled.

Each of these are considered in turn.

A.5.2.1. The Assumptions about the Growth Rates Assumed in the Model

In November 2001, Telestyrelsen wrote to the LRAIC Forum requesting information about howthe growth figures for total co-location had been estimated.

The Forum responded in January 2002 and stated that it had no market forecast data tosubstantiate the values of co-location service demand in future years. The values used havetherefore been made using extrapolations from the Y2000 data supplied by TDC. They alsonoted that their extrapolations took account of the following expectations:

• demand in 2000 is probably well below today’s demand as co-location services were in anembryonic stage then;

• unbundled network service are expected to grow rapidly; and

• further into the future demand growth is uncertain, but the Forum expect some slow downand hence some market saturation effects to show (e.g., possibly no growth after 2003).

7 Cable run/channelling is the additional equipment that must be installed to interconnect the co-location area to

the incumbent operator’s equipment in the building. It therefore links to the hand-over distribution frames and tothe interconnection switch.

Telestyrelsen

28

The Forum concluded by noting that they expect that further work will be carried out to updatethe market demand forecasts for all of the co-location services.

Telestyrelsen agrees that further work is needed to estimate growth rates.

Bottom-up issue-XIV Telestyrelsen would welcome comments from the operators on the estimates for growth of co-locationservices.

A.5.2.2. The justification for the costs that have been inputted into the model

The Forum has used many of the cost inputs in the co-location model having providing limitedjustification for them. There seems to be no supporting justification, for instance, on the level ofcosts for accompanied visits and power. Telestyrelsen therefore believes, that it has not to asufficiently extent been documented that the costs in the co-location model are reasonable.

Bottom-up issue-XV Telestyrelsen would welcome comments on the reasonableness of the cost input in the model.

A.5.2.3. The Technical Features of the Co-location Services Modelled

TDC has raised concerns in their comments to Telestyrelsen about the Forum’s model about theviability of the co-location services modelled in the bottom-up model. TDC have queried, forexample, whether the use of cages are an efficient way of saving on accompanied access arguingthat in practice, access to exchanges will still involve passing TDC equipment.

A.5.2.4. Services Modelled

As stated in section A.5.1 the bottom-up model provides cost estimates of the following co-location services: co-location space, co-location access, co-location traffic, caged co-location andco-location shared and security visit. However, the common guidelines MRP specifies that theservices to be modelled should be those specified in TDC’s standard agreement on co-location(16. march 2000) annex 2. These include:

• Cost per m2 indoors;

• Cost per m2 outdoors;

• Rack positions (one-off costs and yearly costs);

• Power supply (one-off costs and yearly costs);

• Paired cable terminated in the SMP operators main distribution frame (one-off costs andyearly costs);

• Coaxial cable trunk terminated in SMP operators distributor (one-off costs and yearly costs);and

• In-span fiber optic cable (one-off costs and yearly costs)

Although many of these service costs in the bottom-up model may be extrapolated from theunderlying cost input and assumptions, since these are shown at a fairly disaggregated level, the

Telestyrelsen

29

modelled services do not match those required in the common guidelines. More specifically, thebottom-up model does not include the costs of outdoor m2 and in in-span fiber optic cable.Further, the cable lengths on which the costs of the services paired cable and coaxial cable arebased have not been provided.

The service outputs in the co-location model do not accurately match the services provided byTDC. Specifically, Telestyrelsen have not been able to identify the costs of outdoor m2 and in in-span fiber optic cable. The Hybrid model will need to adjust the structure of the model to takethis into account.

Telestyrelsen

30

A.6 Costs in the Bottom-up ModelThis section covers subjects common to the BU-models, including the approaches andassumptions used to estimate replacements costs, annualised costs, operating costs, indirect costs,overhead cost and working capital costs.

A.6.1 Approach and Assumptions to Estimate Replacement Costs

Replacement costs have been estimated either by using costs from the Europe EconomicsAdaptable BU model as a starting point or by direct input from Forum members. Somedocumentation for inputs was provided in the data input reports of November 2001.

A.6.1.1. Methodology to Estimate Replacement Costs

A.6.1.1.1 Replacement costs in the access network

With regard to the access model equipment and installation unit costs may be split into thefollowing two categories:

• Cost of links for trench, duct, copper and fibre that are expressed in DKK per km; and

− cost of trench per km has been estimated on the basis of the costs of trenches overdifferent terrain. The costs cover the whole process from initiation to completion,implying gaining consent, planning, digging, making good (restitution), labour (exceptinstallation), materials (except duct and cable), and any other cost components.

− costs for three types of duct are provided in the model. The most common is 4/5 channelduct, as used in street trench. Mini-duct is assumed from the street to the customer NTPand large duct is assumed to be used close to exchange for many cables.

− unit costs of copper per Km are provided for different pair sizes ranging from 1 to 1,000pairs.

− unit equipment and installation costs of fibre are provided for different number of strandsranging from 12 to 196.

• Cost of nodes for PDP and SPD cabinets, jointing boards and Network Termination PointsNTP that are expressed in Kr per number of cabinets, jointing boards and NTP.− Cabinet costs and costs of jointing boards in PDP/SDP are categorised by size. Unit costs

for each cabinet size and jointing board are estimates of the LRAIC Forum and cover thewhole process from initiation to completion similar to the cost of trench above.

− NTP equipment and installation costs are input for POTS 2 wire, ISDN2 2 wire, xDSL2 wire, fibre, ISDN 30, Leased lines (> 2Mb) and Datanet. NTP equipment costs areNTP alone excluding installation. Installation includes planning, commissioning andjointing/splicing.

A.6.1.1.2 Replacement costs in the core network

Replacement costs in the core model may be divided into infrastructure, transmission, switchingand common site costs.8 The costs included in each of these categories are:

8 See section on operating costs and indirect costs for a more detailed discussion on common site costs.

Telestyrelsen

31

• Infrastructure. Trenching and ducting costs per Km where trenching is subdivided in tosimilar trenching categories used in the access model. Input cost values are from Forummembers.

• Transmission. Cabling (fibre), microwave links, submarine cable links, manholes, splicingbox, optical distribution frames (ODF), SS79, STM multiplexer gateways, STM multiplexerADMs, cross-connects, STM cards electrical, STM cards optical, STM line cards,regenerators, Digital Distribution Frames (DDF), DDF fibre termination point and atomicclock. Input cost values are from Forum members and otherwise modelling team estimates.

• Switching. Costs are divided into RCU and TS and specified for low, medium, high and veryhigh volume for the following equipment types: RCU/TS units, fixed processor, switchblockunit, variable processor unit, software unit and port unit. Further, RCU costs include thecost of line cards. Other costs include signalling points, back-up power supply, power supplyunit, air conditioning unit, security system, site preparation and POI. According to theLRAIC Forum, exchange equipment prices are based on the adaptable bottom-up model. Inthe Forum note “Network Design Justification” of January 2002 a description is provided forthe analysis used to derive values from the adaptable bottom-up model.

Separate installation costs are provided for the above mentioned categories.

A.6.1.1.3 Replacement costs in the co-location model

With regard to replacement costs in the co-location model these primarily include room fit out,Incremental cost of cooling, Guarded room fit out, tie cables, DSLAM splitter, HDF and Cablerun/channelling. Input cost values are from Forum members.

A.6.1.2. Issues for Further Comment

There are three issues for further discussion arising from the approach used by the Forum toestimate replacement cost. These are justification for some equipment prices used, approach todiscounts and other Danish specific factors, and the use of estimates from the Europe Economicsmodel.

A.6.1.2.1 Justifying equipment prices used

As previously stated in a letter of 26 November 2001 “Changes Required to the Bottom-upModel”, the input reports produced by the Forum did not address key documentation issues.Further documentation has since been provided, however it is Telestyrelsen’s view thatdocumentation is still lacking after receiving answers and clarifications provided by the Forum.In this case it is important to stress that interoperability between different suppliers should beensured and additional costs if any should be included A significant part of the equipment usedin the bottom-up model may be regarded as “no name” and as such it is impossible to verifyprices or interoperability.

9 Set to zero in the model.

Telestyrelsen

32

Bottom-up issue-XVI Telestyrelsen believes that replacement costs used in the core model are not documented to a sufficientlyextent (except some of those adapted from the adaptable bottom-up model which may be traced back).A significant part of the equipment used in the bottom-up model may be regarded as “no name” and assuch it is impossible to verify prices or interoperability. Telestyrelsen therefore invites comments on thereasonableness of the cost input in the model.

A.6.1.2.2 Approach to discounts and Danish specific factors

The Forum has provided no documentation of whether equipment prices include volume andtrade discounts or reflect the prices that an efficient operator with the bargaining power of anSMP operator in Denmark would face. The prices derived from the adaptable BU-model appearnot to be adjusted for any Denmark-specific factors, including discounts.

Related to this is the issue of the time horizon for building the network in the bottom-up model.In particular, there is the issue of whether the assumptions made about trenching costs andtrench sharing are consistent with the assumptions that have been made about network roll-out.

A.6.1.2.3 Estimates from the Europe Economics adaptable model

The LRAIC Forum has listed the source of many of their equipment prices as the adaptablemodel developed by Europe Economics. Telestyrelsen believes that this is questionable as theEurope Economics estimates have only been used as an input to the Forum’s own estimates. TheForum has used their own assumptions and some of their own costs – i.e. the cost of line cards –to develop the cost of a hypothetical RCU or tandem exchange. The full cost of an RCU unit inthe bottom-up model, for example, includes the cost of line cards. These were however notmodelled in the adaptable model.

A.6.2 Approach and Assumptions to Estimate Annualised Costs

A.6.2.1. Methodology Used to Estimate Annualised Costs

The bottom-up model reference paper stated that where practical the bottom-up model shoulduse economic depreciation. It also stated that the model documentation should indicate on a caseby case basis for all major asset classes, which of the alternative approaches to depreciation ismost likely to generate a depreciation profile similar to economic depreciation. In addition, thegeneral model reference paper stated that the models should be sufficiently flexible to allowanalysis of the impact of using an alternative method for calculating annualised costs.

The LRAIC Forum did not meet the criteria regarding economic depreciation. In the model anddocumentation of 1 November 2001, they argued that the annuity style approach that they usedinstead was rational, sufficiently accurate, practical and able to be justified. Telestyrelsenrequested that the Forum comply with the criteria and the revised bottom-up model includedmore options with regard to annualisation techniques. However, the model did not attempt toestimate economic depreciation nor did it carry out any “off-line” investigations into the formulathat most closely matched economic depreciation.

Telestyrelsen

33


There are two issues that arise when considering annualised costs. The first of these is theformula used by the Forum to estimate a tilted annuity. The second is the source for some of theparameters used in the model to estimate annualised costs.

A.6.2.2.1 The appropriateness of the PMT function

The PA methodology uses the Excel function PMT to annualise costs. The PMT functionreturns the periodic payment for an annuity based on constant payments and a constant interestrate when payments are due by the end of the year. This translates into the followingannualisation formula:

+

−×

+−

nn CoCSVI

CoC

CoC)1(

111

,

where CoC is the Cost of Capital, n the asset life, I the capital cost (investment) and SV denotesscrap value. To derive the final annualised cost a simple average of the annualisation formula EoYand BoY is calculated equivalent to applying the following formula:

+

−+

×+

+

−××

+−

+1)1(15.0

)1(5.0

11

1nnn CoC

SVCoCI

CoCSVI

CoC

CoC

However, when the PMT formula is used to calculate a tilted annuity the price trend issubtracted from the cost of capital in all places. For example for an end-of-the-year calculationthe PMT function uses the following formula:

−+

−×

−+

−

−nn pCoC

SVI

pCoC

pCoC)1(

11

1

The results obtained are very close, but not the same, to the ones which would be obtained byapplying the standard formula for a tilted annuity (see formula below). The PA annuity leads toan overestimation of the annualised value for negative price trends and an underestimation withpositive price trends compared to the true annualised cost based on the tilted annuity formulaEoY:

+

−×

++−

−nn CoC

SVI

CoCp

pCoC)1(

111

A.6.2.2.2 Source of Parameters

The annualisation calculations in the model require assumptions on the cost of capital, asset pricetrends and asset lives. The basis for the cost of capital calculation is a report by Ernst and Young,which is discussed elsewhere in this report.

Asset pricetrends and asset lives were decided by the Forum on the basis of data submitted byForum members. These appear to be based on judgements rather than by a more systematicapproach, e.g. benchmarks supported by economic depreciation studies. Furthermore, it appearsthat in each case, the value selected was that which was most favourable to the Forum’s case, for

Telestyrelsen

34

example, the longest asset life and the highest assumed rate of price increase. This results inmodelling assumptions, which does not seem plausible in all cases.

A.6.3 Approach and Assumptions to Estimate Operating Costs

In general, operating costs in the bottom-up model are calculated as a mark-up on capital costssince, according to the Forum, it has not been possible to accurately estimate these from firstprinciples. Their argument for this approach is that some activities are usually missed, resultingin an underestimate of costs if activity data is used in a bottom-up model. Therefore the Forumhas opted for an approach where costs are based on estimates by the Forum of the realoperational cost - based on their practical experience.

A.6.3.1. Methodology Used to Estimate Operating Costs

A.6.3.1.1 Operating costs in the access network

In the access model operating costs fall in two classes: One driven by capital expenditure(equipment and installation) and the other one driven by events. When deriving operating costsdriven by events two sets of assumptions are needed.

1. Operating events (per year) per unit of equipment; and

2. Unit cost for each of these events.

This approach has been used for operating costs relating to trench, fibre and copper NTP’s andcopper and fibre cable. For example, the annual operating costs of km trench is calculated as theunit operating cost of 1,000 km trench multiplied by the number of times 1,000 km trenchneeds maintenance (number of events) per year which in turn is multiplied by the total trenchlength in 1,000 km. All these operating costs are therefore driven by key volume indicators,rather than by capital expenditure cost.

A.6.3.1.2 Operating costs in the core network

In the core model there are a number of costs that can be defined as operating costs or indirectcosts. In this section we define operating costs as maintenance. Other operating costs are treatedas indirect network costs are and discussed below.

The approach that the Forum have adopted to estimate maintenance has been to calculate it as amark-up on capital expenditure costs and includes all associated social, travel and managementcosts.

A.6.3.1.3 Operating costs in co-location

In the co-location model all operating costs are calculated as a mark-up on capital costs.


According to the model documentation, operating costs as a percentage of the equipment capitalcost were estimated by Forum members and submitted to the Forum’s consultants. Any absolutecost values submitted through the PA data-clearing process were converted to a percentage, bytaking the ratio of the value to the capital cost.

The question then arises as to whether the operating costs submitted by the Forum areconsistent with the equipment assumed in the bottom-up model. There are two issues toconsider here. First, the operating costs that should be applied need to be the costs associated

Telestyrelsen

35

with operating new equipment, not a mixed vintage of assets. Second, the operating costs neededin the model are those that would be incurred in operating a network of the size and structureassumed in the bottom-up model. These two effects may operate in different ways, but it isunclear how they have been factored into the estimates used in the model.

Telestyrelsen acknowledges the difficulty in satisfactorily modelling of operating costs in abottom-up model. However, only sparse documentation and justification has been provided formark-up’s used in the models. In particular, no justification has been provided for the allowancefor ineffective time when calculating labour costs and no indication is given for the assumedpercentage of labour costs due to ineffective time as required in criteria BU LIII. Thedocumentation simply states that ineffective time has been included in the cost estimates.Further, no indication is given of how values from different Forum members where chosen byPA through the data clearing process.

Telestyrelsen believes that it has not been sufficiently documented that the operating costs havebeen calculated according to MRP guidelines. It is not clear how the operating cost estimates inthe bottom-up model are related to actual operating costs of operators, and how other factors,such as ineffective time have been estimated and included in the ratios.

A.6.4 Approach and Assumptions to Estimate Indirect Costs

A.6.4.1. Methodology Used to Estimate Indirect Costs

The Forum has used different approaches to modelling indirect costs. Many have been modelleddirectly without the use of mark-ups. In the core model these include:

• Accommodation. This refers to the space occupied by one unit of equipment on one sitemultiplied by the cost per square meter. These figures are based on data supplied by TDC asa result of an independent site valuation exercise. The space occupied by one unit ofequipment on one site is estimated for each identified equipment type and an annual cost inKr per m2 for RCU and TS sites is provided.

• Power cost which is the average power consumption in kWh per year per unit of equipmentmultiplied by the cost per kWh. Air conditioning cost is the peak power output per year pernumber of square meters occupied (max kW per m2) multiplied by the annual cost ofrunning an air conditioning unit to dissipate the heat of the maximum kW per m2 emittedby a piece of equipment. Unit costs for equipment related to Power supply, Air conditioning,security system and site preparation are provided for RCU and TS sites. These are defined ascommon site costs and allocated to non-PSTN, co-location, switching, transmission based ontheir individual share of common site capital costs.

• Annual equipment requirements of site maintenance are provided for RCU and TS sites andallocated as common operating costs.

• Security guard costs for TS sites is DKK 300,000 based on the assumption that a full timesecurity guard is needed. For RCU sites an estimate is provided assuming 10% of sites havefull time security person equivalent to DKK 30,000 per RCU site.

The assumptions used in this case are clear and easily audited in the cost input sheet.

Other indirect costs have been modelled using a mark-up based approach, cf. section below.

Telestyrelsen

36

A.6.5 Approach and Assumptions to Estimate Overheads Costs

A.6.5.1. Methodology Used to Estimate Overhead costs

Except for one-off service costs, all service costs in the bottom-up model are uplifted to takeaccount of overheads.

Overhead costs in the bottom-up models are called common business costs that relate tocompany-wide supporting departments. The Forum define them as:

“Costs that are required by an efficient operator with SMP in Denmark, with the scope of services similar toTDC. These costs are common to the businesses of Core, Access, Co-location and other retail services. They arecosts that are required to make the business function, but are not directly related to the services or the network.”

Thus, according to the Forum, the common business costs cover:

• executive (chairman & board and head office);

• business planning and development;

• accounting, finance and audit;

• external relations (relationships with external companies/parties/other operators, corporateadvertising etc);

• human resources (including payroll);

• information management;

• legal;

• procurement; and

• other general administration.

The common business costs used in the model have been computed as the sum of total commonbusiness costs, depreciation cost of common business and capital employed for common businessmultiplied by the cost of capital. Although the Forum states that these costs were estimated forthe entire business including both retail and wholesale costs and that a proportion thereforeshould be excluded to account for retail services, this share is set to zero implying that the inputdata in fact only includes wholesale costs.

Common business costs are then divided by the total cost of the network, including cost ofcapital to yield the uplift factor for share of overhead costs.

Input values in the bottom-up model were obtained from Forum members, who were asked totake account of the MRP guidelines and in particular the need to adjust values to becommensurate with an operator with TDC’s scope of services. The lowest value submitted wasselected.


The estimate used for overheads in the bottom-up model is not broken down into any level ofdetail. It is, therefore, impossible to assess how much of the overhead estimate is made up ofaccounting costs and how much of legal costs (for example). Telestyrelsen acknowledges that anyvalues should reflect the scope of an SMP operator in Denmark. However, as stated in BUcriteria LI, values for overhead costs should also only be included if they are efficiently incurred

Telestyrelsen

37

in building and operating a wholesale core and access network in Denmark. No documentationhas been provided justifying that this is the case.

Further, values used in the model are derived from the values submitted by the operator with thelowest values. This may potentially lead to an underestimation of costs given trade-offs betweendifferent classes of cost; differences in classifications between operators and potentially differencesin network design. In addition, as stated by Forum documentation, one reason for choosing theadopted approach was to gain a more accurate estimate by attempting to estimate the overheadcosts for the entire business including both retail and wholesale instead of attempting to estimatethe common business cost for a fictional wholesale-only business. Since the proportion attributedto retail service is zero in the model the Forum seem to have opted for estimating overhead costsfor the wholesale business in contrary to their explanation.

Bottom-up issue-XVII Telestyrelsen requests information and views of the parties on whether the overhead estimate includesall the necessary costs.

A.6.6 Approach and Assumptions to Estimate Working Capital Costs

A.6.6.1. Methodology Used to Estimate Working Capital Costs

The uplift for working capital is given by the ratio between the cost of working capital and totalcost of the network. The cost of working capital is obtained by multiplying working capital bythe Cost of Capital.

Working capital is obtained by multiplying debtors days by the total cost of the network(including cost of capital) minus creditor days multiplied by the total cost of the network(excluding the cost of capital), divided by 365. With regard to cash the Forum indicate in theirmodel documentation that a percentage increase in the debtor days may be used instead of afigure for the amount of cash a prudent operator requires. Therefore an estimated number ofdebtor days due to cash is added when calculating working capital.

Working capital is calculated according to MRP guidelines. However, assumptions are based onForum member data and no documentation has been provided as to whether they reflect theworking capital of an efficient operator.

Telestyrelsen

38

A.7 Sensitivities in the Bottom-up ModelThis section summaries the results of an analysis of the sensitivity of the bottom-up model tovarious factors. The purpose of this analysis is to provide information on the inputs to thebottom-up model that have major influence on the cost outputs.

The bottom-up model is generally sensitive to changes in the cost of capital, price trends andasset lives when viewed together. However, of the above the most influential single input to themodel is the cost of capital.

Changing many inputs collectively has, as one would expect, significant effect on the outputs ofthe model. For example, changing all input costs (line cards, digging trenches, ducts, copper andfibres) in the access model by 20% implies an increase in service cost of between 11.3% and15.1% for annual usage costs and between 14.7% and 20.0% for expensed costs. The followingsections do not try to map out combinations of different inputs that have significant influence onresults, rather the emphasis is in on discussing individual parameters that may or may not play arole for the outputs of the model.

In order to conduct this analysis, Telestyrelsen constructed a sensitivity tool. This tool enablesTelestyrelsen to analyse a large range of inputs (both individually and collectively) and quantifytheir effects on cost categories, network elements and services. The general impression of thisanalysis is that the bottom-up model is fairly insensitive to a vast majority of inputs when theyare changed individually. An exception to this is the co-location model. Key results are discussedand summarised below. It should be noted that the results presented are based on version 1.2 ofthe bottom-up model (except co-location where version 1.3 has been used).

A.7.1 Sensitivities in the Access Network

Th most influential input on access service results is the cost of capital. The effects on results ofusing a cost of capital of 11% and 13% instead of the 9.3% default are illustrated in the tablebelow.

Table 4: Access model sensitivity to the cost of capital (all other inputs unchanged)

Access service (DKK) Base case New values(CoC 11%)

Change New values(CoC 13%)

Change

Raw copper 2 wire 177.03 208.84 18.0% 247.84 40.4%Raw copper 4 wire 263.60 318.20 20.7% 385.71 46.8%"shared" 2 wire raw copper 95.44 111.69 17.0% 131.59 38.3%Fibre, dedicated link in to RCU or on ADM ring 2,026.55 2,431.79 20.0% 2,952.63 45.7%Dark Fibre – Access (NTP to RCU site) 2,026.55 2,431.79 20.0% 2,952.63 45.7%

With regard to technical parameters in the access model, a few warrant special attention. Theseinclude the level of spares in the copper requirements, route sharing parameters and cost sharingparameters:

• The sensitivity of the level of spares in the copper requirements is generally high. Increasing thelevel of spares from 1.2 to 1.44 (a 20 per cent increase) for SDP - PDP and PDP - RCUroutes and from 1.0 to 1.2 for NTP-SDP has the following effects:

− NTP – SDP: annualised service costs are increased by a maximum of 5.6% for rawcopper 2 wire and a minimum of 3.6% for 2 wire copper – ISDN.

Telestyrelsen

39

− SDP – PDP: annualised service costs are increased by a maximum of 4.2% for raw copper2 wire and by a minimum of 2.7% for 2 wire copper – ISDN.

− PDP – RCU: annualised service costs are increased by a maximum of 4.9% for rawcopper 2 wire and by a minimum of 3.2% for 2 wire copper – ISDN.

• The sensitivity of changes in route sharing parameters is very low. For example, a 20%increase in all route sharing km’s only leads to an average of 1.7% reduction in annual copperservice costs and a 3.7% reduction in annual fibre service costs.

• The sensitivity of changes in the share of cost borne by the access network and the share of costborne by access and core networks – between them the sensitivity to changes is low. For example,the highest reduction in service cost for increases of approximately 20% in all parameters are7% for annual fibre service costs and 3.7% for annual copper service costs.

With regard to unit costs there are two inputs that are of some importance. These include thecost of trench and the cost of copper cables.

• The cost of digging trench implies some sensitivity to the results. For example, increasing alldigging prices (except mini-duct) by 20% implies an average of 1.3% increase in annualcopper service costs and a 3.6% increase in annual fibre service costs. Increasing the cost oftrenching mini duct by 20% increases expensed service cost by between 2.1% and 13.6%.

• The cost of copper cables has some influence, especially driven by cost of 1 pair cables. A 20%increase implies increases expensed service costs of 2 and 4 wire raw copper by 10.1% and13.5% respectively. Cost variations of 2, 5, 10, 20, 30, 50, 100, 250 and 1000 pairs hardlyhave any effects on service costs.

A.7.2 Sensitivities in the Core Network

The single most influential input in the core model is cost of capital. Changing the cost of capitalhas an impact on all annualised costs, the most sensitive being cost categories related to trenchingand ducting. The most sensitive network elements include RCU-TS Tx and TS-TS Tx. Theeffects on service results of using a cost of capital of 11% and 13% instead of the 9.3% defaultare illustrated in the table below.Table 5: Core model sensitivity to the cost of capital (all other inputs unchanged)

Core services (DKK) Base case(averagecost perminute

New values(CoC 11%)

Change New values(CoC 13%)

Change

Local interconnection 0.0165 0.0172 3.9% 0.0180 8.7%Single tandem 0.0232 0.0242 4.2% 0.0253 9.3%Double tandem 0.0289 0.0302 4.5% 0.0318 10.0%

In the following the effects of changing key asset lives, price trends, replacements costs andparameters in the core model are discussed for those inputs that Telestyrelsen has found to be themost sensitive to changes.

• Asset lives and price trends: In general, changing individual asset lives only has very minoreffects on output. One exception is the asset life of fixed processor units for RCU and TS.Here a reduction in the default asset life of 15 years to 10 years increase duration relatedservice costs between 6.1% to 4.0% and set-up related service costs between 6.7% to 0.8%.

Telestyrelsen

40

This significant effect is due to the substantial costs of these units in comparison to otherinputs in the model. Therefore not only asset lives but also the choice of price trends have animpact on the final service costs.

• Replacement costs: With regard to changing replacement cost values the only individual costinput with any significant effect is the full cost of RCUs for all volume categories. Increasingthe cost of a low volume RCU by 20% has the most significant effect on localinterconnection service costs with a 111.7% increase in set-up costs and 3.4% increase induration costs. Similar but not nearly as influential effects can be traced when increasing thecosts of medium, high and very high volume RCUs.

• Parameters: The vast majority of parameters (both technical and allocational) do not give riseto significant changes in output on an individual basis. One exception is the distribution ofnon-PSTN demand parameters.

− The proportion of all non-PSTN traffic sent up to TS sites from RCU sites. Changing thedefault value of 30% to 40% especially affects cost categories related to transmission. Ona network element level the effect is -16.4% on the capital cost of the RCU-TS Txnetwork element and 4.4% on TS-TS Tx. On a service cost level there are no effects onset-up costs. However, the effect on duration related costs varies considerably fromservice to service from -2,8% on Single tandem IC out & in to 4,4% on Numberport -additional I/C.

− The proportion of all non-PSTN traffic sent TS-TS. Changing the default value of 10% to20% also has major influence on cost categories related to TS-TS transmission. On anetwork element level the effect is –17.5% on replacement costs related to TS-TS Tx.On a service cost level there are no effects on set-up costs, however. The effect onduration related costs are negative with the greatest effects being -13.4% and –20.8% onOLO-OLO transit (inter region) and Number of ports - additional I/C respectively. As wouldbe expected there are no effects on local interconnection service costs.

A.7.3 Sensitivities in Co-Location Model

Unlike the two other models the cost of capital is the not the single most influential input in theco-location model. Here replacement cost inputs and demand assumptions have greater influenceon the results.

With regard to replacement costs the effects are summarised in the following:

• Room fit out. A 20% increase in costs results in an increase of 8.14% in annual service costs ofco-location space;

• Incremental costs of cooling. Changing capital costs has only very minor effects;

• Guarded room fit-out. Changing capital costs has a one-to-one effect one caged co-locationservice costs;

• Additional cost for room fit out: per operator. A 20% increase in costs results in an increase of9.78% in the annual service cost of co-location space and a 0,35% increase in one-off costs;

• Tie cable to HDF from MDF (100 Cu pairs). A 20% increase in costs results in an increase of15,96% in the annual service cost of co-location ULL;

Telestyrelsen

41

• Tie cable HDF to DDF coax: A 20% increase in costs results in an increase of 15,24% in theannual service cost of co-location traffic and a 15,51% increase in one-off costs;

• DSLAM "splitter". Increasing capital costs has a one-to-one effect one both annual and one-off service costs of co-location shared;

• HDF (Cu pairs) per 100 pairs. A 20% increase in costs results in an increase of 4.04% in theservice cost of co-location ULL;

• HDF (coax DDF) per 10 pairs. A 20% increase in costs results in an increase of 4,76% in theannual service cost of co-location shared and a 4,49% increase in one-off costs;

• Cable run/channelling. A 20% increase in costs results in an increase of 10,22% in the one-offservice cost of co-location space;

• Cost of electric power per kWh. Doubling he cost from 0.5 to 1 Kr per kWh results in a23,80% increase in the annual cost of co-location space; and

• Co-location planning and training. Changing capital costs has only minor very effects on theannual costs of co-location space.

With regard to demand there is especially one assumption that is of very significant influence.This is the chosen year to analyse service costs. Changing from the default of year 2000 leads tovery significant changes in service costs. This is especially true of co-location LLU and co-location shared with variations of many hundred per cent depending on the chosen year.However, co-location traffic and caged co-location are also effected quite substantially. The onlyservice cost that is generally insensitive to change in demand is co-location space.

Telestyrelsen

42

PART B: REVIEW OF TOP-DOWN MODEL

Telestyrelsen

43

B.1 IntroductionThis part of the report discuss the characteristics of the top-down cost analysis focusing on whatTelestyrelsen perceives to be the main strengths and weaknesses of the analysis undertaken byTDC. Although the reader is assumed to be familiar with the top-down model documentation,in depth knowledge of the analysis and the SAS-model is not required. The reader is not assumedto be familiar with TDC’s response to questions posed by Telestyrelsen during the reconcilliationphase.

Where appropriate Telestyrelsen briefly describes the methodology adopted by TDC - especiallywhere the adopted approach is not clearly described in the documentation. This description bothserve to inform the reader as well to allow Telestyrelsen to verify with TDC that Telestyrelsen hasa correct understanding of key aspects of the analysis.

B.1.1 The Top-down Cost Analyses

TDC’s top-down cost analysis comprises:

• SAS-calculation module which allocates the different cost categories to services;

• Economic depreciation module (in Excel);

• Model documentation; and

• Andersen report on cost of capital.

The original version of the model delivered to Telestyrelsen on 1 November 2001 attributedvalue to fully depreciated assets and consequently did not meet Criteria TD 20. Telestyrelsentherefore requested TDC to modify the model so that fully depreciated assets were not attributedany value. The revised model was delivered to Telestyrelsen on 20 December 2001. TDC hasemphasised that it does not find Criteria TD 20 to be in accordance with the legislation andtherefore considers the first model to constitute the top-down cost analysis.

TDC will provide Telestyrelsen with a final version of the model when PriceWaterhouseCoopershas finalised its audit of the model. The revised model will incorporate a number of correctionsfollowing review comments from Telestyrelsen and corrections arising from the audit.

B.1.2 Key Results

The tables below show the results of the top-down model for the major services.

B.1.2.1. Including the value of fully depreciated assets (1 November 2001)

Table 6: Core results

Interconnection service (øre) per minute per call average per minute

Local interconnection 3.40 5.11 4.95

Interconnection within area 4.93 8.96 7.66

Interconnection between areas 1.67 5.07 3.21

Telestyrelsen

44

Table 7: Access results

Interconnection service Annual cost per year (DKK)

Raw copper 2,106

Dark fibre 9,261

B.1.2.2. Excluding the value of fully depreciated assets (20 December 2001)

Table 8: Core results

Interconnection service (øre) per minute per call Average per minute

Local interconnection -c- -c- -c-

Interconnection within area -c- -c- -c-

Interconnection between areas -c- -c- -c-

Table 9: Access results

Interconnection service Annual cost per year (DKK)

Raw copper -c-

Dark fibre -c-

Telestyrelsen

45

B.2 Overall Comments on Model and DocumentationSome of Telestyrelsens overall concerns with the model are set out below. A more detaileddiscussion of the model is provided in chapters B.3-B.10.

B.2.1 Theoretical Modelling Approach to Capital Costs

A concern for Telestyrelsen with TDC’s top-down model is the extensive use of a theoreticalmodelling approach in respect to asset valuations, i.e., the model starts from assumptions anddimensioning rules rather than TDC’s actual network and costs.

A major strength of a top-down approach, as opposed to a bottom-up approach, is that it isfounded and based on the SMP-operator’s costs of operating in the relevant market and relies ondetails provided by actual cost accounts as opposed to detailed assumptions in bottom-up modelsor corrections to national operating conditions in benchmark approaches. Furthermore a top-down model normally provides a strong audit trail with the possibility of cross checking results,as costs can be traced back to audited accounts.

The top-down model provided by TDC, on the other hand, is in many cases not based on actualquantities and only relies on TDC’s asset register to a very limited extent. The gross assetvaluations used in the model are in many cases the result of the network being modelled fromscratch subject to the scorched node constraint.

A top-down model will always use data which is not in the accounts in order to estimate GRC.On the one hand it needs an inventory of equipment (number of exchanges with number of linecards/ports and so on) and on the other hand it needs price information. In this respect TDC’smodel is not so different from the norm. In many cases however, TDC has estimated equipmentwhich is required rather than taken such information from data bases.

In some cases, in particular for access duct and drop cable, the problem appears to be a reflectionof limitations in TDC’s information systems. In other cases, for example exchange ports, thejustification for TDC’s approach is unclear.

In the case of its annualisation approach, TDC has only used its asset register to a limited degree.According to TDC, the reason for this approach is that the information in TDC’s fixed assetsregister is not reliable. This is due to a number of transfers between accounts and the fact that anumber of entries have been made as net book values.

This naturally raises concerns with regard to the reliability of the results. Also the substantial useof theoretical modelling in the top-down model is not necessarily in line with section 15 (3) ofthe Executive Order on interconnection, which states that “The cost analysis shall start from theprovider's existing cost structure, replacing outdated technological solutions with optimaltechnology.”

In addition to this problem of a more fundamental nature, Telestyrelsen has some more specificconcerns with certain parts of TDC’s theoretical modelling. For example with regard to thereliability of the access-trench calculation, which in one case (for average length of core pairs) isbased on a very small sample.

B.2.2 SAS model and GRC calculations

The actual model delivered by TDC is programmed in SAS and starts from the estimated GRCof the different HCC’s. The underlying GRC calculations are not included but have been carriedout in separate sub-models. The SAS-model then allocates the GRC of the HCCs (input to the

Telestyrelsen

46

model) first to network elements and then to interconnection services, drawing on cost-volumerelationships, volumes and routing factors.

As mentioned in chapter 12 of the top-down MRP “transparent top-down and bottom-upmodels are a prerequisite in the reconciliation process”. As regards the allocation of costcategories to network components and services, the SAS model is reasonably transparent.However, the GRC calculation undertaken in the sub-models are not. Telestyrelsen has beengiven a presentation of the models but has not had access to the actual models. However, inmany cases the methodology used for estimating the GRCs are described in the documentation.On the other hand the actual calculations are not documented. In some cases, the informationprovided in the documentation in combination with other information provided has enabledTelestyrelsen to cross check TDC’s calculations. Based on the information provided,Telestyrelsen has been able to, for example, to generate similar GRC for exchanges, but not toprecisely confirm the values used in either the model or the documentation.

In some cases, e.g. transmission equipment, Telestyrelsen has insufficient information toundertake cross check of asset valuations. Telestyrelsen has therefore requested an equipmentinventory from TDC. Telestyrelsen intends to carry out further checks once this information hasbeen received. This information is also important to refine the reconcilliation calculations fortransmission equipment.

With regard to the validation of TDC’s GRC calculations, thorough auditing by theindependent auditors is therefore of great importance.

B.2.3 Application of CVR to allocate fixed and variable costs

At a very late stage in the process Telestyrelsen has become aware that TDC has used a ratherdifferent approach for allocating fixed and variable costs on the basis of CVR curves thanTelestyrelsen had originally understood to be the case, on the basis of the documentation.

The methodology adopted by TDC implies that the total cost included in the SAS-model inmany cases does not correspond to the total costs estimated and documented for the individualcost categories. This is the case either where the CVR is convex in which the costs in the SASmodel will exceed the relevant GRC for a particular cost category or where the CVR in concavewhere the opposite is the case. An example of a convex CVR is copper distribution cable wherethe cost allocated to network elements is more than DKK -c- million higher than the total GRCfor copper distribution cable. For other cost categories such as e.g. SDH equipment the costsincluded in the model are substantially lower than the total amount of costs associated with SDHequipment. As a whole the adopted methodology implies that the total amount of costs allocatedto network elements is around DKK -c- million lower than the total amount of costs.

The adopted methodology does not ensure that an efficient operator would be able to recover itsforward-looking costs. On the other hand, the methodology may also lead to an efficient SMPoperator recovering more than his forward-looking costs. Hence TDC’s allocation methodologyis not in accordance with the overall principle of setting prices at LRAIC plus a mark-up forcommon costs as described in section 2.2 and 2.3 of the general MRP.

By coincidence the overcompensation (for the access network) approximately matches the undercompensation (for the core network). However, this is simply a coincidence. Moreover prices seton this basis would distort the investment signals for the core and access networks.

Telestyrelsen

47

B.2.4 Utilisation Ratios

In a large number of cases TDC applies the actual utilisation ratios without justifying that theseutilisation ratios also constitute efficient operation (that there is no inefficient excess capacity).

In other cases TDC uses theoretically calculated utilisation rations based on TDC’s planningnorms etc. In many cases these references to planning norms constitute the only justification forthe utilisation ratio applied and often there is generally no comparison with the actual utilisationratios in TDC’s network.

Section 15(3) of the Executive Order states: “The cost analysis shall start from the provider's existing cost structure, replacing outdated

technological solutions with optimal technology. Existing technology/infrastructure elements willthus have to be replaced by new and optimal types of equipment in those cases where theinfrastructure elements in question are outdated, and therefore no longer represent an optimaltechnology.”

As opposed to the bottom-up cost analysis, the top-down model should therefore not model anetwork as it would be built if TDC were to build its network today. In the view ofTelestyrelsen, the optimisation/redimensioning of equipment should only be undertaken if itoptimises costs compared to the existing network. In principle one would therefore expect theutilisation ratios in the modelled network to be at least as high as in TDC’s existing network incases where the existing utilisation ratios are known. In TDC’s existing model this does not seemto be the case in all instances. An example is drop cable where 5 pair cable has been used insteadof the actual cable sizes. This issue will need to be examined in more detail in the hybridmodelling phase.

B.2.5 Building Costs

TDC’s approach to calculating building costs is generally not well documented. Following arequest by Telestyrelsen for further documentation on the calculation of cost per sqm, TDC hassimply informed Telestyrelsen that sqm costs have been calculated as (1+-c-%)*(-c-+publicvaluation/sqm*WACC) where -c-% is a mark-up and the -c- DKK/sqm covers renovationnecessary for the buildings to maintain their value. However, the data, assumptions andcalculation underlying the -c- DKK per sqm and how the public valuation/sqm price has beencalculated has not been described.

Moreover, TDC has made no adjustment for vacant space. According to TDC there is no(inefficient) vacant space. TDC has subsequently informed Telestyrelsen that this view is basedon inspection of 13 selected buildings as well as an evaluation by experts with a detailedknowledge of TDC’s buildings. Given that the network has evolved over time, however, and thatbuildings will often have been dimensioned for equipment, which are different from theequipment currently is use (typically smaller), it is somewhat surprising that there should be noexcess capacity. This will need to be examined in more detail.

B.2.6 Annualisation

According to TDC, the fixed asset register does not allow for a consistent division of assets byvintage. It has therefore not been possible for TDC to adopt the rolling-forward methodologyprescribed by TD 21. Instead, TDC has adopted a single vintage approach and used theNBV/GBV methodology discussed in section 7.1 of the TD MRP. This approach can be clearlyseen from an examination of TDC’s economic depreciation model which effectively looks at the

Telestyrelsen

48

annualisation charge applicable (under a variety of methodologies) for an asset which has thesame average lifetime as implied by the NBV/GBV ratio coming from TDCs FAR. Thisapproach gives rise to at least two potential problems. Firstly, the NBV/GBV ratio is likely toprovide misleading estimates of the average life of assets where there is asset price inflation ordeflation. Secondly, even in the absence of such inflation or deflation, it is unclear whether itwould generate the same results as a multi-vintage approach.

B.2.6.1. Price trends

Most of the price trends used in the top-down model are low/negative. In fact, except forcooling, ventilation and motor vehicles all price trends are negative, cf. table 5 in chapter 8 of thedocumentation. Even the cost of digging which is normally reported to be increasing, is assumedto be falling by 6% per year. Many of the price trends applied also seem to be inconsistent withthe GRC/GBV ratio. For a number of asset categories, TDC has estimated GRCs that aresubstantially higher than the GBV indicating a positive price trend. On the basis of input prices,however, TDC has estimated price trends to be negative. This apparent inconsistency isparticularly critical for access duct, copper cable, copper equipment, power and ventilation,property and PC’s. It should be noted though, that part of the difference between GRC andGBV could be due to the fact that some assets that are still in use are not recorded in the fixedasset register and are therefore not included in the GBV values.

Price trends should be forward-looking, indicating the expected future development inequipment prices. Using the historic development in prices as a proxy may often be the best wayto estimate future price changes. Such a methodology can only be used, however, if the pastdevelopment is believed to continue in the future. According to TDC the falling trenching costsare the result of renegotiating contracts and merging existing contracts into larger contracts.Rather than indicating that trenching costs are likely to decrease in the future, this could,however, also indicate that costs have been inefficiently incurred in the past. This is an exampleof how one should be very careful when projecting past price trends into the future.

Also TDC’s methodology for selecting the time period over which to calculate the past pricechanges seem questionable. TDC does not seem to have applied any consistent methodology forselecting the time period. Sometimes TDC uses a period of 5 years (copper) and sometimes aperiod of 17 years (fibre). The underlying rational for these differences are not justified. In anycase a period of 17 years is by far too long a period for estimating the future development in theprice of optical fibre. In fact the fibre price index shows that fibre prices have been relativelystatic in recent years. In response to this point, TDC argued that this was not the case for largerdimension fibres. However, since the model is mainly based on 12 pair cables Telestyrelsen doesnot find this response convincing.

The very negative price trends imply that TDC incurs a substantial holding loss each year. Underthe FCM methodology this strongly inflates the depreciation charges. In many cases the assumedholding loss even exceeds annual depreciation. The strong impact of TDC’s assumed price trendsare illustrated by the sensitivity analysis provided by TDC on 21 January 2001 on request byTelestyrelsen.

Table 10 illustrates the effect on raw copper costs of changing the price trends for trenching,copper cable, copper equipment, buildings, and motor vehicles. The relevant price trends used inthe top-down model and for the sensitivity analysis are listed in Table 11.

Telestyrelsen

49

Table 10: Sensitivity results for Raw Copper

Weighted average LRAIC, DKK per line

1 November 2001 20 December 2001

Base case price trends*Annuity, Base case asset lives

WACC 14,5%2,099 -c-

Modified price trends*Annuity, Base case asset lives

WACC 14,5%1,688 -c-

Change 19,6% -c-

* See Table 11 below

Table 11: Price trends

Access, Copper TDC values (base case) Sensitivity valuesDuct/Trenching -6 0Copper Cable -5 2Copper Equipment -5 0Buildings -1 0Motor Vehicles 0 0

As can be seen from Table 10, changing the negative price of trench, cable, copper equipmentand buildings from around -5% to around 0%, holding all other variables equal, reduces the costof raw copper with 19-c-% (from 2,099 to -c- and from 1,808 to -c-).

The negative price trends also give rise to another effect. TDC has adopted a so-called “rollingbackward methodology”, where the GRCs estimated (and documented) as end-of-year values for2000 are rolled back to mid-year-prices using the price trends. Depreciation charges are thencalculated on this basis. The more negative the price trend, the higher the GRCs in the modeland the higher the interconnection costs. This also means that the GRCs used in the SAS-modeldeviates from the GRCs in the documentation.

Finally, as will be discussed in more detail in chapter B.4.1, the approach used to estimateeconomic depreciation is not sufficiently robust to provide convincing support for TDC’sselection of annualisaton methodology.

Top-down issue-I

Many of the price trends used in the top-down model do not appear to be forward-looking. Also themethodology used for selecting an estimation period for historic price trends does not seem to beconsistent.

B.2.7 MEA Adjustments

As specified in Criterion TD IV:“The MEA definition that the top-down model should use is that of an asset that can produce thesame services produced by the existing asset at lowest cost, adjusting where possible to reflectdifferences in operating costs, quality, asset lives and space requirements.”

As discussed in 3.2.2. of the top-down MRP, adjustments should be calculated for service qualityand functionality differences to account for the effect these differences will have on the net

Telestyrelsen

50

revenues generated by the asset for each year of the asset’s lifetime. These differences should thenbe discounted by the cost of capital and added to (where the existing asset is superior) orsubtracted from (where the MEA is superior) the valuation. In a similar manner adjustmentsshould be made for differences in operating costs, space requirements and asset life times. Withregard to operating costs, the MRP notes that differences may arise from differences inmaintenance costs, network management costs, and associated indirect costs.

TDC has used the MEA-approach widely. However, TDC has not identified any need for MEA-adjustments. Telestyrelsen finds it unlikely that there should be no indirect cost reductionsassociated with employing modern equipment.

Top-down issue-II

Telestyrelsen would welcome comments from the parties, and the LRAIC Forum in particular,pointing out the type of equipment where non-negligible cost reductions are likely. The respondents arerequested to explain whether the cost reductions stems from reduced operating costs, increasedfunctionality, reduced space requirement etc. Any quantification of the potential cost saving will alsobe appreciated.

B.2.8 Operating Costs

Operating costs have been derived on the basis of TDC’s regulatory business segmented accounts(“forretningsområderegnskabet”) for 2000. This generally seems to be a sound approach sincethese accounts according to TDC are based on the principles of activity based costing (ABC).Whereas TDC has provided substantial documentation on how the total relevant operating costshave been allocated in the top-down model, the documentation provides little support for thetotal amount of costs included in the model.

TDC has conducted an efficiency study of TDC, where TDC is compared with 52 US LocalExchange Carriers. Based on this study TDC concludes that TDC is efficient. Therefore TDChas made no efficiency corrections to the operating costs reported by the business segmentedaccounts. Telestyrelsen has a number of concerns with this efficiency study, which is discussed inmore detail in chapter B.9

B.2.9 Co-location and Interconnection Specific Services

Co-locations services and interconnection specific services are modelled separately. As opposed tothe other interconnection services, the documentation of the TDC’s methodology andcalculations is very limited and there seems to be a potential risk that some costs may have beendouble counted. With regard to building costs TDC has admitted that double counting hastaken place and informed Telestyrelsen that this will be corrected in a revised version of themodel.

B.2.10 Common and Shared Costs

TDC has used a somewhat different methodology for categorising costs as shared and commoncosts than the definition provided by Telestyrelsen. Whereas Telestyrelsen in the General MRPdefines a shared cost to be a cost that is shared between services within an increment and acommon cost to be a cost that is shared between services in different increments, TDC defines ashared cost as a cost that relates to either access or non-access only and a common cost as a costthat relates to be both access and non-access.

In general it is difficult for Telestyrelsen to review the allocation of shared and common costs inthe top-down model as the underlying calculations are not well documented. It is therefore

Telestyrelsen

51

difficult for Telestyrelsen to compare the costs with those costs estimated in the bottom-upmodel at a sufficiently disaggregated level.

Telestyrelsen

52

B.3 Gross Asset Valuation

B.3.1 General Comments on Dimensioning

In parts of the modelled network, dimensioning practice is based upon TDC’s actual network. Inother parts of the network, the dimensioning practice in the model is based on an assessment ofwhat TDC believes is required. The latter approach could potentially lead to rather differentutilisation levels than in TDC’s actual network. This is particularly the case with switches.

The issue of dimensioning is discussed in some detail in the sections on the various plantcategories. However, it is useful to give examples of actual and theoretical dimensioning in themodel. The cable network valued in the model is, with the exception of drop wires, based uponTDC’s actual network. Drop wires are valued on the assumption that each customer is providedwith 5 pairs. Telestyrelsen understand that while this is broadly consistent with TDC’s currentdimensioning practice (although this may change), this does not in all cases reflect thedimensioning policy which has been used in the past. Telestyrelsen understand that thisapproach may result in a rather lower level of utilisation than in TDC’s actual network.

A further example where dimensioning in the model is likely to differ with TDC’s actualdimensioning practice is provided by trunk ports on local exchanges, tandem and otherexchanges. In this case dimensioning is based upon a detailed examination of traffic flowsbetween exchanges taking account of such factors as variations by time of day (in a quitesophisticated way), variations by time of year and network growth. We understand that thisapproach may result in a rather higher level of utilisation than in TDC’s actual network.

As might be expected, dimensioning takes account of growth, modularities and other factors. Inthe case of growth the number of years growth considered depends on the part of the networkand reflect the fact that:

• In some parts of the network it is relatively easy to add additional capacity (e.g. exchangeports);

• In other parts of the network it is much more difficult to add additional capacity (e.g.trench).

In the former case, the dimensioning period is much shorter than the latter.

B.3.2 Overview of TDC’s Network

B.3.2.1. Access Network

The access network is a typical tree-branch structure (with the primary network between MDFand PDP; secondary network between PDP and SDP and final drop). The network ispredominantly copper based. Fibre is only used to a limited degree, specifically for links withcapacity in excess of 2Mbits (2 Mbits links provided over HDSL).

The costs of providing access differ by geotype as a reflection of differences in cable thicknesses(number of pairs); cable lengths and costs of laying the cable. The model uses information on ageotype basis to show these cost differences.

Telestyrelsen

53

B.3.2.2. Core Network

TDC has a conventional network hierarchy: with concentrators, local exchanges, tandemexchanges, international exchanges, IN. There is a 3 level hierarchy in the transmission networkwhich corresponds to some degree with the concentrator, local exchange and tandem hierarchy inthe exchange network. However, the correspondence is by no means complete, a point whichgives rise to some problems in bottom-up modelling.

Concentrators are linked to local exchanges by 2 physically distinct paths although each is onlydesigned to carry a portion of the total traffic. This means that if a local exchange goes downthen some traffic will be lost.

Likewise local exchanges are connected to two tandem exchanges in a given area and potentiallyto tandem exchanges in other areas, where this is warranted by traffic flows. In addition, there area number of local exchange to local exchange routes, again based upon traffic flows. The networkdoes not provide full switch or transmission resilience. Thus, to take an extreme example, if atandem exchange goes down during the busy hour as much as 50% of traffic could be lost.However:

• It is much more likely for an exchange to go down outside the busy hour;

• There is a widespread use of local to local routes and local to tandems in other regions.

For these reasons, a tandem switch failure would probably lead to a lot less than 50% traffic lossin practice. As a result, the exchange network offers a considerable degree of resilience.

There are -c- concentrators in TDC’s network located in 1,578 localities (incl. technical houses).-c- of these concentrators are remote concentrators. According to information provided by TDC,39510 of these locations are technical houses which together house around -c- of the lines in use.There are 146 local exchanges in service plus two test exchanges and 10 tandem exchanges (+2test exchanges). For each of the concentrators, local exchanges and tandem exchanges, two typesof exchange are purchased with rather different cost structures. Costs between the two differsignificantly for concentrators although this has little impact on the valuation because of the mixof exchanges by size category. On the other hand, the costs of one type of local exchange iscurrently far in excess of the other. Nevertheless, by far the majority of local exchanges are of themore expensive make. There is little difference in the cost of the tandem exchanges supplied bythe two manufacturers.

TDC’s transmission network is primarily point to point (with some multiplexing of routes) athigher levels with rings and chains used primarily in the access network. In general, LineTermination Equipment is used in combination with cross-connects at higher level sites whereasADMs tend to be used instead of cross-connects at sites lower down the transmission hierarchy.At some sites WDM equipment is used reflecting shortages of fibre in some areas, although forthe network as a whole there is a large amount of unused fibre.

B.3.2.2.1 Implications of islands

Denmark consists of a peninsula, two main islands and numerous smaller islands. Many of theconcentrators on the small islands are small and in some cases only have a single link to a local

10 This figure – provided in the document “LRAIC Centralbygninger ultimo 2000 – seems to be inconsistent,

however, with the information provided in table 3 in chapter 4 of the documentation, which seems to suggest thatthere are -c- technical houses.

Telestyrelsen

54

exchange. Clearly, the cost of links differs according to whether these are on land or undersea; insome cases radio is used for island connection.

B.3.2.2.2 Resilience

• As discussed, there is a considerable degree of resilience in the exchange network. In additionthe use of diverse routes on the transmission network ensures a considerable resiliencealthough this would be less than in a fully protected fibre ring system. According to TDC,full protection is generally not required by customers (although diverse physical paths is) andis only provided where actually required.

B.3.3 Access Network

Section 2 of TDC’s documentation discusses the valuation of the access network. The focus isentirely on direct network plant. Assets such as land and buildings, computers and transportwhich are associated with the access network are discussed in Section 5 of the documentation.

B.3.3.1. Trench valuation

As a preliminary comment, it is worth clarifying TDC’s trench laying practice. Initial discussionssuggested that cable is buried directly into the ground and not ducted. This is slightly misleadingsince an extra plastic tube is provided to allow for growth in the access network; in the corenetwork plastic tubing is assumed throughout the core network.

B.3.3.1.1 Trench Length

TDC does not have information on trench length within the access network. Therefore theselengths have been calculated as part of LRAIC process. The discussion below initially examinesthe calculation of trench between the Main Distribution Frame and the Secondary DistributionPoint. Further sections consider the calculation of shared trench and trench in the drop network.The calculation of trench length, which is calculated separately for each geotype, is based on thefollowing information sources:

• Number of copper line customers;

• Km cable per customer;

• Utilisation;

• Data base on cable dimensions; and

• Trench kilometre to cable kilometre ratio.

Multiplying the number of customers by the kilometres of cable per customer gives the totalrequired kilometres of copper cable pairs (assuming 100% utilisation). Applying the utilisationrates achieved gives the actual kilometres of copper cable pairs.

The data base on cable dimensions shows the proportion of cable length made up of cables ofdifferent sizes and enables the calculation of kilometres of 50, 100 pair cables and so on. This isapplied to the estimate of actual kilometres of copper cable pairs to estimate the number of pairsand length thereof. This is then converted into an estimate of trench kilometres by applying thetrench kilometre to cable kilometre factor gives an estimate of trench kilometres.

In principle, this methodology appears to be soundly based. However, in practice it could resultin inaccuracies. The first potential problem is that the estimate of kilometres of cable per

Telestyrelsen

55

customer is based on a very small sample. According to TDC this is because it is extremely timeconsuming to calculate kilometres of cable for a customer (an hour per customer). The proposedsample was of 350 customers with 100 from Metropolitan Areas; 100 from Large Cities; 100from Rural A and 50 from Rural B. Customers within each geotype were selected on a randombasis. However, TDC was not able to measure kilometres of cable in all cases and the finalsample turned out to have only 318 observations. Applying normal statistical formulae suggeststhat the true mean could be rather different than the mean estimated from the sample. Indiscussions, TDC has pointed out that these formulae are inappropriate because the distributionof cable length is not a normal one. It has further argued that a better approach would be toexamine the length of cable per customer where half the customers are randomly removed. Forthe case examined by TDC, this resulted in an 8% increase in the estimated length of kilometresper customer. While this is not a huge difference it nevertheless suggests that the estimate oftrench kilometres is subject to a non-trivial margin of error.

It can be noted that other data used in the model is either based on census information or frommuch larger sources (for example, utilisation figures are based on a sample of 10,000 cables foreach geo-type). Nevertheless, given the very big difference in estimated trench length between thetop-down and bottom-up models it may be necessary to examine details of the way themethodology is applied during the hybrid-modelling phase.

The estimate of trench length is based upon TDC’s actual network. Even if the methodology andthe data used to implement the methodology are correct, the resultant length may therefore notnecessarily be that of an efficient operator. As far as Telestyrelsen is aware, there are no publiclyavailable estimates of access trench length for other countries. However, given the very significantdifference with the estimated trench length in the bottom up model, it will be necessary toexamine TDC’s access trench network in further detail during the hybrid modelling phase, e.g.through an examination of actual trench maps.Table 12: Access Trench

Type of Trench Trench Length (kilometres)MDF to Secondary Distribution Point 137,027Trench Shared with Transmission Network 2,570Drop Trench 56,359

The results of this process, trench between the MDF and Secondary Distribution Point, areshown above. Also shown are the estimates of trench shared with the transmission network anddrop trench which are now discussed.

TDC does not know the length of trench which is shared with the transmission network. For thepurposes of modelling it has assumed that -c- of transmission trench in urban areas is shared withthe access network, namely 2,570 kilometres. Telestyrelsen has a number of comments on thisestimate, some of which have been previously discussed with TDC:

• Given that access trench in urban areas would cover all roads, the assumption of -c- sharingappears to be on the conservative side.

• With respect to trenching in other areas, TDC has argued that the routing of access and coretrench tends to be different and further that trench in these parts of the network are laid atdifferent times. While this may be true it is somewhat extreme to assume that there is nosharing at all.

Telestyrelsen

56

• TDC has assumed that there is no sharing with other utilities and has indicated that it is notpractice to share with other organisations. Once again this appears a somewhat extremeposition although Telestyrelsen recognises that trenches may be laid at different timesreducing the scope of sharing to some degree.

Finally, Telestyrelsen believes that the treatment of shared trench in the model is currently notsatisfactory. Essentially, the model involves all shared trench being deducted from the accessnetwork and, by implication, included in the core network. In response to this concern, TDChas stated that in the final revision of LRAIC all shared costs will be treated as common costs.Telestyrelsen views this change as an improvement. While the length of shared trench is assumedto be small, the length of trench for the cable drop is significant11. Unfortunately, themethodology used to calculate trench for drop cable is not well documented in the accesschapter.

B.3.3.1.2 Trench Prices

Trench prices are based on digging agreements as of 1 May 2001. Prices are collected for eachdigging area, cf. table 31 in chapter 2 of the documentation, and include a mark-up for “diggingunder road” and “Rekvisition” which Telestyrelsen interprets as the need to collects diggingpermits from the authorities etc. TDC then calculates average prices for each geo-type, weighingdigging costs with the relative share of exchange areas with a given geo-type in each digging area.

According to TDC trench prices have declined in recent years a factor which is reflected in themodel in two respects:

• GRCs reported in the documentation are generally estimated at year end – within the SASmodel these are converted to mid-year prices. Clearly if prices are decreasing mid-year pricesin model will be higher than year-end prices in documentation.

• The annualisation factor is higher (lower) according to whether asset prices are decreasing(increasing).

It is not clear from the documentation whether TDC has used the prices from 2001 directly asend of year data or whether they have been adjusted with the (negative) price trend. If pricesfrom May has been used to estimate end-of year prices and these GRCs have later been rolledback with the negative price trend (and thus increased), the GRCs would be artificially inflated.

Telestyrelsen is puzzled by the falling (nominal) price trend used for digging. In general,Telestyrelsen would expect trench prices to increase in real terms or at least stay constant in realterms. TDC argues that the nominal (and thus real) price decline reflects improvements inbargaining practice with contractors. In the view of Telestyrelsen, however, this could alsoindicate that TDC’s costs have not been efficiently incurred in the past.

B.3.3.2. Cable Valuation

B.3.3.2.1 Cable Length

The methodology used to estimate the GRC of cable is reasonably clearly described in thedocumentation. A major issue, however, is the utilisation levels of cable.

11 In the model this is somewhat misleadingly referred to as underground drop cable.

Telestyrelsen

57

Table 13: Utilisation levels in access network

The documentation shows the followingutilisation levels (in percentages)

Metro-politan

City Rural A Rural B

Primary and Secondary12 43.6 42.5 40.2 35.5Primary 64.2 62.8 46.2 42.2Drop 25.0 25.0 25.0 25.0

TDC states that its utilisation levels are broadly consistent with planning guides and are those ofan efficient operator. In summary:

• The current planning norm is for secondary distribution points to serve between four andeight houses. This provides a high degree of network flexibility but can only be achieved ifsecondary distribution points are close to the customer and have low utilisation.

• Further, the norm is currently to provide a five pair drop cable per dwelling. This reflects thefact that: I) people move on a regular basis; ii) there has been sustained demand for morethan one pair due to home working. This sustained demand has occurred despite ADSL andISDN; iii) conversion of buildings; iv) it is not significantly more expensive to install five, asopposed to one pair while it is extremely expensive to add extra cable where there is unmetdemand; v) there are modularities in cable size provision although a move to 4 pair cables hasrecently been introduced. Telestyrelsen recognises that these arguments have some meritalthough it also notes that the cable price in the bottom-up model for a 1 pair cable is farlower than that for a 5 pair cable in the top-down documentation. Clearly, this issue willneed to be further examined during the hybrid modelling phase.

• The planning norm is to have 2.5 pairs per customer between primary and secondarydistribution points which implies a maximum of 40% utilisation. In the past less sparecapacity has been provided in certain areas. However, according to TDC, this has resulted inproblems. Telestyrelsen will wish to examine any evidence TDC can provide on this pointduring the hybrid modelling phase.

• Finally, in the primary network the planning norm is -c- of capacity. Given cablemodularities this may result in less than -c- actual utilisation.

B.3.3.2.2 Cable Pricing

The documentation provides a good overview of cable pricing in the model. One point whichshould be emphasised is the allowance for cable wastage. According to TDC there are twosources of cable wastage. The first involves a wastage of 5 metres at each end of a cable. Thesecond involves wastage because cable is provided on cable drums in discrete sections of 500metres. For each cable size TDC has calculated when it is optimal to throw away remainingcable, balancing the cost of cable against the cost of splicing. While it would be potentiallypossible to move unused cable to other areas, in practice this is not rational when the remaininglength is comparatively short.

12 It is unclear from the above whether the first column represents utilisation in the secondary network or whether itis an average of utilisation rates in the primary and secondary network.

Telestyrelsen

58

For 100 pair cable e.g. TDC has estimated the critical length to be -c- meters. In a situationwhere the next stretch is more than -c- meters and there is a less than -c- meters left on the cabledrum, the cable drum is thrown away. TDC has applied this decision rule to a sample ofsubscribers where cable is being rolled out to the subscribers in succession. Using this method forthe sample as a whole, TDC has obtained a total wastage percentage of -c- including both typesof cable wastage.

For 5 pair (drop) cable, the percentage has been determined a little differently. Here it has beenassumed only to splice if there is more than 150 m left on the cable drum. The remaining cablewill always be thrown away if the next stretch is shorter than 150 meters and there is not enoughcable left on the drum. This and the 5 meters wastage at each end of a cable gives rise to asubstantially higher cable wastage percentage of -c-.

Taking into account both the cost of wasted cable and the cost of splicing increase cable prices byapproximately -c- for 5 pair cables and -c- for 1,000 pair cables. Since these figures areconsiderably higher than those for the bottom-up model TDC’s assumption will require furtherexamination during the hybrid-modelling phase.

TDC does not provide sufficient justification for the 5 meters cable wastage at each end or thecritical level for throwing away cable. As these values are critical for the result, they will need tobe examined more carefully. It should also be noted that the wastage percentages have beenderived on the basis of a purely theoretical calculation, which will need to be supplemented withevidence of actual cable wastage. In support of its methodology, TDC has noted that it is noteconomical to move small amounts of cable from one place of the country to another. Whereasthis may well be the case, it is less clear to Telestyrelsen why it should not be economical forTDC to move cables to adjacent areas, in particular for smaller cable sizes.

Top-down issue-III

The parties are invited to provide data for actual cable wastage (the amount of cable disposed as apercentage of total amounts of cable acquired each year).

B.3.3.2.3 Fibre

For unequipped sub-sections of the fibre access network TDC has used 6 fibre cable for dropfibre (“stik”) and 24 fibre for rings, i.e. minimum dimensions are used, c.f. section 2.4.10 of thedocumentation. TDC does not justify that this practice is in accordance with a forward-lookingmethodology.

B.3.3.2.4 Other Issues

TDC does no attribute a scrap value to cable at the end of its lifetime. It argues that the costs oftransporting the cable exceed any potential scrap value.

B.3.3.3. Distribution Points

The approach used to model distribution points is analogous to that for trench lengths in manyrespects. It is based on the number of core pair customers, the number of ‘distributor passings’per customer, utilisation rates for distributors and a distribution of distributor sizes. A finaladjustment is made to take account of the % of distributors, which are underground.

Distributors are (re)dimensioned on the basis of TDC’s actual utilisation ratios which have beenestimated for a sample of 10,000 distributors for each geo-type. If a given distributor has a utili-sation level which is lower than the average utilisation level and it is possible to carry the number

Telestyrelsen

59

of core pairs in use on a lower size distributor, then the size of the distributor is rounded down tothis lower size. Otherwise the MEA is chosen by rounding up the size.

TDC’s network has a very large number of distribution points designed to provide a high degreeof network flexibility. This gives rise to two main issues:

• Is the large number of distribution points justified?

• Are the costs of these distribution points correctly estimated?

The first question will be addressed during the hybrid modelling phase. For present purposes,Telestyrelsen notes the large difference in the number of points between the two models. Withregard to the second question, Telestyrelsen notes that the top-down documentation provideinformation on prices, utilisation and information which allows an estimate to be made aboutthe number of primary and secondary distribution points. However, Telestyrelsen has so farbeen unable to replicate or nearly replicate the figures for distribution points in the SAS model.While it may be the case that there are other elements to the calculation which are missingTelestyrelsen is currently dependent on the results of the model audit for distribution pointGRCs.

B.3.4 Exchanges

B.3.4.1. Introductory Remarks

TDC has developed three separate models to value exchanges, namely:

• A concentrator model, which includes not only the concentrator itself but also the accessfacing ports on the local exchanges.

• An exchange model which covers local exchanges, tandem exchanges and other exchanges

• A model which estimates routing factors for exchanges and also transmission.

TDC uses two types of exchanges at each level of the network hierarchy and these are separatelycosted in the model. It can be noted that there are some significant differences in the costs ofthese types of exchanges at the concentrator level and at the local exchange level. As notedpreviously, the impact of these differences are small at the concentrator level but significant at thelevel of local exchanges. Specifically, the costs of the more expensive make of exchange aresignificantly higher than the other type of exchange. Given that the more expensive make ofexchange is used much more extensively this is an issue Telestyrelsen will pursue during thehybrid modelling phase.

B.3.4.2. Concentrator Valuation

According to TDCs documentation there are:

• -c- Ericsson remote concentrators and -c- Siemens remote concentrators; and

• -c- Erisson host concentrators and -c- Siemens remote concentrators.

Since there are limits to the number of customers who can be housed on a concentrator unit(there are different limits for Ericsson and Siemens concentrators) a site may contain a number ofconcentrator units. As discussed above, a small number of units are housed at technical houses.

For both types of concentrator, TDC segments costs into the following categories:

• Line cards - shown separately for analogue, ISDN2 and ISDN30;

Telestyrelsen

60

• Concentrator ports costs; and

• Rack and other costs.

TDC has supplied cost information for concentrators and also local and tandem exchanges toTelestyrelsen and shown how it can be applied to calculate the cost of an average sizedconcentrator and exchange.

In practice the data is applied at the level of the individual unit. It can be noted that account istaken of modularities in provision, which are important in relation to racking costs and to somedegree in relation to line card costs.

TDC uses its planning norms to estimate BHE traffic requirements and in turn portrequirements (the conversion uses a blocking factor of 0.25%). These norms are:

• 0.05 erlangs per analogue line;

• 0.2 erlangs per ISDN2 line; and

• 7.0 erlangs per ISDN 30 line.

The port requirement for concentrators is also used as the access facing port requirement for localexchanges.

It should be noted that these dimensioning rules are applied on all line cards, including thosethat are not used. Further, an allowance has been made for anticipated growth in the networkbased on a 2 year look ahead. This allowance assumes no growth in PSTN requirements; a 40%growth in ISDN2 and a 45% growth in ISDN30 requirements.

Finally it should be noted that the valuation in the documentation is at end of year prices whilethe valuation in the model is at mid-year prices (this issue also arises in relation to other assets).In order to get from the valuation in the documentation to that in the model two adjustmentsare made:

• a 1.5% mark-up is added to cover spare parts and spare equipment; and

• the end of year price is rolled back using a price trend of -12% for the year (approximately –6% for half the year).

Telestyrelsen has noted that the resultant figure of -c- is still less than the figure of -c- in themodel.

B.3.4.3. Local and Tandem Exchanges

B.3.4.3.1 Valuation

In the case of local, tandem and other exchanges costs are segmented into the followingcategories:

• port costs (other than access facing port costs in local exchanges);

• switchblock costs;

• processor costs; and

• signalling costs.

The calculation of port costs is not well described in the top-down documentation. In essence,the methodology involves a detailed examination of traffic flows between exchanges and

Telestyrelsen

61

dimensions each exchange on the basis of the maximum traffic flow. In practice, the peak flowbetween a given exchange and those to which it is connected may well differ giving rise to thedanger of double counting if individual maximums are summed together. TDC has examinedthe overall flow (between the target exchange and other exchanges) throughout the day and hasdimensioned according to the maximum requirement.

Traffic flows are converted into port requirements using the assumption of a blocking level of0.25 per cent. TDC argues that because traffic may pass through a number of exchanges this isthe level required to result in an end to end blocking ratio of 1 per cent. Certain other aspects ofthis procedure should be noted. Firstly, no extra port capacity is allowed for resilience. Thus,although a local exchange will be parented onto two, or perhaps more, tandem exchanges portsare not dimensioned to allow all traffic to be carried to both exchanges. Secondly, no allowancehas been made for replacement of exchanges. According to TDC it is practice to upgrade ratherthan replace exchanges so an allowance for exchange changeover is not necessary. Thirdly, TDCdoes take into account the fact that traffic levels vary during the course of a week and by time ofyear. Dimensioning is based on the traffic is the 13th busiest hour.

This approach has in many respects much in common with the approach which might beadopted in a bottom-up model and may result in an implied utilisation level. However, twolikely differences with bottom-up models exist. Firstly, dimensioning is not based on the normalbusy hour procedure but instead takes account of differences in busy hour by part of thenetwork. Secondly, the dimensioning matrix is extremely large and contrasts with the muchsimpler approach which is often used in bottom-up models.

Switchblock costs are based on port requirements, excepting the requirement will depend on thetotal number of ports and not just on trunk facing ports. Switchblocks are provided in discreteamounts, (modularities differ between Ericsson and Siemens exchanges), a point which isreflected in the valuation.

According to information provided by TDC the processing cost of Ericsson is independent of thesize of the exchange while processing costs do vary with size (the measure used is ports) for theSiemens exchange.

Signalling costs vary according to the number of signalling links (typically two).

The final class of costs is that for software. TDC makes a distinction between hardware costs,which are dealt with by its exchange models, and software costs, which are not. Software costs arediscussed in Section B.7.6.4.

TDC’s model, in accordance with the reference papers, shows set-up costs and duration costsseparately. Set-up costs are calculated as the sum of processing and signalling costs - with theformer accounting for by far the majority of set-up costs. Duration related costs are estimated asthe sum of port and switchblock costs. Telestyrelsen accepts that this categorisation of costs isappropriate. The example of costs provided is for an average sized exchange. Telestyrelsen hasused this to estimate the proportion of set-up related costs. The resultant proportions of costs aresimilar to those in the model for Ericsson exchanges although in the case of Siemens exchangesthere is a difference, particularly for local exchanges. In this case the costing formula shows that -c- of exchange costs are traffic related whereas the model itself shows -c- of exchange costs beingtraffic related13. The most likely reason for this difference is that the percentage of traffic related

13 The proportion is higher for local than for tandem exchange owing to a combination of the higher number of

ports on tandem exchanges and processing equipment modularities.

Telestyrelsen

62

costs for an average sized exchange may differ from that for the average of all exchanges – TDCwill need to confirm that this is indeed the case.

Top-down issue-IV

Telestyrelsen invites comments on the applied split between set-up related costs and traffic related costs.

B.3.4.3.2 Routing Factors

The routing factor model calculates, the utilisation rate of 2 Mbit/s connections and actualrouting factors. The discussion of routing factors is covered by the documentation. Points whichcan be noted are that:

• The exchange routing factors illustrate the importance of sideways routings, i.e. between localexchanges. According to Telestyrelsen these are used where traffic exceeds -c- erlangs whilelocal exchange to tandem exchange links in other regions are established for traffic flows inexcess of -c- erlangs.

• The routing factors for transmission show the number of hops and have been adjusted totake account of utilisation levels.

B.3.5 Transmission

Transmission costs can be categorised into those which are related to trenching14, optical fibreand transmission equipment. The last of these categories in turn can be broken down into theline systems which are used at the higher levels of the network hierarchy and

TDC’s transmission network consists of 3 separate levels:

• Level 1 consists of 12 nodes in a meshed structure;

• Level 2 around 70 nodes in a meshed structure in 7 different areas; and

• Level 3 – 1,400 to 1,500 nodes in chain or ring structures.

There is some correspondence between this and the exchange hierarchy but the correspondence isnot always a straightforward one.

B.3.5.1. Trench

In some cases actual trench lengths are registered in databases while in other cases lengths aredetermined from build out plans or in the case of technical houses supplied by copper, averagedistances. Total trench length is estimated to be approximately -c- kilometres excluding sea cable.It can be noted that this figure is considerably higher than that in the bottom-up model, evenignoring the impact of trench sharing with the access network.

TDC’s estimate of sharing with the access network was outlined in the discussion of themodelling of the access network. A small amount of trench is also excluded as a result of sharingwith external parties. Telestyrelsen notes that the calculation of sharing between the access andcore networks could be imprecise and a more precise calculation may be required for thedevelopment of the hybrid model.

Trench prices are based on actual costs for 2000. One problem here is that the mix betweenploughing and digging (66% and 34%) in that year may not reflect the mix throughout TDC’s

14 It should be noted that at least for modelling purposes TDC has assumed that all trenches contain plastic tubing.

Telestyrelsen

63

actual network. Unfortunately, TDC has been unable to provide the mix of ploughing andtrenching for earlier years.

B.3.5.2. Fibre

TDC’s calculation of fibre costs follows a similar methodology to that for trench costs.Information supplied in its documentation shows fibre length by dimension of fibre andindicates that TDC currently uses mainly 12 fibre cables. Elsewhere TDC has stated that itscurrent practice is to use:

• -c- fibre cables between Level 1 nodes;

• -c- fibre cables between Level 2 nodes;

• -c- fibre cables between Level 3 nodes.

B.3.5.3. Transmission Equipment

Level 1 and 2 nodes typically use the following equipment:

• STM1/16 line termination systems;

• DXC4/1 cross connects;

• Some ADM equipment; and

• Other equipment such as Optical Distribution Frames.

As noted, full transmission resilience is not provided. However, both Levels 1 and 2 have aminimum of 3 ways out.

Telestyrelsen has not been provided with information on equipment quantities to enable it tofully verify TDC’s calculations. A related problem is that one of TDC’s cost categories, SDH4/1consist of SDH and DWDM equipment. DWDM equipment has distinguished characteristicsand a different role to SDH equipment, which make the aggregation of the two problematic.

Cards for STM1/16 line termination systems terminate traffic at the 155 Mbit/s level. For thisreason cross connects are required to take capacity down to the 2 Mbit/s level required forswitching purposes. Switched capacity at a given node will pass through the cross-connect twice,the first time to be brought down to the 2 Mbit/s level; the second time to be brought back up tothe 155 Mbit/s level. Leased line traffic will pass through the cross-connect once and in somecases may bypass cross-connects. The differing intensity with which different types of traffic usecross-connects and other elements of the transmision network is reflected in TDC’s allocationsbetween switched and leased line capacity. In principle, ADM equipment could be used insteadof cross-connects. However, according to TDC this would increase the cost of the transmissionnetwork where traffic flows are high.

Level 1 and 2 nodes tend to have similar equipment. An exception is that Level 1 sites may useWDM, indicating that there are fibre shortages in some parts of TDC’s transmission network,whereas Level 2 nodes do not.

Traffic flows at Level 3 nodes are lower than at Level 1 and 2 nodes. For this reason ADMs tendto perform the function of cross-connects at Level 1 and 2 nodes.

TDC’s modelled network contains more SDH equipment than in its actual network. This raisesthe question of whether an adjustment to the asset value should be made for differences inoperating costs. No such adjustment is made. According to TDC, while SDH is normally

Telestyrelsen

64

believed to be more cost effective due to the more developed network control, this is related toimproved network quality and flexibility rather than actual cost reductions. The cost of replacinga card on site is the same whether or not it is PDH or SDH equipment. TDC does not have anyfigures showing whether exchanging PDH with SDH implies lower costs in the primaryoperation.

B.3.5.4. Other Issues

TDC has a separate platform for low capacity leased lines (less than 2 Mbit/s). However, theselines share the same trenches and optical fibre.

TDC has provided information on utilisation levels of both fibre and transmission equipment inits documentation. Fibre utilisation is overall 42% although in some cases there are fibreshortages necessitating the use of WDM equipment. Transmission system usage varies from 25%in Level 3, which TDC attributes to modularities of equipment sizes, to 48% at Level 1 nodes.These utilisation levels will need to be further examined during the hybrid modelling phase.

The documentation also provides information on the allocations of costs between switched trafficand leased lines and between different types of switched traffic. The allocation between switchedtraffic and leased lines is shown below:Table 14: Allocation of Transmission Costs between Switched Traffic and Leased Lines

Asset Class Proportion of Costs Allocated to Switched TrafficDuct -c-Fibre -c-Transmission Systems on Major Nodes -c-Transmission Systems on Other Nodes -c-Cross Connects -c-

It can be seen that leased lines are allocated a very large share of transmission costs. Given thevery high leased line capacity in TDC’s network this is as expected. It can also be noted thatthese proportions vary considerably by part of the network with leased lines assuming aparticularly important source of cost on transmission systems covering major nodes. Theseallocations will be examined further during the reconciliation and hybrid modelling phase.

Telestyrelsen also notes that fixed costs for transmission equipment are treated as common costsin the model. This issue is discussed in the Chapter on ABC.

Telestyrelsen

65

B.4 AnnualisationThe Top-Down Model Reference Paper (TD MRP) states that TDC should make indicativeestimates of economic depreciation and compare these annualisation figures produced by tiltedannuities, tilted straight line depreciation and sum of digits depreciation15. The choice betweentilted annuities, tilted straight-line depreciation and sum of digits depreciation should be madeaccording to which produces the best approximation of economic depreciation. In the event thatnone of these methodologies approximate economic depreciation, the estimates of economicdepreciation should be refined and used as the basis for the annualisation charge.

The TD MRP further states that no value should be attributed to Fully Depreciated Assets andthat where accounting methodologies are used, the rolling forward methodology rather thanNBV/GBV ratio should be used to determine Net Replacement Costs.

TDC has developed an economic depreciation model which enable comparison of the differentdepreciation methods with indicative estimates of economic depreciation. TDC has, however,chosen not to follow the criterion relating to Fully Depreciated Assets. In addition, for those assetcategories where an accounting methodology is used TDC has used what is described as the“rolling back” procedure. This chapter discusses the large number of issues associated withannualisation in the context of TDC’s economic depreciation model.

B.4.1 Economic Depreciation Model

TDC’s economic depreciation examines the annualisation charge under four alternativemethodologies: economic depreciation, tilted straight line depreciation, sum of digitsdepreciation and tilted annuities. The model is used in relation to 19 separate classes of assetsincluding: trench, switch hardware, switch software, transmission equipment, land and buildings,IT and motor vehicles.

TDC’s comparisons of the alternative methodologies require assumptions on a number ofdifferent parameters. The most important of these are listed discussed below:

• Rolling Backward Methodology;

• Cost of Capital;

• Rate of price change;

• Economic depreciation asset lives which are consistent with book lives.

• Asset lives;

• Inclusion of fully depreciated assets;

• The use of a representative asset; and

• Zero scrap value.

Telestyrelsen has reservations about a number of the assumptions TDC uses in calibrating itsmodel.

15 In the case of economic depreciation, tilted and sum of digits depreciation the calculation should be inclusive of a

capital cost.

Telestyrelsen

66

B.4.1.1. Rolling Backward Methodology

In the TD MRP, Telestyrelsen advocated the use of the rolling forward methodology wherebythe NRC is estimated by rolling forward both the GRC and accumulated depreciationstatements to take account of asset price inflation or deflation - and where relevant – additionsand writeouts. TDC has instead applied what TDC calls the rolling backward methodology.This involves the following steps:

• Estimate the GRC at the year end and then roll this back to reflect asset priceinflation/deflation as well as additions and write-outs during the year;

• Estimate the current cost accumulated depreciation at the year end as the historic costaccumulated depreciation multiplied by the ratio of GRC to GBV;

• Roll back current cost accumulated depreciation to reflect asset price inflation/deflation aswell as depreciation and write-outs during the year;

• Use the GRC and accumulated depreciation statements to calculate the NRC statement.

A number of points should be made about this approach. Firstly, the end of year NRC is similarto that which would be calculated by applying the NBV/GBV ratio to GRCs and, as argued inthe TD MRP, is biased. Secondly, the annualisation charge within the year will be broadlysimilar to that which would be calculated by the rolling forward methodology. Thirdly, an issuearises as to whether the end of year result should then be rolled forward to the next year orwhether the process above should be used instead. If the latter approach is selected there will be adifference between the closing NRC in the first year and the opening NRC in the following year.Finally, as discussed above, Telestyrelsen has concerns about some of the price indices beingused, which could result in annualisation charges being biased.

Telestyrelsen has discussed the use of the rolling back methodology with TDC and hasquestioned its use in preference to the rolling forward methodology. In response TDC hashighlighted potential limitations in its FAR which, in its view, would result in potentiallysignificant inaccuracies in the rolling forward methodology in practice. Telestyrelsen recognisesthis to be the case.

B.4.1.2. Cost of Capital

TDC uses a cost of capital of 14.5%. The assumptions underlying this estimate are reviewed inthe Reconcilliation report.

B.4.1.3. Rate of Price Change

As pointed out in B.2.6.1 many of the price trends used in the top-down model are low/negative.B.2.6.1 gives examples of individual price trend assumptions and highlights Telestyrelsen’sconcerns regarding these assumptions. It is further noted that using price trends which are toolow or too negative will have two effects in the model:

• Result in overestimates of the annualisation rate.

• Result in overestimates of the mid-year GRC, since end of year GRCs (as calculated byTDC’s sub-models) will be adjusted back to mid-year values by the application of incorrectprice trends.

Telestyrelsen

67

B.4.1.4. Economic Depreciation Lives

Asset lives generated by the application of economic depreciation principles may not beconsistent with book values. TDC refers to a study conducted for the Australian Regulator whichconsiders two methods for dealing with this issue16. The first of these involves assumingreasonable proxies for operating cost levels and trends and letting the economic depreciationcalculation select the appropriate asset lives. The second involves selecting operating cost levelsand trends which are consistent with book asset lives. The NERA report estimates economicdepreciation under both procedures but ultimately selects calculations which are based on bookasset lives. TDC does the same and uses the NERA report in support of doing so. Telestyrelsenhas two comments on this issue:

• There is a third alternative which involves using reasonable estimates of operating cost levelsand trends in combination with the operator’s actual book lives. The basis for this approachis that the physical and actual lives of assets are likely to differ as a result of technologicaluncertainty and other factors. For example, the physical live of duct may well be 60 years ormore but operators typically assign much shorter lifetimes because of technological andperhaps market uncertainty17.

• Book asset lives are (or at least were) generally rather longer for Telstra than for TDC andNERA’s assumptions about operating cost levels and trends, even for the constrained case(where economic depreciation lives and actual book lives are equal) appeared more plausible.As a result Telestyrelsen can better understand NERA’s approach in the case of Telstra thanunderstand TDC’s approach in the present instance.

In TDC’s economic depreciation model, TDC assumes that all operating costs start as -c- of grosscapital cost and rise by a sufficient percentage throughout the asset lives to ensure equality ofeconomic depreciation and book lives. Typically this involves very rapid annual increases inoperating costs often in excess of 30 per cent per annum. Telestyrelsen finds neither theassumptions on starting operating cost levels nor rates of increase very plausible. Replacing thesewith what Telestyrelsen regards as more plausible assumptions does not necessarily affect thechoice of annualisation methodology but does suggest that there may be problems with TDC’sasset lives, even taking account of technological developments and other factors. This issue isfurther discussed below.

B.4.1.5. Asset Lives

Many of TDC’s assets are assigned relatively short asset lifetimes, as shown in the table below:Table 15: Asset Lifetimes

Class of Assets Book LifeFibre Duct (access and transmission) 10 yearsCopper Duct 14.5 yearsCopper Cable 14.25 yearsBuildings 20 years

16 The study was conducted by NERA for the ACCC and can be found on that bodies website (accc.gov.au) at

nera.zip.17 Compared to TDC’s approach this method will lead to a gradual and much more realistic increase in operating

costs over time.

Telestyrelsen

68

Telestyrelsen believes that some of these lifetimes could be inconsistent with plausible outputprice trends, operating cost levels and trends, even taking account of technological and otheruncertainties.

Telestyrelsen also notes that there are very high levels of fully depreciated assets in some parts ofTDC’s network. Some examples are provided in the following table:Table 16: Percentage of Fully Depreciated Assets

Asset RatioFibre Duct (access) -c-Fibre Duct (transmission) -c-Copper Duct -c-Copper Cable -c-18

Buildings -c-

These figures further reinforce the message that, at least in some cases, TDC’s book asset livesmay be too short.

B.4.1.6. Inclusion of Fully Depreciated Assets

Contrary to Criterion 20 of the TD MRP, TDC has included the value of fully depreciatedassets in its model. As a starting point it is useful to consider the implications of including theseassets. In instances where the tilted annuity methodology is used, the inclusion of fullydepreciated assets has fairly straightforward implications. If the GRC of a given asset excludingfully depreciated assets is 500 and is 1,000 if fully depreciated assets are included then the annualcapital charge will be exactly doubled with the inclusion of fully depreciated assets.

The implications of including fully depreciated assets are somewhat less clear cut for otherannualisation methodologies. For example, when:

• Both starting and closing NRCs are calculated using the NBV/GBV ratio; and

• GBVs are calculated exclusive of fully depreciated assets.

the inclusion of fully depreciated assets could, in some circumstances, actually reduce theannualisation charge19.

However, in TDC’s model while the latter assumption is used the former is not20. On the basisof this approach the inclusion of fully depreciated assets will increase the annualisation chargeand the increase is proportionate to the ratio of GRC with and without fully depreciated assets.

Telestyrelsen has asked TDC to provide a run which excludes fully depreciated assets. This runresults in a significant fall in the level of estimated costs.

The impact of excluding fully depreciated assets are presented in Table 17: 18 TDC is unable to distinguish between the relative values of different copper related assets, hence the figures for

access copper duct and access copper cable are the same.19 Including fully depreciated assets will increase GRC and therefore the Operating Capital Maintenance

depreciation charge. However, it also has impacts on the level of holding gains resulting in an FCM depreciationmeasure which may be lower than if FDAs were excluded. The same is true of the overall annualisation charge

20 More precisely, the closing NRC is calculated using a variant of the NBV/GBV ratio but the former is calculatedusing the rolling back methodology. TDC includes the value of fully depreciated assets in GRC but exclude thevalues in GBV. Under this atypical use of the NBV/GBV methodology, the NBV/GBV ratio remains constantand the effect on the annualisation charge is therefor no longer ambiguous.

Telestyrelsen

69

Table 17: Sensitivity results for Raw Copper

Interconnection product1 November 2001

FDA attributed value20 December 2001

FDA not attributed valueChange

Raw copper, average 2,106 -c- -c-Dark fibre average 9,261 -c- -c-Local access 4.95 -c- -c-Access within areaAccess between area

7.669.26

-c--c-

-c--c-

Transit within areaTransit between area

3.219.10

-c--c-

-c--c-

B.4.1.7. The Use of a Representative Asset

For each category of asset examined the model uses a single representative asset which is assumedto have the average live of a typical asset in TDC’s network. As noted in A2.6 this gives rise topotential problems since the average life of assets is calculated according to the NBV/GBVapproach (where the GBV excludes fully depreciated assets). This is likely to give rise tomisleading estimates of asset lives as a result of asset price inflation and deflation and because ofthe importance of fully depreciated assets. Further, even in the absence of these problems someinaccuracies will arise of a result of the approximation impacts of a representative asset.

Given the above, it is worth considering how important these errors may be. Essentially, anyerror in the calculation of the average asset live will affect the estimation of the sum of digitscalculation and also the economic depreciation calculation.

B.4.1.8. Zero Scrap Value

The model assumes that there is zero scrap value. In discussions with TDC it has been pointedout that in the case of certain assets there is a real cost of disposing assets which effectively impliesa negative scrap value. Telestyrelsen recognises this point although it notes that this is lessapplicable in the case of assets such as buildings.

B.4.2 Conclusions

Telestyrelsen has a number of concerns regarding TDC’s annualisation calculations. Inparticular, it believes that some of TDC’s asset lives appear to be overly conservative while someof TDC’s price trends are questionable.

In Telestyrelsen’s view the economic depreciation model is not sufficiently robust to provideconvincing support for TDC’s selection of annualisaton methodology.

Telestyrelsen notes that TDC has chosen to include fully depreciated assets and is aware of theimpact this has on estimated annualised costs. It is also aware of the potential connectionbetween the level of fully depreciated assets and TDC’s asset lives. Telestyrelsen is currentlyconsidering TDC arguments. Because asset valuations are part of the hybrid modelling phase,however, Telestyrelsen will first give a decision on this issue in that connection.

Telestyrelsen

70

B.5 Working CapitalWorking capital is documented in Chapter 7 of the TD-documentation. As a general comment itcan be noted that working capital is small relative to fixed assets. Currently working capital isnegative although when network spares and assets in the course of construction are also includedthe resultant total is slightly positive.

B.5.1 Method

Working capital includes network spares, assets in the course of construction and current assetsand liabilities. Working capital is distributed to HCC’s according to the table below.Table 18: HCC’s related to working capital

Group HCC name HCC no. Amount (DKK)Network spares Access

SwitchingTransmission

Total

126127128

-c--c--c--c-

Assets in course ofconstruction

AccessSwitching

TransmissionTotal

157158159

-c--c--c--c-

Working capital Current assetsCurrent liabilities

Total

133134

-c--c--c-

The business segmented accounts (BSA) 2000 form the input to working capital HCC’s and allfigures are calculated as average of figures at end of year 1999 and end of year 2000.

B.5.2 Potential Problems

There appears to be some contradiction between the fact that inputs to working capital HCC’sare reported to be taken from BSA 2000 and the fact that figures are reported to be averages ofBSA 1999 and 2000. While figures generally seems to be calculated as averages of BSA 1999 and2000, it is not evident on what basis (BSA 2000 or average 1999-2000) the allocation keys havebeen derived. The same approach for both figures and allocations keys should be chosen for thepurpose of consistence.

B.5.2.1. Network Spares

Network spares are included at acquisition cost regardless of the year of acquisition. This maynot be appropriate for equipment that is subject to substantial price changes.

It will need to be confirmed that only the efficient stock of network spares is included. Networkspares that cannot be used because they are technologically outdated, broken, or wasted shouldbe regarded as inefficient and not be included.

B.5.2.2. Assets in the Course of Construction

Assets in the course of construction are included at book values. Although assets are normally notdepreciated before they are put into service, it is not clear from the documentation whether this

Telestyrelsen

71

is the case in the top-down model or not. Likewise actual cost are used as CCA where assets donot take a long time to put in place. However, TDC does provide any figures indicating howlong time the assets are in course of construction.

It is not possible to verify the resulting cost allocation in Table 7 from the allocation keyspresented in Table 6, probably because “Faste kredsløb” (leased lines) is included in Table 7.However, it is not described how “Faste kredsløb” has been allocated to access, transmission andswitches.

B.5.2.3. Working Capital

Working capital assets and liabilities are taken from the business segmented accounts level 2 and3. The nature of the different types of working capital is not described.

Current assets includes fixed assets (excluding tangible fixed assets) and stock. It will need to beverified that no double counting takes place.

B.5.3 Compliance with Criteria

The SAS model and the documentation include calculations of working capital and are in thisrespect compliant with Criterion CG 24.

All working capital figures are calculated as an average of end of year 1999 and end of year 2000in compliance with Criterion TD 22.

Telestyrelsen

72

B.6 Operating CostsOn the basis of an internally conducted efficiency study, TDC concludes that it is an efficientorganisation . In consequence, no efficiency adjustments are required to operating costs actuallyincurred. Telestyrelsen has a number of concerns with this efficiency study, which is discussed inmore detail in B.9

Operating costs have then been derived on the basis of the TDC’s regulatory businesssegmentation accounts (“forretningsområderegnskabet”) for 2000. In general this seems to be asound approach since these accounts according to TDC are based on the principles of activitybased costing (ABC). Please refer to chapter B.7 for a more detailed discussion on the costallocation principles adopted by TDC.

TDC has provided substantial documentation on how total operating costs have been allocatedin the top-down model. However, the documentation does not provide enough details abouthow total operating costs were found in the first place.

Telestyrelsen has therefore decided to undertake a more detailed examination of the operatingcosts included in the model to ensure that only relevant costs are included and that there are nodouble counting of costs. Given that certain costs such as the cost of co-location services, changein the number plan etc. have been estimated outside the actual top-down model there is a risk ofpotential double counting.

Telestyrelsen

73

B.7 Allocation of costsThis chapter examines the use of ABC approaches in the modelling. The scope of the chapter isbroad in the sense that it considers not only the allocations to components but also the treatmentof shared and common costs – as such it comments either explicitly or implicitly on CVRs in anumber of places.

Strictly speaking shared and common costs by their very nature cannot be allocated orapportioned to components. However, the use of a mark-up regime within the model means thatall costs are ultimately ‘allocated’ or ‘apportioned’ to one or another component. It is thereforeimportant to ensure that the allocations or apportionments of common and shared costs arejustifiable while accepting that alternative allocations or apportionments are possible.

A discussion of ABC is important for two related reasons. Firstly, many of TDC’s cost categoriesprovide services which are not related to TDC’s network division. For example, TDC’s humanresources division ‘services’ both TDC’s network division and other parts of TDC. Therefore it isnecessary to ensure that network components do not include (or do not omit) costs which arerelated to network services. Secondly, it is necessary to ensure that the pot of costs which isallocated or apportioned to the network is allocated across different components in anappropriate manner.

TDC’s documentation provides some information to assist Telestyrelsen in carrying out boththese tasks. However, particularly in respect of the former task, the documentation is notcomplete. For example, Section 6 shows the extent to which certain network operating costs areeliminated from the calculation of core and access costs since, for example, they are related tooutpayments to other operators. What it does not do is show how the network pot of costs isobtained in the first place. Nor does it provide sufficient information on the way in which costsare allocated or apportioned to network components.

To address this problem Telestyrelsen has discussed a range of issues with TDC and sent TDCquestions on these issues. In addition, TDC has given Telestyrelsen presentations covering:

• The determination of network related costs;

• The methodology used to allocate and apportion network operating costs between differentcomponents.

While both these presentations gave a clearer idea of TDC’s approach they did not providesufficient information on the way in which allocation and apportionment bases were derived.Telestyrelsen has therefore sent a further request to TDC asking for details on these bases.

On the basis of the information provided in the documentation, the model and in TDC’ssubsequent responses, Telestyrelsen believes that there could be number of problems with TDC’smethodologies. These include:

• Differences between the allocations and apportionments for the capital and operating costsassociated with certain assets. In general the allocations and apportionments appear better forcapital costs than for operating costs.

• Inconsistencies between the level of common and shared costs within capital costs and relatedoperating costs. In general, Telestyrelsen would expect that where a capital cost contains aproportion of common or shared costs, the associated operating cost (e.g. the associatedmaintenance expenditure) would obtain at least that proportion of common or shared costs.After all if a 50% reduction in transmission requirements reduces transmission equipment

Telestyrelsen

74

requirements by only 25% it seems unlikely that transmission operating costs would reduceby more than 25%.

• Inconsistencies between the level of common and shared costs for related classes of operatingcosts. For example, a 10% increase in output volumes is supposed to increase corporate anddivisional overhead costs by 10% but to have no impact on non-exchange related buildingcosts. This appears highly implausible.

B.7.1 Treatment of other increments - shared and common costs

The MRPs require the TD-model to distinguish between at least three increments: Access, Coreand Other increments. The latter includes among other things the retail increment, theinternational increment and the mobile increment, cf. sections 3.3. of the general MRP.

Criterion CG 5 specifies that “The models should identify the costs that are common betweenthe other increments and the core and access networks”. Criterion TD 34 specifies that “Themodel should show the cost of other increments grouped together”.

In principle, TDC does not meet these Criteria. When defining and allocating shares andcommon costs, TDC only distinguishes between two increments: “Access” and “Non-Access”.

Telestyrelsen defines a shared cost as a cost that is shared between services within an incrementand a common cost to be a cost that is shared between services in different increments. TDC, onthe other hand, defines a shared cost as a cost that relates to either access or non-access only and acommon cost as a cost that relates to be both access and non-access.

As a minimum TDC should have split the increment “non-access” into two “Core” and “Otherincrements”. Cost that were then shared within Core (Other increments) should then only havebeen distributed via mark-up to network elements in Core (Other increments)

Moreover, TDC in some cases label some costs as common costs although they in the view ofTelestyrelsen actually are shared costs. For example it is not clear to Telestyrelsen why theoperating costs related to SDH equipment have been treated as a common costs when mostSDH equipment is used in the core network only21.

The calculation and allocation of shared and common cost in the SAS model is discussed inAnnex 1 of the TD-documentation, section B.1.3.

B.7.2 Differences between Figures in Documentation and Model

The GRCs shown in the documentation generally differ from those in the model, in some casesby significant margins. One reason for these differences is that figures in the documentation tendto be based on end-of year prices whereas those in the model tend to be at mid-year prices. Afurther reason is that figures in the model may contain an allowance for spares. In some casesadjusting for these two factors brings the two sets of figures into alignment but this is not alwaysthe case. For example, there is a small unexplained gap for concentrator costs. In the case ofdistribution cable the gap is far larger. TDC will therefore need to provide sufficient informationto reconcile the figures in the documentation and the model.

At a very late stage it has become clear that the inconsistency may be due to a more fundamentalproblem with TDC’s methodology for allocating fixed and variable costs to network elements onthe basis of CVR. This is discussed in the next section.

21 Part of the reason is that SDH equipment is used for ISDN 30 in the access network.

Telestyrelsen

75

B.7.2.1. Application of CVRs to allocate fixed and variable costs

Telestyrelsen has become aware that TDC uses a rather different approach for allocating fixedand variable costs on the basis of CVR curves than Telestyrelsen had originally understood onthe basis of the documentation22. So far Telestyrelsen has been under the impression that TDC’sapproach corresponded to Telestyrelsen’s recommendations.

Telestyrelsen now understands that TDC allocates variable and fixed costs in the following way:Figure 1: Allocation of fixed and variable cost in the top-down model

CVR-curve B CVR-curve A

CVR-curve C

Network element Yx%

100%

Volume

Costs

100%

YB

YC

YA

Fixe

d c

osts

Var

iable

cost

sfo

r net

wor

k el

emen

t Y

For a given HCC, TDC first allocates the variable costs to network elements: TDC starts from100% of the volume and reduce the volume by the amount used by network element Y (x %).On the vertical axis, TDC then reads of the associated (“incremental”) costs to be allocated tonetwork element Y. For a CVR-curve like CVR-curve B this will give an “incremental” cost ofYB. TDC then continues with the next network element that uses the given HCC at hand,starting over from 100% of volume. A consequence of the approach is that the total amount ofvariable costs allocated to network elements may be higher as well as lower than the total amountof variable costs. If the CVR curve is concave (B), the sum of the allocated costs will be less thanthe total whereas if the CVR curve is convex (C), the sum of the allocated costs will be more thanthe total.

In principle there is nothing wrong with this methodology, which has also been discussedbetween TDC and Telestyrelsen.

The problem is that TDC has then allocated fixed cost without taking this over/under allocationof variable costs into account. TDC has simply allocated the total amount of fixed costs inproportion to the amount of variable costs that have been allocated to network elements

As discussed in the General MRP section 2.2 and 2.3, a price set on the basis of LRAIC will notallow an efficient operator to recover his forward-looking common costs. Criterion CG 2

22 Telestyrelsen’s recent understanding is based on a letter dated 27 February 2002, where TDC addresses a question

posed by Telestyrelsen regarding inconsistencies in the GRC calculation for “Access copper distribution cable”.

Telestyrelsen

76

therefore specifically states that the model(s) should allow coverage of common costs. Thisshould be done via a mark-up, cf. Criterion TD 37.

TDC has chosen to adopt a similar (mark-up) approach for shared costs, which Telestyrelsenunderstands and accepts. However, shared and common costs should be those parts of total costswhich are not directly allocated to network elements, ie the difference between total costs andallocated (variable) costs – not the total fixed costs.

As an example of the implications of TDC’s methodology, it can be mentioned that the totalamount of costs for distribution cable allocated to network elements (unterminated twistedcopper pair by geo-type) are more than DKK -c- million higher than the total amount of costsassociated with copper distribution cable.23 For other cost categories such as e.g. SDH equipmentthe costs included in the model are substantially lower than the total amount of costs associatedwith SDH equipment. As a whole the adopted methodology imply that the total amount of costsallocated to network elements is around DKK -c- million lower than the total amount of costs24.

In Telestyrelsens view TDC’s allocation methodology is not in accordance with the overallprinciple of setting prices at LRAIC plus a mark-up for common costs as described in sections2.2 and 2.3 of the general MRP. By coincidence the overcompensation (for the access network)approximately matches the under compensation (for the core network). However, this is simply acoincidence. Moreover prices set on this basis would distort the investment signals for the coreand access networks.

B.7.3 Access Costs

TDC’s treatment of access costs creates confusion in a number of respects. Firstly, the modelcontains separate categories for copper and fibre duct without providing an explanation of howthe separate costs are derived. Duct/cable ratios are used to estimate duct lengths for both copperand fibre, but the documentation does not discuss the extent to which copper cable and fibrecable share duct. Secondly, access related leased lines costs are treated in an unsatisfactorymanner. In the GRC calculations where leased lines share assets with subscriber lines, costs tendto be bundled together e.g. for access network duct – copper, although there are exceptionsHDSL equipment. For operating costs a different approach is adopted with, for example, accesscopper maintenance costs being shown separately for subscriber lines and leased lines. What isunclear is what happens to the access related portion of leased lines for overhead categories suchas divisional and corporate overheads. In Telestyrelsen’s view, it would have been better to createa separate network component for access leased lines.

B.7.4 Exchange Costs

The major allocation and apportionment issues associated with exchanges are i) thedetermination of the relative levels of access and core related costs for the concentrator; ii) the

23 TDC has estimated the CVR curve for distribution cable to be convex. When estimating this CVR, TDC has

kept 10-pair cable as the minimum cable size when scaling down the number of customers. Cable sizes go up to2800-core pairs in certain areas. According to TDC, costs therefore fall relatively much in the beginning whenscaling down volume. Telestyrelsen is somewhat surprised by this finding and notes that if 10 pair copper cable iskept as a minimum it should be treated as a fixed costs. Telestyrelsen therefore questions the identified shapewhich suggests that there should be decreasing returns to scale for distribution cable at the margin. At first sightthis appears counter-intuitive. For example, the cable prices documented by TDC in table 26 of chapter 2indicates slightly decreasing cost per cable by cable size, suggesting a concave CVR curve. It should be noted thatthe concave/convex issue only relates to variable costs.

24 The figures of DKK -c- million and DKK -c- million appears from TDC’s letter of 27 February 2002.

Telestyrelsen

77

determination of the relative levels of duration and set-up costs; iii) the determination ofcommon and shared costs.

TDC’s exchange capital costs are based on actual exchange prices paid by TDC. These pricesshow separately the costs of access related elements (line cards); core related elements (ports,signalling, processing). In principle, this provides a sound methodology for the costing work.However, there appear to be some problems with the way the information is used:

• TDC’s model shows that approximately -c- of costs are considered to be common. TDC’scosting formula differentiates between line card costs, which are clearly access related; portcosts, which are clearly traffic related and cabinet costs, which are used by both lines andcalls. Common costs should be that portion of cabinet costs which is still incurred with nearzero output. According to the documentation (Table 12, Section 3) this amounts toapproximately -c- of concentrator costs, a slightly higher figure than shown in the model.

• Telestyrelsen finds the treatment of exchange operating costs unconvincing. TDC hasseparate maintenance and repair categories which are in turn split into pay and non-payelements. All these costs are considered to be core related. A subsequent cost category, whichappears much further down the model output sheet than (HCC results) the above categories,is entitled Switching linecards. Telestyrelsen presumes that this category includes on the onehand both maintenance and repair elements relating to line cards and on the other hand bothpay and non-pay elements relating to line cards.

• It can be noted that the proportion of core related operating costs for concentrators is muchhigher at around -c- than it is for capital costs (approximately -c-). Further, one would expectsome costs to be categorised as being common but this is not the case. Further investigationof this is needed.

B.7.5 Transmission Costs

TDC has developed detailed allocation bases for transmission capital costs based on the way inwhich that network is used by different services. Separate bases have been developed for:

• Trench and fibre;

• SDH equipment at major nodes and radio;

• SDH equipment at other nodes; and

• Cross-connects.

These bases suggest that different services use different parts of the transmission network indifferent ways. For example, switched traffic accounts for a considerably smaller percentage of theusage of SDH equipment at major nodes (-c-) than at other nodes (-c-). While Telestyrelsenintends to investigate these bases further it believes that these seem to provide an appropriate wayto determine the costs of individual network components.

Based on the documentation provided so far, Telestyrelsen however has some concerns withregard to the treatment of fixed costs. For transmission equipment on major routes theproportion of fixed costs is -c-. In addition to which there are non-linearities in the relationshipbetween these costs and output resulting in -c- of total costs being allocated as fixed costs. Thesefixed costs are treated as common costs in the model on the grounds that transmissionequipment is used to support ISDN30. Specifically, an ISDN 30 node is connected via a localexchange and if the customer does not live in an area covered by the local exchange the

Telestyrelsen

78

connection has to be forwarded to that exchange. Therefore a small proportion of transmissioncosts are access related and therefore fixed costs for transmission are treated as common costs.

Telestyrelsen could understand this procedure if common costs were allocated at the level ofindividual cost categories but believes that it will result in too many costs being allocated to theaccess network if common costs are allocated over the totality of core and access costs.

Telestyrelsen also notes that the allocation of operating costs for transmission does not seem to beconsistent with that for any class of capital cost. For example, -c- of transmission operating costsare allocated to the core network which is greater than for any category of capital cost while -c- ofoperating costs are considered to be common which is less than any category of capital cost.Telestyrelsen believes that there should be consistency between the treatment of operating andcapital costs25. This implies that:

• Shared or common costs related to operating costs cannot be lower than for that class ofcapital costs which has the lowest share of shared or common costs. The exact proportion ofshared or common cost should depend on the intensity with which different classes of capitalcost generate the need for operating costs and on whether there are any economies of scale orscope within operating cost functions themselves.

• The proportion of operating costs allocated to particular components can in general not belower than the lowest figure for different classes of capital cost. A possible exception arises ifthere are significant economies of scale or scope in the operating cost related activities.

B.7.6 Other Costs

B.7.6.1. Land and Buildings

TDC has used square metres as the driver for allocating building costs to the telecoms networkand within that network to access, transmission and exchanges. While this is not an ideal driver,since square metres may not be perfectly correlated with value, Telestyrelsen believes that it is anacceptable base. However, TDC has not provided sufficient detail on how the network relatedproportion of support functions within its business is determined. Nor has TDC provided a clearpicture of how the respective space usage of access, transmission and exchanges has beendetermined. For those parts of buildings which are occupied with network equipment such adetermination is straight forward. However, this is not the case within office buildings.

Once each of the three categories of component specific cost has been determined costs are thendistributed using GRC as the cost, taking account of common and shared costs.

TDC has informed Telestyrelsen that the cost volume relationships for all classes of land andbuildings are horizontal. It has further been assumed that a proportion of costs equal to the totalsquare metres occupied by exchanges buildings is allocated to exchange and other equipmentbased on a key for allocating costs from the business segmentation accounts. Telestyrelsen hasbeen provided with details of that allocation key.

A potential problem with this approach is that while it may be true that the CVR for exchangerelated and perhaps other equipment related buildings is horizontal this is not the case for officerelated land and buildings where costs will change with volumes. Telestyrelsen finds it surprising

25 This does not necessarily imply that the percentages allocated for operating costs should be the same as that for

capital costs.

Telestyrelsen

79

that TDC argue that divisional and corporate overheads contain no fixed costs whereas officeland and building costs are entirely fixed.

B.7.6.2. Motor Vehicles/IT

Motor vehicle capital costs are classified as ‘Motor Vehicles – Network Operations Related’.These costs are assumed to be entirely fixed, an assumption which is hard to justify. One wouldexpect that motor vehicle costs are primarily related to maintenance, repair, provision andinstallation functions and should be classified as component specific, shared or common in thesame way as some combination of these functions. Since the underlying functions are not 100%common it is hard to understand why the GRC for Motor Vehicles should be considered ascommon. It is also difficult to understand why Motor Vehicles Operating Costs should beconsidered to be entirely variable.

B.7.6.3. Switch Planning

Switch planning costs are assumed to be entirely shared. Given that these costs must be assumedto also include the costs for concentrator planning, the majority of which are access related,Telestyrelsen cannot understand this assumption.

B.7.6.4. IT Costs

IT capital costs are treated under a number of different cost categories. Hardware, cabling andPC costs are, according to TDC, considered to be switch related. Telestyrelsen finds thisassumption hard to understand. Software costs are considered to be either common (support andSDH) or shared (network software). Telestyrelsen also do not find these assumptions fullyconvincing. If hardware, cabling and PC costs can be allocated on the basis of switch costs, onewould expect software costs could also be allocated on this basis. Further, it is hard to believe thatthese costs are entirely fixed, whereas TDC elsewhere has assumed that switch software costs areentirely variable.

IT operating costs are assumed to be completely shared. In Telestyrelsens view this assumption isunwarranted for three reasons. Firstly, IT services are used by both access and core services and ifit is indeed the case that these are entirely fixed, such services should be considered to becommon rather than shared. Secondly, Telestyrelsen believes that IT costs are not invariant tovolumes. An increase in output will, for example, increase the number of staff within anorganisation thereby increasing both capital and operating IT costs. Finally, Telestyrelsen notesthat IT hardware capital costs are assumed to be component specific on the grounds that they aredriven by switch costs. While Telestyrelsen doubts whether switch costs are an appropriate driverit also believes that it is not correct to allocate hardware capital costs between components and totreat operating costs as being shared.

B.7.6.5. Divisional and Corporate Overheads

A large proportion of TDC’s cost basis is described as being divisional and corporate overheads.In a separate presentation, TDC has provided Telestyrelsen with some details about the costswhich underlie these categories and has outlined the various steps involved in allocating andapportioning these costs to the telephone network. However, TDC has not provided details onthe allocation and apportionment bases themselves. Thus, Telestyrelsen is unclear on the basesused to determine whether these costs are related to the telecoms network; nor is it clear how thecosts within the telecoms network have been allocated.

Telestyrelsen

80

A further concern is that since a considerable proportion of capital costs are considered to befixed, one would expect a similar, if not higher, share of operating costs to be fixed. A simpleexample will illustrate this point. If it is assumed that a doubling of transmission capacityrequirements results in a 50% growth in transmission equipment requirements it is hard to seewhy it should result in more than a 50% growth in transmission operating cost requirements.However, this in turn will mean that personnel, finance and other ‘overhead’ costs related totransmission are unlikely to increase by more than 50%.

B.7.7 Summary/main issues

Telestyrelsen has a number of reservations about the use of cost allocation principles in the modeland the shape of some of the CVR curves. It believes that:

• some of the assumptions regarding shared and common costs are questionable;

• relationships between capital and operating costs and between related classes within capitaland operating costs are not convincing;

Telestyrelsen

81

B.8 Co-location and Related ServicesCo-location and related services are included in Chapter 9 of the TD-documentation.

The chapter documents the cost of the following services:

• co-location;

• numbering plan and prefix;

• installation cost related to raw copper and operator pre-selection;

• points of interconnection; and

• interconnections capacity (30-grupper).

In the remainder of this chapter, references to tables, etc. are made to Chapter 9 of the LRAIC-documentation, unless otherwise stated.

B.8.1 Methodology

The services are modelled separately (stand alone) as a supplement to the main LRAIC-model(and are therefore not part of the LRAIC SAS-model).

The starting point for the cost calculations are the cost documentation for standard offers whichaccording to the law shall be submitted to Telestyrelsen.

The presentation is at a very aggregated level. As a result, the underlying assumptions and sub-calculations are generally missing.

The calculations are based on historical costs and volumes (year 2000). According to TDC, thesecalculations are updated and adjusted in accordance with LRAIC-requirements, e.g. assets valuesare adjusted according to the replacement cost principle, and square meter prices for co-locationare based on public valuations for year 2000.

Staff related OPEX includes the cost of housing and other staff related cost. A general overheadof -c- has been added to the cost of all other (non-staff related) cost components. The overheadfigure has been calculated on the basis of TDC’s business segmentation accounts for the year2000.

B.8.2 Potential Problems

Telestyrelsen notes, that the documentation which forms the basis for TDC’s calculations hasnot been prepared with the objective of documenting LRAIC-costs and hence it may bequestioned whether the calculations fully comply with the LRAIC-methodology.

Another potential problem resulting from the modelling of certain cost separately is thepossibility of double counting some of the cost, e.g.:

• the sqm. in the co-location calculations may be the same sqm. as in exchange buildings in theLRAIC SAS-model;

• the cost of the network planning system Panter, i.e. a total of -c- of the annual cost of Panterare allocated to interconnection specific cost. However, the cost of Panter may already bepartially or fully allocated in the LRAIC SAS-model; and

Telestyrelsen

82

• staff related OPEX, i.e. the resources consumed in the Net- and Wholesale-division, may alsohave been accounted for in the LRAIC SAS-model.

Telestyrelsen is currently closely examining the cost documentation included in chapter 9 of thetop-down documentation, both in connection with TDC submissions of standard offers and inconnection with compliance with the special LRAIC-methodology. To the extend thatTelestyrelsen finds it necessary, TDC will be asked for further cost documentation.

B.8.3 Compliance with Criteria’s

CG 26 states that co-location cost for interconnection and for access lines should be shownseparately. TDC has shown only one type of co-location cost covering both interconnection andaccess lines.

CG 26 states that housing (sqm. cost) should be showed separately for each exchange. It hashowever been recognised that these cost are shown per geo type instead, and that is what TDChas done. Otherwise the co-location model is in accordance with CG 26.

Telestyrelsen

83

B.9 Efficiency Study

TDC has undertaken an efficiency study based on Data Envelopment Analysis (DEA), insupport of its claim that its costs, and operating costs in particular, are efficiently incurred. Thestudy compares TDC’s efficiency with that of 51 US Local Exchange Carriers. TDC present 4different sub-analyses with different input and output parameters and different assumptionregarding economies of scale. In all of these analyses TDC obtains a score of (at least) 100. TDCtherefore argues that it should be considered to be an efficient operator.

However, the efficiency study suffers from a number of shortcomings and weaknesses. Some ofthe most important ones are:

1. Total reliance on DEA

2. Use of US LECs as the only benchmark operators

3. No adjustments are made to put TDC and the US LECs on a comparable basis

4. No sensitivity analysis on assumptions made by TDC

5. No weight restrictions are imposed. This means that there is a danger that an operator mayappear to be efficient as a result of putting a large weight on one or few factors and assigninga low (or zero) weight to all other factors.

6. Excessive reliance on the output “main switched minutes”.

These problems are discussed in more detail in the following sections.

B.9.1 Total Reliance on DEA

In addition to simple unit cost (ratio) analysis and basic (OLS) regression analysis, there are twomain ways of estimating relative efficiency across firms: Stochastic Frontier Analysis (SFA) andData Envelopment Analysis (DEA). Whereas SFA uses econometric techniques26, DEA usesmathematical programming27. Neither of these techniques are free from criticism and thereforeefficiency studies normally use both, with the results of each method being used as a cross-checkon the results of the other.

TDC, however, has relied entirely on DEA as its single benchmarking method for assessingefficiency. This is highly problematic as DEA has a number of drawbacks. For example, unlessrestrictive assumptions are imposed on the model, DEA is incapable of assessing the efficiency ofcompanies if they are operating in different environments or their inputs/outputs lie on the edgeof the data set. In particular, DEA may assume that these firms are absolutely efficient merely

26 SFA addresses the main problem with standard regression analysis, namely the assumption that the calculated

residuals (deviations from the calculated “average”) can be fully attributed to relative inefficiency. This is highlyproblematic since cost differences among other things may also be the result of measurement errors or otherfactors. SFA decomposes the residuals between “error” and “genuine inefficiency”. The technique draws on theinsight that errors are equally likely to increase or reduce measured costs for any operators where as inefficiencieswill only increase costs. In the case of inefficiencies the distribution of the residuals will therefore be skewed (actualcosts exceeds that estimated by the regression). By measuring the extent of this skew to the distribution, SFAcalculates the relative importance of observation errors vs. genuine inefficiencies. On the basis of this analysis theestimated regression line is then adjusted “downwards” by the estimated amount to form a so-called “efficiencyfrontier”. The distance of the companies cost relative to this frontier will then provide information on which tobase an estimate of the company’s inefficiency.

27 The DEA approach is described at length in section 10.4 of the top-down model documentation.

Telestyrelsen

84

because it is unable to identify a peer group with which these companies can be compared.28 Byway of contrast, Stochastic Frontier Analysis is able to take into account the impact of suchvariables on the frontier, by assuming a particular functional form for this.

The main problem with DEA is that it is inherently subjective. That is, there is no statistical wayto test the robustness of the formulated model, nor is there a way to test whether importantvariables have been omitted. Repeating the analysis with and without some variables does nothelp: it is known that adding variables never decreases DEA scores, so any company that appearson the frontier in simple models is bound to stay there in more complex models. As a result,there are no objective means of judging whether the results are reliable.

A further problem with DEA is that it assumes that all deviations from the frontier are due toinefficiency. This is equivalent to the argument that the data are free from statistical “noise”. Inpractice, this makes the analysis vulnerable to measurement errors and outliers. DEA scores arealso sensitive to input or output specification and to the size of the sample. Increasing the samplesize either reduces or, at best, keeps constant, the efficiency scores. Increasing the number ofoutputs and inputs without increasing the number of companies will inflate efficiency scores,because the dimensions in which a particular company can be unique (and hence shown to beefficient) increase.

Another potential drawback of DEA is that in many cases it can result in most of the firms lyingon the efficiency frontier, depending on the number of observations, the nature of outputs andthe returns to scale. This is because DEA does not assume any particular functional form for thefrontier. At the extreme, an organisation with no peers is regarded as efficient. As a result, it maybe difficult to distinguish relative efficiency29.

In summary, DEA can be a useful technique for measuring relative efficiency, but its suitabilitywill depend on the specific problem in question – for example, whether random influences donot heavily affect the data or whether the effects of omitted variables are less of an issue.Therefore, other techniques should be used as a cross-check on the results provided by DEA.

B.9.1.1. Use of US LECs as the only benchmark operators

TDC is compared with Incumbent US Local Exchange Carrier (LEC). The environment inwhich these firms operate is somewhat different than it is for TDC. First of all, LECs are onlyproviding local access provision whereas TDC is a full-scale network provider. Secondly, thecompetitive pressure on the LEC is substantially lower than the competitive pressure faced byTDC and other European operators. It may therefore be questioned whether these LECs (or eventhe best of them) represent a relevant benchmark for efficiency. TDC notes that the US LECs arenormally thought of as setting the international benchmark for efficiency. As noted by TDCthere is another possible reason why the US LECs are often used for benchmark analysis: dataavailability. Telestyrelsen is more convinced by the latter explanation and considers that it wouldbe desirable to supplement the US data with data from European operators. Telestyrelsen,however, accepts that such data is unlikely to be readily available.

28 This is exactly what happened in the case of BT’s efficiency study, which was prepared on behalf of OFTEL, and

also referred to by TDC in the model documentation. As a result, DEA was subsequently abandoned in favour ofSFA.

29 In fairness it should be noted that 51 observations (operators) normally will be sufficient for an analysis with only4 outputs and one input, provided that firm-specific factors are taken into account.

Telestyrelsen

85

B.9.2 No adjustments made to place TDC and US LECs on a comparable basis

TDC does not seem to take into account differences between the environment in which it andthe LECs are operating. Important differences could e.g. be that:

• The service offering of LECs is likely to differ from that of TDC. The traffic mix maytherefore not be readily comparable.

• Calling rates are substantially higher in the US than in Denmark. Local calls are often free inthe US.

• Line density is lower in the US than in Denmark. One would therefore expect costs, andaccess costs in particular, to be substantially higher in the US30.

• Terrain may be different. Denmark is comparatively flat which implies that the cost ofconstructing the network will be lower all other things being equal.

• Wage rates and other costs will differ between US and Denmark.

As noted, a number of these factors are likely to work to TDC’s advantage although otherfactors, such as calling rates, may operate in the other direction. A priori it is difficult to assessthe overall impact of these factors making it all the more important that they are taken accountof in the analysis.Another potential problem is that TDC is comparing the total costs of the modelled (optimised)network with the total actual costs of the LEC. Even though TDC comes out as an efficientoperator (possibly due to low capital costs) this does not necessarily imply that TDC’s operatingcosts are efficiently incurred. TDC actually presents an analysis where operating costs has beenused as input instead of total costs. However, this analysis is only performed under theassumption of variable returns to scale where it is much easier to become efficient than under anassumption of constant returns to scale. TDC does not explain the rationale for this approach. Itwould have increased the validity of TDC’s analysis if the comparison has also been made underthe assumption of constant returns to scale.

Finally, it is not even clear whether the compared cost categories are indeed comparable. Forexample, it is unclear whether the LECs have been calculated on an analogous basis to those ofTDCs31. Also it is worth mentioning that costs are calculated on the basis of book values. It isunlikely that the LECs will have applied depreciation principles that are similar to those of TDC.For example assets lives tend to be higher in the US. The asset values may therefore not bedirectly comparable, particularly as LEC data is shown in historic cost terms32.

30 TDC has compared line densities (provided by BT for the US since they are not readily available) with the

efficiency score. TDC concludes that there is no relationship between efficiency and line density. Telestyrelsen issomewhat sceptical of this conclusion, however. All international experience (including TDC’s geo-typecalculations) indicates that access costs are highly related to line density.

31 LEC’s data contains a number of cost categories which are clearly predominantly, if not entirely, network relatedwhile others are retail related. However, a number of other cost categories are both network and retail related.TDC will have needed to make an implicit or explicit assumption about the extent to which these costs arenetwork and retail related.

32 TDC may have made some adjustment to put data in CCA terms but this is not discussed in the documentation.

Telestyrelsen

86

B.9.3 No Sensitivity Analysis on Assumptions made by TDC

TDC must have made assumptions (or, at least, has followed NERA’s assumptions) on a numberof important factors that are difficult to extract from publicly available data, such as therelationship between US and Denmark leased line tariffs, traffic volumes and routing factors. Itappears that no sensitivity analysis was undertaken to assess the reliability of the results uponthese assumptions.

B.9.4 No Weight Restrictions

One potential strength of DEA is that it is capable of assigning its own weights to inputs andoutputs, recognising that different companies in practice may need to attach different weights tooutputs. However, this can also be a major weakness of DEA, if the weights are not appropriatelyapplied. Since all companies can adopt their most favourable weights, most of them may appearto be efficient. This is e.g. illustrated in Table 3 of section 10.6 of the documentation where 17out of 52 companies appear to be 100% efficient.

An operator may be labelled as efficient by DEA by being specialised in the production of asingle output. This will be the case even if the operator is highly inefficient in producing all otheroutputs. One way around this problem is to assign appropriate restrictions on the parameters toensure that no false efficiency can be achieved by putting a large weight on one or few factors andassigning a very low (or zero) weight to all other factors. Weight restrictions are normallynecessary to ensure that no (relevant) output is totally ignored from the estimate of efficiency.However, no such restrictions have been applied by TDC in this study, on the grounds that todo so would imply a certain degree of subjectivity in the analysis33.

Confronted with this critique, TDC has provided Telestyrelsen with an “example” (restated inTable 19 below) which, according to TDC, “proves that TDC is not efficient due to anartificially high weighting of a single output”.Table 19: Example of weightings for TDC in analysis A with constant returns to scale

Input and output variable Data for TDC Weightings forTDC

Recalculation to virtual weight(percentage)

Total Cost (Mill. USD) 764 0.0013092 100Local Switched Minutes (mill.) 42,019 0 0Main Switched Minutes (mill.) 8,506 0.000025 21.3Total Access Lines (thousands) 3,254 0.0001 32.5Fiber Cable Kilometres (thousands) 16 0.029576 46.2

In this example the production of local exchange minutes is attached zero value even thoughlocal exchange minutes are an important outputs of TDC and a highly important output of theLECs. No output should be ignored in such way in the determination of efficiency, especiallynot when it constitutes such a large proportion of total output. It is particularly problematic sincethe output of local exchange minutes is relatively low for TDC compared with those of the USLECs34.

33 It is true that such restrictions introduced a certain degree of subjectivity. However, it is worth noting that DEA

will always be subjective (e.g. through the choice of input and output).34 In fairness it should be noted that TDC notices there are also examples of weightings in which local exchange

switched minutes are weighted positively and where TDC still is efficient. This may not say much however. With

Telestyrelsen

87

The weight examples provided by TDC, however, indicate that the inclusion of weightrestrictions would probably not do much in the programming problem solved for TDC, and istherefore less important than Telestyrelsen anticipated on the basis of the information providedin the top-down documentation. However, the weight examples provided by TDC instead havepointed out a much more fundamental problem with the analysis that was not apparent. Theproblem will be discussed in more detail in the next section.

B.9.5 Excessive Reliance on the Output “Main Switched Minutes”

As mentioned above a principal weakness of DEA is that an operator may be labelled efficient bybeing relatively efficient (or just specialised) in the production of a single input. In the case ofTDC the 100% score is achieved, in all but one analysis, by being efficient in the production ofmain switched minutes.

It is therefore important to examine whether main switched minutes represents a fair parameteron which to compare TDC with the US LECs. The average call of a LEC uses a main switch to amuch lower extent than the average call of TDC, an observation confirmed by TDC. The factthat TDC produces a large number of main switched minutes for a given total cost may thereforenot necessarily imply that TDC is efficient. It may simply indicate that TDC faces a higherdemand for main switched minutes (and possible a lower demand for other outputs such as e.g.local exchange minutes).

This leads to the main problem with the efficiency study, highlighted by the weight examplesprovided by TDC. In defence of the approach taken, TDC emphasises the point that there is nota single set of weights that make TDC efficient but rather a continuum of weights. In relation tothe weight example listed above, TDC e.g. notes that “TDC is the best of all to produce mainswitched minutes, and could therefore become efficient, even if it is only main switched minutesthat is attributed a weight. Thus a continuum of weights exists which will make TDC efficient inthe analysis.”

As it appears from TDC’s example, the weight that TDC attaches to main switched minutes is21 per cent. This is surprising, as one would expect the model to attach the highest weight to thisoutput given that TDC is the best of all to produce main switched minutes. By the phrase“continuum of weights” TDC appears to imply that, in fact, TDC will seem to be efficient aslong as the weight attached to main switched minutes ranges from 21 per cent to 100 per cent,regardless of what happens to the weights of the remaining outputs. TDC seems to interpret thisfinding as providing confidence in TDC’s conclusion.

In the view of Telestyrelsen, on the other hand, the vast range of values that can be attributed tothe weight attached to main switched minutes and still make TDC seem efficient is veryproblematic and leads to questions about the robustness of TDC’s methodology. Normally, onewould expect an efficient unit to become inefficient when the weights are changed sodramatically. The fact that a 100% efficiency score can be achieved is almost reflecting the vastlydifferent environments in which the companies are operating.

The problem can best be explained by looking at the following (fictitious) worked example:

Suppose one has data from 4 companies: 3 LECS, denoted by A, B and C and TDC. Theinput/output data is set out in Table 20.

a Variable Returns to Scale specification (compared to constant returns) for example more companies tend toappear efficient

Telestyrelsen

88

Table 20: Input and output variable in worked example

Input and output variable A B C TDC

Total Cost 100 102 113 109

Local Switched Minutes 22 24 33 15

Main Switched Minutes 31 29 17 89

Access lines 26 40 24 14

DEA now uses mathematical linear programming techniques to find the set of weights thatmaximise the efficiency score of each firm, subject to the constraint that none of the remaininghas an efficiency score greater than 100% at those weights. Therefore it solves the same problem4 times, one for each firm. We concentrate on TDC since this is of more interest to us.

Consider the following set of weights for the three outputs:

• Local Switched Minutes: 0 per cent

• Main Switched Minutes: 100 per cent

• Access Lines: 0 per cent

Then the companies’ index values are:

7.81109

0*14100*890*15 =++=TDC

31100

0*26100*310*22=

++=A

4.28102

0*40100*290*24=

++=B

15113

0*24100*170*33=

++=C

Since at this set of weights, TDC achieves the highest value for the DEA index, one will concludethat TDC is efficient.

Suppose now that we gradually decrease the weight attached to main switched minutes. First wedecrease the weight by 30 percentage points and increase the weights attached to the remainingtwo outputs by 15 percentage points each. The impact on the companies’ index value are shownin Table 21.Table 21: Efficiency score at different weights

Efficiency score Weights (Local Minutes, Main Minutes , Access lines)Company (0%, 100%, 0%) (15%, 70%, 15%) (30%, 40%, 30%) (39%, 22%, 39%)

TDC 81.7 61.2 40.6 28.3A 31.0 28.9 26.8 25.5B 28.4 29.3 30.2 30.7C 15.0 18.1 21.2 23.0

Since at this set of weights, TDC also achieves the highest score, DEA will conclude that TDC isefficient again.

Telestyrelsen

89

Next, we decrease the weight attached to main switched minutes by another 30 per cent. As weobserve, TDC’s score deteriorates substantially although it remains efficient compared to theothers.

Finally, the weight attached to main switched minutes takes the value of 22 per cent. Now, TDCstops being efficient since at this set of weights, B achieves a higher score.

Therefore, there is a continuum of weights (ranging approximately from 25.5 per cent to 100 percent for main switched minutes) that make TDC appear efficient. This, however, does notnecessarily mean that TDC is truly efficient. All it tells us is that TDC appears to be relativelygood in producing a particular output. However, the reasons why it is efficient remainsunknown.

If TDC is good in producing main switched minutes purely because of environmental factorsspecific to Denmark, then this should not be attributed to efficiency. In reality, efficiency is ameasure that is controllable by the firm in the long run. All firms can become efficient becausethey share the same production function. This principle underlies the methodology of comparingdifferent firms. If one believes that companies do not share the same production function, thenclearly any comparison is meaningless.

Given the similarities in the types of businesses, there is indeed a rationale for comparing theLECs with TDC, after allowing, however, for some factors that are specific to each firm andespecially to TDC in relation to the remaining companies (as mentioned above there are anumber of important differences between the environments in which TDC and the US LEC’sare operating). Since this crucial step is omitted, it is not possible to characterise TDC as efficienton this basis. Since environmental factors are not controllable by the companies, they should beexcluded from any measure of efficiency.

What the efficiency study shows is not efficiency, but a composite measure of explainingdeviations among firms, part of which is due to environmental factors - that are inherentlyuncontrollable by firms - rather than efficiency.

To further reinforce this point it is worth considering the one analysis where a somewhatdifferent approach, namely Analysis B where local exchange and main exchange minutes arecombined. In this case, TDC also obtains a score of 100%, not because of its switchingefficiency, but because of its performance on fibre kilometres. However, this analysis suffers fromat least two major flaws. Firstly, fibre kilometre performance is to a large degree a factor which isoutside the company’s ability to influence since it is highly correlated with population density.Indeed, we are aware of other studies which have treated fibre (or sheath) kilometres as anenvironmental variable designed to capture population dispersion. Secondly, given theconsiderably higher cost of producing local than main switch minutes there is a good case forattaching a higher weight to the former. A recent study by NERA35 attaches twice the weight tolocal switch minutes than to main switch minutes. While these precise factors are somewhatsubjective the underlying principle is, in our view, sound. Imposition of such a weighting systemmay reduce TDC’s relative performance.

B.9.6 Conclusion

Telestyrelsen has a number of concerns with the efficiency study provided by TDC for thereasons outlined above. Telestyrelsen does not find that the efficiency study has proven TDC tobe an efficient operator - not in absolute terms nor in relative terms. It should be noted that

35 “The comparative efficiency of BT”, July 2000, p. 32.

Telestyrelsen

90

TDC has provided Telestyrelsen with a copy of the data set which was used in TDC efficiencystudy. This data set will be used in Telestyrelsens further investigation.

Telestyrelsen

91

B.10 General Costing Issues

B.10.1 Volumes

As prescribed by the MRPs, TDC estimates costs on the basis of traffic data for 2000. Stockssuch as the number of customers and number of circuits are estimated for end-of-year 2000.Volumes are documented in annex 3 of the documentation.

As TDC did not sell any “dark fibre” in 2000, TDC has calculated costs on the basis of km offibre cable used by TDC.

B.10.2 Margins for Growth

The top-down documentation does not show the anticipated growth per annum for each servicefor one, three and five years as prescribed by Criterion CG 8. TDC has later informedTelestyrelsen that the network has been built according to existing planning norms, but has beendimensioned to meet the growth rates published by Telestyrelsen.

Different planning horizons have been used for different platforms. For example the horizon islonger for access than for transmission since it is much quicker to install new equipment than itis to dig down cables. TDC’s network (actual and modelled) is generally dimensioned on thebasis of existing planning norms.

For switching, the growth rates applied for ISDN2 and ISDN30 are 40% and 45% respectively,which according to TDC corresponds to a two-year planning period. No growth has beenassumed for PSTN. The growth is included in the dimensioned traffic used for calculating inter-exchange connections as well as for determining the cost of line cards. No growth has beenassumed for IC connections.

For access, generally 3-5 year growth rates are taken into account. According to TDC, however,it is more relevant to examine current installation practice, which ensures flexibility in the accessnetwork and results in the lowest costs in the long run.

For transmission, a planning horizon has been used which meets the requirements for supplydelivery times and which in TDC’s experience gives a reasonable upgrade frequency in thenetwork and includes those delivery times on equipment and installation times that TDC has.For line systems this is done by including capacity for a working layer. The time horizon isdifferent on different stretches depending on growth. At stations with large growth it may be ½-1year. In the thin parts of the network it may be substantially longer (maybe 3-5 years).

B.10.3 Routing Factors

For each type of call, TDC has estimated routing factors indicating the use of switching elementsand transport elements per call and per minute. This has been done in a separate routing factormodel. The resulting routing factors are listed in table 24 of chapter 3.

The routing factors for calls and traffic differ for many call types. Whereas a call between twodifferent exchange areas will always be signalled up to and processed by a tandem exchange, forexample, the call may be set-up directly between two local exchanges. As a result the tandemswitch will be used more intensively for calls than for traffic. With regard to the IN switch, “InAdvanced” services (by definition) uses an IN switch for the entire duration of the call, where asother calls only use the IN switch 1,2 or 3 times per call.

Telestyrelsen

92

Criterion TD 35 requires the SMP operator to “provide information (for all calls) showing thepercentage of calls following a particular routing pattern (e.g. 2 RSSs, 1 local exchange processor,2 remote to local transmission links).”.

Such information has not been provided by TDC. In many cases, however, it is possible forTelestyrelsen to reconstruct these figures from the routing factors. TDC has also providedinformation on the distribution of traffic for different types of transmission link, cf. table 23 inchapter 3 of the documentation.

One interesting thing to note here is the significant of LS-LS links.

For Interconnection calls within areas, for example, TDC has reported a routing factor for trafficof (RSS/SSS; LX;TX) = (-c-)

Such a routing factor can only be explained by the following distribution of traffic:Table 22: Inferred traffic distribution for interconnection traffic within area

Percentage of traffic Route (switching elements)

-c- SS-LX-LX(IC)36

-c- SS-LX-TX-LX(IC)

-c- SS-LX(IC)

-c- of IC calls within area are thus conveyed directly between two local exchanges. For IC callsbetween areas the same percentage appears to be at least -c- with such a call using only -c- tandemexchange on average37.

Based on table 23 of chapter 3, -c- of the traffic between LXs in different interconnection areasseem to be carried on direct (logical) LX-LX links. For traffic between LXs located in the sameinterconnection area, -c- of traffic seem to be carried directly between the LXs, whereas only -c-pass a TX. Only -c- of total traffic pass through two TXs.

Routing factors for transmission and signalling have been estimated somewhat differently fromthe standard approach. TDC has divided transmission links into three categories (jumps/“hops”):jumps within layer 1 of the transmission network; jumps within layer 2 of the network (incl.jumps from layer 1 to layer 2); and jumps within layer 3 of the network (incl. jumps betweenlayer 2 and 3. The jumps in layer 1 are typically longer than jumps in layer 2 and 3, which allother things being equal implies higher costs. On the hand jumps in layer 2 and 3 arepredominantly being handled in lower level transmission systems, which all other things beingequal implies higher cost for these jumps. These opposite effects are reflected in the relative costsof the different types of jumps. Taken these factors into account along with other factors such asutilisation ratios, TDC finds that the cheapest distance is between the LX and TX, due to highcable size and the fact that the TX normally lies in Layer 1. The distance between the LX and theSS on other hand is relatively expensive due to small cable sizes and the fact that the link typicallyuses layer 3.

For each leg/link the number of hops are turned into what TDC refers to as the number of“equivalent hops” by weighting the hops according to their relative cost. When costing servicesthese number of equivalent hops are then multiplied by the cost per equivalent hop.

36 LX(IC) denotes a local exchange dedicated as IC exchange37 The percentage is likely to be even higher as some calls may use two tandem exchanges.

Telestyrelsen

93

Although TDC’s methodology generally seems to be sound, it is very difficult for Telestyrelsen toevaluate the figures reported by TDC, as the underlying calculations are not transparent. It is alsodifficult to compare the routing factors reported by TDC with those of other operators, asTDC’s routing factors bear no resemblance to actual routing factors. The routing factors fortransmission and signalling are shown in table 24 of chapter 3 in the documentation. As opposedto the switching routing factors, these routing factors cannot be interpreted as the times thenetwork element is used. Instead they seem to indicate the number of “equivalent hops”.

characteristics of the top down and bottom up cost analyses · pdf...

Documents