PPDR kanada

Ref: 35971-214 – Public .

Model documentation for the

Norwegian Post and

Telecommunications Authority

Mobile cost model

version 8 final (v8F)

27 May 2013

Ref: 35971-214 – Public

.

Ref: 35971-214 – Public .PUBLIC

Contents

1 Introduction 1

2 Conceptual approach for the NPT v8F model 4

2.1 Summary of recommendations from the NPT v7.1 model 4

2.2 Operator-related conceptual issues 5

2.3 Technology-related conceptual issues 9

2.4 Service-related conceptual issues 18

2.5 Implementation-related conceptual issues 20

3 Demand forecasting 26

3.1 LTE demand forecasting 26

3.2 OTT traffic 30

3.3 Updates of historical demand parameters 33

3.4 Updates of forecast demand parameters 34

4 Calculations related to the EC/ESA Recommendations 38

4.1 Structure of the generic operator calculation 38

4.2 Generic operator input derivations 40

4.3 The Pure LRIC calculation 44

4.4 The LRIC and LRIC+++ 46

5 Mobile network design 47

5.1 Pure LRIC in-fill adjustments 47

5.2 HSPA upgrades 48

5.3 UMTS Ethernet backhaul deployment 50

5.4 Spectrum licences 51

Annex A Excerpts from the v7.1 model documentation

A.1 Coverage

A.2 Radio network: Channel kit (CK) and carrier requirements

A.3 Backhaul transmission [this section has been superseded by the information presented in

the main body of this report]

Annex B Model adjustments from v7.1 to v8D

B.1 Model corrections

B.2 Revised input parameters and other decisions

Mobile cost model version 8 final (v8F)


Annex C Operator submissions in developing v8F

C.1 Comments related to the demand calculations

C.2 Comments related to the network calculations

C.3 Comments related to the costing calculations

C.4 Process-related comments

Annex D Model adjustments from v8D to v8F

D.1 Model corrections

D.2 Revised input parameters and other decisions

Annex E Expansion of acronyms

Mobile cost model version 8 final (v8F)


Copyright © 2013. Analysys Mason Limited has produced the information contained herein

for the Norwegian Post and Telecommunications Authority (NPT). The ownership, use and

disclosure of this information are subject to the Commercial Terms contained in the contract

between Analysys Mason Limited and NPT.

Analysys Mason Limited

St Giles Court

24 Castle Street

Cambridge CB3 0AJ

UK

Tel: +44 (0)845 600 5244

Fax: +44 (0)1223 460866

[email protected]

www.analysysmason.com

Registered in England No. 5177472

NOTE: [] marks the removal of confidential information.

Mobile cost model version 8 final (v8F) | 1


1 Introduction

The Norwegian Post and Telecommunications Authority (NPT) has determined prices for mobile

termination in Norway by means of the long-run incremental cost (LRIC) method since 2007.

In 2006, a bottom-up long-run incremental cost model (v4) was constructed and finalised for NPT

by Analysys Mason Limited (Analysys Mason), with the aim of calculating the cost of voice

termination for the GSM mobile operators in Norway. In 2009, this model was upgraded to

include 3G technologies and a ‘Pure LRIC’ calculation, and the final version (v7.1) was issued in

September 2010. This version (the NPT v7.1 model) currently forms the basis of wholesale mobile

termination price regulation of Norwegian mobile operators.

In late 2012, NPT contracted Analysys Mason to undertake a further upgrade of the NPT v7.1 model

used to set the prices for mobile termination in Norway. This report documents the final version

eight (v8F) of the mobile LRIC model (the NPT v8F model) issued for national consultation in

summer 2013. The NPT v8F model builds on from the draft version 8 model (the NPT v8D model)

that was issued for consultation in March 2013.

Since the NPT v7.1 model was finalised, the Norwegian mobile market has evolved in several ways,

all of which have been reflected in the design of the NPT v8F model:

Two mobile virtual network operators (MVNOs), Tele2 and Network Norway, formed a joint-

venture company (Mobile Norway) in order to deploy a 2G and 3G network. Mobile Norway

has deployed significant infrastructure from 2010 onwards. During the second half of 2011,

Tele2’s parent company (Tele 2 Sverige AB) acquired Network Norway. Therefore, these two

MVNOs are now both owned by the same company.

3G networks and services have continued to evolve for the mobile network operators, and both

Telenor and TeliaSonera (i.e. NetCom) have since launched Long Term Evolution (LTE, or

4G) networks.

Over-the-top (OTT) services (such as mobile IP telephony and mobile VoIP, and similar

services for SMS) are becoming more widespread and are therefore likely to affect the demand

forecasts for circuit-switched traffic within the existing model.

In April 2011, the European Free Trade Association (EFTA) Surveillance Authority (ESA)

released a Recommendation for the costing of termination rates,1 which is analogous to that

published by the European Commission (EC) in May 2009.2 In particular, the Recommendation

requires the consideration of a ‘generic’ operator and ‘pure’ incremental costing.

1 See http://www.eftasurv.int/media/internal-market/ESAs-Recommendation-on-termination-rates.pdf

2 Commission Recommendation of 7 May 2009 on the Regulatory Treatment of Fixed and Mobile Termination Rates

in the EU (2009/396/EC). Available at http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:L:2009:124:0067:0074:EN:PDF



NPT intends for the v8F model to inform its future decisions on wholesale termination regulation.

This document describes the v8F model for industry consultation.

A schematic of the NPT v8F model is shown below in Figure 1.1. The operator-specific inputs are

used to calculate the inputs for a generic operator as discussed in detail in Sections 4.1 and 4.2.

The model uses these demand and network design inputs in the calculation of operator

expenditure, which is then depreciated and allocated using routeing factors to give the unit costs

by service for the three actual mobile network operators and the modelled generic operator.

Figure 1.1: Model schematic [Source: Analysys Mason, 2013]

The remainder of this document is laid out as follows:

Section 2 describes our reconsideration of the conceptual approach from the NPT v7.1 model

to give the concepts used in the NPT v8F model

Section 3 describes the changes made to the demand forecasting in the NPT v8F model

Section 4 describes the calculations included to reflect the EC/ESA Recommendations

Section 5 describes additional changes made to the modelled network design.

Market evolution

Operator demand

Migration profiles

Routing factors

Demand drivers

Technical inputs

Network design

algorithmsOperator network

Operator

expenditure

Unit costs

Price trends

Economic

depreciation

Service cost

Netcom inputsTelenor inputsMobile Norway

inputs

Generic operator

inputs

Input Calculation OutputKEY:

WACC



The report includes a number of annexes containing supplementary material:

Annex A provides excerpts from the v7.1 model documentation that describe aspects of the

network design that have since been revised

Annex B provides an overview of the key changes made to the NPT v8D model

Annex C considers the operator submissions in developing the NPT v8F model including our

analysis and response

Annex D provides an overview of the key changes made to the NPT v8F model

Annex E provides an expansion of the acronyms used within this document.



2 Conceptual approach for the NPT v8F model

The document Conceptual approach for the upgraded incremental cost model for wholesale

mobile voice call termination, 1 August 20093 (‘the 2009 concept paper’) was developed as part of

the previous LRIC process and contained the recommendations on which the NPT v7.1 model was

based, covering both the bottom-up calculations and the subsequent top-down reconciliation. This

section describes some revisions to these recommendations, which we believe are required.

The conceptual issues previously considered are classified in terms of four modelling dimensions:

operator, technology, service and implementation.

The remainder of this section is set out as follows:

Section 2.1 reaffirms the conceptual issues associated with the NPT v7.1 model and

distinguishes those that require additional consideration

Section 2.2 deals with conceptual issues related to the definition of the operator to be modelled

Section 2.3 discusses conceptual issues related to the technologies employed

Section 2.4 examines conceptual issues related to the service definitions

Section 2.5 explores conceptual issues related to the implementation of the model.

2.1 Summary of recommendations from the NPT v7.1 model

The 2009 concept paper was developed as part of the previous LRIC process and established the

principles for the NPT v7.1 model. The paper included 17 recommendations that will form the basis of

the NPT v8F model, but are being reconsidered due to recent developments in the Norwegian market.

Figure 2.1 summarises the recommendations that will require either minor rewording or significant

revision in the NPT v8F model. All other recommendations remain unchanged.

Figure 2.1: Conceptual decisions for the NPT v7.1 model [Source: Analysys Mason, 2013]

Conceptual issue Recommendation from the v7.1 model Reconsider?

[1] Structural implementation Bottom-up, reconciled against top-down

information

Reword

[2] Type of operator Actual operators with a hypothetical third

network operator

Reword

[3] Size of operator Actual size of operators with a hypothetical third

network operator

Reword

[4] Radio technology standards 2G and 3G, as needed to reflect actual

operators

Revise

[5] Treatment of technology

generations

Included within the model explicitly Revise

3 See http://www.npt.no/marked/markedsregulering-smp/kostnadsmodeller/lric-

mobilnett/_attachment/1803?_ts=1390fd7ef91




[6] Extension and quality of

coverage

Reflect historical and expected future coverage Reword

[7] Transmission network Actual transmission networks as far as possible Reword

[8] Network nodes Apply scorched node, optimised for efficiency Reword

[9] Input costs Mixed approach based on actual/average costs Reword

[10] Spectrum situation Include capability to capture actual or

hypothetical allocations, as well as licence fees

Revise

[11] Service set Both voice services and non-voice services Revise

[12] Wholesale or retail Apply a 75:25 split of overhead costs No change

[13] WACC Apply NPT’s mobile operator WACC No change

[14] Depreciation method Economic depreciation No change

[15] Increments Calculate LRIC, Pure LRIC and LRIC +++ costs Reword

[16] Years of results All relevant past and future years (i.e. from 1992) No change

[17] Mark-up mechanism Use equi-proportionate mark-up (EPMU) No change

In the subsequent sections, modifications to recommendations are highlighted in red.

2.2 Operator-related conceptual issues

The conceptual issues revisited in this section are shown in Figure 2.2.

Figure 2.2: Decisions on the operator-related conceptual issues taken for the NPT v7.1 model [Source:

Analysys Mason, 2013]


[1] Structural implementation Bottom-up, reconciled against top-down

information

Reword

[2] Type of operator Actual operators with a hypothetical third

network operator

Reword

[3] Size of operator Actual size of operators with a hypothetical third

network operator

Reword

The operator-related recommendations are relevant to the modelling of two actual operators and a

hypothetical third operator in the NPT v7.1 model. These have been reworded for the

NPT v8F model to apply to the three actual operators and a generic efficient operator.

2.2.1 Structural implementation

There are two main ‘directions’ for modelling the costs of the mobile network operators: bottom-

up or top-down modelling. There is also a third alternative: a combined approach (usually called a

hybrid model) can be adopted in which the bottom-up model usually ‘leads’ the calculation, and

the top-down model supplies complimentary and valuable reference data points. It is necessary to

define the modelling approach at the beginning of the project, prior to the collection of data, since

this choice determines what will eventually be possible with the model – e.g. cross-comparison of

operator data, investigation of alternative hypothetical operators.



Developing an understanding of the costs of the different mobile operators in the Norwegian

market can be achieved by being able to model, and parameterise, operators’ networks and

demand differences within a common structural form (i.e. a bottom-up model). A bottom-up

model also has the benefit that it can be circulated (without any confidential operator information)

to all industry parties, including non-mobile operators. This transparent circulation facilitates

industry discussion of the approach taken to demand and network modelling. In addition, operator-

specific models can be discussed bilaterally with each mobile party.

In order to make appropriate decisions regarding price regulation for the Norwegian market, NPT

will need to understand the actual costs that each operator faces. Although a top-down model can

produce actual costs, it lacks the ability to explore operator differences with certainty or

transparency. Therefore a hybrid model is most likely to satisfy NPT’s requirements to:

achieve industry ‘buy-in’ to the approach

provide reassurance to the operators that the model replicates not only their networks, but

more importantly their overall costs

enable accurate understanding of operator cost differences

have a tool that can be used to explore price-setting issues.

A hybrid model demands information from market parties on both network and cost levels.

However, the information demands for a hybrid model is only marginally more extensive than

would be needed for just a bottom-up or top-down approach.

NPT believes that bottom-up data will be relatively straightforward to source from operators’

management information (e.g. demand levels, network deployments, equipment price lists), and

top-down data should be available from financial accounting departments, usually with some

requirement for pre-processing stages.

NPT believes that the modelling approach that will deliver the most benefits and relevant information

for its costing and price-setting activities will be a hybrid model, ‘led’ from the bottom-up direction.

This bottom-up led hybrid model essentially means that the top-down part of the hybrid model is less

onerous for all parties, and refined for the purpose of being used as inputs to a bottom-up model:

It is not necessary to construct stand-alone, top-down models capable of full service costing

and depreciation (since the bottom-up model is capable of this).

The model and industry discussions are not hindered by opaque and confidential top-down

calculations (since the bottom-up model can be discussed more freely with market parties).

The top-down ‘model’ can be condensed to simply a presentation of suitably categorised top-

down accounting data, against which the bottom-up model can be reconciled.

The recommendation established in 2009 only needs rewording in order to capture the inclusion of

a Mobile Norway-specific calculation and a generic operator, as well as the removal of references

to the “third operator” calculation.

Recommendation 1, reworded: Develop a bottom-up cost model reconciled against top-

down accounting data for the three actual network operators and a generic efficient operator,

resulting in a hybrid model.



2.2.2 Type of operator

The choice of operator type to be modelled feeds into NPT’s decision on pricing for suppliers in

Market 7. However, the choice of operator type for cost modelling purposes, as outlined here, does

not preclude NPT from adopting an alternative basis for pricing. As a result, this costing and

pricing conceptual issue has been separated into its constituent costing and pricing parts. This

section of the conceptual approach refers to the type(s) of operator to be costed in the model.

The main options for operator type are outlined below.

An actual operator: this reflects the development and nature of an actual network operator

over time, and includes a forecast evolution of the operator in order to develop long-run costs.

This type of model will aim to identify the actual costs of the operators being modelled, and

should result in the most accurate quantification of the operators’ cost differences. An

operator-specific, top-down reconciliation can be carried out with this type of model. This type

of model can also be used to reflect average or hypothetical operators, by adjusting various

input parameters.

An average operator: by adopting an average operator approach, the cost model will merge

inputs, parameters and other features of actual Norwegian network operators to form an

average operator cost model. As a result, it may be harder to explore, identify and quantify the

cost differences between the network operators, and reconciliation of a bottom-up model

against top-down data must be carried out at an average level.

A hypothetical operator: this type of model aims to generate only the cost level which would

be achieved by a hypothetical operator in the market, usually a hypothetical new entrant. As

such, this type of model is focused on defining the demand inputs, network design and cost

levels that the hypothetical operator would experience, and therefore determines the cost base

of the hypothetical operator. Because of the hypothetical nature of this model, it is more

difficult to explore and quantify the differences between each actual operator’s costs and the

hypothetical set-up. Top-down reconciliation of a bottom-up model must also be carried out in

a discontinuous manner. The “generic operator” as is described by the EC/ESA

Recommendation, can be seen as a type of hypothetical operator.

The choice of operator type affects two main outcomes of the modelling work:

the level of understanding NPT can gain on the costs of each actual network operator (and in

particular differences in costs between operators)

the ability of the model to cope robustly with alternative operator choices when it comes to

determining the operator specification and network specification of cost-oriented mobile

termination prices.






Recommendation 2, reworded: Adopt an actual operator costing for Telenor, TeliaSonera4

and Mobile Norway, which can accurately determine the costs of each actual network operator

and robustly explore individual cost differences between these three mobile operators. The

model will also be populated to calculate the costs of a generic efficient operator in Norway.

This generic operator is not intended to reflect any of the actual mobile network operators,

but is intended to be generically applicable to the cost of mobile termination in Norway.

should be applicable to Mobile Norway as the third infrastructure operator. Actual MVNOs will

also be included.

2.2.3 Size of operator

One of the major parameters that define the cost of an operator is its market share. It is therefore

important to determine the evolution of the market share of the operator over time. In addition to

market share measured on a subscriber basis, we also include the volume and profile of traffic that

the operator is carrying within the scope of operator size.

The parameters that are chosen to model operator market share over time have a strong effect on

the overall level of economic costs calculated by the model (in a mobile network, share of traffic

volume is more significant than share of subscribers). These costs can change significantly if

short-term economies of scale (such as network roll-out in the early years) and long-term

economies of scale (such as fixed costs of spectrum fees) are fully exploited. The more quickly the

operator grows, the lower the eventual unit cost will be.




Recommendation 3, reworded: Consistent with Recommendation 2, the actual size of the

three actual incumbent infrastructure operators should be modelled according to historical

market development, with a forecast size for each operator. The scale of the generic efficient

operator will also be forecast. It is expected that this forecast market development will

reflect both subscriber and volume equalisation at some point in the future, although at a

network level, if Mobile Norway is modelled with a coverage significantly lower than 99%

population, then we expect to model an unequal share of traffic by network. NPT will also

consider the likelihood of a fourth UMTS infrastructure player entering the market in the

near future, and whether any demand scenarios are relevant for exploration in this respect.

The actual scale of MVNOs will also be modelled.

4 This operator will continue to be referred to as NetCom in the v8F model.



2.3 Technology-related conceptual issues

In this section, we describe the technological aspects of the model: radio technologies and generations,

network coverage and transmission topology, scorched-node calibration, equipment unit costs, and the

spectrum of the modelled operators. The issues revisited in this section are shown in Figure 2.3.

Figure 2.3: Decisions on the technology-related conceptual issues taken for the NPT v7.1 model [Source:



[4] Radio technology standards 2G and 3G, as needed to reflect actual

operators

Revise

[5] Treatment of technology

generations

Included within the model explicitly Revise

[6] Extension and quality of

coverage

Reflect historical and expected future coverage Reword

[7] Transmission network Actual transmission networks as far as possible Reword

[8] Network nodes Apply scorched node, optimised for efficiency Reword

[9] Input costs Mixed approach based on actual/average costs Reword

[10] Spectrum situation Include capability to capture actual or

hypothetical allocations, as well as licence fees

Revise

2.3.1 Radio technology standard

Mobile networks have been characterised by successive generations of technology, with the most

significant progress being the transition from analogue to digital (GSM), and the subsequent migration

to UMTS. A further migration of traffic to LTE networks is also beginning to occur in Norway.

There are 4 main options for the radio technology standard that is explicitly included in the model:

GSM only This approach attempts to construct cost estimates based on the mature

current technology, which is then assumed to remain in operation in the

long run. A GSM-only approach can be considered conservative because it

may not reflect any productivity gains that might be expected from a move

to next-generation technology – although proxy treatments for the next

generation can be suitably applied to the GSM-only construct.

Including

analogue in past

years

It is possible to make allowances for higher-cost (but nevertheless valid)

technologies in earlier years – such an allowance would involve calculating

technology-specific costs and producing a weighted average cost per

terminated minute (reflecting the balance of minutes carried on analogue

and GSM). However, analogue services are no longer offered in Norway,

and so the weighted average cost would not take into account an analogue

component, and efficient forward-looking costs will be unaffected by

historical analogue operations.



Including UMTS Including UMTS explicitly has added complexity and model detail, and

produces a lower eventual cost estimate in the situation where voice

termination costs are migrating to a lower-cost UMTS technology. The

bottom-up model is significantly more complex as a result of including

UMTS and requires additional supporting top-down cost data for UMTS.

Including

advanced

technologies in

future years

Today’s UMTS (third-generation) networks are characterised by active (but

evolving) high-speed data services (HSDPA and HSUPA).

In the coming years, two additional technologies will continue to be deployed

in Norway:

(third-generation) UMTS900, which utilises re-farmed 900MHz

frequencies to provide wider area coverage than can be achieved with

the current 2100MHz UMTS frequencies.

(fourth-generation) LTE deployments at 2600MHz and other

frequencies – this technology requires a new air interface to be

deployed (as well as new user equipment). However, once deployed,

this technology will allow both significantly increased data traffic

throughput and proper5 mobile voice over IP.

From the perspective of mobile termination regulation, the modern-equivalent technology should

be reflected – that is, the proven and available technology with the lowest cost over its lifetime.

Twenty years ago, the modern-equivalent technology for providing mobile telephony was

analogue (NMT).

At the time of the original cost modelling work in 2006, NPT considered that GSM was primarily

the efficient technology for providing voice termination. Currently, all Norwegian mobile

networks provide both GSM and UMTS voice and data services, and migration of traffic from

GSM to UMTS is proceeding in some way. All UMTS networks in Norway offer HSDPA services

as standard.

At the current point in time (2013), and given the current focus of the model on voice termination,

NPT continue to believe that it is not necessary to explicitly model LTE in principle. This decision

can be attributed to the uncertainty over key aspects of LTE network deployment:

the long-term coverage expected for the networks

the relevant spectrum allocations

the extent of infrastructure sharing between operators.

5 That is, LTE mobile handsets will not have a circuit-switched LTE transmission mode, and voice will be carried over

the air interface as packetized IP traffic.



The forecast increase of LTE services will have an impact on the traffic carried over the 2G and 3G

networks. However, the NPT v8F model does not need to explicitly model the network design for LTE

to consider this, although it does need to consider the voice, SMS and data services that are carried over

2G/3G networks. It may also need to implicitly consider some sharing of infrastructure costs between

(2G+3G) today’s main networks, and future (2G+3G)+LTE networks, for example by applying some

percentage profile of LTE demand into the routeing factors used for cost allocation.

Both the EC and ESA Recommendation indicate that a “model for mobile networks should be

based on a combination of 2G and 3G employed in the access part of the network”, which supports

the approach proposed above.

The recommendation established in 2009 will therefore be revised as follows:

Recommendation 4, revised: Use a model which reflects the operators’ actual GSM and

UMTS networks from 1993 onwards. The model should contain actual GSM traffic and

subscriber volumes and reflect the prices paid for modern-equivalent GSM equipment in

each year. The model should also contain existing UMTS subscribers, traffic, HSPA data

and network equipment, since all Norwegian mobile operators are using UMTS network

infrastructure. The rate of migration from GSM to UMTS will be projected from the latest

actual status of the mobile operators. Deployment of UMTS900 is anticipated in the

situation that GSM networks are shut down. LTE traffic and networks will not be explicitly

modelled, however migration of voice, SMS and high-speed data services to an LTE

network will be included, and some sharing of infrastructure costs to LTE demand may be

included using a proportionate cost allocation to LTE.

2.3.2 Treatment of technology generations

Modelling a single technology network in a long-run cost model provides a simplification of the

multi-technology reality. Mobile network generations are only expected to remain valid for a finite

number of years – a long-run cost model effectively makes predictions of parameters in perpetuity.

Therefore, as operators manage the migration of demand and subscribers from one generation to

the next, so too can an LRIC model make corresponding parametric assumptions.

Three particular areas appear most significant in the context of mobile termination costing:

Migration of

traffic

The migration of traffic from one network to another affects the output profile

produced by the network assets of each technological generation. This changes

the level of unit costs over time for each generation, irrespective of depreciation

method6. The long-run cost from a single technology that can be operated in

perpetuity will be lower than the long-run cost of a technology with a finite

6 Although, of course, the choice of depreciation method determines when and how unit costs change as a result of

migration.



lifetime (provided there are assets which have a higher lifetime output7).

However, a single technology model will not necessarily capture any

productivity gains from moving to the next technology, such as higher system

capacity or greater service demand. Therefore, a single-technology, long-run

cost may be higher than the blended average cost from improving generations

of a mobile cellular technology.

What is important from a cost modelling perspective is to understand the

implications of modelling a single technology network and single technology

demand for the level and timing of cost recovery when contrasted with the

multi-technology situation faced by real mobile operators.

Proxies for change Proxies for factors that change from one generation to the next may be

applied in a cost model to mimic the effects of successive technology

generations. As introduced under ‘migration of traffic’ above, successive

generations of cellular technology can be expected to have measurable

output rises8. Also, the cost per unit of capacity is likely to reflect continued

technological improvement9. The key issue for an LRIC model is

consistency: modelling continual levels of demand growth without

technological evolution (and vice versa) would appear to be inappropriate.

Economies of

scope

A number of network and non-network costs will effectively be shared by

successive generations of technology – in these instances it will be possible to

extract the same (or greater) utilisation from an asset irrespective of the rate or

existence of migration. Certain network assets fall into this category: for

example, base station sites may continue to be rented from one generation to

the next, backhaul transmission may be transparent to 2G and 3G traffic,

business overhead functions will support both technology generations, etc.

Given these economies of scope between technology generations, service

costing for certain assets should be independent of migration.

As discussed in Section 2.3.1 above, it is proposed to model LTE implicitly in the NPT v8F

model, which affects the treatment of technology generations. The recommendation established in

2009 will therefore be reworded as follows:

7 Which is likely to be the case, if there are long-lived assets which are technology specific (e.g. a licence fee).

8 This has been observed for analogue to GSM, and is expected for GSM to UMTS.

9 For example, analogue to digital, TDMA to W-CDMA.



Recommendation 5, reworded: Consistent with Recommendation 4, adopt a consistent set

of long-run forecast parameters: in particular, GSM volumes and GSM equipment prices,

and UMTS volumes and UMTS equipment prices. An increasing proportion of voice traffic

is being carried on UMTS networks in Norway, and migration of data users from GPRS to

UMTS/HSPA networks also results in a (significantly) greater proportion of data traffic

being carried on the next-generation technology. Next-generation technologies should also

enable higher total volumes of voice and data traffic to be carried. According to the current

rate of migration to UMTS, it appears that operators are migrating more slowly than forecast

in the original model. This suggests that the original expectation of GSM shut-down in 2015

is unlikely to be achieved. Therefore GSM shut-down is projected for at least 2020. While

the model considers 3G technology in perpetuity, migration from UMTS to LTE has been

added into the demand calculations.

2.3.3 Extension and quality of coverage

Coverage is a central aspect of network deployment and of the radio network in particular.

Appropriate coverage assumptions to apply to the modelled operator can be determined through

the following questions:

How should historical coverage be reflected?

How far should geographical coverage extend in the long run?

How fast should the long-run coverage level be attained?

What quality10

of coverage should be provided, at each point in time?

The definitions of coverage parameters have two key implications for the cost calculation:

Level of unit costs

due to present

value (PV) of

expenditures

The rate, extent and quality of coverage achieved over time determine the

present value (PV) of associated network investments and operating costs.

The degree to which these costs are incurred before demand materialises

represents the size of the ‘cost overhang’. The larger this overhang, the

higher the eventual unit costs of traffic will be.

Identification of

network elements

and common costs

that are driven by

traffic

In a situation where coverage parameters are relatively large, fewer network

elements are likely to be dependent on traffic. This reduces the sensitivity

of the results to assumed traffic algorithms.

Furthermore, common costs are generally incurred when costs remain fixed

in the long run. With larger coverage parameters specified for an operator,

increasing proportions of network costs are invariant with demand and

hence likely to be common costs.

10

By quality of coverage we are specifically referring to the density of radio signal – within buildings, in hard-to-reach

places, in special locations (e.g. airports, subways, etc.).



For the operator-relevant conceptual issues discussed in Section 2.2, the recommendation

established in 2009 requires rewording to take into account the move from two actual operators

and a hypothetical third operator in the NPT v7.1 model to three actual operators and a generic

efficient operator in the NPT v8F model.

Recommendation 6, reworded: Consistent with Recommendation 2, actual historical

levels of geographical coverage and coverage quality for the three actual network operators

should be reflected in the model. A forecast for future geographical coverage should be

applied in the model, consistent with operators’ planned coverage expansions. Planned

improvements in coverage quality should also be reflected in parts of the network that are

not driven by traffic. A national coverage profile will be applied to the generic efficient

network operator. The GSM and UMTS coverage profiles of the mobile networks should be

modelled separately, taking into account UMTS900 which is being used for eventual full

national coverage by 3G.

2.3.4 Transmission network

A number of factors affect the choice of transmission network used by an operator. These include:

historical demand and network evolution

forecast demand and network evolution

build or buy preference of individual mobile operators

availability of new generations of transmission technology from alternative providers

range and price of wholesale transmission services.

During the upgrade of the model, it will be necessary to analyse differences in network transmission to

carry traffic from the base stations, and to connect switching sites with backbone capacity.

All differences between the modelled operators’ actual networks will have associated cost

implications. Therefore, it will be necessary to identify material transmission differences and

explore the method and rationale for selecting the chosen network transmission.

The recommendation established in 2009 only requires a rewording to capture the fact that the

generic operator can then use the transmission methods as modelled for the actual operators.

Recommendation 7, reworded: Consistent with Recommendation 2, each operator’s

actual transmission network should be modelled, identifying material differences in the

choice, technology or cost of transmission elements but aiming to adopt an efficient, modern

and standardised modelling approach where possible. This standardised approach will then

be applied to the generic operator.



2.3.5 Network nodes

A mobile network can be considered as a series of nodes (with different functions) and links

between them. Of these node types, the most important are sites for base stations, sites for

BSCs/RNCs and sites for switching equipment. In developing algorithms for these nodes, it is

necessary to consider whether the algorithm should and does accurately reflect the actual number

of nodes deployed. In situations where the operators’ network is not viewed as efficient or modern

in design, or where network rationalisation is planned, the model may be allowed to deviate from

the operators’ actual number of nodes. This aspect may be highlighted when looking at GSM and

UMTS networks – since later equipment tends to have a higher capacity and is therefore more

likely to be located in fewer, larger switching sites.

Specification of the degree of network efficiency is a crucial regulatory costing issue, and one

which is sometimes encompassed by the application of a ‘scorched-node’ principle. This ensures

that the number of nodes modelled is the same (exactly or effectively, as required) as in reality,

albeit with modern-equivalent equipment deployed at those nodes. This is coupled with the

commonly held view that mobile networks are generally efficiently deployed and operated due to

infrastructure competition. The main alternative is the ‘scorched-earth’ principle, which allows the

number and nature of nodes modelled to be based on a hypothetical efficient network, even if it

deviates from operational reality.

Adopting a scorched-node principle requires an appropriate calibration of the model, to ensure

node counts correspond with reality. This ensures that the level of assets in the model is not

underestimated due to factors that are not explicitly modelled. The application of network node

adjustments indicates the network efficiency standards which will define the level of cost recovery

allowed through regulated charges.

While the recommendation established in 2009 remains fully applicable regarding the three actual

network operators, it requires some clarification with regard to the treatment of the generic

efficient operator.

For the generic operator, we wish to reasonably reflect the network nodes of the actual operators in

Norway and, as such, we are not using a scorched-earth approach. Instead, we have defined

particular generic operator inputs using the values of the actual operators. As these actual operator

values have been derived using the scorched-node principle, the generic operator will implicitly

reflect the scorched-node principle.

Examples of the generic operator inputs derived in this manner include cell radii for coverage

sites, cell radii for in-fill sites and the number of switching locations.



Recommendation 8, revised: Consistent with Recommendation 2, adopt actual network

designs in terms of numbers of network nodes. The starting point for this will be submitted

data on the number and nature of nodes in operators’ actual networks, which we shall

validate for high-level efficiency with our expert view. In the radio network, we suggest

applying a scorched-node calibration to ensure that the model can replicate operators’ actual

deployed site counts: this effectively ensures that radio network design parameters which

are not modelled explicitly are implicitly captured in the model. The efficient nodes for the

generic efficient operator are defined using the values of the actual operators. As these

actual operator values have been derived using a scorched-node principle, the generic

operator will implicitly reflect the scorched-node principle.

2.3.6 Input costs

To calculate the costs of a mobile network using a bottom-up incremental cost model, the unit

costs of different types of network equipment are a required input. There are four general

approaches, discussed below, that could be taken in defining input costs:

Actual cost This method allows the identification of the unit costs applicable to each

operator in order to develop two complete sets of equipment cost data. This

method, whilst comprehensive, can result in difficulties when trying to

understand reasons for overall cost differences between operators, since there

may be no cross-references between unit costs when populating the two models.

Lowest cost The mobile operators in Norway have strong incentives to purchase and operate

their network equipment at the lowest possible cost. Therefore, it is reasonable

to assume that the price paid by any operator for a given unit of equipment will

be the lowest possible price that the operator could pay, and using any lower

value will result in the operator being unable to recover its full costs. Using the

lowest unit costs carries the risk of underestimation of costs, since:

one operator might have access to lower unit costs that cannot be

replicated by the other operator

a lower unit cost in one category might be balanced by a higher unit cost

in another

the efficient unit cost might not necessarily be the lowest, as there are

other considerations involved in a real purchasing decision (e.g. ties to

maintenance contracts, vendor selection, etc.).

Highest cost Using the highest unit costs has the same potential problems as using the

lowest unit costs, but leading to a risk of overestimating cost.



Average cost Given the staggered nature of network deployment, the price paid for any given

unit of equipment by each operator at any given time will naturally vary.

However, the discipline of competition in the retail market should mean that all

operators aim to minimise their costs over the long term. Therefore, using

averaged unit costs should produce an efficient overall network cost.

A further advantage of using average costs is that it avoids adhering

dogmatically to a particular principle (e.g. lowest or highest cost), which can

be unreasonable under certain circumstances, and instead provides a

reasonable, practicable alternative.

The recommendation established in 2009 has been reworded for the NPT v8F model to apply to

the generic operator rather than the third operator.

Recommendation 9, reworded: Given the practical and regulatory difficulties of

accurately and unambiguously defining the lowest cost base for an operator, we recommend

a mixed approach based on actual and average costs. Our starting point for assessing the

level of input costs will be the actual costs incurred by the operators – informed by data

submitted by the operators. Where it can be shown that unit costs equate closely to the same

functional network elements (e.g. a BSC of the same capacity), we shall endeavour to use

average costs applicable to all operators. Where it can be shown that each operator has a

materially different unit cost base (e.g. in the price of a suite of equipment from a particular

vendor), then operator-specific actual costs will be adopted. Efficient unit costs will need to

be estimated for the generic operator model, without revealing confidential operator data.

2.3.7 Spectrum situation

Actual mobile operators’ spectrum allocations – in terms of amount11

, band12

and any fees13

paid –

and use of their allocated spectrum, are likely to differ. Some of these differences may be assessed

to be outside of the operators’ control – e.g. restrictions on the availability and packaging of

spectrum over time.

Any cost differences arising from these spectrum allocations or use should be understood and

estimated, and could be taken into account in the cost basis of regulated prices if appropriate (and

significant). This involves understanding how the differences in operators’ spectrum result in

different network deployments, how these are best captured and parameterised in the model, and

ultimately what the resulting cost differences are. The benefit of being able to model the actual

spectrum of the operators is that it greatly assists manageable scorched-node calibration of a

bottom-up network design with actual data, and reconciliation of calculated costs with actual costs.

11

Amount of paired MHz, less guard bands.

12 PGSM, EGSM or DCS.

13 One-time or recurring fees, including duration of any licence payment.



Alternatively, some hypothetical amount of spectrum could be defined –this would require a clear

understanding of the cost differences between this hypothetical allocation and the actual operator

allocations. It would be possible to attempt to construct a purely hypothetical spectrum model without

clear reference to actual operator factors. This hypothetical approach could, for example, be defined

assuming that the generic operator has an “average” allocation of spectrum in the Norwegian market.

The recommendation established in 2009 has been revised in order to specify the methodology used for

the allocation of spectrum to the generic efficient operator in the NPT v8F model. Additional text has

been included to explicitly consider the principle of future licence renewals. In the v7.1 model, all

licences were renewed periodically, with renewal fees assumed to increase with inflation.

Recommendation 10, revised: Develop a model capable of capturing the network and cost

differences due to the actual operators’ spectrum allocations, through modification of a

small number of key parameters. It is expected that spectrum differences are negligible for

Telenor and TeliaSonera. Generic spectrum allocations will be developed/defined for the

generic operator. Our principled position with regard to future licence auctions/renewals is

not to pre-empt any future expected value or allocation and therefore to retain the current

modelling approach of regularly repeating the existing spectrum allocations and applying

inflation-increasing payments.

2.4 Service-related conceptual issues


Figure 2.4: Decisions on the service-related conceptual issues taken for the NPT v7.1 model [Source: Analysys

Mason, 2013]


[11] Service set Both voice services and non-voice services Revise

[12] Wholesale or retail Apply a 75:25 split of overhead costs No change

2.4.1 Service set

The treatment of economies of scope achieved by the actual voice and data operators depends on

whether the modelled operator offers non-voice SMS, GPRS, EDGE and HSPA services to its

subscribers. Economies of scope arising from the provision of these services across a shared

infrastructure should result in a lower unit cost for voice services where total traffic volumes are

higher. The standalone network costs (e.g. hardware and software) incurred by the operators – and

therefore likely to be reflected in the model – implicitly include the support for non-voice services.

Assessing both voice and data services in the model increases the complexity of the calculation and the

supporting data required, and should result in a lower unit cost for voice services due to economies of

scope. Conversely, however, excluding costs relevant to non-voice GSM services (and developing a

standalone voice cost) can also be complex. In Norway, some non-voice services (e.g. SMS and GPRS)

are reasonably proven services rather than emerging services. In the case of HSPA, traffic volumes



have grown rapidly – therefore a conservative approach to forecasting future data traffic may be

appropriate if suggested economies of scope are significant (subsequently strongly reducing the

economic cost of voice on the basis of an uncertain data traffic forecast).

Recommendation 11 as established in 2009 refers only to conventional GSM and UMTS services.

It therefore requires revision to indicate that the NPT v8F model now includes forecasts for

additional services, namely:

LTE data megabytes

OTT variants of voice services

OTT variants of SMS services.

Recommendation 11, revised: The modelled operator should provide data services (SMS,

GPRS, EDGE, HSPA and LTE) alongside voice services. The modelled operator will

additionally provide OTT variants of voice and SMS services that will be carried over the

network as high-speed data (HSPA and LTE). The associated economies of scope will be

shared across all services, although care will be taken where uncertain growth forecasts

significantly influence the economic cost of voice. The approach to allocating costs between

voice and UMTS data services (particularly HSPA) will be carefully examined during the

implementation of Recommendation 15 (choice of increment) since there is likely to be a

much larger proportion of traffic from data services in today’s networks (compared to four

years ago when data accounted for less than 5% of network traffic).

2.4.2 Wholesale or retail

In a vertically separated model, network services (such as traffic) are costed separately from

retail activities (such as handset subsidy or brand marketing). Business overheads are then marked

up between network and retail activities, and the wholesale cost of supplying mobile termination is

only concerned with the costs of the network plus a share of business overheads.

In a vertically integrated model, retail costs are considered integral to network services and

included in service costs through a mark-up, along with business overheads.

To date, NPT has identified its market analysis as that relating to the wholesale call termination market.

As such, NPT intends to consider only those costs that are relevant to the provision of the wholesale

network termination service in a vertically separated business. However, costs that are common to

network and retail activities will be recovered from wholesale network services and retail services. This

will be treated as a mark-up on the LRIC (though excluded by definition from the Pure LRIC).

A vertically separated approach results in the exclusion of many non-network costs from the cost of

termination. However, it brings with it the need to assess the relative size of the economic costs of

retail activities in order to determine the magnitude of the business overheads to be added to the

incremental network costs.



The Recommendation as established in 2009 has been left unchanged.

Recommendation 12, unchanged: Consistent with the original model, we propose to

maintain the indirect cost treatment of business overhead expenditure. This allocation results

in an approximately 75:25 split between network and retail activities respectively. In the

upgraded model, retail costs will not be remodelled; instead the 75:25 split of overhead

costs will be applied as an exogenously defined cost allocation.

2.5 Implementation-related conceptual issues


Figure 2.5: Decisions on the implementation-related conceptual issues taken for the NPT v7.1 model [Source:



[13] WACC Apply NPT’s mobile operator WACC No change

[14] Depreciation method Economic depreciation No change

[15] Increments Calculate LRIC, Pure LRIC and LRIC +++ costs Reword

[16] Years of results All relevant past and future years (i.e. from

1992)

No change

[17] Mark-up mechanism Equi-proportionate mark-up (EPMU) No change

2.5.1 WACC

The appropriate level of return to be allowed on regulated services is a standard aspect of

regulatory cost modelling. The level of WACC has a direct, material effect on the calculated cost

of termination, but it does not need to be applied in the model until the final costing stages. The

Recommendation as established in 2009 does not need to be changed.

Recommendation 13, unchanged: Update NPT’s mobile operator WACC calculation.

2.5.2 Depreciation method

The model for mobile network services will produce a schedule of capital and operating

expenditures. These expenditures must be recovered over time, ensuring the operator can also earn a

return on investment. There are four main potential depreciation methods:

historical cost accounting (HCA) depreciation

current cost accounting (CCA) depreciation

tilted annuity

economic depreciation.

Economic depreciation is the recommended approach for regulatory costing. The table below

shows that only economic depreciation considers all potentially relevant depreciation factors.



Figure 2.6: Factors considered by each depreciation method [Source: Analysys Mason, 2013]

HCA CCA Tilted annuity Economic

Modern-equivalent asset (MEA) cost today

Forecast MEA cost

Output of network over time

Financial asset lifetime 14

In a mobile network cost model where demand varies over time (e.g. for an actual operator), results

produced using tilted annuity will differ significantly from economic depreciation. The difference

between HCA and CCA depreciation is inclusion of modern-equivalent asset prices – which is

applied in the calculation as supplementary depreciation and holding gains/losses. The difference

between HCA and CCA is generally uninteresting, in the light of more significant differences

between HCA and economic depreciation.

Economic depreciation is a method for determining a cost recovery that is economically rational,

and therefore should:

reflect the underlying costs of production: i.e. modern-equivalent asset (MEA) price trends

reflect the output of network elements over the long run.

The first factor relates the cost recovery to that of a new entrant to the market (if that market were

competitive), which would be able to offer the services based on the current costs of production.

The second factor relates the cost recovery to the ‘lifetime’ of a mobile business, in that investments

and other expenditures are in reality made throughout the life of the business (especially large, up-

front investments) on the basis of being able to recover them from all demand occurring in the

lifetime of the business. All operators in the market are required to make these large up-front

investments and recover costs over time. These two factors are not reflected in HCA depreciation,

which simply considers when an asset was bought, and over what period the investment costs of the

asset should be depreciated.

The implementation of economic depreciation to be used in the model is based on the principle that all

(efficiently) incurred costs should be fully recovered, in an economically rational way. Full recovery of

all (efficiently) incurred costs is ensured by checking that the PV of actual expenditures incurred is

equal to the PV of economic costs recovered. An allowance for a return on capital employed, specified

by the WACC, is also included in the resulting costs.

14

Economic depreciation can use financial asset lifetimes, although strictly it should use economic lifetimes (which

may be shorter, longer or equal to financial lifetimes).



The Recommendation established in 2009 will remain unchanged.

Recommendation 14, unchanged: NPT intends to retain the original model’s economic

depreciation calculation to recover incurred network expenditure over time, with a cost

recovery in accordance with MEA price trends, network output over the long run, and the

discount rate. In addition, for comparative purposes only, a straight-line accounting

depreciation calculation will also be applied in the model. Further details of economic

depreciation are supplied in the Annex, but operators have the opportunity to comment on

the implementation of economic depreciation in the draft model released to industry during

this consultation process.

2.5.3 Increments

Increments in a cost model take the form of a service, or set of services, to which costs are

allocated, either directly (for incremental costs) or via a mark-up mechanism (if common costs are

to be included). Specifically, the model constructed is used to gain an understanding of how costs

vary, or are fixed, in response to different services. This enables costs to be identified as either

common or incremental. In final costing stages, common costs may be marked up onto the relevant

increments.

The size and number of adopted increments affects the complexity15

of results and the magnitude16

of the marked-up incremental costs.

Incremental costs should in practice be determined by calculating the difference in costs with and

without the increment present. Subsequently, calculating the difference in costs with and without

combined increments would determine the precise structure of costs that are common to the various sets

of increments. An incremental costing approach that runs through this complete set of small increment

permutations can give rise to very complex results, which must be resolved carefully to ultimately

identify marked-up incremental costs. However, calculating the incremental cost of only a single

increment simply requires the model to calculate with and without the defined increment.

Where increments include more than one service, rules will need to be specified to allocate the

incremental costs to the various component services. These allocation rules could be on the basis

of average loading, peak loading or some other method. Increments which combine

distinguishable services such as voice traffic, SMS traffic and GPRS traffic will need carefully

assessed routeing factors for allocating costs to the services – since in this combined increment

approach it is through routeing factors, rather than network algorithms, that non-voice service

incremental costs are identified.

15

More increments = more calculations required of the model and more common costs (or a larger aggregate common

cost) to deal with in the mark-up.

16 Through the mark-up mechanism.



Most of the costs associated with a mobile network are driven by traffic (i.e. it is the marginal

increase in traffic that drives the marginal increase in cost). However, this is not the case for a

subset of network costs that are driven by the number of subscribers. These costs typically include

the visitor location register (VLR) and home location register (HLR), which principally function as

databases of subscribers and their locations, plus the switching costs associated with the service of

periodically updating the location of all active subscribers.

Whilst the network cost of updating the HLR and reporting the location of handsets is dependent

on subscriber numbers, there is a precedent in Europe for recovering these costs through received

calls (which should therefore include on-net voice and also SMS delivery). This is because

location updates and interrogating the VLR/HLR for subscriber location are only required for

terminating traffic – and can be considered a common cost for all terminated traffic.

The magnitude of incremental costs, and costs common to increments, depends on the interaction of the

number and nature of increments with the cost functions of network elements. More complex

increments will require network design algorithms that are cognisant of relevant volume components.

Applying a combined traffic increment implies focus on the routeing factors which share out

traffic costs – particularly the degree to which SMS and data traffic load the network (or are

accommodated by it in other ways, such as channel reservations).

Applying small increments implies a focus on the network design algorithm at the margin, and the

degree to which capacity-carrying elements vary in the long run with the variance of different

traffic types. The NPT v7.1 model is capable of calculating the incremental costs of wholesale

termination (which we have referred to as the “Pure LRIC”) by either:

including or excluding technical network design adjustments

applying economic depreciation to the avoided cost of termination traffic, or calculating the

difference of the economic depreciation when including or excluding termination traffic.

Any combination of these two effects can currently be calculated. However, the NPT v8F model is

intended to focus on the Pure LRIC calculation that includes network adjustments and

calculates economic depreciation of the avoided costs.

The Recommendation established in 2009 has been reworded to emphasise this approach.

Recommendation 15, reworded: In order to supply NPT with the range of potential costs,

which it may apply to wholesale termination regulation, the model should calculate both

LRIC+++ and LRIC results. Accordingly, the original model LRIC+++ method will be

updated to include the relevant UMTS aspects, whilst the ESA Recommendation will be

used to define an avoidable cost calculation (‘Pure LRIC’) approach to the wholesale mobile

termination service. In the Pure LRIC case, we shall explore the sensitivity of the result to

the technical assumptions that are applied in the model to estimate the difference in costs

without mobile termination volumes. Specifically, it will be possible to include appropriate

network design adjustments. Economic depreciation will be applied to the avoided

expenditures of terminated voice.



2.5.4 Years of results

There are three options for timeframes for the calculation:

One year only

(e.g. 2009)

This approach can simply compare costs today with prices today.

Forward-looking

only (e.g. 2009

onwards)

A forward-looking calculation is capable of answering questions about the

future, but is difficult to reconcile with the past (and therefore, potentially,

the present).

All years

(e.g. 1992

onwards)

Having a calculation for all years will make it easier to use full time-series

data and consider all costs over time. It provides the greatest clarity within

the model as to the implications of adopting economic depreciation

(compared to other forms of depreciation).

The calculation of mobile termination costs in particular years provides a range of information:

current-year costs can be compared to current-year prices

forecast costs can be used to define RPI-X price caps

a full time series of costs can be used to estimate windfall losses/gains due to a change from

historical to accounting cost paths and provides greater clarity as to the recovery of all costs

incurred from services over time.

Analysys Mason’s experience of bottom-up LRIC models, and their use in conjunction with top-down

information, indicates that a full time-series model provides:

the greatest clarity and confidence in results, particularly when it comes to reconciliation

against historical top-down accounting data

the widest range of information with which to understand how the costs of the operators vary over

time and in response to changes in demand/network evolution

the opportunity to include additional forms of depreciation (such as accounting depreciation)

with minimal effort.

The Recommendation established in 2009 has been reworded to capture the current approach used.

Recommendation 16, reworded: NPT proposes to adopt a full time-series model that

calculates the costs of all three actual operators from their GSM launches in 1993 and 2009

(and capturing the first GSM expenditure in 1991 and 1992 where relevant), following on to

UMTS deployments in 2001 and beyond. The model will therefore be able to calculate

operators’ costs in current and future years, giving NPT the greatest understanding of cost

evolution and flexibility in exploring pricing options. The third generic operator will be

modelled according to a recent entry date, in a full time-series approach that considers its

current and future all years of operation after launch.



2.5.5 Mark-up mechanism

The specification of an LRIC+ model will result in certain cost components being classified not as

incremental, but as common costs. Common costs are those costs required to support one or more

services, in two or more increments, in circumstances in which it is not possible to identify which

specific increment causes the cost. Such costs do occur in mobile networks (and more extensively

in mobile business overheads). However, depending on the maturity of the network, they may not

be as significant as in a fixed network. These common costs need to be recovered from services in

some way, generally by using a mark-up on incremental costs in an LRIC+ situation.

Two main methods for mark-up mechanism are put forward and debated in the context of mobile

termination costing:

Equal

proportionate

mark-up (EPMU)

In this method, costs are marked up pro-rata to incremental costs. It is

simple to apply, and does not rely on developing additional supporting

information to control the mark-up calculation. EPMU has been applied by

Ofcom and PTS in their previous mobile cost calculations.

Ramsey pricing,

and its variants

Ramsey pricing is a targeted common-cost mark-up mechanism which loads

the burden of common-cost recovery on those services with low price elasticity

(thus least distorting consumer consumption and welfare away from the

optimal). Variants exist on Ramsey pricing methods which take into account

operator profit (as opposed to welfare) maximising incentives, or additional

effects such as network externalities. Supplementary information is required by

these approaches to control the mark-up algorithms.

The choice of mark-up mechanism affects the resulting marked-up unit costs, particularly where

non-equal mark-ups are applied, and especially if common costs are large. This choice therefore

directly influences the cost-oriented price for mobile termination.

The Recommendation established in 2009 will not be changed.

Recommendation 17, unchanged: NPT proposes to apply an equi-proportionate mark-up

(EPMU) for network common costs and the network share of business overheads in the

LRIC+++ calculation.



3 Demand forecasting

Since the development of the NPT v7.1 model, certain new trends and technologies have emerged

in the mobile market, which have required a reassessment of the modelling of the demand

forecasts. The most significant of these changes have been:

the global development of fourth generation (4G, or LTE) mobile networks and devices

an increase in adoption of over-the-top (OTT) services, such as mobile IP telephony and

mobile VoIP, and similar services for SMS, by mobile users as an alternative to traditional

voice and SMS messaging.

The details of our adaptations of the demand forecasts to encompass these changes are discussed

below:

Section 3.1 discusses the implicit modelling of LTE traffic and services in the NPT v8F model

Section 3.2 discusses the inclusion and modelling of OTT traffic.

In addition, data from NPT, operators and publicly available data sets has been used to update

demand parameters in the NPT v8D/F models:

Section 3.3 details the updates made to historical demand parameters

Section 3.4 discusses the changes made to forecast demand parameters.

3.1 LTE demand forecasting

Following the development of the NPT v7.1 model in 2010, both Telenor and TeliaSonera have

begun deploying LTE networks, rolling out in a number of the main cities in Norway.17

The ESA Recommendation18

explicitly states that “the bottom-up model for mobile networks

should be based on a combination of 2G and 3G employed in the access part of the network”.

Therefore, the NPT v8F model does not directly model the costs of LTE services.

In the Swedish mobile LRIC model, PTS modelled an urban LTE network to consider the impact

of aspects such as the sharing of network costs of sites or backhaul transmission between 2G/3G

and LTE networks. The final version of the model issued in July 201119

indicated that, for a 2G/3G

network operator:

17

Based on Telenor’s website, LTE coverage is currently limited to parts of Oslo, Stavanger, Trondheim, Bergen,

Kristiansand and near Frederikstad. Based on TeliaSonera’s website, coverage appears to be more extensive and includes much of the coast of south-east Norway and other areas in central Norway.

18 See http://www.eftasurv.int/media/internal-market/ESAs-Recommendation-on-termination-rates.pdf.

19 See http://www.pts.se/upload/Remisser/2011/Telefoni/10-8320-pts-mobil-lric-final-model.zip.



including an urban LTE network covering 30% of the population reduced the LRIC+++ of

mobile termination by approximately 5% in the long run

including an urban LTE network covering 30% of the population had almost no impact

(<0.5%) on the Pure LRIC of mobile termination.

From these results, we have concluded that the considerable additional complexity of

implementing an LTE network design, in addition to the existing 2G/3G network designs, is not

proportionate to the impact of LTE networks. Therefore, we do not explicitly model the network

design for LTE, though we do consider its share of voice, SMS and data services. We also assume

the possibility of a percentage of LTE demand to be passed into the routeing factor table for shared

infrastructure supporting 2G+3G and LTE network layers (effectively, radio sites). This estimates

the effects of cost sharing between services.

The migration of voice, SMS and high-speed data services to an LTE network was added into the

demand calculations in the NPT v8D model and retained for the NPT v8F model. The changes

made and calculations used for deriving the services carried over the LTE network are found on

the D3_M8D and D3_M8F worksheets and are discussed in more detail below.

3.1.1 Voice and SMS demand forecast updates

The previous NPT v7.1 model’s voice and SMS forecasts were derived on a total volume,

technology-neutral basis. To account for the proportion of voice and SMS that will be moved

across onto the LTE network in the future, a similar methodology is used as for the migration of

services from 2G to 3G networks. As a result, the total voice/SMS traffic across all technology

generations remains largely the same as in the NPT v7.1 model, with only the proportion of this

traffic for each generation updated in the NPT v8F model.

The traffic demand calculations have been updated by introducing 3G to LTE voice and SMS

migration profiles to the D3_M8D and D3_M8F worksheets of the model. These migration profiles

are derived such that voice and SMS traffic on the 3G network remains largely stable throughout

the forecasting period. A start date of 2015 has been used for the beginning of migration of voice

and SMS services to LTE, given the likely timescale for VoLTE and IMS deployment by

operators.

These new traffic migration profiles feed into the calculations of 3G and LTE voice demand

forecasts as can be seen in Figure 3.1 below, with the specific calculations as follows:



This structure is replicated for the calculation of SMS traffic across technology generation.

Figure 3.1: Illustration of the voice traffic calculations in the NPT v8F model [Source: Analysys Mason, 2013]

3.1.2 Cost sharing with LTE

The model is capable of including a proportion of LTE megabytes in the routeing factors of

network assets which are likely to support 2G, 3G and LTE radio infrastructure (effectively radio

sites). This has the effect of reflecting (in a lower LRIC and LRIC+++ result) greater economies of

scope which can be anticipated by 2G+3G+LTE combined network infrastructure.

As the LTE network is not explicitly modelled, the pure LRIC of the wholesale voice termination

increment in a 2G+3G+LTE network model is not calculated (this result is effectively only

calculated in the 2G+3G case present in the network design algorithms).


2G voice 3G voice LTE voice

3G to LTE voice

migration profile

2G to 3G voice

migration profile

Digital mobile

penetration

Population

Market subscriptions

Operator market

shareOperator

subscriptions

Minute/sub/month

(excluding OTT)

Minutes/year

(excluding OTT)

NPT market traffic

information

Operator OTT

proportions



3.1.3 High-speed data demand forecast updates

While the NPT v7.1 model contained forecasts for total market voice minutes and SMS, the

forecast high-speed data traffic considered only megabytes carried over 3G (specifically, the

HSPA) networks20

.

Inputs for the total high-speed data traffic (across all technologies) have been added to the model

with these figures derived from historical operator data on the proportion of total mobile

broadband traffic carried over LTE and year-on-year growth in data usage per connection derived

from NPT market data . Our modelled migration of data traffic to the LTE network begins in 2009,

reflecting the launch of TeliaSonera’s LTE network.

The LTE high-speed data traffic is therefore calculated as the difference between this total

NPT v8F model forecast and the existing NPT v7.1 model HSPA forecast, which has been forecast

to remain stable from 2016 onwards, as illustrated in Figure 3.12. The details of this calculation

methodology can be seen in Figure 3.2 below.

Figure 3.2: Illustration of the high-speed data traffic demand calculations in the NPT v8F model [Source:


20

Low-speed data is assumed to be those megabytes carried over the GPRS/EDGE and UMTS R99 networks.


High speed

subscriptions

3G high-speed data

megabytes

LTE high speed data

megabytes

LTE data migration

profileTotal high speed

megabytes

OTT voice and SMS

megabytes

High-speed

subscriber data

usage

HSPA subscriber

data usage

Total high-speed

data growth



3.2 OTT traffic

OTT services are carried by third-party clients using data bearers. This traffic is therefore not

interconnected via voice gateways since it is carried as data bits. Therefore, operators do not

necessarily know the minutes/messages carried as OTT. OTT services are expected to become

more widespread in Norway and are therefore likely to affect the demand forecasts of circuit-

switched traffic within the model.

In the future, substitution may occur for conventional mobile voice and, similarly, usage of data

messaging could increase at the expense of conventional SMS usage. This means that more

voice/messages are likely to be carried as data bits in the network.

In our consideration of OTT voice and SMS traffic, we have continued to forecast total voice

usage by service, and total SMS by service. We have then separated the OTT voice and SMS

traffic out from the total, technology-neutral, traffic projections in the model using a modelled

proportion of this traffic in each year that is carried by OTT. This proportion is derived from

operator data, as well as information from NPT’s “The population's use of electronic

communications in 2011” survey.21

Figure 3.3 below indicates that in 2011 few users surveyed made regular use of OTT services. This

suggests very low current levels of take-up for OTT services in Norway, and this conclusion is

supported by operator data. Use of OTT services is expected to increase rapidly however, with the

proliferation of smartphones and development of various OTT services such as iMessage,

GoogleTalk, FaceBook messaging, mobile Skype, etc.

Frequency of OTT service

use

Voice services Messaging

services

Figure 3.3: Results of the

NPT “The population's

use of electronic

communications in 2011”

survey [Source: NPT,

2011]

Daily 0.40% 2.96%

Weekly 2.67% 2.54%

Other usage 4.95% 9.07%

No usage 91.98% 82.52%

Therefore, we have used a conservative forecast for OTT take-up in the NPT v8F model, with a

slow increase in the proportion of voice/SMS traffic carried as OTT to 15% in the long term.

The OTT traffic is then converted to high-speed megabytes and included in the modelled service

demand as HSPA and LTE traffic. The conversion rate used for the OTT voice traffic (average of

30kbit/s and 100kbit/s) is derived from the Skype figures for both the minimum and recommended

download and upload speeds for calling, as reported on its website.22

Meanwhile, the conversion

factor for OTT SMS traffic uses the bytes per SMS factor from the NPT v7.1 model.

21

See http://data.norge.no/data/befolkningens-bruk-av-elektroniske-kommunikasjonstjenester-2011.

22 See https://support.skype.com/en/faq/FA1417/how-much-bandwidth-does-skype-need.



The calculated megabytes of data traffic are then included into the high-speed data traffic

forecasts. However, the relationship between OTT voice traffic and high-speed data traffic is not

one-to-one, with an on-net call requiring both an upload and a download of the call data for each

party. Similarly, both incoming and outgoing calls require the network to upload and download the

call data. The resulting data traffic flows can be seen in Figure 3.4 below.

Figure 3.4: Data traffic generated by OTT voice calls [Source: Analysys Mason, 2013]

As a result of the difference in treatment of data traffic to voice traffic, the OTT voice traffic is

included in the high-speed data forecasts as follows:

SMS traffic, conversely, behaves in a similar manner to data traffic, and an on-net OTT SMS is

both downloaded and uploaded once, while an incoming OTT SMS is downloaded once and an

outgoing OTT SMS uploaded once. These traffic flows are shown in Figure 3.5 below.

Own operator

network

Own operator

network

On-net calls

Incoming or

outgoing off-

net calls

KEY:Downlink

high-speed data

Uplink

high-speed data

Off-net

traffic



Figure 3.5: Data traffic generated by OTT messages [Source: Analysys Mason, 2013]

The treatment of OTT SMS traffic means that we use the following formulae in our mapping of

OTT SMS traffic onto the modelled high-speed demand forecasts:

The changes made and calculations used for deriving the OTT traffic and megabytes are found on

worksheets D3_M8D and D3_M8F, and illustrated for voice traffic in Figure 3.6 below. The

structure of the equivalent calculations for OTT SMS traffic is identical.

Own operator

network

Own operator

network

On-net SMS

Incoming

SMS

Own operator

networkOutgoing

off-net SMS

KEY:Downlink

high-speed data

Uplink

high-speed data

Off-net

traffic

Off-net

traffic



Figure 3.6: Illustration of the OTT voice calculations in the NPT v8F model [Source: Analysys Mason, 2013]

3.3 Updates of historical demand parameters

Historical demand updates for the years 2009–12 were provided both in the NPT market data and

by operators in response to data requests. These were used in the update of historical demand

parameters undertaken for the NPT v8D model, and retained in the NPT v8F model. The data

received was used in the assignment of mobile service providers to a host mobile network operator

(MNO) i.e. Telenor, TeliaSonera or Mobile Norway.

A new demand data worksheet (D3_M8D) was included in the NPT v8D model and updated in

order to align demand inputs (mainly for Telenor, TeliaSonera and Mobile Norway) with the NPT

market data. An additional demand data worksheet (D3_M8F) was included in the NPT v8F model

to allow transparency of further demand modelling changes made, following operators’ comments

on the NPT v8D model. The model can be configured using any of the four market calculations.

Details of the demand parameters that have been updated since the NPT v7.1 model are shown in

Figure 3.7 below.


OTT voice HSDPA

megabytesOTT voice HSUPA

megabytes

OTT minutesVoice to data

conversion factorSkype voice bit rates

Operator

subscriptions

OTT minutes as % of

total

Incoming

minutes/sub/month

including OTT

Total outgoing voice

including OTT

Total outgoing

minutes/sub/month

including OTT

Total incoming voice

including OTT



Figure 3.7: Historical demand parameters in the NPT v8F model updated for the years 2009–12 [Source:


Input Operators updated Source used for data update

Digital mobile penetration Total mobile market NPT market data

Market share of high-speed

subscriptions by operator

All operators NPT market data

Mobile broadband penetration Total mobile market NPT market data

Market share by operator All operators NPT market data

Outgoing voice minutes per

subscriber per month

All operators NPT market data

On-net minute proportion All MNOs NPT market data

Incoming voice minutes per


All MNOs Both operator data and NPT

market data

Outgoing SMS per subscriber

per month

All MNOs Operator data

On-net SMS per subscriber per

month

All MNOs Operator data

Incoming SMS per subscriber

per month

All operators Both operator data and NPT

market data

Low-speed data MB per


All operators Operator data

Mobile broadband HSDPA

megabytes per high-speed

subscription per month

All operators Operator data

3.4 Updates of forecast demand parameters

The population year-end historical data and forecasts have also been updated for the

years 2009–41 using data from Statistisk Sentralbyrå (SSB).23

This forecast is consistent with that

contained in NPT’s v1.6 fixed model.

These changes made to the 2009–12 parameters in the NPT v8F model discussed in Section 3.3

have resulted in revisions being made to some of the long-term demand forecasts. These are shown

in Figure 3.8 below.

23

See http://www.ssb.no/befolkning/



Service v7.1 monthly

forecast

v8F monthly

forecast

Figure 3.8: Modelled

long-term forecast

endpoints in the NPT

v7.1 and v8F models

[Source: Analysys

Mason, 2013]

Digital mobile

penetration

115% 115%

Mobile broadband

penetration

20% 20%

Originated voice 240 min/sub/month 190 min/sub/month

Incoming voice 125 min/sub/month 110 min/sub/month

Incoming SMS 50 SMS/sub/month 50 SMS/sub/month

Outgoing SMS 70 SMS/sub/month 70 SMS/sub/month

Low-speed data

usage

10 MB/sub/month 100 MB/sub/month

HSPA data usage 1000 MB/sub/month 1400 MB/sub/month

The most significant of the updated forecasts since the NPT v7.1 model are described in more

detail below, namely:

Section 3.4.1 discusses the modelled population forecasts

Section 3.4.2 discussed the modelled mobile broadband penetration forecasts

Section 3.4.3 discusses the modelled high-speed data traffic forecasts.

3.4.1 Population

The long-term projected endpoint of 5 million used in the v7.1 model was exceeded in March 2012,

according to SSB.23

The v8F model projects that the population will continue to grow, reaching 6.461

million in 2041, rather than stabilising at 5 million as was assumed in the v7.1 model. As stated above,

this is consistent with the population forecast used in NPT’s v1.6 fixed model.

Figure 3.9: Population

forecasts in the NPT

v7.1 and v8F models

[Source: Analysys

Mason, 2013]

0

1

2

3

4

5

6

7

Ye

ar-

en

d p

op

ula

tio

n (

mill

ion

)

v7.1 v8F



As a result of this increased population forecast, total voice traffic carried by the mobile networks in the

NPT v8F model is actually higher than that in the v7.1 model despite the reduction in voice usage per

subscriber per month described in Figure 3.8. While there is population growth forecast across all

Fylker, it is more rapid in Fylker such as Oslo and Akershus as can be seen in Figure 3.10 below.

Forecast compound annual

growth rate (CAGR) to 2020

Fylker Figure 3.10: Compound

annual growth rate of

population forecasts by

Fylker for the years

2012–2020 [Source:

Statistisk Sentralbyrå

(SSB), 2013]

1.51% – 1.00% Akershus, Aust-Agder, Buskerud,

Oslo, Rogaland

1.01% – 1.50% Hordaland, Møre og Romsdal,

Østfold, Sør-Trøndelag, Vest-Agder,

Vestfold

0.51% – 1.00% Finnmark, Hedmark, Nord-Trøndelag,

Oppland, Sogn og Fjordane,

Telemark, Troms

0.01% – 0.50% Nordland

3.4.2 Mobile broadband penetration

While the long-term demand forecast for mobile broadband penetration in both the v7.1 and v8F

models remains the same at 20%, penetration in the v8F model is forecast to fit with the updated

data points for the years 2009–12 and reaches this endpoint more slowly. The modelled year-on

year growth rates for mobile broadband penetration after 2012 has been cross-checked with NPT

market data. The mobile broadband penetration forecasts are shown in Figure 3.11.

Figure 3.11: Mobile

broadband market

penetration in the NPT

v7.1 and v8F models

[Source: Analysys

Mason, 2013]

0%

5%

10%

15%

20%

25%

Pe

ne

tra

tio

n

v7.1 mobile broadband v8F mobile broadband



3.4.3 High-speed data traffic

The growth forecasts for total high-speed data traffic discussed in Section 3.1.2 result in a rapid

increase in modelled data traffic. While the proportion of this high-speed data traffic carried by

LTE networks increases during the modelled time period, as shown in Figure 3.12, the long-term

demand forecasts for HSPA services in the NPT v8F model are set such that traffic over HSPA

remains stable at approximately 20 billion megabytes from 2016 onwards.

Figure 3.12: High-

speed data

consumption in the NPT

v8F model [Source:


0

10

20

30

40

50

60

70

80

Hig

h s

pe

ed

da

ta M

B (

bill

ion

)

Total high-speed data (v8F) HSPA (v8F)



4 Calculations related to the EC/ESA Recommendations

Both EC24

and ESA25

have released recommendations regarding the costing calculations for

mobile termination rates. A number of adjustments have been made to both the NPT v7.1 model

and the NPT v8F model to consider these recommendations:

Section 4.1 outlines the structure of the addition to the NPT v8F model of a generic operator

Section 4.2 discusses the definition of the inputs for this generic operator

Section 4.3 describes the ‘Pure LRIC’ calculation included in the NPT v8F model.

Section 4.4 sets out the existing LRIC and LRIC+++ calculations in NPT’s model.

4.1 Structure of the generic operator calculation

The NPT v8F model models a generic Norwegian operator in addition to the three actual MNOs.

The modelling of a generic operator is outlined in the ESA released recommendation for the

costing of termination rates, which recommends modelling an efficient-scale operator (by

implication, not an actual operator). This is very similar to the EC’s Recommendation of

May 2009.26

To create a generic operator calculation, the inputs can be determined as a function of the inputs

from the actual MNOs (Telenor, TeliaSonera and Mobile Norway). These actual operator inputs

have been calibrated and reconciled to the most recent year of available operator data (2011 or

2012) and are related to:

demand e.g. subscribers, traffic

network design e.g. cell radii, mix of backhaul topologies

costs e.g. unit capex, cost trends, lifetimes, etc.

The generic operator can be calculated by choosing ‘Generic_operator’ on the A0_CTRL

worksheet. The inputs are then selected on the A9_M, A6_NtwDesSlct, A8_UtilIn, and

D4_CostBase worksheets. When distributed to operators and published, the generic operator inputs

will be given as ‘pasted values’ in the Excel worksheets. This is because the worksheets with

confidential operator-specific data (where the generic operator input calculations are undertaken)

are redacted prior to distribution of the model. However, as part of the redaction, we leave a note

of formulae used to generate the generic operator inputs beside each selected input cell, so that it

can be inspected by industry parties.

24

See http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:L:2009:124:0067:0074:EN:PDF.

25 See http://www.eftasurv.int/media/internal-market/ESAs-Recommendation-on-termination-rates.pdf.

26 COMMISSION RECOMMENDATION of 7 May 2009 on the Regulatory Treatment of Fixed and Mobile Termination

Rates in the EU (2009/396/EC). Available at http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:L:2009:124:0067:0074:EN:PDF.



As the generic operator’s inputs do not exactly reflect any specific MNO (but rather rounded

average or standardised inputs) and all operator confidential data is redacted, the model is suitable

for distribution to all industry parties.

In Figure 4.1 below, we illustrate how the actual operator-specific inputs feed into the generic operator

input profile and how any of these input profiles can be used to parameterise the NPT v8F model. The

NPT v8F model produces LRIC, LRIC+++ and Pure LRIC service cost outputs for the three actual

operators and the generic operator. The model as shown below has been delivered to NPT.

Figure 4.1: Structure of the full model as delivered to NPT [Source: Analysys Mason, 2013]

The NPT v8F model has been redacted in order to be published by NPT. Figure 4.2 below shows

the material that will be publicly available.

Figure 4.2: Structure of the redacted NPT v8F model as published by NPT [Source: Analysys Mason, 2013]

Pure LRIC calculation

(as in the previous model,

uses the difference

between 2 model states)

Generic input

profile

[Generic]

Network cost

calculation

engine:

Uses selected

operator inputs to

drive model

LRIC calculation

(as in the previous model)Run network

cost model

with all traffic

Run network

cost model

with all traffic

except

termination

volume

Input profile

[Telenor]

Input profile

[Netcom]

Input profile

[Mobile

Norway]

NPT v8F model

LRIC+++ calculation

(as in the previous model)

Pure LRIC calculation

(as in the previous model,

uses the difference

between 2 model states)

Generic input

profile

[Generic]

Network cost

calculation

engine:

Uses selected

operator inputs to

drive model

LRIC calculation

(as in the previous model)Run network

cost model

with all traffic

Run network

cost model

with all traffic

except

termination

volume

NPT v8F model

LRIC+++ calculation

(as in the previous model)



4.2 Generic operator input derivations

As in the case of the third network operator modelled in the v7.1 model, certain inputs for the

generic operator need to be chosen in principle (in particular, certain inputs cannot always be

defined as a function of those of the actual operators). Our proposed definition of these inputs is

described in Figure 4.3 below. The derivation of the generic operator inputs for coverage and

subscriber market share is described in more detail in Sections 4.2.1 and 4.2.2 respectively.

Figure 4.3: Key input values for the generic operator [Source: Analysys Mason, 2013]

Input Value Comments

Radio

technologies

2G, 3G and LTE networks

– although the LTE network

is not explicitly modelled

All three MNOs in Norway currently use both 2G

and 3G technologies, and all three have spectrum

available(permanently or temporarily) for LTE

technologies

Operator

network

deployment

2012 asset purchase for a

2013 network launch with

immediate scale

Reflects the cost constraints that would exist in a

fully contestable market (where costs are set by an

operator that can reach the immediate scale of an

existing operator)

Subscriber

market share

Average value based on

number of networks (35%

for voice, 33% for data);

achieved immediately

Defined as 1/N, where N = the number of

comparable mobile coverage networks in Norway,

representing an efficient operator’s market share

Coverage profile Almost ubiquitous GSM

and UMTS population

coverage (almost 100%)

Reflects coverage of other national network

operators

2G/3G network

shutdowns

2G shutting down in 2020

and 3G in perpetuity

Reflects assumptions established in the NPT v7.1

model

Core network

technologies

All IP core from launch Modern equivalent asset for core networks

Transmission

technologies

Backhaul topologies

currently used by operators

Reflects actual 2G technologies used by Norwegian

operators. 3G backhaul is assumed to be Ethernet

from launch

Service set Full range of 2G and 3G

voice, SMS and data

services as currently

modelled beginning at

launch

Reflects service set currently offered by actual

Norwegian operators. LTE service volumes are

forecast, though not costed. It is also possible to

include LTE volumes in the LRIC and LRIC+++ cost

allocation.



The specific approach used to derive other input values for the generic operator is documented

below in Figure 4.4.

Figure 4.4: Other input values for the generic operator [Source: Analysys Mason, 2013]

Input derivation Name of input

Average of MNOs

(rounded)

Low-speed data user proportion

Voice, SMS and low-speed data migration profiles for 2G to 3G/3G to LTE

Spectrum allocations and payments

Coverage and in-fill cell radii

Air interface and network blocking probabilities

Unit capital/operating expenditure per network element

Call attempts per successful call and average call duration

Proportion of weekday traffic in a year

2G voice, 3G voice and HSPA traffic demand per geotype

Voice, SMS, low-speed data and high-speed data busy hour proportions

Type of site proportions across owned tower, third-party tower and third-party roof site

Proportion of 2G sites available for 3G NodeB upgrade

BTS capacity

Proportion of NodeBs with HSDPA 7.2 and HSUPA activated

2G and 3G repeater requirements

Proportion of sites that use microwave backhaul

BSC/RNC capacity (TRXs)

MSC coverage, CPU parts and port parts inputs

MGW/MSS/MSC parameters

BSC, RNC and core network locations

Traffic routeing across national backbone transmission links

HLR parameters

SMSC/MMSC/GSN parameters

Network layer shutdowns (2G radio, 3G radio, layered core)

Common operator

inputs

Asset lifetimes and planning periods

Capital and operating cost trends

Sum of operator

2G, 3G and LTE

values, multiplied

by generic operator

market share and

relevant 2G to 3G

and 3G to LTE

migration paths

Digital subscriptions – year end

High speed data subscriptions – year end

2G/3G/LTE incoming, on-net and outgoing off-net voice

2G/3G/LTE incoming, on-net and outgoing off-net SMS

MMS

2G/3G low-speed data traffic (GPRS/R99)

3G/LTE high-speed data traffic (HSPA/LTE)

Assumed to be

inactive

TSC locations

% national minutes which are also transited across transit layer (if present)

Backhaul 64kbit/s link channel threshold

Access nodes per cluster node

Legacy core network layer shutdown

Year in which GSM operator stops overlaying additional sites

Set to launch year

of the network,

2012

Network layer activations

Launch of 3G coverage network

Year that MSCs are made 3G-compatible



4.2.1 Generic operator coverage

The generic operator is assumed to have both a GSM and UMTS coverage network. The model

assumes almost ubiquitous GSM population coverage for the generic operator using 900MHz

spectrum. This is made up of both wide area and in-fill coverage, with 80% of coverage being

wide area, and the remaining 20% being in-fill coverage.

The generic operator is assumed to deploy an UMTS coverage using both 2100MHz and 900MHz

spectrum. This coverage is assumed to be for 99.99% of population, with the corresponding area

coverage shown by Fylke below in Figure 4.5.

Fylker UMTS area coverage Figure 4.5: Comparison

of total UMTS area

coverage by Fylke

[Source: Analysys

Mason, 2013]

Oslo, Østfold, Vestfold 90–100%

Akershus, Hedmark, Møre og

Romsdal, Nord-Trøndelag, Sogn

og Fjordane

80–90%

Rogaland, Telemark, Vest-Agder 70–80%

Aust-Agder, Buskerud, Nordland,

Sør-Trøndelag, Troms

60–70%

Finnmark, Hordaland, Oppland,

Svalbard

50–60%

4.2.2 Generic operator subscriber market share

As stated in Figure 4.3, the assumed generic operator market shares are derived as the average

value based on the number of coverage networks in Norway.

For the market share of voice, it is assumed Norway is covered by 2.85 networks, with

Telenor/TeliaSonera attaining almost 100% coverage and Mobile Norway assumed to attain 85%

population coverage in the long run. This value is in line with the efficient coverage level derived

as part of the June 2012 recommendation published by NPT.27

These assumptions give a generic

operator voice market share of 100%/285%= 35.1%.

For the market share of data, we first of all account for the 450MHz operator ICE28

, which has

approximately a 5% market share. On the remaining 95% of the market, we then use the same

calculation as for voice to give a generic operator data market share of (100%–

5%)/(100%+100%+85%) = 33.3%. We remove ICE since it is unlikely that their current spectrum

holdings will allow them to carry a significant market share of data traffic in the long-term.

27

See http://www.npt.no/marked/markedsregulering-smp/marked/marked-7/_attachment/2346?_ts=139b9c2a05b.

28 According to http://www.ice.no/privat/dekning.aspx, ICE currently has approximately 90% population coverage.

http://www.ice.no/privat/dekning.aspx



4.2.3 Generic operator subscriber demand calculations

In the NPT v8F model, the generic operator demand calculations are designed to follow the same

overall flow as the actual operator’s calculations. This should allow for easier comparison between

the generic operator and the actual operator and allows for easier sensitivity testing of parameters.

However, we have not replicated the specific calculations used in defining the actual operators’

inputs over time, but rather maintained consistency between historic and forecast input definitions.

For example, while an actual operator’s input is defined using actual data for historic periods and

an extrapolation to an end point for future periods, the generic operator’s input is defined as a

function of the actual operators’ data consistently across both historic and future periods.

Figure 4.6: Calculation flow of actual operators which is replicated for the generic operator [Source: Analysys

Mason, 2013]

Mobile operator

market share

Mobile subscribers by

operator

Population

Mobile penetration

Mobile broadband

(MBB) penetration

MBB operator market

share

MBB subscribers by

operator

Outgoing voice per

sub

On-net voice

proportion per sub

Total outgoing

(excl. OTT)

Total on-net

(excl. OTT)

OTT proportion

Incoming fixed/

mobile/ international

voice per sub

Total incoming

(excl. OTT)

Total low speed dataLow speed usage per

sub

Total HSPA / LTE

megabytes

High speed usage per

subOTT voice and SMS

megabytes

On-net / outgoing /

incoming SMS per sub

Total on-net / outgoing

/ incoming SMS

(excl. OTT)

Input CalculationKEY:



4.3 The Pure LRIC calculation

The NPT v7.1 model was updated in early 2009, when a draft version of the EC Recommendation

was available. The Pure LRIC implemented at the time is set out in Section 7.2 of the NPT v7.1

model documentation.29

In April 2011, the ESA subsequently released its own Recommendation.

Both Recommendations specify that only the costs ‘avoided when not offering voice termination’

are allocated to the voice termination service, with wholesale termination to be treated as the ‘last’

service in the network. In addition, it is specified that non traffic-related costs (such as subscriber

costs), network common costs and business overhead costs are not to be allocated to the end result.

To calculate the Pure LRIC in the NPT v8F model requires that the model is run twice: once with

wholesale mobile terminated voice and once without. Clicking on the “Run Pure” macro button on

the A0_Ctrl worksheet will result in the model calculating twice, with the necessary information

from both runs stored as values on the D1_PureLRIC worksheet. The Pure LRIC of termination is

then calculated as shown in Figure 4.7.

Figure 4.7: Calculation of Pure LRIC [Source: Analysys Mason, 2013]

The difference in both capex and opex (the avoidable expenditures) is determined from the two

model calculations, and economic depreciation is then applied to this difference. This is run

separately for capex and opex, in order to apply their respective cost trends. The Pure LRIC of

termination in each year is then calculated as the ratio of total economic cost in that year divided

by total (avoided) terminated minutes.

29

See http://www.npt.no/marked/markedsregulering-smp/kostnadsmodeller/lric-

mobilnett/_attachment/1804?_ts=1390fd85d55.

Run model with

all traffic

Run model with

all traffic except

termination

increment

volume

Expenditure with

termination

Output profile

with termination

Expenditure

without

termination

Output profile

without

termination

Difference in

expenditure

Difference in

output

Capex and opex

trends

Economic cost of

difference in

expenditure

Total economic

cost of the

difference

Pure LRIC

per minute

Termination traffic

volume




In calculating the Pure LRIC, the modelled network design assumptions reflect some of the

consequences of the modelled network carrying a lower traffic loading over its lifetime when

termination is excluded. The Pure LRIC calculation has been further refined in the modelling in terms

of two technical adjustments detailed below. This is because a pure LRIC calculation is based on the

technicalities of the cost model at the margin (in response to a small increment of traffic).

These technical adjustments can be de-activated in the model calculation, giving an alternative

“purest” LRIC calculation, as was similarly described in Section 7.2 of the NPT v7.1 model

documentation.

4.3.1 Technical adjustments to the network design to increase traffic sensitivity

The calculation has been adjusted to include specific traffic sensitivity in parts of the network

design where assets are not avoided (i.e. not avoided in the network model calculations), but where

it can be expected that assets would be avoided in the case of a real network dimensioned for no

termination traffic.

These adjustments in the network calculation alter how asset counts are calculated when excluding

voice termination, and as such increase the modelled avoidable cost and thus the Pure LRIC.

The adjustments in the NPT v8F model are:

a smaller-scale deployment of GSM and UMTS in-fill sites

a slight increase in the 3G cell radii for the six most urban Fylker.30

This accounts for the “cell

breathing” effect in UMTS, where a lower assumed traffic loading in the long term (such as

the entire removal of wholesale terminated voice) can allow for a larger planned coverage cell

radius.

4.3.2 Technical adjustments to the costing calculation to include non-traffic-sensitive costs

The Pure LRIC calculation has also been adjusted to include costs from certain assets that are not

dimensioned to be traffic-sensitive, but where it can be expected that costs would be avoided in the

case of a network dimensioned for no termination traffic. For example, this includes wholesale-

related costs from assets such as the network billing system, intelligent network (IN) platform and

the network management system (NMS).

The NPT v8F model includes the functionality to include part or all of the calculated LRIC per unit of

output (i.e. excluding all mark-ups) for the selected assets as an additional contribution to the Pure

LRIC. The methodology for this is shown in Figure 4.8. The routeing factors by asset for the 2G and

3G voice termination services are used to calculate the LRIC contribution per minute for a 2G

terminating minute and 3G terminating minute respectively. The voice migration profile is then used to

derive a blended contribution per minute, which is added to the calculated Pure LRIC.

30

These include Akershus, Hordaland, Oslo, Østfold, Rogaland and Vestfold.



Figure 4.8: Calculation

of an additional

contribution to the Pure

LRIC to capture non

traffic-sensitive costs

[Source: Analysys

Mason, 2013]

4.4 The LRIC and LRIC+++

The LRIC and LRIC+++ approaches are calculated in the same way as for the NPT v7.1 model,

consistent with the previous approach in Europe for fixed and mobile termination costing.

For the LRIC, as detailed in Section 7.1 of the NPT v7.1 model documentation,31

the average

incremental costs of traffic are defined in aggregate, then allocated to various traffic services using

routeing factors.

The LRIC+++ is then derived using three equi-proportionate cost-based mark-ups. These include:

network common costs (including the mobile coverage layer)

location updates

administrative overheads.

These three costs are shown below in Figure 4.9 as the blue, white and purple boxes, respectively.

Figure 4.9: Illustration

of the costs relevant to

the LRIC+++ [Source:


31


mobilnett/_attachment/1804?_ts=1390fd85d55.

Relevant assetsRouting

factors

Migration profile

LRIC per 2G

incoming minute

LRIC per 3G

incoming minute

Additional

contribution per

minute

LRIC per unit of

output


Traffic incremental costs

Additional radio sites, BTS/NodeB,

additional TRX/carriers, higher-

capacity backhaul links,

BSC/RNC, switches, etc.

Su

bs

crib

ers

HLR

, LU

Mobile coverage network

Radio sites, BTS/NodeB, first

TRX/carrier, backhaul link, NMS,

licence payments, etc.

Network share of business overheads



5 Mobile network design

For full details of the network design in the NPT v7.1 model, please refer to Section 4 and

Annex A of the NPT v7.1 model documentation.32

The majority of the mobile network design

remains the same as in this model. The small numbers of changes that have been made are

described below.

5.1 Pure LRIC in-fill adjustments

For full details of the original cost model coverage and in-fill design, please refer to Section 4.1.2,

Annex A.1.1 (for 2G) and Annex A.2.1 (for 3G) of the NPT v7.1 model documentation.

In-fill sites are used to fill in the gaps in 2G and 3G wide-area coverage and improving the

contiguousness of the network. They consequently have a lower cell radius, reflecting the smaller

uncovered areas which these cells satisfy – with in-fill also acting partially to provide capacity in

areas of overlap with the initial coverage layer. In the absence of termination traffic, in-fill sites

would likely be rearranged to reduce their capacity function and increase the coverage function as

shown below in Figure 5.1.

Figure 5.1: Wide-area GSM coverage and infill [Source: Analysys Mason, 2013]

32


mobilnett/_attachment/1804?_ts=1390fd85d55

Infill sitesCoverage sites

Infill layout without terminationInfill layout with termination

Infill covered with two sites Infill covered with one site



For this reason, the model conceptually assumes that if x% of traffic is removed from the 2G

network by excluding termination, then x% of in-fill sites can be removed from the 2G network.

This is captured in the model by assuming a larger cell radius for the in-fill sites under the model’s

‘no termination traffic’ case. This is calculated using the equation below, where is the proportion

of traffic excluded and is the cell radius multiplier

√

The v8F model includes a GSM multiplier for the in-fill radius pre-2006 and post-2008, reflecting

the changing proportion of terminating traffic in the 2G network with the launch of 3G and LTE

radio networks.

A UMTS in-fill multiplier is also included in the model, however is assumed constant across the

model, given the traffic reduction effect seen in the UMTS network is reduced compared to GSM.

This is due to the 3G network’s requirement to also support mobile broadband (HSPA) traffic,

leading to proportionally much less UMTS traffic avoided when termination traffic is excluded

from the network than under the 2G network.

In addition in the v8F model has been adjusted to remove the cell breathing effect from 3G in-fill

sites. This is for two reasons: because in-fill sites by definition fill in gaps, therefore significantly

over-lap the other (wide-area) coverage sites; secondly, by reducing the number of in-fill sites in

response to the reduced load, the average load on the in-fill cells would remain the same (i.e. they

would not ‘relax’). For the avoidance of doubt, the effect of cell breathing is retained for the wide-

area coverage sites, deployed in “Stage 1” and “Stage 3” 3G coverage in the model.

5.2 HSPA upgrades

For full details of the HSPA network design, please refer to Annex A.2.3 of the NPT v7.1 model

documentation.33

In addition to the four grades of HSPA deployed in the NPT v7.1 model, three

further HSPA software upgrades have been included in the NPT v8F model. The grades that are

now modelled are shown below in Figure 5.2.

HSDPA grades HSUPA grades Figure 5.2: Grades of

HSPA modelled [Source:

NPT v8F model, 2013] 3.6Mbit/s 1.46Mbit/s

7.2Mbit/s 5.76Mbit/s

14.4Mbit/s

21Mbit/s

42Mbit/s

33





In the NPT v7.1 model, all 2100MHz and 900MHz UMTS sites were deployed with a minimum

grade of HSDPA3.6 using a single shared carrier. In addition, a proportion of sites could then be

assumed to be upgraded to HSDPA7.2 (and then subsequently HSDPA14.4). In the NPT v8F

model, all Fylker are upgraded to an HSDPA grade in a specified year of activation. An equivalent

approach is used for HSUPA deployments.

Figure 5.3: HSDPA element deployment [Source: Analysys Mason, 2013]

The conversion factor for converting HSPA data megabytes to voice-equivalent minutes is used in

the routeing factor table to allocate costs between voice and data services. This factor has been

updated in the NPT v8F model to reflect the new modelled grades of HSPA and is calculated

based on the “weighted-average channel data rate”. This is defined by the total amount of traffic

carried at each HSDPA and HSUPA grade over the modelling period, as described in Figure 5.4.

Figure 5.4: Modelling flow of HSDPA and HSUPA conversion factors [Source: Analysys Mason, 2013]

HSDPA grade

activation dates

Total number of

3G sites

Sites at each

HSDPA grade

Effective HSDPA

rate per NodeB

Carriers at each

HSDPA grade

HSDPA grade

deployed in each

year


Additional IP

overheads

HSDPA/HSUPA

conversion to

minutes

Traffic carried at

each grade over

time

Channel elements

required per grade

Channel data rate

per grade

Weighted average

channel data rate

Proportion

HSDPA/HSUPA

Data rate per

grade




5.3 UMTS Ethernet backhaul deployment

For full details of the original cost model backhaul network design, please refer to Annex A.1.4

(for 2G) and Annex A.2.4 (for 3G) of the NPT v7.1 model documentation.

As with the NPT v7.1 model, 3G backhaul is assumed to be logically and physically separate from

2G backhaul in the NPT v8F model. The NPT v8F model contains the option to deploy Ethernet

backhaul links for 3G backhaul. The network design for 2G backhaul and non-Ethernet 3G

backhaul is unchanged from the previous NPT v7.1 model.

The NPT v8F model splits the 3G backhaul requirements into microwave and leased-line

backhaul. The proportion of each of these categories that are Ethernet is then calculated using a

migration profile (specified by Fylke and over time for each operator). Tunnel sites are treated

separately and are assumed to migrate to Ethernet backhaul using their own profile.

The Ethernet links can vary in speed depending on the amount of traffic (including voice, R99 and

HSDPA traffic) per site by Fylke. The Ethernet backhaul is dimensioned as either 20Mbit/s or

50Mbit/s links, based on the average busy-hour traffic throughput per site in each Fylke. The

number of site links is then aggregated across geotypes by speed.

The necessary number of Ethernet ports is also dimensioned for the 3G network both for voice and

wireless Ethernet links, in terms of 10Mbit/s ports.

The different configurations for 3G backhaul, and their corresponding port requirements, are

shown below in Figure 5.5.

Figure 5.5: 3G backhaul

physical configuration

[Source: Analysys

Mason, 2013]

Non-Ethernet

Indoor/Tunnel

sites

Up to

n×NodeB per

RNC

Fibre

backbone

Ethernet

microwave

Ethernet

leased line

RNC

E1 E1

1×E1

Non-Ethernet

leased line

Non-Ethernet

Microwave

RNCETH

ETH



5.4 Spectrum licences

The spectrum allocations and spectrum fees for the 2100MHz spectrum licences have been

updated for the NPT v8F model in line with the results of the auction in November 2012,34

with a

renewal period of 20 years. Furthermore, both of Mobile Norway’s 2100MHz licences are

renewed separately.

NPT does not intend to take a position on future spectrum auctions and their outcomes. Therefore,

the 900MHz/1800MHz licences are modelled as being renewed every 12 years, with costs

increasing by the relevant level of forecast inflation. The 2100MHz licences are modelled as being

renewed after 20 years.

34

See http://www.npt.no/aktuelt/nyheter/2-ghz-auksjonen-avsluttet

Ref: 35971-214 – Public . PUBLIC

Annex A Excerpts from the v7.1 model documentation

For full details of the network design in the NPT v7.1 model, please refer to Section 4 and

Annex A of the NPT v7.1 model documentation.35

Revisions have been made to the network

design concerning in-fill coverage, HSPA software upgrades (and subsequent channel kit

requirements), as well as the determination of the 3G backhaul.

For reference, the relevant sections of the NPT v7.1 model documentation are provided below.

A.1 Coverage

Coverage was determined on the basis of the radio database of each of the operator’s networks, as

submitted to the NPT as part of the data request. From this database, the area covered at a signal

strength of –94dBm was calculated: this strength represents approximate outdoor coverage.

Coverage calculations were made for the following sets of frequencies:

GSM900

GSM1800

GSM900+GSM1800 (i.e. GSM)

UMTS.

Indoor coverage, in terms of area and population, reflected by a higher signal strength, is

commensurately lower, though not used to drive network deployment in the model except for the

initial coverage of the third entrant operator reflected in the v7.1 model.

In the initial network roll-out years, additional sites are assumed to be rolled out to maximise the

area covered, with little or no overlap between cells. In the later years, sites are deployed for in-fill

purposes. These sites fill in the gaps in wide-area coverage and improving the contiguousness of

the network. They consequently have a lower cell radius, reflecting the smaller uncovered areas

which these cells satisfy. This concept is shown in Figure A.1 for the operators’ GSM networks:

35



Mobile cost model version 8 final (v8F) | A–2


Figure A.1: Wide-area

GSM coverage and in-

fill [Source: Analysys

Mason]

The coverage profile of the GSM network is defined for each operator, on the basis of 900MHz

frequencies, using the inputs from the original (v4) model. These inputs have been updated for the

period 2005–2008 using outputs of the GSM coverage recalculation performed by NPT in 2009.

For the modelled UMTS networks, the approach to wide-area and in-fill coverage has been

modified slightly. The UMTS model assumes the following roll-out process:

“wide-area” coverage of the urban areas in each Fylke is deployed using 2100MHz spectrum

“in-fill” coverage of the urban areas in each Fylke is deployed using 2100MHz spectrum

rural coverage in each Fylke is deployed using UMTS900 equipment, on the assumption that as

GSM frequencies become unloaded, they can be re-farmed for a 900MHz UMTS deployment.

Therefore, the first two parts of this coverage roll-out are similar to the GSM network algorithm,

albeit with alternative parameters reflecting the proportion of population (and hence area) and cell

radius used for the deployment. The third part of UMTS coverage aims to replicate GSM coverage

in order that the GSM network may be shut-down.

Coverage site

Infill site

Time

Are

a c

ove

red

(k

m2)

rcoverage

rinfill

Approx

year

1997

80

:20

are

a r

ule



This concept is shown in the figure below for the operators’ UMTS networks. Although the roll-out

using 2100MHz spectrum only covers approximately 25% of the Norwegian land area, it reaches more

than 90% of the population. UMTS900 is then used to increase the UMTS coverage to equal to GSM

coverage, but only covers the remaining 5–10% of population across a large area of the country.

Figure A.2: UMTS coverage and in-fill [Source: Analysys Mason]

The parameters determining these calculations can be found in the NtwDesBase and NtwDesSlct

worksheets.

A.2 Radio network: Channel kit (CK) and carrier requirements

Channel kit requirements are calculated separately for voice/R99 and HSPA, by first calculating

the channel element (CE) requirements.

CK requirements for voice/R99

To calculate channel element (CE) requirements for voice and R99 data, the inputs required are:

total voice and R99 BHE traffic by Fylke

total NodeB sectors and sites by Fylke (as previously calculated)

channel element utilisation.

Figure A.3 shows a flow diagram describing the calculation of CE/CK required. Having calculated

the number of CEs deployed at each site, the number of carriers required can then be calculated

directly.

Time

Are

a c

overe

d (

km

2)

e.g. 25% area of all Fylke

e.g +60% area of Fylkee.g. +2% area of all Fylke

TimeP

op

ula

tio

n c

overe

d (

km

2)

e.g. 90-95% population of all Fylke

~+1% population of all Fylke

e.g. +5% population of all FylkeKey

2100MHz urban wide area

2100MHz urban infill

900MHz rural



Figure A.3: Channel kit

deployment [Source:

Analysys Mason]

The blended site sectorisation across the whole 3G network, by Fylke, is calculated as a first step.

The Erlang demand per NodeB sector is then derived and converted into a CE requirement per

sector using the Erlang B table. This calculation accounts for both CE utilisation and soft

handover. The CE requirement per site is then calculated using the blended sectorisation and

assuming that a minimum number of 64 CEs are activated on every NodeB. The number of CEs

required is obtained by multiplying the number of sites and the CE requirement per site.

The number of carriers required, first per site and then in total, can then be calculated according to

a maximum number of CEs deployed per NodeB (128).

CK requirements for HSPA [this section has been superseded by the information presented in

the main body of this report]

Four grades of HSPA are deployed in the model: HSDPA 3.6, HSDPA 7.2, HSDPA 14.4 and

HSUPA1.5. Each is assumed to be activated in the network from a particular year onwards. It is

assumed that HSDPA 3.6 is deployed at every NodeB from launch, whereas HSDPA 7.2 are only

deployed at a proportion of sites in each Fylke. In addition, HSDPA 14.4 is deployed a certain

number of years after HSDPA 7.2 is deployed as an upgrade to some sites. Channel elements for

HSDPA and HSUPA are calculated separately, as shown below in Figure A.6.

CE utilisation

UMTS2100MHz

coverage sites (F, t)

Blended sectorisation

(F, t)

Average UMTS2100MHz

sectorisation (F)

Radio network blocking

probability

UMTS900MHz

coverage sites (F, t)

Average UMTS2100MHz

sectorisation (F)

Soft handover

R99 BHE per sector

(F, t)

R99 BHE (F, t)

CE required per site to

meet R99 BHE, including

soft handover (F, t)

Minimum CE deployed

per site (F)

CE deployed per site

(F, t)

Total 3G sites (F, t)

CE deployed (F, t)

Maximum CE per

NodeB (F)

Carriers deployed per

NodeB (F, t)

Carriers deployed (F, T)


CE deployed for HSPA

(F,t)Total CE deployed (F, t)

Total CK deployed (F, t)CE per CK



Figure A.4: HSPA

channel element

deployment [Source:

Analysys Mason]

Within each Fylke, the model makes the distinction between sites with HSDPA 3.6 and

HSDPA 7.2/HSDPA 14.4.

The model also includes a cross-check to ensure that the deployed HSDPA capability (in terms of

average HSDPA rate per NodeB) can support the offered throughput (in terms of average HSDPA

busy-hour throughput per NodeB) in all Fylker in all years. This is illustrated below in Figure A.5.

The cross-check assumes underutilisation of HSDPA channel elements that is greater than R99 channel

elements. This is because of the greater difference between the cell loading at its maximum and the

loading of the average busy-hour for HSDPA compared with that for voice and R99 data. As a result,

an average to peak BHE loading of 200% is used in deriving HSDPA CE utilisation.

Figure A.5: Verification

of sufficient HSPA

deployment [Source:

Analysys Mason]

This cross-check is linked into the Ctrl worksheet and is highlighted in red if the check fails.

HSDPA years of

activation (flavour, F)

Proportion of sites with

HSDPA7.2 activated (F)


Proportion of sites with

HSUPA activated (F)

Sites deployed with

HSDPA3.6 activated

(F,t)

Sites deployed with

HSDPA7.2 (or 14.4)

activated (F,t)

Sites deployed with

HSUPA activated (F,t)

Minimum HSUPA

channels per NodeB

HSUPA years of

activation (F)

Minimum HSDPA

channels per NodeB

(flavour)

HSDPA CE deployed

per NodeB (F, t)

HSUPA CE deployed

per NodeB (F, t)

HSDPA peak load in

data busy-hour (t)

Distribution of HSDPA

traffic (F)

HSDPA throughput

(F, t)Total 3G sites (F, t)

HSDPA throughput per

site (F, t)HSPA CE utilisation

Sites deployed with

HSDPA7.2 (or 14.4)

activated (F,t)

Sites deployed with

HSDPA3.6 activated

(F,t)

Effective HSDPA rate

per NodeB (F,t)

Checks that deployed capability is greater than throughput



A.3 Backhaul transmission [this section has been superseded by the information

presented in the main body of this report]

The calculation of the number of backhaul links and the corresponding number of E1 ports

required is set out in Figure A.6. Requirements for voice/R99 and HSPA are dimensioned

separately. The capacity requirements and then the means of deployment are calculated separately.

Figure A.6: UMTS

backhaul calculation

[Source: Analysys

Mason]

Step 1: Capacity requirements

The number of E1s required per site is calculated to fulfil the capacity requirements for a backhaul

link. The requirements for voice/R99 and HSDPA are considered separately, calculating:

120 channels per E1 for voice/R99 CE

blended Mbit/s requirements for HSDPA/HSUPA, including an overhead for IP.

It is assumed that backhaul requirements for HSDPA are provisioned according to the speed in

Mbit/s. In addition, we assume that HSUPA uses the uplink backhaul capacity already deployed

for HSDPA, i.e. there is no backhaul deployment dedicated to HSUPA.

The effective capacity per E1 is calculated for voice. The number of E1 links required per site is

obtained by simply dividing the circuits per site with the effective capacity per E1 link.

Total number of NodeBs

(F, t)

Proportion of sites using

microwave (F)

NodeBs connected using

(8Mbit/s) microwave

links (F, t)

R99 Mbit/s required per

site (F, t)Link utilisation

16kbit/s channels per E1

Blended HSDPA Mbit/s

required per site (F, t)

R99 CE required per site

(F, t)

Data IP overhead

Sites deployed with

HSDPA3.6 activated

(F,t)

Sites deployed with

HSDPA7.2 (or 14.4)

activated (F,t)

Sites deployed with

HSUPA activated (F,t)

HSUPA Mbit/s required

per site (F, t)

Total Mbit/s required per

site (F, t)E1 capacity

E1s required per site

(F, t)

E1s per microwave

backhaul link

Microwave links required

(F, t)

Leased line E1s required

(F, t)

NodeBs connected

using leased lines (F, t)

Total number of NodeBs

(F, t)



Step 2: Backhaul network design algorithms

There are two types of backhaul to be considered in the network: microwave (8Mbit/s links) and

leased lines. The percentage of sites which have microwave backhaul is an input into the model.

The number of microwave 8Mbit/s backhaul links is set to be a minimum of one per site. The

model allows for more than one 8Mbit/s link per site. The number of E1 units occupied in each

8Mbit/s microwave link is calculated.

The number of sites using leased lines calculated as the difference between the total sites and the

total sites using microwaves. The total number of E1 leased lines required is the product of the

total number of NodeB sites using leased line and the number of E1 required per site (from Step 1).

Tunnel repeater sites are assumed to use only E1 leased lines and hence are added to the leased-

line requirement of the macro NodeB layer.


Annex B Model adjustments from v7.1 to v8D

In this section, we describe the adjustments made to the structural calculations of the v7.1 model,

including the correction of some minor aspects of the model.

The step-by-step changes made in the model in order to make the adjustments discussed below are

detailed in the change log.

B.1 Model corrections

Corrected high-

speed connections

On the D3_M8 worksheet, the market share calculations for 2009–11 for

Telenor’s and TeliaSonera’s high-speed connections were corrected to

account for the portion of the market covered by ICE Nordisk.

Corrected effective

average HSDPA

rate calculation

On the B6_NwDes worksheet, the effective average HSDPA rate

calculations were changed to only include the additional sites at each rate

when calculating the weighted average.

Corrected

calculation for “In-

fill cell radius –

after 2008”

On the A6_NtwDesSlct worksheet, the IF() statement adjusting the in-fill

cell radius after 2008 was corrected so that the correct value was obtained

for the Pure LRIC case.

B.2 Revised input parameters and other decisions

Developed Mobile

Norway calculation

Inputs were developed for a Mobile Norway-specific calculation, in the

same way as Telenor and TeliaSonera, allowing for asset calibration and

cost reconciliation against Mobile Norway’s submitted data.

These Mobile Norway-specific inputs have been derived from data received

from Mobile Norway, Tele2, Network Norway and NPT, and can be found

in the following worksheets: D3_M8, B7_LifeIn, A8_UtilIn,

A4_NtwDesBase and D4_CostBase.

Changed third

operator to generic

operator

The hypothetical third operator entry was completely reworked to be a

calculation for a generic operator (in addition to the three actual MNOs).

The demand, network and cost inputs for the generic operator are derived

from the inputs used in the model for the Telenor, TeliaSonera and Mobile

Norway-specific calculations.

The changes made and calculations used for deriving the generic operator

inputs are discussed in more detail in Section 4.2 above.

Mobile cost model version 8 final (v8F) | B–2


HSPA site

upgrades

On the B6_NwDes worksheet, the number of sites which are upgraded with

different HSDPA and HSUPA rates, has been changed from just 2100MHz

sites to all 3G sites. This is to reflect operator information which shows that

both 2100MHz and 900MHz NodeBs are equipped with faster HSPA grades.

Added HSDPA21

and HSDPA42

software assets

Two additional upgrades of HSDPA were added to the modelled asset list,

representing 21Mbit/s and 42Mbit/s respectively:

the definition of minimum channel deployments, speeds, and network

deployment timing of these assets were added to the A4_NtwDesBase

worksheet

routing factors for the new assets were added to the B9_RF worksheet

links to the asset counts were included on the C02_FullNw worksheet.

the calculation of the number of HSDPA assets required in the network

was added to the B6_NwDes worksheet, in a similar fashion to the other

HSDPA assets

in addition the HSDPA deployment profiles, HSDPA average throughput

and checks, and the LMA requirement calculations, were updated on the

B6_NwDes worksheet to include the new assets.

The new HSDPA calculations are discussed in more detail in Section 5.1.

Added a second

HSUPA upgrade

asset

A second HSUPA upgrade was added to the modelled asset list, in a similar

fashion to the HSDPA assets described above. In addition, the calculation of

HSUPA channels per NodeB was deleted from the model.

The new HSUPA infrastructure is discussed in more detail in Section 5.1.

Updated operator

market shares

Based on NPT market data, the market shares of ‘registered and hosted

subscribers (excluding telemetry)’ and ‘high-speed data subscriptions by

operator’ have been updated for each operator in 2009–12.

Updated Telenor/

TeliaSonera

network design

inputs

As part of the calibration of the Telenor/TeliaSonera operator models for the

period 2009–2012, several revisions were made to their operator-specific

inputs on the A4_NtwDesBase and A8_UtilIn worksheets. In particular, the

inputs related to the 3G coverage of these operators were updated.

Updated Telenor/

TeliaSonera unit

costs

As part of the reconciliation of the Telenor/TeliaSonera operator models for

the period 2009–2012, several revisions were made to the inputs on the

D4_CostBase worksheet. In particular, the unit costs assumed for the

following assets were all reduced:

layered core equipment

HSPA software upgrades

radio site (opex only).

The lifetimes for radio sites were also extended by five years.



Updated capex/

opex cost trends

As part of the reconciliation of the Telenor/TeliaSonera operator models for

the period 2009–2012, several revisions were made to the inputs on the

C01_CostTrends worksheet. In particular:

the capex trends were revised in 2009–2012 for BTS, NodeB, channel

kit, sites/ancillary, BSC and microwave backhaul

the opex trends were revised for BTS/NodeB, BTS/NodeB transmission,

network billing system, IN and radio sites.

Introduced

migration to LTE

network for voice,

SMS and high-

speed data traffic

Migration profiles for voice, SMS and high-speed data services to an LTE

network have been added.

Inputs for the total high-speed data traffic (across all technologies) have

been added to the model, with the LTE high-speed data traffic being derived

as the difference between this total forecast and the established HSPA

forecast. It is assumed, however, that the previous NPT v7.1 model’s minute

and SMS forecasts were derived on a total, technology-neutral basis, with

the new NPT v8D model’s 3G voice and SMS traffic profiles consisting of a

proportion of this total traffic profile.

The changes made and calculations used for deriving the services carried

over the LTE network are found on the D3_M8 worksheet and are discussed

in more detail in Section 3.

Adjusted voice and

SMS forecasts to

account for OTT

traffic

A consideration of OTT voice and SMS traffic has been added to the model,

separating this traffic out from the total (technology-neutral) traffic

projections. This OTT traffic is then converted to high-speed megabytes and

included in the modelled demand as HSPA/LTE traffic.

The changes made and calculations used for deriving the OTT traffic and

megabytes can be found on the D3_M8 worksheet and discussed in more

detail in Section 3.2 above.

Update of operator

demand data for

2009–12

In 2012, Analysys Mason updated a copy of the v7.1 model for a separate

piece of work for NPT related to asymmetric termination.36

The updates to

the demand calculation, as described in Section 3.1 of the technical report,37

were incorporated into the NPT v8D model.

Operator data and NPT market data for the period 2009–12 have been used

to update the demand in these years for each operator. These can be found in

a new market calculation worksheet in the NPT v8D model (D3_M8). More

detail on these demand-related updates can be found in Section 3.3.

36

See http://www.npt.no/marked/markedsregulering-smp/marked/marked-7/anmodning-om-omgj%C3%B8ring



Update of demand

forecasts

The updated historical data received for the period 2009–12 from the

operators and NPT has resulted in some forecast volumes diverging from

those forecast in the v7.1 model. As a result, we have updated the forecasts

to give a more realistic projection of demand. The specific forecasts updated

are discussed in more detail in Section 3.4.

Addition of

Ethernet backhaul

for 3G sites

The backhaul calculations were enhanced to include Ethernet backhaul

options for 3G sites. This reflects changes in the Norwegian mobile

networks towards increased deployment of IP backhaul, as a result of

increasing HSPA speeds and LTE deployments. The new network design

calculations are described in more detail in Section 5.3.

2G and 3G

spectrum licences

Following the November 2012 auction, both the spectrum allocations and

spectrum fees for the modelled 2100MHz spectrum have been updated.

Update of the

modelled WACC

The WACC in the NPT v8D model has been altered to the value that has

been calculated in parallel to the model update.

37

See http://www.npt.no/marked/markedsregulering-smp/marked/marked-7/_attachment/2349?_ts=139b9c507e1


Annex C Operator submissions in developing v8F

C.1 Comments related to the demand calculations

C.1.1 2G data traffic

NPT assumes a GRPS usage of 70 Mio MB for 2013. TDC finds this usage tremendously

low. In Denmark, GPRS/EDGE is used in areas without sufficient 3G coverage and

constituted 10% of the total data traffic in 2012, whereas in the draft model GPRS/EDGE

constitutes only 0.65% in 2013. Such a low percentage requires an almost 100% 3G/LTE

geographical coverage, and this is not the case in Norway, nor for the generic operator.

TDC requests NPT to increase the 2G data traffic to 10% of the total data traffic for 2012 and let

the volume decline over the following years, depending on the modelled 3G/LTE roll-out.

Analysys Mason’s response

Analysys Mason has investigated the historical proportion of data that is GPRS/EDGE in the

Norwegian market using data provided by the MNOs for the year 2012. The actual operator data

indicates 5.2% of all Norwegian mobile data traffic in 2012 is carried as GPRS/EDGE, which is

more in line with the 10% figure quoted for Denmark by TDC.

The v8D model contained the correct volume of total low-speed data (including R99 on the 3G

network), but the migration profile for low-speed data had not been updated from the v7.1 model.

Analysys Mason has therefore revised the low speed data migration profile for the MNOs in the

v8F model for the years 2009–2012, using operator data where provided.

The forecast migration profile from 2013 to the 2G network shutdown at the end of 2020 has also

been adjusted so that operator absolute GPRS/EDGE traffic volumes remain stable year-on-year.

We have applied a stable traffic profile (as opposed to steadily declining traffic) because GPRS

data traffic is generated by all subscribers in the market (2G, 3G and mobile broadband) so will be

largely unaffected by underlying technology migration. The impact of these adjustments to the

migration profile on the modelled GPRS/EDGE data traffic is shown in Figure C.1 below.

This gives a GPRS usage for the generic operator of approximately 282 million megabytes for

2013, which is equivalent to nearly 3% of all data traffic in 2013 decreasing to 1% in 2020 before

2G shutdown. This steady volume, but declining percentage, balances the impact of rapidly

increasing growth in data traffic combined with increased usage of 3G/4G networks to carry the

majority of traffic.

Mobile cost model version 8 final (v8F) | C–2


Figure C.1: Total

market GPRS/EDGE

megabytes in the NPT

v8F model [Source:


The changes to the modelled low-speed data migration profile detailed above result in a small

reduction in the model results, as illustrated for the generic operator in Figure C.2 and Figure C.3

below. As can be seen, this change in the low-speed data migration profile results in a reduction in

the LRIC+++ and LRIC for the generic operator.

Figure C.2: Generic operator results of the v8D

model basecase [Source: Analysys Mason, 2013] Figure C.3: Generic operator results of the v8D

model with updated low-speed data migration profile

[Source: Analysys Mason, 2013]

While investigating the point raised by TDC, an issue was identified with the formula used to

calculate the volume of LTE upload data carried by each operator. Specifically, the formula

incorrectly calculated the HSUPA data traffic rather than LTE data traffic. This formula has been

corrected in the v8F model. This may have distorted TDC calculation of the proportion of data

traffic that is carried as GPRS/EDGE in 2013.

0.0

0.2

0.4

0.6

0.8

1.0

1.2

To

tal m

ark

et G

PR

S/E

DG

E m

eg

ab

yte

s (

bill

ion

s)

0.153 0.145

0.133 0.124

0.084 0.077

0.070 0.063 0.088

0.078

0.065 0.054

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0.18

0.20

Mo

bile

vo

ice

te

rmin

atio

n c

osts

(N

OK

no

min

al)

LRIC+++ LRIC Pure LRIC

0.131 0.126

0.119 0.113

0.072 0.067 0.062 0.058

0.088 0.078

0.066 0.054

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0.18

0.20

Mo

bile

vo

ice

te

rmin

atio

n co

sts

(NO

K n

om

ina

l)




C.1.2 Market calculation structure

Telenor states that the logic in the model for arriving at the demand scenario is reversed in

the generic model compared to the operator specific model. This difference is not

documented. This problem has been acknowledged by NPT/AM (mail March 18). Since the

difference is undocumented Telenor has spent valuable time figuring out why the generic

model not is functioning as expected, and worse, the difference makes it impossible to do

sensitivity analysis across the operator specific and the generic model in a consistent way.

NPT/AM has not explained the reason for choosing a different modelling structure for the

generic operator. From an operator perspective it is very hard to see the benefit in choosing

different design for the two models.

Telenor acknowledges that it is possible to modify the demand sensitivity array on sheet

A9_M. However, it is necessary to ensure consistency in assumed market share, on net

proportion, inbound volumes etc. This consistency cannot easily be ensured by the demand

sensitivity parameters. This is in contrast to the specific assumptions on the same parameters

on sheet D3_M8, but as stressed above, and acknowledged by NPT and AM, those

parameters does not play a role for the generic operator.

Telenor would like to reiterate that the model package, due to the lack of documentation is

misleading, and furthermore, that the differences in modelling structure makes it hard to do

consistent sensitivity analysis.


As noted by Telenor, the NPT v8D model has a different order of calculation flow in the market

sheet for the generic operator than for the actual operators.

As detailed in Section 4.2, the generic operator does not have its own actual input data, but rather

inputs are defined as a function of the input data supplied by the actual operators. In the case of

market demand these inputs are frequently derived from the sum of the actual operator’s total

market data, so as to allow explicit consideration of the generic operator modelled market share on

the total traffic.38

In the NPT v8D model the difference in calculation order of the generic operator was intended to

demonstrate the point at which the generic operator modelled market shares were used to derive

the specific generic operator traffic levels. We appreciate that this made some types of sensitivity

testing more difficult for the generic operator than for the actual operators.

38

For example, in the case of the generic operator’s ‘incoming, on-net and outgoing off-net voice’ the derivation is

detailed as ‘sum of operator 2G, 3G and LTE values, multiplied by generic operator market share and relevant 2G to 3G and 3G to LTE migration paths’. This is opposed to the actual operators’ whose data was input on ‘minutes per subscriber per month’ basis as detailed in Figure 3.7 in the v8D documentation.



Following Telenor’s comments we have adjusted the logic of the generic operator calculation on a

copy of the D3_M8D worksheet to create the D3_M8R worksheet, to match the calculation flow of

the actual operators. Differences in the point at which certain inputs are now rounded means that

some of the generic operator market traffic levels have fractionally changed from the NPT v8D

model – though these changes impact the generic operator LRIC, LRIC+++ or Pure LRIC of

termination by less than 0.015 øre in all cases and as such we consider this effect negligible. This

is shown below in Figure C.4.

(Nominal

NOK)

LRAIC,

rounded

2dp

Pure LRIC,

rounded

2dp

LRAIC,

rounded

1dp

Pure LRIC,

rounded

1dp

Figure C.4: Comparison

of 2014 cost results at

different levels of

decimal rounding

[Source: v8D/v8R model,

2013]

M8D 0.08 0.09 0.10 0.10

M8R 0.08 0.09 0.10 0.10

We have used this new structure as the basis for the market calculation for the v8F model

(D3_M8F worksheet).

C.1.3 GSM traffic

It is Tele2’s understanding that this first sensitivity test is showing the impact of changing

the assumptions for the treatment of traffic carried over the GSM network.

The v8D model basecase is assuming that GSM bearer services will shut down in 2020

(Model Documentation, Figure 4.3, p. 41). The different assumptions used for the sensitivity

is GSM in perpetuity [1a] and steady GSM traffic until shutdown [1b].

The impacts of these different scenarios on the model results are material.

The specific assumptions used in the three scenarios shown above are not explicitly

described. In scenario [1a] the GSM (2G) bearer services will continue in perpetuity,

whereas in the basecase and scenario [1b] GSM (2G) bearer services are shutdown in its

entirety from 2020.39

Tele2 is of the opinion that a conservative approach must be used. Consequently the

basecase assumptions for 2G shut down should not be used, since the basecase assumptions

are less conservative than both scenarios [1a] and [1b].

Furthermore, Tele2 finds that both the basecase and the [1b] scenario are not compatible

with the ESA Recommendation, which states:

“the bottom-up model for mobile networks should be based on a combination of 2G and 3G

employed in the access part of the network” (para 12).

39

Tele2 assumes that 2G is assumed to shut down in 2020 also in Scenario [1b], although this is not explicitly stated.



From the Model Documentation we understand that the total voice/SMS traffic across all

technologies remains largely the same as in the LRIC v7.1 model and that the distribution of

traffic to each technology is using the same migration patterns used for migration from 2G

to 3G. The difference between the three scenarios used in test 1 is the portion of traffic

carried on the 2G network (i.e. traffic not migrated to 3G or LTE technologies).

The base case scenario assumes a linear distribution of voice and SMS traffic until 2019 as

described in AM Industry meeting presentation p. 70:

AM has not presented a similar presentation of the distribution of voice and SMS traffic

which is used in scenario [1a]. For scenario [1b] AM has assumed "flat levels of GSM

traffic until shutdown".

Since the traffic assumptions for scenario [1a] has not been presented it is difficult to depict

the absolute traffic volumes used by AM in scenario [1a]. Tele2 kindly asks the NPT or AM

to present the exact assumptions used by AM in scenario [1a].

As a general comment, Tele2 does not agree with the basecase assumption (linear

reductions). It is Tele2’s experience that the reduction of 2G-traffic has occurred in a less

aggressive pace than what earlier market forecasts have predicted. This supports a more

conservative assumption on the reduction of 2G-traffic.

Tele2 reserves further comments on the 2G traffic volume assumptions, until such

information has been provided.


We expand the paragraph of the ESA Recommendation reference referenced by Tele2:

“The cost model should be based on the efficient technological choices available in the time frame

considered by the model, to the extent that they can be identified. Hence, a bottom-up model built

today could in principle assume that the core network for fixed networks is Next-Generation-

Network (NGN)-based. The bottom-up model for mobile networks should be based on a

combination of 2G and 3G employed in the access part of the network, reflecting the anticipated

situation, while the core part could be assumed to be NGN-based.”40

We believe that a GSM shutdown is compatible with the ESA Recommendation. The ESA

Recommendation says that the access part of the network should be based on a combination of 2G

and 3G technologies (as the v8D model basecase is), but does not specify the mix of traffic that

these technologies may carry over time. Additionally given the beginning of the ESA paragraph,

we believe there is no requirement in the ESA recommendation for operators to be modelled using

2G technology beyond the point at which use of the existing network infrastructure is efficient

given the ‘anticipated situation’.

40

See http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=CELEX:32009H0396:EN:HTML



Tele2 had particular questions about the traffic assumptions made for Scenario [1a] in the

“Sensitivity tests of NPT’s v8D LRIC model” document issued in March 201341

(“the sensitivity

document”), in which GSM traffic is maintained in perpetuity. These traffic assumptions are

shown, alongside the result of the scenario, in Figure C.6 below.

Figure C.5: Generic operator results of Scenario [1a]

as presented in Sensitivity tests of NPT’s v8D LRIC

model [Source: Analysys Mason, 2013]

Figure C.6: Voice traffic assumptions for Scenario

[1a] [Source: Analysys Mason, 2013]

Tele2 is of the opinion that NPT should be more conservative in setting the voice migration profile

from 2G to 3G technologies. This would mean voice traffic remaining on the higher-cost 2G

network for longer while the 3G unit costs would simultaneously rise due to the network carrying

less traffic and being unable to take advantage of the economies of scope and scale associated with

carrying greater traffic volumes. Thus voice termination costs would rise as they fail to benefit

from the steady migration from old technology (2G) to new technology (3G).

The historical migration profile of Telenor/NetCom indicates that there has been a slowing in

migration, but this appears to be temporary. Assuming migration to 3G by 2020 is not

unreasonable given the current migration profiles of these two operators. We propose that MN can

also migrate to 3G by 2020, given the small size of their 2G network.

Defining a different migration profile for the generic operator is hard to justify: we believe that

continuing to use the average of these actual operator projected migration profiles is reasonable.

The later discussion in the section C.2.1 about setting a dynamically efficient set of costs is also

relevant here.

41


mobilnett/_attachment/6462?_download=true&_ts=13d39ae3c84



C.1.4 Migration to LTE

Tele2 notes that AM has presented two alternative scenarios to the basecase scenario. The

impacts on the LRIC+++ results are significant in both scenarios, whereas the impact on the

LRIC and the pure LRIC results is only significant in scenario [3b].

The basecase assumption is that 20% of voice migrates to LTE in the long term. In scenario

[3a] 60% of voice migrates to LTE in the long term. In scenario [3b] assumes 3G shutdown

in 2030 and all voice migrates to LTE.

It is difficult to see what specific volume which is assumed to be carried on 2G networks in

scenario [3a] and [3b] (in the short term), and we assume that this is the same as the basecase

scenario (as depicted in the charts cited above from AM Industry meeting presentation p. 70).

It is difficult to understand why the results are showing such significant impacts in the years

2014–2017 when the assumptions seem to be affecting long term distribution of traffic.

Tele2 therefore asks the NPT to provide further information about these specific

assumptions, and in particular why the impact on the results is so significant.

TeliaSonera agrees that migration of UMTS to LTE has to be added into the demand

calculations.

TeliaSonera makes reference to section 3.1 in the model documentation that is received

regarding ”LTE demand forecasting”, where it is given an explanation of how voice, SMS

and high-speed data services has been added into the demand calculations in the v8D model.

In the proposed model 2015 has been used as a start date for the beginning of migration of

voice and SMS services to LTE. TeliaSonera is of the opinion that such migration actually

will take place earlier than 2015.

TeliaSonera expects that VoLTE will be introduced earlier than 2015. It is anticipated that

will be rolled-out rapidly and with good coverage soon after the auction that will take place

later this year, and in that view the demand forecasts and start date implemented in the v8D

model seems to be too conservative as TeliaSonera sees it.

TeliaSonera will in this connection draw the attention to article of 6 December 2012 in

Teknisk Ukeblad, where Telenor among others says:

“LTE er rundt tigangeren raskere enn på 3G og vi ser en datavekst på rundt 40 prosent

årlig. Derfor er det viktig både for oss og mobilbrukerne at vi får nytte av det mye raskere

nettet. CS Fallback gjør at LTE når får med seg telefonene også. Allerede nå ser vi at to

prosent av pakkedata går på LTE-nettet. En av tre vil ha LTE-dekning allerede i år og i

2015 vil det være ni av ti, sier sjefen for Telenor Norge, Berit Svendsen.”42

42

See http://www.tu.no/it/2012/12/06/telenor-apner-lte-for-mobiltelefoner




Tele2 has requested further information regarding the assumptions used in the calculations of the

sensitivities related to voice migration from 3G to LTE, Scenarios [3a] and [3b]. These scenarios

looked at the impact of 60% of voice migrating to LTE in the long term and 3G radio network

shutdown in 2030, with all voice migrating to LTE respectively. The results of these sensitivities

are reproduced below in Figure C.7.

Figure C.7: Generic operator results of Scenario [3a] and Scenario [3b] as presented in the sensitivity

document [Source: Analysys Mason, 2013]

The voice traffic volumes carried across the 2G, 3G and 4G radio networks in the basecase and

Scenarios [3a] and [3b], are illustrated in Figure C.8 below.

Figure C.8: Voice traffic volumes under the basecase, Scenario [3a] and Scenario [3b] LTE voice migration

assumptions [Source: Analysys Mason, 2013]

The reason for the significant impact of these scenarios on the v8D model output for the years

2014–2017 is due to the significant difference in voice traffic volumes carried by the 3G network

over its assumed lifetime. This is represented by the 3G voice areas in Figure C.8 above. The

amount of voice traffic carried on the 2G/3G networks is smaller in Scenario [3a] (and

significantly smaller in Scenario [3b]) than the basecase. Economic depreciation takes the traffic

carried over the whole network lifetime into account when determining the cost recovery profile.



Therefore, even though the differences in voice volumes up to 2017 are not too significant, the

cost recovery profile in those years is still heavily influenced by the more significant differences in

2G/3G voice in the longer-term.

TeliaSonera supports their argument that LTE roll-out should begin earlier and more rapidly with

reference to a December 2012 article from Teknisk Ukeblad which quotes the head of Telenor

Norway as saying:

“Already, we see that two per cent of the packet data goes on LTE network. One in three will have

LTE coverage already this year and by 2015 there will be nine out of ten.”

We have run an additional sensitivity, Scenario [3c] in which we have moved the start of voice

migration to LTE forward by 2 years to 2013 as shown in Figure C.9.

Figure C.9: Generic

operator 4G voice

migration profile in the

v8D model basecase

and Scenario [3c]

[Source: Analysys

Mason, 2013]

The result of this sensitivity test are shown below in Figure C.11, which when compared with the

results of the v8D basecase in Figure C.10 indicates that the impact on the model outputs for the

generic operator of moving the start of LTE network roll-out forward by two years to 2013 are

minimal.

0%

5%

10%

15%

20%

25%

4G

vo

ice

mig

ratio

n p

rofile

Base case Scenario [3c]




model basecase [Source: Analysys Mason, 2013]

Figure C.11: Generic operator results of the v8D model

with LTE migration profile start date brought forward by

2 years, Scenario [3c] [Source: Analysys Mason, 2013]

We consider that this purported 2% migration refers to packet data rather than voice; given the

source article states that “Problemet har vært at LTE-nettet ikke har noen telefonitjeneste ennå.

Det vil ta et år eller to før vi får det som går under betegnelsen VoLTE - Voice over LTE.” It is

unlikely that 2% of voice is being carried on the LTE network. Therefore, we do not believe that

the 4G voice migration profile should be adjusted as TeliaSonera suggest.

Assuming a higher proportion of voice migrates to 4G in the long term has two main consequences:

the 2G/3G networks carry less voice traffic, which increases the LRIC results

implicitly, the proportion of voice assumed to be carried on 4G in the near future (to 2017) is

higher, with the unit cost of 4G voice not calculated within the cost model and thus not known.

The overall cost of voice termination is a blend of the costs of 2G/3G/4G voice traffic. A

conservative migration of voice to 4G reduces the proportion of the 4G component for which there

is uncertainty over the specific unit cost. In principle, the long-run unit cost of 4G voice should be

lower than that of 3G voice because of the even greater sharing of costs with higher data traffic

and the use of single IP air interface as opposed to channelized R99 versus HSPA layers.

NPT does not have a calculation of 4G voice costs in order to assess whether it is in fact inefficient

to migrate voice to 3G and to adopt an alternative situation of migrating voice to 4G technologies.

On this basis, it is reasonable to continue with the current 2G to 3G migration until such time as it

is clear that a new technology is carrying a material proportion of voice traffic within the short- to

medium-term period around when NPT is setting wholesale price regulation.

0.153 0.145

0.133 0.124

0.084 0.077

0.070 0.063 0.088

0.078

0.065

0.054

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0.18

Mo

bile

vo

ice

te

rmin

atio

n c

osts

(N

OK

no

min

al)


0.155 0.146

0.135 0.125

0.084 0.078

0.070 0.064 0.090

0.080

0.067

0.056

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0.18

Mo

bile

vo

ice

te

rmin

atio

n c

osts

(N

OK

no

min

al)




In addition, a faster migration of voice to 4G networks would also implicitly assume that VoLTE-

compatible handsets become widely available and are used in the near future. There is no evidence

that this is the case.

We see no reason to allow a more rapid migration of voice to 4G if operators claim that the costs

of VoLTE are high. It would be inefficient from the perspective of the operators paying mobile

termination rates to support migration to a higher cost (voice) technology. Hence we believe that it

is reasonable to retain the approach of the v8D model and assume that migration commences in

2015 with 20% of voice (excluding OTT) migrating to 4G in the long-term.

► Shared costs with 4G

In addition to the above considerations, we believe it is reasonable to assume a degree of site cost

sharing between 2G, 3G and 4G traffic. Specifically, we include a proportion of 4G megabytes in the

cost allocation routing matrix for relevant site costs. This proportion can be estimated in two ways:

Firstly, using the degree to which LTE sites are to be co-located with 2G/3G sites. We believe

that operators will seek to extensively co-locate 4G infrastructure on existing sites nationwide

after the deployment of widespread LTE technologies (in particular following the upcoming

800MHz auction).

Secondly, considering whether there are any higher coding rate efficiencies to be obtained from

LTE technologies meaning that 4G megabytes are ‘more efficient’ than HSPA megabytes on a

like-for-like comparison. In the long-term, this appears plausible, although the degree to which

advanced technologies such as MIMO are used to carry traffic remains to be seen.

On the basis of the two points above, we believe it is reasonable to apply a 50% factor to LTE

megabytes within the routeing calculation, using the functionality set up in the v8D model (where

the factor was assumed to be zero).

C.1.5 Traffic share in Tele2’s network

Tele2 states that in order to reflect the market shares used in the scenarios, another set of

assumptions needs to be made, namely the share of Tele2’s and Network Norway’s traffic

carried in Mobile Norway’s network.

In the v8D model, it is assumed that 85 % of Tele2’s total traffic will be produced in Mobile

Norway’s network. Tele2 has explained to the NPT and AM that, approximately [] of the

traffic is carried in Mobile Norway’s network when Mobile Norway’s network reached 65%

population coverage.

[]

Consequently Tele2 do not agree with the assumption of 85% traffic share with 85%

population coverage. In Tele2’s view, all scenarios should be updated to reflect Tele2’s own

estimates for on-net share in Mobile Norway’s network.




In response to Tele2’s comment, we have sensitivity tested the inclusion of a multiplier in the

calculation of the Mobile Norway market share in the v8F model, such that it is scaled in

proportion to their network coverage. This multiplier is constructed using a relationship between

the third operator network coverage and the proportion of traffic carried in their network, in line

with data received from Tele2. This relationship is illustrated in Figure C.12 below.

[] Figure C.12:

Relationship between

network coverage and

the proportion of traffic

carried in own network

for Tele2 as used in

sensitivity [Source:


The impact of this multiplier is to reduce the network share of Tele2, while increasing that for the

other two MNOs, Telenor and NetCom. The shares of the three operators in both the v8D and the

model with the network coverage relationship implemented can be seen in Figure C.13 below.

Figure C.13: Market

share of MNOs in the

v8D and model with the

relationship between

network coverage and

the proportion of traffic

carried in own network

models implemented

[Source: Analysys

Mason, 2013]

While this change in the modelled market share of the three MNOs can be considered important, it

has minimal effect on the outcome of the generic operator calculation. As shown in Figure C.15

below, the differences in the values of the generic operator LRIC, Pure LRIC and LRIC+++ as a

result of the sensitivity test are minimal. The key impact of the revised market shares on the

generic operator is an adjustment in the market share weighted average inputs used by the generic

operators in its market demand calculations.

0%

10%

20%

30%

40%

50%

60%

70%

80%

Ma

rke

t sh

are

Telenor updated Telenor v8D

NetCom updated NetCom v8D

MobileNorway updated MobileNorway v8D




model [Source: Analysys Mason, 2013] Figure C.15: Generic operator results of the v8D

model with the updates to the MNO market shares

implemented [Source: Analysys Mason, 2013]

While Tele2 does not believe Mobile Norway’s assumed traffic share is appropriate given the 85%

long-term coverage forecast, Analysys Mason believes that these concerns have in fact been

remedied through the v8F model change undertaken to densify the Mobile Norway radio network

from its v8D level as detailed in Section C.2.7. As a result of this model adjustment, the v8F

modelled network is better able to carry a proportion of traffic corresponding to its population

coverage. Therefore, we believe that it is reasonable to retain the approach of the v8D model,

where a network operator with x% population coverage can carry x% of their own traffic.

C.1.6 Market share for Tele2 and Network Norway

The generic operator model assumes 99% coverage. AM has explained that the coverage in

the generic operator model is assumed to be national and not the average of actual operators.

Tele2 has understood that the LRIC-model assumes that Tele2’s traffic in Mobile Norway’s

network is assumed to be equal to the population coverage and that this is modelled using

market share for the operator. In the basecase the market share calculation is derived from the

assumed coverage of all MNOs. In the model, AM has assumed that Tele2 and Network

Norway attain 33.3% retail market share in the long-run.

Tele2 is of the opinion that the market share cannot be modelled based on these simple

assumptions. In particular, the national roaming tariffs will have a very strong impact on the

possibility to gain further market shares for Tele2 and Network Norway as long as Tele2 and

Network relies on national roaming. As mentioned above in section 3.4.2 a relatively large

share of Tele2 and Network Norway’s combined traffic will be produced by national roaming

in TeliaSonera's and/or Telenor's network (where the national roaming tariffs – and not the

cost of using Mobile Norway’s network - will determine the cost).

0.153 0.145

0.133 0.124

0.084 0.077

0.070 0.063 0.088

0.078

0.065

0.054

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0.18

Mo

bile

vo

ice

te

rmin

atio

n c

osts

(N

OK

no

min

al)


0.154 0.145

0.134 0.124

0.084 0.077

0.070 0.063 0.088

0.079

0.066

0.055

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0.18

Mo

bile

vo

ice

te

rmin

atio

n c

osts

(N

OK

no

min

al)




In a report from professor Frode Steen and professor Øystein Foros from 200843

, the first-

mover advantages for Telenor and TeliaSonera (and the corresponding late-mover

disadvantage for Tele2) are discussed. Foros and Steen (2008) held that there are significant

effects of the first mover advantages that will have an impact on the market shares for new

entrants, such as Tele2 and Network Norway, who has entered the market long after Telenor

and TeliaSonera. Foros’s and Steen’s findings can be summarised as follows44

:

When NPT compares prices (telepriser.no), they find that the non‐MNOs generally have

lower retail prices than the two MNOs. However, the two incumbent MNOs have

significantly higher market shares than the service‐based entrants.

The service‐based entrants’ high marginal costs (due to the fact that wholesale prices are

average cost-based) provide a competitive advantage for the two incumbent MNOs. We

observe that Telenor and NetCom offer what is known as “friends or family” tariffs,

where on‐net traffic is priced very low or at a price of zero. This new development

potentially has two effects in the retail market that both work to the benefit of the early

entrants. First, the operator with the largest network will, all other things being equal,

becomes the preferred operator. A more indirect effect of the friends or family pricing is

that it seems as if customers are coordinating themselves to ensure that their own micro

network of contacts and friends end up in the same network. Birke and Swann (2005)45

suggest that micro‐network effects are significant in customers’ choice of operator.

The third‐mover disadvantage for a third facility‐based provider is significantly higher

than NetCom’s second‐mover disadvantages in 1993. In 1993, NetCom and Telenor both

started from Greenfield with respect to GSM. The entry conditions will be significantly

different for a third MNO that now enters the market. National roaming will be an

essential input, and the conditions for national roaming will to a large extent determine to

what extent the third entrant has the ability to be a viable and aggressive rival to the

incumbents. By nature, the incumbents may have incentives to endogenously increase the

entry barriers with respect to national roaming (see discussion above).

If the Norwegian customers have somewhat similar preferences in their adoption choice as

was found in Birke and Swann (2005), this suggests that the incumbents, Telenor and

NetCom, also have an advantage when it comes to quality. Interestingly, this has become a

topic in negotiations with Network Norway. Telenor will not allow Network Norway to have

roaming on both networks simultaneously since Network Norway in this way would actually

have better quality in terms of coverage than both Telenor and NetCom have individually.

43

See “Do first-mover advantages call for asymmetric regulation in the Norwegian mobile market?”, July 2008,

available on http://www.nettvett.no/ikbViewer/Content/108743/Tele2%20Bilag%204%20-%20offentlig.pdf

44 See "The Norwegian mobile market: How to achieve more facility based competition, Professor Øystein Foros and

Professor Frode Steen, March 2010.

45 See Birke, D. and G.M.P. Swann, 2005. Social networks and choice of mobile phone operator, The University of

Nottingham Business School, Industrial Economics Division, Occasional Paper Series, 2005‐14.



Correspondingly, Telenor is not willing to accept Network Norway’s marketing of a

“package” that communicates to customers that by subscribing to Network Norway they will

have both the partial network of Network Norway and the roaming option on Telenor’s

network. This suggests that the early entrants still look at coverage and quality as significant

factors in their marketing. Moreover, Telenor still communicates to end‐users their advantage

related to better quality and coverage in their marketing.

The market statistics published by the NPT show that the market shares remain stable for

Telenor and TeliaSonera over time, a fact that supports the findings in Foros Steen (2008).


Tele2 has raised concerns that Mobile Norway will reach 33% retail market share in the long-run.

Given the use of this retail market share forecast in determining the traffic share carried on Mobile

Norway's network they have recommended considering the validity of this assumption. This

principle was established in the v4 model, based on a hypothetical forecast of a balanced three-

player market, and was considered plausible by Telenor at the time.46

In recognition of Tele2’s concerns, we have run a sensitivity test for reduced forecast growth in

Mobile Norway’s market share so as to reach a lower level market share, while simultaneously

increasing the market shares of the other modelled MNOs (Telenor and NetCom), as shown in

Figure C.16.

Figure C.16: Market

share of MNOs in the

v8D and model with

slower growth in Mobile

Norway market share

[Source: Analysys

Mason, 2013]

46

http://eng.npt.no/ikbViewer/Content/Model_documentation_v4.pdf?documentID=50981

0%

10%

20%

30%

40%

50%

60%

70%

80%

Ma

rke

t sh

are

Telenor updated Telenor v8D

NetCom updated NetCom v8D

MobileNorway updated MobileNorway v8D



Changing the assumed 33% retail market share has a significant effect on the Mobile Norway

costs, but a very limited effect on the generic operator as shown in Figure C.18 below.


model [Source: Analysys Mason, 2013] Figure C.18: Generic operator results of the v8D model

with the slower growth in Mobile Norway market share

implemented [Source: Analysys Mason, 2013]

As with the sensitivity test in Section C.1.5, this impact on the generic operator arises because several

generic operator demand values are calculated from a weighted average of actual operator inputs.

Therefore, changing the MNO market share has an impact (albeit small) on the generic operator traffic.

Given the forecast has previously been established and the exact development of the future market

is unknown, we consider it reasonable to retain the v8D assumption of a 33% long-term retail

market share for Mobile Norway in the v8F model. This assumption of equal market share is also

strengthened by the improved quality of Mobile Norway’s network in the v8F model.

C.1.7 OTT traffic

TeliaSonera makes reference to section 3.2 in the model documentation – regarding “OTT

traffic.” From TeliaSonera’s view point the model use too conservative forecast as regards

proportion of traffic (voice/SMS) carried by OTT – 15% in the long term.

TeliaSonera expects that a larger share than 15% of traffic will be carried by OTT. Services

like Home from Facebook will, as TeliaSonera sees it, trigger a larger proportion of traffic

carried by OTT than anticipated in the, and TeliaSonera expects that such traffic will

increase rapidly.47

In addition to launch of new services, it is anticipated that new terminals

and more user-friendly applications affect the size of such traffic.

47

See http://www.digi.no/914562/vil-dominere-mobil-opplevelsen

0.153 0.145

0.133 0.124

0.084 0.077

0.070 0.063 0.088

0.078

0.065 0.054

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0.18

0.20

Mo

bile

vo

ice

te

rmin

atio

n c

osts

(N

OK

no

min

al)


0.154 0.145

0.134 0.124

0.084 0.077

0.070 0.063 0.089

0.079

0.066

0.055

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0.18

0.20

Mo

bile

vo

ice

te

rmin

atio

n c

osts

(N

OK

no

min

al)





The relationship of the OTT minute proportion to the 2G and 3G voice traffic is given below (with

the same relationship also applying to SMS):

As discussed in Section 3.2, in the NPT v8D model a conservative forecast was assumed for OTT

take-up, with OTT traffic accounting for 15% of all voice/SMS traffic in the long term – resulting

in the remaining 85% of all market minutes being split between 2G, 3G and LTE voice.

The article referred to by TeliaSonera, discusses the launching of the Facebook Home app, an

application that “kan skifte ut låse- og hjemmeskjermen til enkelte eksisterende og kommende

Android-mobiler med en Facebook-orientert opplevelse”. While this article touches on the

potential implications for users of the Facebook messenger service, there is no direct comment on

to what extent this will impact the overall OTT messaging and voice traffic, and specifically

whether this will cause it to increase beyond the growth levels already considered. As such we do

not consider this article to be relevant in defining the forecast OTT traffic.

We highlight the Norwegian retail market appears to be increasingly moving towards selling ‘all

inclusive bundles’, with Telenor48

, NetCom49

and Tele250

each having introduced subscriptions

where the customer pays for the required amount of data, and all voice and SMS usage is included

for free. This makes it less attractive for subscribers to use OTT services given the advantage of

normal voice calls,51

and may act to limit the growth of OTT services in the future.

In response to the concerns raised by TeliaSonera about the proportion of voice and SMS traffic to

be carried by OTT we have run a sensitivity, Scenario [4c], in which we consider the impact of

quadrupling the proportion of OTT traffic to 60% in the long run – where only 40% of voice

traffic in the market will be left as traditional voice minutes to be split across 2G, 3G and LTE.

48

For example Telenor’s Komplett plans (XL, L, M+ and M). See http://www.telenor.no/privat/mobil/mobilabonnement/

49 For example NetCom’s Smart plans (Super, Pluss, Basis and Mini). See https://netcom.no/privat/mobilabonnement

50 For example Tele2’s Fastpris plans (XL, L, and M). See https://www.tele2.no/abonnement.aspx

51 These include using the phones native (and optimised calling interface), the receiving party not having to own/run

specific software, and the general higher quality of service ensured through CS calling.

https://netcom.no/privat/mobilabonnement



The change in the results of v8D model under this scenario are not as dramatic as might be

expected from such a large increase in the proportion of voice and SMS traffic being carried by

OTT. This can be attributed to a reduction in the number of assets required to build the network to

carry the traffic, and therefore a reduction in the network costs.

We highlight that the traffic carried via OTT effectively has a zero termination charge (on both an

LRIC+++ and pure LRIC basis) because the cost is paid by the mobile party data bundle. As such,

voice termination carried via OTT can be considered to give no contribution to the blended cost of

termination. When we blend together the LRIC+++ and pure LRIC results with the OTT

proportion at zero cost the average costs for terminated traffic fall, as shown in Figure C.19 below.

Figure C.19: OTT blended results of the generic operator v8F model [Source: Analysys Mason, 2013]

In this scenario, it can be argued that a zero termination charge for OTT voice could be efficient,

and as such we consider it possible that MNOs may decide to move to bill-and-keep for voice

termination traffic at a point in time when the suggested OTT proportion becomes material.

However overall, we consider that given the current low usage indicated by NPT’s survey from

2011, and uncertainty in the development of these services, that a conservative forecast remains

appropriate. Therefore we propose to retain the v8D model assumption, with 15% of all voice

carried as OTT in the long-term in the v8F model.

0.00

0.05

0.10

0.15

0.20

Mo

bile

vo

ice

te

rmin

atio

n c

osts

(N

OK

no

min

al)

LRIC+++ (v8F) LRIC+++ (update) LRIC+++ (update blend)

LRIC (v8F) LRIC (update) LRIC (update blend)

Pure LRIC (v8F) Pure LRIC (update) Pure LRIC (update blend)



C.1.8 Market share for generic operator

TeliaSonera refers to section 4.2.2 regarding Generic operator subscriber market share.

Market share for generic operator is based on an assumption that TeliaSonera and Telenor

attaining almost 100% coverage and that Mobile Norway will attain 85% coverage in the

long run. These assumptions give a generic operator voice market share of 35.1%.

From TeliaSonera’s point of view, there is no basis for the assumption that Mobile Norway

will attain 85% coverage in the long run. As TeliaSonera have observed, it is only NPT that

have mentioned that 85% coverage probably will be most efficient coverage for the third

operator. The model should therefore use 75% coverage for Tele2 in the calculations of

generic operator voice market share.


As mentioned by TeliaSonera and discussed in Section 4.2.2 above, the long-term population

coverage assumption for Mobile Norway is set at 85% for the v8D model giving a generic operator

voice market share of 100% ÷ (100% + 100% + 85%) = 35.1%.

Analysys Mason notes that Mobile Norway’s population coverage is expected to exceed 75% by

[], and as such TeliaSonera’s suggestion that this 75% population coverage level should be

assumed for Mobile Norway in the long-term does not appear reasonable. However, to assume

100% population coverage without actual evidence would be inconsistent with NPT’s asymmetric

pricing decision taken in June 2012.52

The analysis undertaken by NPT in June 2012 indicated that

it would appear to be efficient for the third operator to achieve 85–90% population coverage.

Therefore we believe retaining the v8D model assumption of achievement of 85% population

coverage in the long-term in the v8F model to be appropriate.

52

See http://www.npt.no/marked/markedsregulering-smp/marked/marked-7/_attachment/2346?_ts=139b9c2a05b

http://www.npt.no/marked/markedsregulering-smp/marked/marked-7/_attachment/2346?_ts=139b9c2a05b



C.2 Comments related to the network calculations

C.2.1 Shut down of the 2G network

TDC finds it problematic to include 2G network roll-out for a generic operator that launch

network in 2013. First of all a new operator will not invest in an obsolete technology that

will last for only seven years. This short technology cycles further means that 2G-specific

investments need a pay-back time (seven years until 2020) which is shorter than the

equipment lifetime (typically up to 8 years). A required payback time shorter than the

equipment lifetime is not an efficient use of the equipment and causes unnaturally high

yearly depreciations in order to have fully depreciated assets in year 2020. The fewer years

for depreciation the higher the yearly cost becomes. TDC cannot see how such investment

policy will be chosen by an efficient operator. The effect is demonstrated in slide 70 in the

"Industry presentation for NPT – 2013 update of NPT's mobile LRIC model".

Given NTP's approach of modelling a 'today' entry data, the problem will aggravate in the

coming years when the yearly update displaces the entry date to 2014, 2015 etc. whereby

the pay-back period is reduced. TDC finds that the problem must be solved by letting the

2G-assets be partly depreciated before the entry year. TDC requests NPT to implement such

a change, where the 2G capex is reduced to a share of 7/8 for a 2013-launch.

Telenor observe that in the LRIC model it is assumed that the 2G network is switched off in

2020. This assumption is essentially different from Telenor’s network strategy where the

plan is to maintain the 2G network at least until 2025. Telenor does not consider it viable to

switch off the 2G network []. Another reason for this strategy is that Telenor has

obligations to serve the M2M market. Telenor has promised the M2M market that their

equipment base can work until 2025. It takes time to gradually migrate the M2M equipment

base (a lifetime of ten years required due to high fixed costs of terminals) to LTE and [].

Telenor assumes that other mobile network operators in Norway have similar strategies to

maintain the 2G network beyond 2020.

As the 2G network is switched off and spectrum is refarmed to LTE, the spectrum portfolio

in the generic model should match this assumption e.g. spectrum in the 1800 MHz band

should be zero for GSM usage after the 2G switch-off date.

Main point: It is more realistic that the 2G network will be switched off around 2025 and

this should be reflected in the generic model.

Tele2 notes that with reference to the statement in the ESA Recommendation para 12, AM

concludes that "the cost model needs to directly model the costs of LTE services".

Tele2 is of the opinion that the ESA Recommendation means that the NPT must use a

combination of 2G and 3G as long as Tele2 plans to build 2G sites and use these after 2020.



The Generic operator model assumes a 2G to 3G deployment and migration profile, equal to

the average of the three MNOs. Tele2 has no information about the migration profile

assumed for the other two MNOs, but is of the opinion that the migration profile for Mobile

Norway cannot assume a full migration from GSM to 3G/LTE as soon as 2020. []. Tele2

does not assume that the planned 2G sites will be decommissioned by 2020. Tele2 is of the

opinion that it is difficult to predict how soon all 2G handsets will be replaced. Terminals

used for machine-to-machine are more difficult to replace than handsets and the population

of such terminals may suggest continued use of GSM-technology beyond 2020.

TeliaSonera agrees, in principle, with the proposal regarding adjustment of

Recommendation 5 regarding “Treatment of technology generations”. TeliaSonera is

however of the opinion that it is not realistic that GSM shut-down will take place already in

2020. Even Analysys and NPT seem to be unsure with regard to GSM shut-down due the

statement: “GSM shut-down is projected for at least 2020.” From TeliaSonera’s point of

view proposed GSM shut-down in the model have to be prolonged due to the good coverage

of the GSM networks and to high use of GSM, hereunder the use of GSM to M2M services

such as alarms etc.


As noted by TDC the generic operator launches its network in 2013, despite having deployed the

assets in 2012. This leads to only seven years of cost recovery, despite the equipment having

completed a whole modelled ‘lifetime’ of eight years. However, it is intended that the network’s

first year (2012) should be an ‘installation year’ in which the (work in progress) sites are only fully

deployed and come online at the end of the 2012 period. As such this means that the eight-year

lifetime should be completed by the end of period 2020 (rather than in 2019 as in the v8D model).

The v8F model has been adjusted to allow the intended eight years of cost recovery by extending

the 2G migration profile so that it reaches 0% in 2021, as shown in Figure C.20. The v8F model

has also been adjusted to ensure that the 2G assets deployed in 2012 are not replaced in 2020, and

the opex (which previously extended into the 2020 period) does not change.



Figure C.20:

Comparison of v8D and

updated v8F model 2G

to 3G migration profiles

[Source: Analysys

Mason, 2013]

Telenor, TeliaSonera and Tele2 have each suggested that the GSM network should be maintained

beyond 2020, in line with their plans, particularly in regards to providing a service for their

machine-to-machine (M2M) customers. This is a reasonable premise. We believe however, that

these (mostly fixed) terminals could be served with a minimal 2G wide area coverage network

beyond 2020, as shown in Figure C.21. In addition, we consider that the costs of maintaining this

type of minimal 2G network specifically to fulfil M2M contracts are not relevant to the costs of

mobile voice termination, especially if voice has largely migrated away from 2G.

Figure C.21: Example

of generic operator 2G

BTS requirements to

continue basic network

for M2M [Source:


Excluding the considerations of M2M, we have run an additional sensitivity looking at moving the 2G

network shutdown back to 2025. The results of this sensitivity are shown in Figure C.23 below.

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Sh

are

of vo

ice

tra

ffic

on

2G

v8D v8F

0

1

2

3

4

5

6

BT

S (

tho

usa

nd

)

2G coverage BTS 2G capacity BTS



FigureC.22: Generic operator results of the v8D


Figure C.23: Generic operator results of the v8D model

with 2G radio network shutdown moved back to 2025,

Scenario [1e] [Source: Analysys Mason, 2013]

This sensitivity [1e] clearly shows that the unit costs of voice rise with the suggested five-year

extension to the GSM network. This will be caused by additional asset replacements (e.g.

replacing eight-year lifetime BTS for another five years) combined with generally declining

(migrating) voice traffic leaving the GSM network. Therefore, we do not consider this scenario of

a five-year extension to GSM to reflect a reasonably efficient solution for the costs of voice traffic.

We believe that Telenor’s comment that shutdown of the 2G network is not viable until the current

level of 2G coverage []. This aligns with TeliaSonera’s comment that 2G shutdown should be

delayed due to the advantage of the 2G network’s current high levels of coverage across Norway.

However, our understanding is that the 2G service coverage is likely to be replaced by the 3G network,

given the greater penetration of 3G handsets and the significantly higher current 3G network coverage.

The v8F model is consistent with this requirement with each operator having a full 3G network, with

2G equivalent coverage, deployed by the point of 2G network shutdown – as such ensuring no loss in

coverage due to 2G shutdown, and ensuring 2G shutdown does not need to be [].

In previous LRIC modelling decisions, NPT has adopted the principled position of reflecting a

reasonable, steady migration profile from legacy 2G technology onto emerging (and now proven) 3G

technology. This steady migration is seen as dynamically efficient from the perspective of voice

termination traffic costs. To adopt a materially longer 2G network lifetime would be to change this

position in principle and result in a situation where NPT sets medium- and long-term voice termination

costs on the basis of voice traffic remaining on a higher cost 2G network for much longer.

0.153 0.145

0.133 0.124

0.084 0.077

0.070 0.063 0.088

0.078

0.065 0.054

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0.18

0.20

Mo

bile

vo

ice

te

rmin

atio

n c

osts

(N

OK

no

min

al)


0.166 0.155

0.142

0.130

0.090 0.083

0.074 0.067

0.097 0.086

0.072 0.059

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0.18

0.20

Mo

bile

vo

ice

te

rmin

atio

n c

osts

(N

OK

no

min

al)




In conclusion, a five-year extension to the 2G network combined with declining 2G voice traffic is

not considered efficient, and it is possible with appropriate network optimisations for operators to

maintain a (thinner) M2M network for longer without causing voice service costs to rise. Thus, we

propose to maintain our basecase approach of considering 2G shut-down at the end of 2020,

combined with the modification based on TDC’s comment for the 2G assets to ensure eight years

of productive output.

We have investigated the impact of the 2G radio network shutdown on operator’s spectrum

portfolios and found that, in the v8D model, there are no 2G licence fees included as either capex

or opex after the 2G radio network shutdown in 2020. However, where 2G spectrum usage was

shown after 2020 in the v8D model, the v8F model has been corrected to show no usage. This

investigation also raised an issue with the v8D model not charging UMTS900 periodic spectrum

fees until after 2G network shutdown for both Telenor and NetCom (though the generic operator

and Mobile Norway were not affected). This has been corrected in the v8F model so that all

operators incur the appropriate level of 3G spectrum costs.

For the avoidance of doubt, the citation made by Tele2 is incorrect. The presentation states

“therefore we do not believe that the cost model needs to directly model the costs of LTE

services”.53

We also highlight that the ESA Recommendation states that the “model for mobile networks

should be based on a combination of 2G and 3G employed in the access part of the network”, not

that both technologies must persist for the entire model duration. Therefore, we believe that this

supports the 2G network shutdown approach used in the v8D model and which is retained in the

v8F model.

C.2.2 GSM in-fill cell radius

TDC notes that NPT presents a model result with a pure LRIC results higher than the LRIC

result for 2014. Such result clearly indicates that the pure LRIC model is not implemented

correctly. TDC requests NPT to revise the model or at least give an intuitive explanation for

pure LRIC costs higher than the LRIC costs.

TDC finds that the high pure LRIC cost is a result of a mistaken implementation of the

GSM-infill deployment, (ref. slide 60). TDC acknowledge that a correction to the modelling

of GSM in-fill sites might be appropriate when avoided cost is to be calculated. TDC

however finds that the implementation in the model overrate the effect of GSM sites that

can be avoided.

53

See http://www.npt.no/marked/markedsregulering-smp/kostnadsmodeller/lric-mobilnett/_attachment/6459?_download=true&_ts=13d39ac4a4d



The correction to the number of GSM in-fill sites is modelled in 'A6_NtwDesSlct' cell 124,

where a 'GSM in fill radius multiplier for pure LRIC' is defined. This multiplier increases

the in-fill radius 20% in the current draft model when pure LRIC is calculated, meaning that

the average GSM in-fill site is able to cover an area (1,20^2=1,44) 44% larger when

terminated traffic is not served by the network. TDC finds this increase of coverage area

undocumented in the model.

Given the volumes and routing factors used in the model (sheet 'TtlEleOut'), the total traffic

in the GSM net with and without terminated traffic can be calculated.

Figure C.24: Total GSM

traffic with and without

termination [Source:

TDC, 2013]

It is seen that the traffic falls to app. 75% (=3.7 bio/4.9 bio) when terminated traffic is

excluded. In other words will the GSM in-fill site be able to serve a 33% (1/0.75) larger

area, when terminated traffic is excluded. This served area equals a 'GSM infill radius

multiplier for pure LRIC'of 1.15 instead of 1.20 used in the current draft model.

The factor has significant effect on the pure LRAIC result. NPT is therefore requested to

change the factor to 1.15. NPT should note that the factor will change further, if the amount

of 2G data traffic is altered.

Telenor notes that, as for 2G indoor coverage; the pure LRIC calculation takes into account

that operators would roll out less 2G infill sites in the hypothetical scenario without the

terminated volume. The economic argument for this adjustment is similar to the economic

arguments for taking coverage into account (see section 6 below). Infill sites are rolled out if

there is expected to be sufficient traffic and thus revenues to cover capex and opex. In the

hypothetical scenario where the terminated volume is removed there will be less traffic and

revenues per infill site and hence, some sites that are profitable at the outset become

unprofitable as the terminated volume is removed. Telenor agrees with NPT/AM that this

effect should be taken into account. Whether 1.2 is the correct number or not – is hard to

assess. It depends on a set of assumptions related to a hypothetical scenario, but for Telenor

it seems like this parameter value represents the lower limit of what the 2G infill multiplier

should be.



In Telenor's view the model is inconsistent with respect to infill sites and we see no

justification as to why 2G and 3G infill sites are treated differently in the model. The

argument for having an infill multiplier on 3G should be even stronger than on 2G since the

2100MHz frequencies gives poor indoor coverage and each infill site will cover a small

number of potential customers and hence, the proportion of marginal infill sites is larger on

3G compared to 2G. Thus the number of 3G infill sites will be more sensitive to removal of

the terminated volume, relative to 2G infill sites.

An infill multiplier on 3G should be in addition to ‘cell breathing’. The reason for taking

cell breathing into account for 3G is that this effect is driven by traffic load alone. Since the

hypothetical scenario implies that a proportion of traffic is removed, network dimensioning

will have to be adjusted accordingly.

Main points are (i) the 2G infill multiplier captures a relevant aspect of avoidable costs and

(ii) there should have been a similar infill multiplier, also for 3G infill sites (in addition to

‘cell breathing’).

TeliaSonera refers to section 4.3 in the model documentation. In section 4.3.1 regarding

Technical adjustments to the network design to increase traffic sensitivity it is mentioned

that among adjustments in the v8D model are “a smaller-scale deployment of GSM in-fill

sites”. From TeliaSonera’s point of view the reduction of GSM in-fill sites must be

explained as the proposed reduction will affect in-door coverage in the network.


The assumed reduction in 2G in-fill sites mentioned in Section 4.3.1, as raised by TeliaSonera, was

documented in a presentation delivered to industry in 200954

, regarding the v6 model. In-fill sites

“fill in the gaps in wide-area coverage and improving the contiguousness of the network. They

consequently have a lower cell radius, reflecting the smaller uncovered areas which these cells

satisfy.”55

We note that in addition to this coverage function, the in-fill sites also partially overlap

with the existing wide-are coverage network, adding additional capacity to the coverage layer.

Analysys Mason accepts TDC’s points regarding the GSM in-fill radius multiplier for pure LRIC,

and we have therefore made updates to the v8F model to take this into consideration. We have

analysed the generic operator 2013 GSM traffic in the v8F model, with GPRS/EDGE megabytes

updated as set out in Section C.1.1, as shown in Figure C.25. This means that the generic operator

GSM traffic falls by approximately 20% when termination is excluded.

54

See http://eng.npt.no/ikbViewer/Content/113898/Updated%20LRIC%20model_industry%20slides_241109.pdf,

slide 34 of 95

55 NPT’s mobile cost model version 7.1; Model documentation;

http://www.nettvett.no/ikbViewer/Content/113945/Model%20documentation%20for%20NPT%20with%20responses_Public_011209.pdf

http://eng.npt.no/ikbViewer/Content/113898/Updated%20LRIC%20model_industry%20slides_241109.pdf



Services on the

GSM network

Routeing

Factor

Traffic

including

termination

Traffic

excluding

termination

Figure C.25: Total

generic operator 2013

GSM traffic with and

without termination in the

v8F model [Source:


2G on-net voice 2.0000 1 040 850 647 1 040 850 647

Incoming voice to

2G from other

networks

1.0000 1 224 785 833 0

Outgoing voice from

2G to other

networks

1.0000 1 245 086 194 1 245 086 194

2G on-net SMS 0.0017 602 245 235 602 245 235

Incoming 2G SMS

from other networks

0.0009 474 870 160 474 870 160

2G Outgoing SMS

to other networks

0.0009 505 057 597 505 057 597

GPRS/EDGE

megabytes

5.5721 281 972 487 281 972 487

Minute equivalents

(traffic × RF)

6 124 661 968 4 899 876 135

The model conceptually assumes that if x% of traffic is removed from the 2G network by

excluding termination, then x% of in-fill sites can be removed from the 2G network. This is

captured in the model by assuming a larger cell radius for the in-fill sites.

Therefore, if we assume that 20% of in-fill sites are removed when termination is excluded, the

appropriate GSM in-fill radius multiplier for pure LRIC in the v8F model should be set at 1.12.

This is calculated using the equation below, where is the proportion of traffic excluded and is

the cell radius multiplier

√

We note that the majority of in-fill sites will remain when termination traffic is excluded, and as

such are still able to fulfil the other purposes of ‘filling in’ wide-are coverage gaps and providing

additional indoor coverage (as per TeliaSonera’s comment). While a reduction in the number of in-

fill sites for capacity reasons may lead to a marginally reduction in indoor coverage, we note this is

likely to only be equivalent to the marginal reduction of indoor coverage when ‘normal’ capacity

sites are reduced.

The v8F model has been updated such that a different multiplier can be used for the GSM in-fill

radius pre-2006 and post-2008, reflecting the changing proportion of terminating traffic in the 2G

network with the launch of 3G and LTE radio networks. This is particularly relevant in our

considerations of Telenor and NetCom, who operate 2G networks before 2006.



The ‘GSM in-fill radius multiplier for pure LRIC - 2008 onwards’ has been set to 1.10. This is

because the proportion of GSM traffic removed from the total market over the period from 2008 to

2G network shutdown (2008–2020), is approximately 19%. Therefore the GSM in-fill radius

multiplier, calculated using the formula above, is approximately 1.10. A similar calculation on the

cumulative total market GSM traffic pre-2006 gives a removed proportion of 32%, corresponding

to a GSM in-fill radius multiplier of approximately 1.20. Based on this we have maintained the

1.20 value from the v8D model for the “GSM in-fill radius multiplier for pure LRIC - up to 2006”

parameter.

For the same reasons as set out in the GSM network, it is possible to take the view that not all

UMTS in-fill sites are likely to be required when the UMTS traffic load is lower (although there

may be reasons why we believe this effect is less pronounced in UMTS networks such as the need

to retain support for HSPA traffic density). Consequently, we have added a UMTS in-fill radius

multiplier for pure LRIC in the v8F model in the same manner as that used for GSM in-fill sites.

We have analysed the total market UMTS traffic in the same manner as for GSM, as shown in

Figure C.26 below.

Services on the

UMTS network

Routeing

Factor

Traffic

including

termination

Traffic

excluding

termination

Figure C.26: Total

market cumulative UMTS

traffic with and without

termination in the v8F

model (billions) [Source:


3G on-net voice 2.4000 153 153

Incoming voice to

3G from other

networks

1.2000 199 0

Outgoing voice from

3G to other

networks

1.2000 175 175

3G on-net SMS 0.0017 120 120

Incoming 3G SMS

from other networks

0.0009 87 87

3G Outgoing SMS

to other networks

0.0009 118 118

R99 megabytes 5.6000 130 130

Mobile broadband

megabytes -

HSDPA

0.9102 630 630

Minute equivalents

(traffic × RF)

2120 1880

This analysis shows that the cumulative total market UMTS traffic falls to approximately 89% when

voice termination is excluded. This means that we can assume that 11% of UMTS in-fill sites can be

removed when voice termination is excluded. This leads to a UMTS in-fill cell radius multiplier in the

v8F model of 1.06. This multiplier is much closer to 1 than that derived for GSM above, as there is

proportionally much less UMTS traffic avoided when termination traffic is excluded from the network,

due to the requirement to also support mobile broadband (HSPA) traffic.



Telenor states that the reduction of in-fill sites as a result of the exclusion of terminating traffic is

necessary as the building of these sites cannot be justified on a cost (opex and capex) basis when

the traffic volumes are reduced, simultaneously reducing the revenues per site. This reduced

revenue will result in sites being unprofitable and hence Telenor claims the in-fill sites should be

reduced to a point where all those remaining in the model are profitable. While we agreed that in-

fill sites would likely be rearranged in the absence of termination traffic to reduce their capacity

function and increase the coverage function, we do not accept Telenor’s point about economic

viability, as we do not consider it relevant given that many uneconomic sites are deployed in

operator’s networks for ‘coverage’ reasons. The economics of viability are also skewed by the

willingness to pay for voice, termination and data.

We furthermore reject Telenor’s suggestion that there should also be a cell-breathing effect applied to

in-fill sites. This is for two reasons: because in-fill sites by definition fill in gaps, therefore significantly

over-lap the other (wide-area) coverage sites; secondly, by reducing the number of in-fill sites in

response to the reduced load, the average load on the in-fill cells would remain the same (i.e. they

would not ‘relax’). As such the v8F model has been adjusted to remove the cell breathing effect from

3G in-fill sites. For the avoidance of doubt, the effect of cell breathing is retained for the wide-area

coverage sites, deployed in “Stage 1” and “Stage 3” 3G coverage in the model.

C.2.3 Effects of Norwegian topography and population density

Telenor states that the pure LRIC calculations have to some extent taken into account the

indoor coverage – see Section 5 above. In addition, the pure LRIC calculations must take

into consideration the effects of the Norwegian topography and population density.

Telenor agrees that a properly calculated pure LRIC (avoidable cost LRIC) will ensure

prices that are efficient in a static sense. However, Telenor is of the view that in Norway,

where investments to a large extent is coverage driven (as opposed to capacity driven) one

has to take dynamic considerations into account. Termination services are one of many

services being produced on the mobile networks. In a static perspective, given that a

network already is rolled out, avoidable cost will result in efficient pricing. However,

networks (platforms) will over time be upgraded and/or replaced. An efficient, regulated

price must also give sufficient incentives to make such investments. By ignoring coverage

costs, pure LRIC will not ensure such incentives.

The population density of Norway is only one quarter of the least populated countries of

continental Europe.56

Telenor believes that it will be a mistake to use the costing blueprint

developed for the densely populated continental Europe in Norway without further

considerations.

56

The population density of Norway is, 16/km2, whereas the population density for all the countries in continental

Europe is above 90/km2. Germany and UK has a population density of 230 and 260 respectively. Source: http://en.wikipedia.org/wiki/List of sovereign states and dependent territories by population density



When deciding on extending coverage or not, a mobile operator will compare the revenue

generated by a new base station to the costs of investing and running the base station. Only

if revenues are higher than costs, the base station will be built. Now, in the hypothetical

scenario where the terminated volume is removed, the volume of services assumed to be

produced on the base stations will decrease. Hence, base stations, that at the outset was

marginal, will become unprofitable. Thus in the hypothetical scenario without terminated

volumes, the network will cover a smaller area. This effect is evident. The challenge is of

course to assess the size of the effect. It is disappointing that NPT, contrary to what both

NPT and the Ministry said in 2009/2010 have not attempted to assess this effect.57

Telenor would suggest, until better evidence is presented, to assume that one should remove

5% of coverage area for all geotypes under the hypothetical scenario where the terminated

volume is removed.

Main points are (i) The implications on network coverage of deploying the avoidable cost

concept (pure LRIC) needs to be analysed (ii) A pragmatic approach to compensate for

national justification is to include some coverage sites in the pure LRIC calculation.


In the v8F model, as detailed above in C.2.2, both 2G and 3G in-fill sites are removed when

termination traffic is excluded. This site removal is modelling by assuming a larger cell radius (as

detailed above) for the 2G in-fill reducing the number of sites deployed. We highlight that this

adjustment is equivalent to operators covering less area with cell radii kept constant, and as such

captures the effects noted by Telenor.

However, we do not believe that the economic argument presented by Telenor applies to wide area

coverage since this already reflects a general mix of (highly) economic and (highly) uneconomic

sites. Furthermore, given Norway’s seasonal traffic distribution (e.g. many sites deployed in

coastal Southern Norway that are quite empty in the winter periods; sites deployed along long

sparsely populated roads) Telenor’s arguments about the efficient level of coverage are not clear-

cut. For example, Oslo-resident boat owners will value the ability to receive calls whilst offshore

in the summer period, and the call externality arising would lead to lower termination rates.

57

It should also be noted that when the EU commission did its impact assessment in 2009, they analysed effects of

cutting MTRs down to 2.5 Eurocent, which is approximately 0.2 NOK, i.e. roughly 3 times higher than the estimated costs for the generic operator. Hence the calculation results from the NPT model is way below the levels foreseen, and assessed by the commission. See: EU, 2009, COMMISSION STAFF WORKING DOCUMENT accompanying the COMMISSION RECOMMENDATION on the Regulatory Treatment of Fixed and Mobile Termination Rates in the EU, Brussels, 7.5.2009, SEC(2009) 599.



C.2.4 Rollout period

In the LRIC model it is assumed that the generic mobile network is rolled out within one year.

Based on own experience it is Telenor's opinion that a mobile operator needs more than one

year to roll out a nationwide network. The ongoing roll out of the third mobile network in

Norway also indicates that it is necessary with more than one year roll out period. Telenor

suggests that the roll out period for the generic operator is changed to at least two years.


As discussed in Section 4.2, the generic operator is assumed to purchase assets in 2012, prior to a 2013

network launch with immediate scale. We note that no actual operator’s deployment would have the

certainty of immediate scale to an equi-proportional market share, and so the deployment profile seen

under the generic operator would not be expected to match actual operator’s deployment profiles, such

as that undertaken by the third operator. As explained extensively in this document, the generic

operator reflects certain key principles about roll-out and market share development.

The model already includes planning periods for the purchase of new network equipment for all

operators, for example the model buys generic operator’s sites nine months before required

deployment. This means that in the v8F model the generic operator purchases 3256 sites (including

owned and third party sites) in 2011 and an additional 2155 sites in 2012 ready for traffic demand

applied in 2013, and as such the site purchase is already spread across two years as Telenor

suggested. We consider that an efficient operator would not purchase assets two years and nine

months ahead of traffic requirements and that in fact the single year ‘head start’ in deploying the

coverage network is appropriate given the principle of immediate scale.

C.2.5 Voice over LTE (VoLTE)

Telenor states that LTE (4G) is a data only system and it will not support circuit switched (CS)

voice. Thus, alternative systems are required. VoLTE is expected to become the standard for

providing carrier class voice services on LTE networks. Such solutions are not yet rolled out.

Until VoLTE is operational one has to deploy so called <<CS fall-back>> implying that

before making a phone call from an LTE enabled handset, the handset will go off the LTE

network and connect to the 3G or 2G network. There are two drawbacks with this solution.

Firstly, the user may experience delays as the handset goes off and on networks. This delay

will not necessarily be significant for the user. Secondly, as the terminal goes off 4G, all

data sessions will be terminated and then typically be re-established on 3G. Disconnecting

and re-establishing connections in the middle of e.g. a movie or a game is expected to result

in a bad user experience. Furthermore the data speed will typically go down implying that

some types of activities will terminate. If the call has to be handled in 2G the data-

connection will be resumed for the period of the call. This is expected to result in a

significant negative effect on the user experience with 4G terminals.



VoLTE will solve the problem described above by providing carrier class voice by use of

packet switched telephony on the LTE network. Hence voice will be carried by the data

only LTE network. The implication will be that the terminal can remain on the LTE

network even when calls are made or received. The negative effects described above will

accordingly be avoided.

Costs related to VoLTE will typically be related to call servers, gateways, and systems for

providing QoS throughout the network. It is reason to believe that the sending party pays regime

will be maintained also for VoLTE, i.e. receiving calls will be free of charge, whereas the

originator pays a price that must cover the costs of the call end to end. Notice that this is in

contrast to so called OTT voice where the recipient typically pays for airtime on the receiving

end (for OTT we accordingly have a mix of sending and receiving party pays).

LTE is optimised for efficient delivery of data services. This is opposed to 2G and 3G that

to a large extent is optimised for voice. The implication is that it is not unlikely that the

(incremental) cost of producing voice on LTE is higher than the incremental cost on 2G and

3G. Hence it can be a serious mistake to disregard VoLTE costs and VoLTE revenues in the

LRIC context.

With respect to termination it is notable that the argument for not relying on CS-fallback is

related to receiving calls. When making a call the user can deliberately pause the data session,

do the phone call, and then resume the data sessions on LTE. Thus for making calls it can be

argued that CS-fallback would suffice. This is in contrast to receiving calls. It is in this case that

data sessions on LTE become abruptly interrupted. Thus it can be argued that costs related to

VoLTE are driven by inbound calls and costs should be allocated accordingly.

Of course, inbound calls consist of on net calls and inbound calls from other networks. In an

LRAIC context the relevant increment would accordingly be on net voice traffic + inbound

voice traffic from other networks (but notice that on net calls only count once, since the

originating leg is disregarded). The purest approach to avoidable cost in this context is to

consider the cost difference between running the system at actual output, and the cost of

running the system without inbound traffic from other networks.

Specifically with regards to use of a generic operator model with balanced calling patterns

and with 34% market share, only 34% inbound traffic will be on-net. The remaining 66% is

terminated traffic. Thus, when doing the hypothetical exercise of removing terminated

volume, the obvious question is whether the one-third of inbound traffic being on net would

be sufficient to induce the MNO to invest in the VoLTE platform in the first place. Telenor

believes not. There must be a critical volume of calls where users derive benefit from the

VoLTE platform to make the investment economical.

Telenor is accordingly of the opinion that costs related to VoLTE should be accounted for

and furthermore, that there are strong reasons to allocate this cost to termination increment.



Main points:

costs of providing voice over LTE (VoLTE) should be included in the cost base

there are strong arguments in favour of allocating a majority of the VoLTE costs on the

termination increment.


Telenor’s response raises many varied points and future expectations.

► Point 1: The costs providing VoLTE should be included in the cost base

Firstly, while we acknowledge Telenor’s point that VoLTE will likely require additional

investment and cost related upgrading certain network infrastructure, the additional platform costs

of providing VoLTE should be allocated generally to voice traffic carried by the 4G network not

2G or 3G voice. 4G voice traffic (comprising either OTT or VoLTE is a very small proportion of

traffic within NPT’s period of regulatory interest) and so any contribution from VoLTE platform

costs will also be small.

Secondly, VoLTE will share significant radio resources with fast mobile broadband services,

where this sharing exists in a common air interface. This contrasts with separate air interfaces for

3G voice (R99 QPSK CE) and HSPA data (16-64QAM codes). Thus, we would expect that the air

interface costs of VoLTE will gain significantly from the shared economies of scale in the LTE

radio layer, with a significantly lower cost per megabyte of voice traffic to reflect this. As the

amount of megabytes in the 4G network rises (as industry commentators expect) then there will

also be significant economies of scale in many cost elements – e.g. in shared high-capacity IP

backhaul, which will be occupied mainly by data traffic.

Given these points, we reject Telenor’s suggestion that VoLTE costs need to be included and will

lead to higher costs.

► Point 2: The majority of the VoLTE costs should be allocated to the termination increment.

Furthermore to the above points, we note that if VoLTE is already expected to be more costly than

2G or 3G voice then Telenor should begin to seek ways of reducing that cost to the significant

benefit of incoming callers, or not deploy 4G until it can recover any higher costs from its 4G

customers rather than incoming callers (e.g. implement a form of ‘receiving party contributes’ for

any additional cost of traffic higher than the currently calculated 2G+3G cost, especially if Telenor

considers that it has only incurred the VoLTE system cost for receiving calls). As such if Telenor’s

comment on the high cost of VoLTE is true, then efficiency principles further dictate that current

VoLTE costs should not be allocated to the termination increment even for 4G voice.



With regard to the issue of interrupting data sessions with incoming calls on CS-fallback (CSFB),

and managing the quality of experience, we suggest that:

Telenor delays the introduction of circuit voice on 4G terminals if CSFB is considered too

detrimental to user experience, or seek to implement a call signalling solution which enables a 4G

handset user to decide whether to accept an incoming call during a data session, or to reject the call

to voicemail if they are occupied on the internet, streaming or other data usage. This approach

would also reveal whether there is a call externality (i.e. whether 4G handset users value the

acceptance of an incoming call, even if they are otherwise occupied on a data session).

Telenor explains clearly to 4G customers (if the problem remains in the future) prior to the

sale of a 4G handset that incoming voice calls may terminate users’ data sessions and the

opportunity to be called on a 4G phone number may result in a lower quality of experience.

However, Qualcomm has stated in its recent overview of CSFB that for recent releases (i.e.

Release 9) CSFB additional call setup delay is reduced to just 0.5 seconds.58

This matches details

by Nokia Siemens Network’s that “the call setup time with CSFB is only a fraction longer than in

current 3G networks”.59

Therefore, while we appreciate that operators will likely choose to move

from CSFB to VoLTE in the long term, we do not believe the reason is just “related to receiving

calls” as Telenor claims (given these can already be done under CSFB), but rather related to other

network and consumer factors (such as interrupting data sessions).

In conclusion, given the principle of not explicitly modelling 4G assets and costs, and the fact that

VoLTE (and all related infrastructure) is specifically related to providing 4G voice services, we have

not adjusted the v8F model to include VoLTE costs or to allocate these costs to the voice increment.

C.2.6 UMTS-only network

Tele2 notes that AM has presented one alternative scenarios to the basecase scenario. The

impacts on the LRIC+++ results are significant. Scenario [2a] models a UMTS network

scenario (i.e. no 2G). Tele2 is of the opinion that Scenario [2a] is not realistic.


To be clear, the scenarios included in the sensitivity presentation are provided only as sensitivity

tests.

58

See http://www.qualcomm.com/media/documents/files/4g-world-2012-csfb.pdf

59 See http://www.nokiasiemensnetworks.com/sites/default/files/document/cs_fallback_brochure_final.pdf

http://www.qualcomm.com/media/documents/files/4g-world-2012-csfb.pdf



C.2.7 Number of sites and population coverage

Tele2’s plans include more than []. This means that Tele2 will have achieved more base

stations by the end of 2013, than the model assumes in aggregate. Mobile Norway’s network

will reach 75% population coverage in []. Mobile Norway has planned to extend its

network to 100% coverage. The current plan is that this population coverage will be reached

in []. We understand that AM has included the following number of sites in the model:

[] 3G coverage sites; B6_NwDes rows 1022/1069

[] 2G coverage sites (co-located with the 3G sites); B6_NwDes row 77

[] tunnel sites; B6_NwDes row 1095

Tele2 finds that the model assumption underestimates the number of sites required to obtain

a sufficient quality to attract subscribers in competition with Telenor and TeliaSonera. This

means that the market share assumptions (i.e. 28% market share) used in the model is not

realistic.

Tele2 has asked AM how would it affect the model if the number of sites required to achieve

85% population coverage is [] (closer to Tele2’s estimate). AM has responded that:

"This would assume either smaller UMTS cell radii by Fylker or higher area coverage by

Fylker than are currently modelled. This would increase the number of UMTS coverage

sites of both Mobile Norway and (indirectly) the generic operator (whose cell radii are an

average of the actual operators)."

Tele2 is of the opinion that the Tele2 specific model must be updated to reflect this.

In Tele2’s opinion, the effect of increasing the number of sites in Tele2’s network, would

result in an higher number of sites in aggregate (i.e. the sum of all Telenor, TeliaSonera and

Tele2’s sites). Logically, this would increase the aggregate cost and we therefore expect this

to have a significant effect on the generic operator model, i.e. with increased costs and

consequently increased output results for LRIC+, LRIC and pure LRIC.

Tele2 does not have the opportunity to test how this will affect the generic operator model

since Tele2 can only change the input assumption in the Tele2 specific model. We kindly

ask the NPT, or AM, to present revised results based on the current modelling approach

with the revised input. In order to reflect Tele2’s current roll-out plans, we suggest that a

new scenario, where 100% coverage for Mobile Norway’s network is occurring, somewhere

between [] is tested to show the impact on the result on the generic operator model.

It should be noted that this assumption is depending on Tele2’s access to the frequencies in

the 800MHz, 900MHz and 1800MHz. Since these frequencies will be auctioned in 2013 it is

not possible for Tele2 to provide any firm estimates until the auctions have been concluded.




In its comments Tele2 raises two points regarding Mobile Norway’s network which we have

investigated in turn.

In the v8D model, Mobile Norway is considered to reach long term population coverage of 85%

and a market share of 28% (= 85% × 33% retail share). However, considering Tele2’s comments

that Mobile Norway’s aim is to reach 100% population coverage over the timescale of the model,

we have further investigated Mobile Norway becoming a national operator over the long term.

To achieve this, we have adjusted Mobile Norway coverage forecasts so that the stage 1 3G

2100MHz wide area coverage is set as []. We note that this is considerably higher than was

previously assumed. The total 3G coverage was adjusted to give a population coverage of []

which generated the same site numbers as stated by Tele2. In addition, to keep Mobile Norway’s

actual historic data calibrated, the operators ‘Stage1: 3G coverage completion lag after 2012’ was

increased to five years and the ‘UMTS900 coverage completion lag’ was increased to four years.

These changes led the Mobile Norway model to deploy []. Whilst this leads to a reasonable

increase in the cost of termination for Mobile Norway, it has a marginal effect on the generic

operator. Since all three actual operators were now considered national, this led to an increase in

the generic operator’s market share to 33% of voice traffic.

The results of this change in the Mobile Norway coverage in the v8D Mobile Norway model can

be seen in Figure C.28 below.

Figure C.27: Mobile Norway results in the v8D model

basecase [Source: Analysys Mason, 2013]

Figure C.28: Mobile Norway results in the v8D model

with Mobile Norway as a national operator [Source:


[] []

Below, in Figure C.30, we demonstrate the impact this change has on the generic operator’s results.



Figure C.29: Generic operator results in the v8D


Figure C.30: Generic operator results in the v8D

model with Mobile Norway as a national operator


However, as noted in Section C.1.8, assuming 100% population coverage is inconsistent with the

network analysis that informed NPT’s pricing decision June 2012, and as such the coverage level

is retained at 85% as in the v8D model.60

In relation to Tele2’s second point, currently the v8D model reaches [] sites under the 85%

coverage assumptions. We have investigated the changes required to achieve the level of [] sites

as proposed by Tele2, including both reducing the UMTS cell radii by Fylker and increasing the

proportion of coverage by Fylker compared to the levels assumed under the v8D model.

In consideration of adjusting the cell radii, we set Mobile Norway’s cell radii in each Fylke’s equal

to the minimum of the three operators’ calculated cell radii. Even with this change we were unable

to reach the levels of sites proposed by Tele2, and note that [].

In consideration of adjusting the coverage area, we do not think is it appropriate to adjust the total

3G coverage (as this level is defined by the 85% coverage population) so instead we have

investigated increasing the UMTS2100 forecast population coverage, capped in each Fylker to not

rise above the UMTS2100 national population coverage levels used for the generic operator. This

resulted in overall national population coverage of [] for the UMTS2100 network, and achieved

Tele2’s proposed site levels. We note that this affects the Pure LRIC even less than the change of

Tele2 to a national operator, and has negligible effect on the generic operator’s costing.

60

See http://www.npt.no/marked/markedsregulering-smp/marked/marked-7/_attachment/2347?_ts=139b9c3911d

0.153 0.145

0.133 0.124

0.084 0.077

0.070 0.063

0.088 0.078

0.065

0.054

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0.18

Mob

ile v

oic

e te

rmin

ation

costs

(N

OK

no

min

al)


0.158 0.149

0.137 0.127

0.086 0.079

0.071 0.065

0.088 0.078

0.066

0.054

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0.18

Mob

ile v

oic

e te

rmin

ation

costs

(N

OK

no

min

al)




Information and new understanding of the Mobile Norway network gained in the recent years, now

that the network is fully operational, indicates that it does appear necessary for Mobile Norway to

‘densify’ its coverage network. This is particularly because it does not have its own nationwide

900MHz GSM network to ‘fall back on’ and voice and data service customer experience is

improved by having better contiguous coverage within its own network.

This increase in the UMTS2100 coverage would also act to increase the quality of the 3G network

and as such negates Tele2’s point that a modelled low quality 3G network would be unable to

attract the sufficient subscribers to reach the 28% market share in the v8D model.

C.2.8 Coverage for generic operator

TeliaSonera makes reference to section 4.2.1 in the model documentation, where it is

assumed that generic operator will establish a UMTS network covering 99.9% of the

population.

Existing UMTS network gives coverage to less than 99.9% of the population. TeliaSonera is

therefore of the opinion that it is unrealistic to assume coverage of 99.9% in the model for

the generic operator, and that this coverage assumption must be reduced.

Further is TeliaSonera of the opinion that it must be reflected in the model that the “last”

coverage percentages are relatively more expensive to roll-out than the “first” percentages.

Deployment of UMTS to 99.9% of the population will require new transmission to a large

number of base stations. This must be taken into considerations if coverage for generic

operator is going to be reduced in the model. Modelled network cost for the generic operator

is as we understand based on average costs from the operator-specific models, and without

adjustment of the network costs the modelled costs for generic operator will be too high.


The generic operator is modelled to match the average existing UMTS2100 coverage seen in

actual operators in 2012 (as per the principles of immediate scale), therefore these coverage levels

are by definition realistic.

The UMTS900 is intended to extend 3G coverage to reach the same coverage level as 2G in the

long run, as was detailed in the previous model document61

, and as such we do not believe it is

necessary to adjust the final level of 99.9% population coverage for the combined ‘Forecast 3G

coverage area (2100MHz+900MHz)’. However, following TeliaSonera’s comment we have

adjusted the time required to reach the ‘Forecast 3G coverage area (2100MHz+900MHz)’ of

99.9% to eight years in the v8F model, equivalent to the 2G shutdown point, which matches the

timeline used for each of the actual operators.

61

See Section 4.1.2: “UMTS900 is then used to increase the UMTS coverage to equal to GSM coverage”



We acknowledge TeliaSonera’s point that covering the final percentage of population is more

expensive than covering the initial percentages. This effect is already taken into account in the

model through conversion of population coverage into area coverage, and site deployment being

based on this area coverage. Therefore, in line with TeliaSonera’s comments, the final percentage

of population coverage does account for a significant increase in the required area coverage

(beyond that required for initial percentages), and hence a large increase in the required number of

sites and therefore cost.

Given this ‘increased population coverage cost’ effect is already included in the model, and the

generic operator coverage has only been adjusted to further reflect the actual operators coverage

deployments in v8F, we consider that the average operator site costs are still appropriate for the

generic operator.

As can be seen in Figure C.32, this change has a small impact on the v8D results.




model with 3G forecast end point changed to 2020


0.153 0.145

0.133 0.124

0.084 0.077

0.070 0.063

0.088 0.078

0.065

0.054

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0.18

Mob

ile v

oic

e te

rmin

ation

costs

(N

OK

no

min

al)


0.156 0.147

0.135

0.124

0.086 0.079

0.071 0.064 0.088

0.079

0.066

0.054

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0.18

Mob

ile v

oic

e te

rmin

ation

costs

(N

OK

no

min

al)




C.3 Comments related to the costing calculations

C.3.1 Economic depreciation

Telenor has repeatedly raised our concern over the method deployed for allocating costs over

time as part of our consultation responses to the previous revisions of the model. The method

for allocating costs over time is claimed to be ‘economic depreciation’. A more suitable

description would be ‘intertemporal average cost’. One of the drawbacks with the deployed

method is that historic events are decisive for setting future termination rates. The EU

recommendation on terminations rates' is clearly stating in the preamble, paragraph 13;

Taking account of the particular characteristics of call termination markets, the costs of

termination services should he calculated on the basis of forward-looking long-run

incremental costs (LRIC) ". (Our emphasis.) Telenor is of the opinion that the methodology

deployed in the Norwegian model should have been discarded for this reason. It should be

noted that the generic operator is assumed to start operation in 2012. Hence there is no history

for the generic operator, and the model is by construction forward looking for this operator.

Furthermore, since Tele2 has a relatively short history the error in their model is presumably

small. Below we will provide an illustration that highlights the problem within the operator

specific models for Telenor and NetCom.

Version 8 of the NPT model calculates two flavours of a pure LRIC measure. The preferred

measure of pure LRIC seems to be to deploy the AM economic deprecation formula on the

avoided cash flow. It seems to be preferred since it is indicated in bold on the A0 Ctrl sheet and

because the slide deck, page 57, from the operator meeting, March 1, states that "focus" is on this

measure. For short we call it pLRIC1. The second measure of pure LRIC unit costs is calculated

by taking the difference in economically depreciated costs with and without the terminated

volume. We call this unit cost pLRIC2. Notice that both measures of pure LRIC ensure that

"avoided cost" are approximately covered in NPV terms over the lifetime of the network.

To illustrate how calculation results depend upon historic events we have done some

alterations on historic numbers in the Telenor model the period 1994–2001. Hence, all

changes are at least ten years old. We have made four changes to the model:

1. We assume that the digital mobile penetration for the period 1994–2001 is identical

to the penetration in 2002; 85.3%

2. We assume that the total outgoing voice minutes per subscriber per month for the

period 1994–2001 is identical to the volume of outgoing voice in 2002; 94 minutes

3. We assume that the on net proportion for the period 1994–2001 is identical to the on

net proportion in 2002; 50%

4. We assume that the total incoming minutes per operator for the period 1994–2001 is

identical to the incoming minutes in 2002; 1 939 345 000



We have accordingly replaced the actual, gradual increase in volumes with an immediate volume

equal to 2002, and then deployed this for the period 1994–2001.62

In a forward looking model,

changing historic numbers, more than ten years old, should of course not affect the calculation

result. However, as the illustration below reveals, it has a dramatic effect.

Figure C.33: Comparing pure LRlCs, Telenor model, AM basecase and hypothetical increased volumes for

the period 1994–2001 [Source: Telenor, 2013]

[]

The illustration above demonstrates that the estimated pure LRICs are a function of historic

events. Changing historic assumptions have large impact on cost estimates for later periods.

Thus, they are not forward looking. This is directly inconsistent with the EU

recommendation cited above. A natural consequence of this observation is that the operator

specific models starting in 1992 cannot be used for calculating future termination rates.

Inspection of calculation results from the generic model presented at the CTRL sheet

suggests that the difference between the two measures of pure LRIC is close to zero up until

2019. However, for 2020 the difference is significant. Below we have plotted the two

measures of pure LRIC for the generic operator from 2019 and onwards. As we can see, the

difference is significant. In 2041, the pLRIC1 is almost twice as high as pLRIC2.

Figure C.34: Pure LRIC for the generic operator [Source: Telenor, 2013]

[]

As we can see, the difference is significant, also for the generic operator. In the slide deck

presented by AM on the operator meeting March 1, they did not elaborate on the two

different measures of pure LRIC. The slide deck (p. 57) states that "focus" is on pLRIC 1,

not pLRIC2, without any further discussion. Both estimates can of course not be correct.

Telenor urges NPT/AM to explain the difference and provide some arguments as to why

focus is on one over the other.

Main points are (i) The Telenor and NetCom operator specific pure LRIC model results are

not applicable for setting termination rates (ii) NPT/AM must explain the difference

between the two, significantly different, calculated measures of pure LRIC and present

arguments in favour of the measure they recommend.

62

It should be noted that there may be better ways to implement the changes (i.e. also change assumptions on the

dimensioning side), but the point we are making is that the future pure LRlCs are history dependent. This conclusion is of course robust for improvements on the 1994 - 2002 assumptions. The point is that if we change history, the future will also change. Thus the model is dependent upon historic events and thus fails to be forward looking.




Telenor’s long-past historic adjustments to the model, while interesting and highlighting the point

regarding the depreciation, simply do not reflect the market which existed at that time. Instead,

Telenor should focus on changing the assumptions used in the forward-looking period, to show the

effects of plausible future changes on the cost results using different depreciation methods.

Changing historic facts is not plausible, irrespective of whether the chosen depreciation method

relies upon such time periods in doing its calculation.

Because we have a multi-year model and NPT is setting a multi-year price control, it is necessary

to undertake a form of ‘inter-temporal average’. This averaging method brings multi-year stability

as otherwise, the different incremental costs in each year could give rise to pure LRIC results that

do not vary in an intuitive fashion, possibly becoming even zero or negative as avoided traffic over

time leads to re-phasing of large asset deployments.

Below we repeat part of our response from ‘NPT’s mobile cost model version 6’ in 2009.63

“Analysys Mason rejects Telenor’s assertion on the lack of attention in this area: time has been

spent understanding the interaction between economic depreciation and pure incremental costs in

our mobile cost models, including NPT’s v5.1. Only in simple or certain types of models (e.g.

single year, static depreciation calculation) will the pure LRIC calculation of avoidable costs

function in a uniform way. In models that are developed to be more complex (e.g. reflecting the

actual time-evolution of operators, using non-uniform demand profiles) then calculating the

avoidable cost of one particular service will require investigation.

As a result of our work in this area, we have concluded that whilst the full-time-series economic

depreciation method fits well with the calculation of LRIC+++ costs over the lifetime of the

modelled mobile businesses, it is less suited to calculating avoidable costs in some situations. In

particular, this unsuitability arises when the avoidable increment of demand is not a uniform

proportion of demand over time (as is the case with wholesale mobile termination supplied over

Telenor’s and NetCom’s networks). As Telenor has submitted, this results in (undesirable)

increased inter-temporal effects, which means that while costs may be (overall) lower without

wholesale termination, cost recovery is also moved forwards/backwards in time according to the

profile of residual demand applying to each network element. With data services more important

in the later years, this can mean that costs without wholesale termination are postponed more into

the future relative to the all-service calculation. As such, unconstrained pure incremental costs

can be very low or negative in later years.

63

See

http://www.nettvett.no/ikbViewer/Content/113945/Model%20documentation%20for%20NPT%20with%20responses_Public_011209.pdf



Therefore we believe that an alternative calculation of pure LRIC is required. We adopted an

alternative approach in the draft v5.1 cost model issued to industry parties. In this situation, the pure

LRIC is calculated from the (present value) difference in network expenditures arising from the

removal of the wholesale termination volume, constrained over time so that the underlying equipment

price trends apply also to the pure LRIC components of cost (just as these price trends apply to the

LRIC+++ components of cost). We consider it reasonable that the calculated pure LRIC is directly

constrained by the equipment price trends for the same reason that the LRIC+++ should reflect the

(declining) underlying costs of supporting traffic volumes with network equipment.

It is plausible (counter to Telenor’s suggestion above) that the removal of mobile termination

volumes causes certain costs to be higher than otherwise: as an example, we could consider radio

sites and ancillary costs, which have an increasing site cost trend (as in the revised v6 model).

Removing termination volumes may result in some sites being deployed later in time – with rising

site costs, it could indeed appear (in particular years) that costs were higher without termination.

However, over the lifetime of the network, as in our pure LRIC calculation, this effect does not

dominate the overall cost calculation and the pure LRIC is calculated to be a small (positive)

value per minute when compared to the LRIC+++.”

Therefore, as stated before, “we believe that our primary method of calculating the pure LRIC is

reasonable and justified, and a suitable basis on which NPT could consider setting pure-LRIC-

based rate regulation.”

The table below indicates the difference in the pure LRIC of voice termination when calculated

with both the “economic costs of the difference” (Method A, as proposed for the v8D model) and

the “difference in the economic costs” (Method B, as suggested by Telenor in the 2009 update). In

particular, we emphasise that Telenor’s proposed implementation gives rise to negative values for

the pure LRIC of voice termination in later years (between 2018-2020). We would also observe

that the two methods give very similar answers for the generic operator in the period 2014–2020.

Figure C.35: Comparison of model outputs in the period 2014–2020 using two methods of pure LRIC [Source:

NPT v8F model, 2013]

Modelled operator Results comparison

Telenor and NetCom Method B at least 2.5 øre lower

Mobile Norway Method B at most 1 øre higher

Generic operator Method B at most 0.5 øre higher

For the avoidance of doubt, we propose to use Method A for the pure LRIC implementation.



C.3.2 Cost trends

On sheet CO1_CostTrends, the development in unit prices is documented. It is assumed that the

price of most types of equipment will continue to fall over the modelling period, whereas the cost

of site acquisition and buildings are assumed not to change. These assumptions are based on

forecasts and cannot be verified. Telenor is of the opinion that the cost trend assumptions chosen

by NPT/AM are inconsistent. One must either:

i. Assume that all prices remain constant over the modelling horizon.

ii. Or assume that there are different price trends for different goods and services;

Based on experience there is reason to believe that the price of electronics will decrease.

Based on experience there is reason to believe that the price for land and labour will increase.

Hence, the modelling assumptions should be changed, either one should assume an increasing price

trend for buildings and site acquisition, or one should assume constant price of all types of capex.


To clarify, the cost of site acquisition and buildings in the v8D model are only constant in real

terms. This means the model assumes that the amount actually paid by operators (as recorded in

their accounts) will increase over time in line with inflation. In addition, we believe that Telenor’s

comment that “the cost trend assumptions chosen by NPT/AM are inconsistent” does not hold

given that our cost trends do “assume that there are different price trends for different goods and

services” – with capex assets grouped into many separate categories, each defined on the

C01_CostTrends worksheet, and opex cost trends defined on a per asset basis on the same

worksheet.

However we believe that Telenor may also be commenting that they consider the adjustment on

the v8D cost trends to be too low for buildings and site acquisition costs – an issue that was

discussed in the previous modelling project. To set the v6/v7 model cost trends, two sources of

data were used from Statistics Norway: the construction cost index for road construction, and the

construction costs index for works in offices and commercial buildings64

between 2000 and 2009.

To investigate Telenor’s comment we have reconsidered these cost trends using the updated series

to the present. The compound annual growth rate (CAGR) of these indices, in real terms, for both

the 2000–2009 and 2009–2012 periods are summarised below in Figure C.36.

64

http://www.ssb.no/bkiror_en/



Figure C.36: CAGR in real terms for both cost trend periods, by index [Source: Analysys Mason calculated

from Statistics Norway, 2013]

Index Category 2000-2009 CAGR65

2009-2012 CAGR

Road

constructions

Total costs 2.1% 2.4%

Labour costs 2.9% 2.1%

Works in

commercial

buildings

Total costs 3.7% 2.2%

Administrative expenses 1.9% 1.4%

Labour costs 2.5% 1.8%

Materials 4.8% 2.5%

Based on the analysis of these trends, we have therefore amended the v8F model cost trends to

2018 (matching the principle of forecasted trends as used in the v7.1 model) as follows:

site acquisition and civil works now have a 2009–2018 capex trend of +1.5% (based on the

trend for labour and administrative costs)

radio ancillary equipment now have a 2009–2018 capex trend of +2.5% (based on the cost

trend of materials)

We note that this has not significantly affected the calibration or reconciliation of the actual

operators and has a limited effect on the generic operator costing of LRIC, LRIC+++ and Pure

LRIC and shown in Figure C.23.


model basecase [Source: Analysys Mason, 2013] Figure C.38: Generic operator results of the v8D

model with updated cost trends [Source: Analysys

Mason, 2013]

65

These CAGRs have changes slightly since the v7.1 model documentation as both the final 2009 inflation and index

values have been revised since the v7.1 documentation was written. These changes however do not affect any of our previous conclusions.

0.153 0.145

0.133 0.124

0.084 0.077

0.070 0.063

0.088 0.078

0.065

0.054

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0.18

Mob

ile v

oic

e te

rmin

ation

costs

(N

OK

no

min

al)


0.1540.146

0.1350.125

0.0840.078

0.0700.064

0.0880.079

0.066

0.055

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0.18

Mo

bile

vo

ice

te

rmin

atio

n c

ost (N

OK

no

min

al)

LRIC +++ LRIC Pure LRIC



C.3.3 Reflection of actual roaming tariffs in the MTR price cap

During the meeting with the NPT and AM on 26 February 2013, Tele2 asked how Tele2 and

Network Norway’s roaming traffic was modelled, and at what cost. AM stated that NR tariffs

are assumed based on a growing traffic volume in Tele2’s network and that there is an equal

increase in the retail market share for Tele2 and Network Norway. AM has assumed that NR

tariffs will be reasonable, but was not very specific on the actual levels of the tariffs.

Further, Tele2 understood that AM explained that AM assumed that Tele2/Network Norway

will be able to negotiate “a reasonable cost level” for the traffic not covered by Mobile

Norway’s network.

AM explained in the meeting that there were two alternative solutions to model NRA costs. The

first was to model a scenario with 85% coverage (i.e. the basecase scenario in test 7). The second

was to model a scenario with 100% coverage (i.e. Scenario [7b]). In both scenarios it could be

assumed that the generic operator has one-third of the market (implying a different set of costs).

In order to understand this better, Tele2 therefore asked a follow-up question, asking AM to

explain where, in the model, the NR tariff assumptions are shown and to elaborate on what

AM consider as a reasonable annual cost level for 15% (assumed by AM in the model) of

Tele2’s traffic when the OPEX related to 85% of the traffic is approximately MNOK 460 in

the long run?

AMs response to the first question was:

"[..] To be clear, the cost of national roaming traffic is included in the model in terms of

what it costs the incumbents to carry that traffic. The assumed Tele2/Network Norway

national roaming traffic is split between Telenor and TeliaSonera and that traffic is added

to their operator-specific models. Therefore, the cost of that traffic is part of the overall

long-run costs incurred by the incumbents.

The price paid by Tele2 to the incumbents for the national roaming traffic is not included in

the model. Instead, the cost calculated for Tele2 is only its own cost for carrying its own

network traffic. We do not model the transfer charges internal to the market for national

roaming traffic (origination and termination and data)."

AMs response to the second question was:

"The cost model and various other sources could be used to establish "a reasonable cost

level" for national roaming traffic. However, it is not the purpose of the project to calculate

this result. As described, the model takes the approach of taking all traffic and splitting it

between the three networks, such that national roaming traffic is contained within the cost

calculation of the incumbents. The implication is that all the traffic within all three network

models is 'long-run cost based'.".



Tele2 understands that a large proportion of terminated minutes to Tele2 and Network

Norway’s subscribers (i.e. the roaming share) will be included in the model and counted at

the LRIC cost for TeliaSonera and/or Telenor. It follows from this, that a large portion of

Tele2 and Network Norway’s forecasted traffic (derived from the assumed retail market

share for the generic operator) will not be calculated with a cost that reflects the actual NR

tariffs paid to TeliaSonera and Telenor.

Should the price cap be determined based on the generic operator model, this means that

Tele2 will have a price cap for terminated minutes that may be below the actual NR tariffs.

The model results (regardless of whether it is the LRIC+++, LRIC or pure LRIC) show a

price level which is significantly lower than the current NR tariffs.

The NPT has itself assumed in its decision on the 26th June 2012 that neither Telenor, nor

TeliaSonera, will have incentives to offer national roaming at reasonable tariffs, if at all.

Tele2 will therefore be put at a serious competitive disadvantage, compared to Telenor and

TeliaSonera, as long as there is a delta between the actual NRA tariffs and the regulated

MTR price cap.

[]


The issue of national roaming sits outside the remit of the LRIC model. NPT is aware of Tele2’s

comments on the relationship between national roaming and MTRs.

C.3.4 Licence costs

Tele2 is of the opinion that the licence costs assumed in the model are underestimated.

Significant spectrum allocations will take place in 2013/2014. Although it is not possible to

predict the outcome of the auction, it must be safe to assume that all MNOs included in the

generic operator model will seek to acquire the frequencies in the 800MHz, 900MHz and 1800

MHz bands that will be auctioned in 2013/2014. It is a significant amount of the total spectrum

allocated for mobile services and the expected auction proceeds should be included in the model.

[]

TeliaSonera refers to section 5 regarding Mobile network design. With regard to renewal of

spectrum licences, it is mentioned in section 5.3 that for the 1800 and 900 MHz bands

licences will be renewed every 12 year and the licences in 2.1 GHz band every 20 years.66

66

Jf. SDs dokument ”Høring om tildeling av frekvenser i 800 MHz-båndet herunder spørsmålsstillinger knyttet til

tildeling av frekvensressurser i 900 og 1800 MHz-båndene” av 20. mai 2011



TeliaSonera would like to draw your attention to the Ministry’s hearing in 2011 regarding

allocation of available spectrum in the 800, 900 and 1800 MHz bands where it is proposed that

new licences in the 900 and 1800 MHz bands will have duration for 15, and not 12 years.


The May 2011 Høring om tildeling av frekvenser i 800 MHz-båndet herunder spørsmålsstillinger

knyttet til tildeling av frekvensressurser i 900 og 1800 MHz-båndene consultation document referred to

by TeliaSonera discusses licence duration in Section 5.4. For the avoidance of doubt, while discussed

in the consultation document, the 800MHz band due to be auctioned in 2013/14 are not within the

scope of the v8F model, as they are not relevant to the modelled 2G/3G networks.

While this document recommends a licence duration of 15 years, the Ministry has decided that the

upcoming spectrum permissions will have a 20-year licence period: “Tillatelsene i samtlige tre

frekvensbånd skal tildeles med varighet på 20 år.”67

However, the fees paid historically were for licences of a particular duration. If we revise the

periodicity of the licences, then the licence cost must be revised (in some way) to reflect the value

of the longer lifetime of the licence. While this adjustment is simple in theory (using the

discounted value of additional years), the result is to ensure any increase in value should offset the

increase in lifetime, and result in the same overall cost per unit of traffic. Therefore, for simplicity,

we recommend leaving the currently modelling approach unchanged from the v8D model, and as

such the model represents actual historic data.

Similarly, in reaction to Tele2’s comments, we do not believe any adjustments to the modelling to

reflect uncertain predictions of the 2013/14 auction outcome would be appropriate and have

chosen to retain the approach to licences from the v8D model. As shown by recent auctions in

Europe, it is not safe to rely on prior expectations for the payments which may be bid for

spectrum.

67

See: http://www.npt.no/teknisk/frekvensauksjoner/planlagte/auksjon-14-790-862-

mhz/_attachment/7072?_download=true&_ts=13e174ca875



C.3.5 Cost assumptions for sites

The model assumes that nearly all Mobile Norway sites should be third-party sites. Tele2

understands that the average capex/opex for a third-party is found in cells

'C04_UnitCapex'!X103:X114 and 'C07_UnitOp'!X103:X114 respectively.

Tele2 finds that these costs are too low compared to the actual costs.

The costs for Mobile Norway’s own sites (roof top og GF) must be adjusted from

approximately [], and the cost for shared sites must be increased to approximately []

per site to reflect that Telenor is charging Mobile Norway for all relevant costs on site

modifications (in some instances new masts must be established) required to meet Mobile

Norway’s requests for co-location. The above figure is based on an estimate that

approximately [] of all co-location requests results in modifications costs that Mobile

Norway must cover.

In addition, following a request for a description on how Tele2 has arrived at the cost values stated

by Tele2:

Following NPT’s question, Mobile Norway reviewed the estimates of the average costs for

greenfield (GF), Roof-top (RT) and shared sites (SS). The results show a higher cost for

Greenfield and Site sharing than the figures presented in our response [shown above].

Mobile Norway’s own cost estimates show an average cost per site as follows:

Figure C.39: Blended average site cost for Mobile Norway; all figures in NOK [Source: Tele2, 2013]

[]

‘A’ shows the average costs for radio-planning, site acquisition and civil works.

‘B’ shows the average cost when the cost for upgrade have been included. The upgrade

costs relate to the upgrade of all existing sites (2G on 3G sites and 3G on 2G sites) as well

as expected future upgrades (to three to five systems) based on the current average price for

upgrades in the vendor contracts.

The estimates show a blended average for all of the [] (existing and planned) in Mobile

Norway’s network plan. The estimates are based on the historical costs for the [] sites

already build and the expected cost for the additional [] sites.

The costs have been allocated to Greenfield (GF), Roof Top (RT) and shares sites (SS)

according to the existing, and expected, relative share of each category of the total number

of sites.




We have investigated the additional site cost information supplied by Tele2. While Tele2 believe

that their modelled site costs should be higher, our consideration of their supplied data has

suggested that some assets also need reducing in cost to reach their average unit cost figures.

Tele2 has supplied a single point unit cost, representing a blended average site cost across time.

We have compared this to a weighted average created using modelled asset purchase timings

(under the Mobile Norway national population coverage case) multiplied by the nominal site costs

in each year. We assume that Tele2’s site costs for ‘Greenfield’ map to the average of the v8D

model’s 2G-only and 3G-only ‘owned’ assets, whereas ‘Roof top’ and ‘Shared sites’ are both

considered to map to the v8D model’s 2G-only and 3G-only ‘3rd party’ assets.

The supplied ‘A’ category costs are believed to relate to the sum of the ‘site acquisition and civil

works’ and ‘ancillary/permits’ asset costs, whereas the difference between the ‘B’ and ‘A’

category (i.e. the upgrade part of the costs) relate to the ‘2G/3G ancillary/permits’ assets.

On this basis we have updated the Mobile Norway asset unit base costs to match the data supplied

by Tele2. This change also affects the generic operator, resulting in new site costs as shown below.

2013 unit costs

(NOK nominal) v8D model v8F model

Figure C.40: Generic

operator ‘site acquisition

and civil works’ and

‘ancillary/permits’

combined site costs,

[Source: Analysys

Mason, 2013]

2G-only owned tower site 1 113 000 1 196 000

2G-only 3rd

party site 643 000 637 000

2G/3G owned tower 351 000 311 000

2G/3G 3rd

party site 351 000 311 000

3G-only owned tower site 1 344 000 1 196 000

3G-only 3rd

party site 559 000 614 000



C.4 Process-related comments

C.4.1 Impact assessment of introducing pure LRIC

In 2010 the NPT stated that:

“161. NPT will make a new calculation based on pure LRIC by the end of 2013. In the course

of the next regulatory period, NPT will also do a more detailed impact assessment by only

including traffic-related costs and exclude all coverage costs from the LRIC calculation in

Norway. Since LRIC model uses conservative estimates for traffic growth and development of

equipment costs, etc., NPT deems using LRIC as a basis for price regulation in the next

period will not pose any likely risk of undercoverage. If NPT were to find it appropriate to

introduce pure LRIC after 2013, these results would likely be considerably lower than the

present results based on LRAIC.” 68

AM has explained that the model will reveal the amount of coverage related costs which

allows interested parties to run the model where the coverage adjustments are not included.

Tele2 is of the opinion that pure LRIC is not suitable to reflect the specifics of the Norwegian

mobile market. Tele2 reserves further comments until the NPTs assessment is made available

to Tele2.

These comments are considered to be out of scope of the model documentation and discussion and

will be dealt with as part of NPT’s price decision.

C.4.2 Determination of price caps - Generic model cost or highest specific operator cost

In previous M7 decisions the price caps have been determined based on the operator with the

highest costs. Tele2 is of the opinion that price cap regulation must secure cost recovery and that

the same principles must be used, i.e. that the operator with the highest cost must be used.

Given that all MNOs now are regulated to a symmetric MTR level, this determination will

have a significant impact on the competition in the market.



68

The English text is from the NPTs notification to ESA on 25 August 2010 and is repeated in NPTs decision from

27th September 2010 paragraph 166.



C.4.3 Consultation on conceptual paper

TeliaSonera states that the conceptual issues in connection with the modelling could be classified

in terms of four modelling dimensions; operator, technology, service and implementation.

In the documentation it is stated that existing operator-related conceptual issues only have

been reworded, c.f. section 2.2. Although only minor adjustments have been added to the

operator-related issues, it could from TeliaSonera's angle, and with reference to

Recommendation 2, specifying that “The generic operator is not intended to reflect any of the

actual mobile network operators, but is intended to be generically applicable to the cost of

mobile termination in Norway”, be questioned why it is necessary to develop operator-specific

models in addition to a cost model for a generic operator.

Taking into account the above mentioned, combined with the scale of resources that applies

for both operators and at the authority with regard to development of the operator-specific

models, and the fact that this is the third time that operator-specific models have been

developed, TeliaSonera is of the opinion that it is no longer required to develop such models.

With regard to technology-related conceptual issues, it is stated in the documentation that

previous approaches in the model related to radio technology standards, treatment of

technology standards etc. have to be revised.

TeliaSonera do not have any comments to the proposed revision of Recommendation 4 regarding

“Radio technology standards”, only to the proposal of how LTE/4G is used in the model.

In principle TeliaSonera agrees with the proposal regarding adjustment of Recommendation 5

regarding “Treatment of technology generations”. TeliaSonera is however of the opinion that

it is not realistic that GSM shut-down will take place already in 2020.

As regards service-related conceptual issues it is necessary to revise Recommendation 11 in

order to include forecasts for services as LTE and OTT in the model. TeliaSonera is however

of the opinion that the proposed forecasts in this connection is too conservative.

Finally, and when it becomes to implementation-related conceptual issues, is it TeliaSonera’s

understanding that only minor changes of previous approaches is proposed. Anyway, and with

reference to what previously said NPT69

, and to what that both TeliaSonera and Telenor called

attention to in its notes to NPT, TeliaSonera finds it necessary to remind of the need for a

separate hearing regarding conceptual approaches and a separate consequence analysis

regarding the use and calculation of pure LRIC.



69

Jf. vedtak 27. september 2010 avsnitt 166 hvor det bl.a. heter at: ”PT vil foreta en ny beregning basert på ren LRIC innen utgangen av 2013. I løpet av neste reguleringsperiode vil PT også gjøre en nærmere konsekvensutredning av kun å inkludere trafikkrelaterte kostnader og ekskludere akke dekningskostnader fra LRIC-beregningen i Norge. …”


Annex D Model adjustments from v8D to v8F

In this section, we describe the adjustments made to the structural calculations of the v8F model.

D.1 Model corrections

Corrected

expenditure flag to

include planning

period

On the B8_DemIn worksheet, the ‘2G radio’ and ‘Core’ expenditure flags

(which limit the years for both opex and capex expenditure) were corrected

to include the relevant planning periods for each category. No change was

made to the ‘3G radio’ and ‘Layered core’ expenditure flags. This change

only affects the calculations for Mobile Norway and the generic operator.

Adjusted 2G

migration profiles

to include traffic in

2020

On the D3_M8F worksheet the 2G to 3G voice, SMS and low-speed data

migration profiles were adjusted for all operators to reach 0% in 2021 (one

year later than in the v8D model).

This change reduced the gradient of the 2G migration profile from 2013

onwards, slightly increasing the total amount of traffic using the 2G network

and reducing the amount of traffic using the 3G network.

Corrected two asset

unit costs

On the D4_CostBase worksheet, the ‘3G-only 3rd party site

ancillary/permits’ and ‘HSUPA upgrade per NodeB (32CE, 2×SF4)’ asset

unit costs were corrected for each operator.

Corrected operator

LTE upload

demand data

On the D3_M8F worksheet, the ‘LTE upload data’ category was corrected

in for each operator’s market demand matrix so that it selected the correct

proportion of data when split between LTE and HSUPA.

Corrected the

treatment 2G and

3G spectrum

relating to 2G

network shutdown

On the A4_NtwDesBase worksheet, the ‘900MHz spectrum used in GSM

network’ input for Mobile Norway was updated to vary with 2G radio

network shutdown.

Additionally, the ‘UMTS900 spectrum (total frequencies)’ calculations for

the three MNOs were corrected to take into account the refarming of the

900MHz spectrum to 3G services after the 2G radio network shutdown.

The ‘UMTS 900 periodic fees, nominal, capex’ calculations for Telenor and

NetCom were corrected to allow 3G licence payments while the 2G network

was in service.

Mobile cost model version 8 final (v8F) | D–2


D.2 Revised input parameters and other decisions

Revised low speed

data migration

profile

On the D3_M8F worksheet, the low speed data migration profile for the

MNOs has been revised for the years 2009–2012, using operator data where

provided.

The forecast migration profile from 2013 to the 2G network shutdown in

2020 has been adjusted such that operator GPRS/EDGE traffic volumes

remained constant.

Reordered the

generic operator

calculations

On the D3_M8R (and D3_M8F) worksheet, the generic operator

calculations were adjusted to match the calculation flow of the actual

operators.

The changes made and new calculations used for deriving the generic

operator inputs are discussed in more detail in Section 4.2.3 above.

Revised the GSM

in-fill radius

multiplier

On the A6_NtwDesSlct worksheet, a ‘GSM in-fill radius multiplier for pure

LRIC’ has been added both for the periods pre-2006 and post-2008 (with

2007 taken as the average between these), reflecting the changing

proportion of terminating traffic carried in the GSM network.

While the “GSM in-fill radius multiplier for pure LRIC - up to 2006”

parameter remained at 1.20, as in the v8D model, the “GSM in-fill radius

multiplier for pure LRIC - 2008 onwards” has been set to 1.10 – with the

“In-fill cell radius - 2008 onwards” calculation updated to reflect this new

parameter.

Included a UMTS

in-fill radius

multiplier

On the A6_NtwDesSlct worksheet, a ‘UMTS in-fill radius multiplier for

pure LRIC’ has been added, acting in the same manner as that used for

GSM in-fill sites in the v8D model.

This multiplier was set to 1.06, based on the reduction in UMTS traffic to

approximately 89% of the total when termination traffic is excluded.

Adjusted the

Mobile Norway

coverage

On the A4_NtwDesBase worksheet, the Mobile Norway UMTS2100

forecast coverage was adjusted to better reflect Mobile Norway’s site

progression. This change results in the Mobile Norway model deploying a

denser 3G network, and hence being better suited to reach its 33% long-term

market share forecast and carry traffic proportion equivalent to its coverage.

Revised the generic

operator years to

reach forecast 3G

coverage area

On the A4_NtwDesBase worksheet, the parameter for the generic operator

for years to reach the ‘Forecast 3G coverage area (2100MHz+900MHz)’

was adjusted to eight, bringing it in line with the equivalent parameters for

Telenor and NetCom.

Mobile cost model version 8 final (v8F) | D–3


Adjusted historic

cost trends

On the C01_CostTrends worksheet, the capex cost trends for ‘Site

acquisition and civil works’ and ‘Radio ancillary equipment’ were adjusted

for the years 2009–2013. These cost trends were adjusted to +1.5% and

+2.5% respectively, using data on cost trends for labour, administrative

costs and materials.

Adjusted Mobile

Norway’s unit costs

On the D4_CostBase worksheet, Mobile Norway’s site asset unit costs were

updated to reflect newly supplied data.

Adjusted NetCom’s

cell radii in

Hordaland

On the A4_NtwDesBase worksheet, the NetCom parameter for the 3G

900MHz coverage cell radius in Hordaland was adjusted to closer match the

radii seen in other geotypes and by other operators.


Annex E Expansion of acronyms

2G Second generation of mobile telephony

3G Third generation of mobile telephony

4G Fourth generation of mobile telephony

BSC Base station controller

BTS Base (transmitter) station

EC European Commission

ESA EFTA Surveillance Authority

EDGE Enhanced data for global evolution

EPMU Equi-proportionate mark-up

ETH Ethernet

GPRS General packet radio system

GSM Global system for mobile communications

GSN GPRS serving node

HLR Home location register

HS(D)(U)PA High-speed (downlink) (uplink) packet access

IMS IP Multimedia Subsystem

IN Intelligent network

IP Internet Protocol

LMA Last mile access

LRIC Long-run incremental cost

LTE Long-term evolution

LU Location update

Mbit/s Megabits per second

MGW Media gateway

MHz Megahertz

MMSC Multimedia message service centre

MNO Mobile network operator

MSC Mobile switching centre

MSS MSC server

MVNO Mobile virtual network operator

NMS Network management system

NPT Norwegian Post and Telecommunications Authority

NodeB Denotes the 3G equivalent of a BTS

PTS Swedish Post and Telecom Authority

R99 Release-99

RNC Radio network controller

SMS Short message service

SMSC SMS centre

TRX Transceiver

TSC Transit switching centre

UMTS Universal mobile telecommunications systems

WACC Weighted average cost of capital

Engineering

PPDR kanada