Comparative techno-economic evaluation of LTE fixed wireless access, FTTdp G.fast and FTTC VDSL network deployment for providing 30 Mbps broadband services in rural areas

Telecommunications Policy xxx (xxxx) xxxx

Contents lists available at ScienceDirect

Telecommunications Policy journal homepage: www.elsevier.com/locate/telpol

Comparative techno-economic evaluation of LTE fixed wireless access, FTTdp G.fast and FTTC VDSL network deployment for providing 30 Mbps broadband services in rural areas Nikos Ioannou, Dimitris Katsianis∗, Dimitris Varoutas Department of Informatics and Telecommunications, Faculty of Science, National and Kapodistrian University of Athens, Athens, Greece

A R T IC LE I N F O

ABS TRA CT

Keywords: LTE FWA FTTdp G.fast FTTC VDSL Digital agenda for Europe State aid NGA broadband

In rural areas in Europe, the deployment of High-Speed Broadband access networks lags far behind the urban and suburban areas due to difficulties of fiber rollout in the final meters. FTTC VDSL is the most widespread technology used for NGA network deployments in Europe but the coverage of VDSL networks is still limited in rural areas. FTTdp networks using G.fast have been proposed as a cost-effective and future-proof alternative to FTTH and FTTB especially in rural areas where FTTC and VDSL cannot always deliver service speeds of 30 Mbps which is the minimum bandwidth defined in the European Digital Agenda as a target to be met by 2020. However, the new target of EU Commission for Gigabit Society and 100 Mbps connections, upgradable to 1 Gbps, for all households in 2025 are unlikely to be achieved by VDSL Vectoring and other copper-based technologies. On the other hand, Fixed Wireless Access (FWA) networks based on LTE technology can be used as a “last mile” solution to provide high-speed broadband access to areas where fixed broadband is limited. LTE technology offers high-speed connections able to support internet browsing and IP services, while it can theoretically support up to 300 Mbps depending on network load and sharing. Thus, it can be considered a viable alternative to other fixed network solutions especially when considering future upgrades to 5G networks that promise gigabit speeds per user. In this paper, a techno-economic study is performed to assess the feasibility of an FWA network deployment based on LTE technology in comparison to FTTdp G.fast and FTTC VDSL network rollout for delivering service speeds of 30 Mbps in rural areas. A variety of different population density, competition and regulation policy scenarios is considered. Cash flow results are presented and standard financial indexes for the business cases are discussed. The results are being assessed through a sensitivity and risk analysis to determine the most influential factors on the return on the investment. Furthermore, the (non) profitability of these cases and the subsidization needed from structural funds are analyzed. The results are aimed to contribute to the debate over network evolution scenarios among academia, industry, regulators, policy makers and governments.

1. Introduction The European Commission's Digital Agenda for Europe (DAE) defined the broadband targets to be met by European Member States by 2020. Some of the main DAE objectives are the following: a) all citizens should have access to broadband speeds of at least

∗

Corresponding author. E-mail addresses: ΝikosΙ[email protected] (N. Ioannou), [email protected] (D. Katsianis), [email protected] (D. Varoutas).

https://doi.org/10.1016/j.telpol.2019.101875 Received 31 October 2018; Received in revised form 29 August 2019; Accepted 17 September 2019 0308-5961/ © 2019 Elsevier Ltd. All rights reserved.

Please cite this article as: Nikos Ioannou, Dimitris Katsianis and Dimitris Varoutas, Telecommunications Policy, https://doi.org/10.1016/j.telpol.2019.101875

Telecommunications Policy xxx (xxxx) xxxx

N. Ioannou, et al.

30 Mbps and b) 50% of households should have broadband subscriptions with speeds of at least 100 Mbps. The second objective was expected to be achieved through fixed Next Generation Access (NGA) networks which would be deployed mostly in urban and suburban areas. In reality, 100% population coverage of 30 Mbps broadband proved to be a challenging objective especially for rural areas of Europe where the deployment of NGA networks lags far behind the urban and suburban areas due to the low population density and the high cost of deployment of fiber networks. The European Commission reported (European Commission, 2017b) that in 2017 the average NGA coverage in rural areas of Europe was 46.9% compared to the 80.1% of overall NGA coverage mostly led by NGA network deployment in urban and suburban areas. However, according to the recent report of European Court of Auditors (European Court of Auditors, 2018) rural coverage is still far below the total coverage and in 14 Member States the 30 Mbps coverage is less than 50% while in 20 countries is less than 70%. In the European Commission's reports, it is stated that VDSL (Very-high-bit-rate Digital Subscriber Line) has the highest coverage growth and availability compared to other fixed broadband technologies in rural areas. Although this may be true, as an upgrade of the existing ADSL networks, VDSL can neither always provide speeds of 30 Mbps in rural areas where the average cable distance between the cabinet and the households is more than 1 km nor does it offer a future upgrade path for such long distances. Fiber to the Distribution Point (FTTdp) networks using G.fast have been proposed as a cost-effective alternative to FTTH and FTTB (Schneir & Xiong, 2016b) especially in rural areas (Schneir & Xiong, 2016a) in an attempt to reduce the copper cable length and provide higher broadband speeds compared to FTTC technologies. In addition, G.fast requires less power than VDSL technologies and the Reverse Power Feeding feature is expected to help reduce OPEX costs for distribution point installations. Another important indicator mentioned in the European Commission's reports is that 4G LTE coverage has reached more than 89.9% in rural areas. In contrast, fixed NGA networks and fast broadband coverage remain problematic in rural areas (less than 50%) and DAE objectives are unlikely to be achieved (European Court of Auditors, 2018). Therefore, Fixed Wireless Access (FWA) networks based on LTE technology can be used as a “last mile” solution to provide high-speed broadband access to areas where fixed broadband is limited (Furuskär, Rao, Blomgren, & Skillermark, 2011; Jha & Saha, 2016; Ovando, Pérez, & Moral, 2015). LTE technology offers high-speed connections able to support internet browsing and IP services, while it can theoretically support up to 300 Mbps depending on network load and sharing. Thus, it can be considered a viable alternative to other fixed solutions, especially when considering reuse of existing infrastructure. Another factor pointing toward FWA is that the EU Commission's Communication on Gigabit Society set new objectives for 2025 which include access to networks offering a download speed of at least 100 Mbps, upgradable to 1 Gigabit for all European households, rural and urban. However, fixed NGA coverage is still very low in rural areas and VDSL Vectoring networks are unlikely to be sufficient for the Gigabit Society ambitions for 2025 when speeds upgradable to 1 Gbps are required. The high coverage of the existing LTE networks and the very promising capabilities of 5G networks indicate that LTE FWA could be a quick step towards future 5G FWA and gigabit wireless networks. The importance of 800 MHz band to the deployment of high-speed wireless internet services has also been recognized by European Commission since 2010 by proposing technical rules to ensure that radio communications equipment using the 800 MHz band, can be used efficiently for wireless broadband networks, such as LTE or WiMAX (European Commission, 2010b). Moreover, in 2016 the European Council set the year 2020 as the deadline for reassigning 700 MHz band for wireless broadband services under harmonized technical conditions in order to promote the take-up of 4G and the future rollout of 5G networks (Council of EU, 2016). The increased coverage offered by these lower frequency bands is a key factor when considering investments in sparsely populated areas. 4G FWA networks have already been deployed in Belgium and the Netherlands. In the case of central Greece, the rollout of LTE FWA for delivering 30 Mbps in rural areas was subsidized using EU funds. Many European operators have launched fixed wireless broadband services as an alternative or an upgrade to traditional DSL broadband. Orange in Spain already offers 4G home broadband up to 50 Mbps. Bouygues Telecom in France offers “4G Box” service, Vodafone in Germany “GigaCube” service and T-Mobile in the Netherlands used FDD + TDD LTE in order to launch “4G for home” service. Other operators like T-Mobile in Austria and in the Czech Republic and Sunrise in Switzerland have also launched similar fixed wireless services. In Ireland, mobile operator Imagine launched a fixed wireless service delivering NGA speeds in rural and suburban areas. Operator Hi3G Denmark entered the fixed wireless broadband market in 2019 by delivering speeds of up to 71 Mbps over its 4G network. Lastly, in July of 2019, Orange Romania deployed a multi-vendor 5G FWA network as a real-world trial, which was the first in Europe. The telco had access to the 26 GHz mmWave band for the trial period, and used Massive MIMO and beamforming from the access points. According to Orange Romania, they achieved coverage beyond 1 km at 1 Gbit/s speed for a single user and aggregated cell downlink throughputs of 3 Gbit/s with four simultaneous users. Low demand and uncertainties related to revenues and the high CAPEX of NGA investments are the main factors that could lead to market failure when attempting to achieve DAE objectives in European rural areas. In view of this, the European Commission has stated that “DAE objectives cannot be reached without the support of public funds” (European Commission Communication, 2013). In addition, according to the European Commission's report (European Commission, 2013), the amount required to achieve the 2020 broadband targets was estimated at up to 250 billion euro in 2013. The European Investment Bank also estimated that half of the costs are expected to be used for investments in rural areas, where 20% of the population lives, although it is expected that the reuse of existing infrastructure could bring the rollout costs down. In view of this, various funding institutions have been used, those offering grants like European Regional Development Fund (ERDF) and European Agricultural Fund for Rural Development (EAFRD) and those that offer loans like European Fund for Strategic Investments (EFSI) and European Investment Bank (EIB). However, only about 15 billion euro were made available to Member States from the EU for supporting broadband investments for the 2014–2020 programme period (European Court of Auditors, 2018). 2

Telecommunications Policy xxx (xxxx) xxxx

N. Ioannou, et al.

As shown by the recent report of European Court of Auditors (European Court of Auditors, 2018), two of the five examined Member States may achieve the 30 Mbps target by 2020 but rural areas remain problematic in most Member States. Private public partnership (PPP) initiatives can greatly contribute to increasing NGA coverage as shown in Hungary, Ireland and Italy. Four countries, Hungary, Ireland, Italy and Poland identified financing gap and used EU funds (mostly from ERDF and EAFRD) along with support from the national budget. However, only Hungary and Italy allocated sufficient funds for all households after allocating the financing gap. In most of the Member States, public support comes mostly from the national budget and to less extent from EU funds, with the exception of Hungary and Poland. According to EC guidelines for State aid (“EU Guidelines for the application of State aid rules in relation to the rapid deployment of broadband networks,” 2013), the CAPEX of a broadband investment can be subsidized in the case of white NGA areas where NGA networks do not at present exist and where they are not likely to be built within 3 years by private investors. In these areas, it is unlikely that there will be infrastructure competition and there is a high chance of market failure. However, in the case of grey NGA areas (only one NGA network is in place or is being deployed in the coming 3 years and there are no plans by any operator to deploy a NGA network in the coming 3 years) more detailed analysis is required for State intervention while in black NGA areas (at least two NGA networks of different operators exist in a given area or will be deployed in the coming 3 years) State aid is permitted only under certain conditions. Moreover, EC guidelines for State aid encourage spectrum (re)allocation to promote wireless internet access for underserved areas and EU decisions on radio spectrum policy programme (European Parliament and Council, 2012) encourage spectrum sharing, but subsidization of the spectrum has not been discussed yet. A number of different studies have been conducted to assess the feasibility of NGA investments in European areas. Schneir & Xiong (2016a); (2016b) have compared FTTH, FTTB, FTTdp, FTTdp-street, FTTC finding that FTTdp-street leads to significant cost reductions compared to FTTH and FTTB especially in rural areas. However, according to their findings VDSL (CO-VDSL and FTTC) was the most feasible option but not applicable in all types or rural areas (< 1 km cable length). The authors acknowledged that further studies should be conducted to assess the cost advantages of fixed wireless LTE networks compared to fixed NGA networks. Ovando et al. (2015) have evaluated the feasibility of LTE FWA networks for covering of 75.3%–93.3% of the Spanish population in rural areas. The analysis showed that the feasibility of the investment is largely dependent on service take-up (demand) and that infrastructure sharing has a great impact on cost savings. Frías, González-Valderrama, & Martínez (2015) have compared the costs of FTTH and LTE FWA networks for providing 30 Mbps in rural areas in Spain finding that in municipalities with less than 1000 inhabitants is more feasible to deploy LTE networks. In Feijóo, Ramos, Armuña, Arenal, & Gómez-Barroso (2018), authors have used NUTS3 level geographical data and data from various relevant reports and academic studies to estimate the investment costs for different technologies (FTTH, FTTB, FTTC, LTE) and for different geotypes in order to estimate the investment gap to meet digital agenda targets. The paper's main conclusion was that most EU countries are far from achieving the DAE targets and additional public subsidies and funds from private parties or Private Public Partnership schemes are needed to meet the set targets. Other studies including Falch and Henten (2010) and Nucciarelli, Castaldo, Conte, & Sadowski (2013) focus on public-private partnerships for upgrading broadband investments. These studies examine the ups and downs of public-private interplay and show that quite early governments had understood the significant role of the public sector in funding broadband investments. In summary, there is a lack of cost studies that compare fixed wireless and the various fixed NGA networks in different rural areas of Europe while addressing the issue of State aid and desirable tariffs. In this paper, a techno-economic study is performed to assess the feasibility of an FWA network deployment based on LTE technology in comparison to FTTdp G.fast and FTTC VDSL network rollout for delivering service speeds of 30 Mbps in rural areas. How does different population density, competition and regulation policy correlate with project profitability? A variety of different scenarios are considered within the paper. These scenarios include country groups with different rural population densities (medium, low, high), the three different technologies (LTE, VDSL, G.fast) as well as various regulation and policy driven scenarios/options which could contribute to the feasibility of the NGA investment within realistic assumptions. The results are being appraised through a sensitivity and risk analysis to determine the most influential factors on the return on the investment. We also discuss the (non) profitability of these cases and the subsidization needed from structural funds. The paper focuses only in rural areas where it is already apparent that the first DAE objective will not be met in most of the European countries by next year (2020). The results aim to contribute to the debate over network evolution scenarios among academia, industry, regulators, policy makers and governments especially when considering the current progress of NGA and LTE coverage in rural areas of Europe and the capabilities of 5G mobile networks as a future upgrade to current mobile infrastructure. The remainder of this paper is organized as follows: In section 2, we present the main assumptions and scenarios that define the business case of this study along with the methodology used for modelling and dimensioning of each technology. Demand forecasts and illustrated in section 2.1 followed by geographical classification and area characteristics are given in section 2.2. The technologies and the regulation policy scenarios under consideration are presented in section 2.3 and 2.4 respectively. In section 2.5, the techno-economic methodology is presented including the main modelling assumptions. Results regarding investment costs, expected revenues and sensitivity analysis are illustrated in section 3 4as well as risk assessment and investment subsidization results in section 3.3. Concluding remarks are provided in section 4. 2. Business case definition and methodology In this section, the main aspects of the business case of the study are discussed as well as the methodology used. Initially, the demand forecasting and the corresponding data and assumptions will be presented. The demand curve derived from the forecasting process will be used as one of the main inputs of the technoeconomic model. Secondly, the geographical framework and the most 3

Telecommunications Policy xxx (xxxx) xxxx

N. Ioannou, et al.

Fig. 1. Demand forecasts in rural areas using the Tonic model.

important characteristics of European rural areas will be discussed based on the data of Eurostat. The main geotype input parameters for dimensioning of the network will be calculated for three groups of European countries with different rural population densities. Moreover, we present the three different technologies (LTE, VDSL, G.fast) and architectures (FWA, FTTC, FTTdp) and their technical parameters which will be modeled. The different policy and regulation scenarios considered in the techno-economic model will be discussed in section 2.4. These scenarios are essentially policy options and choices that could contribute to the feasibility of the investment depending on the network technology deployed (technical scenario) and the market condition in the area (grey or white NGA area). Finally, economic and cost assumptions and the parameters of the techno-economic model will be presented in section 2.5. 2.1. Demand forecasts The demand forecast is based on data of NGA coverage and subscriptions of households in European rural areas for the years 2012–2017 (European Commission, 2018). Three diffusion models were employed for data fitting: Logistic, Gompertz and Tonic (Olsen et al., 2006). Although the differences between Tonic and Gompertz were not significant (less than 0.5% on a yearly basis), Tonic model showed to fit the aforementioned data with lower error values compared to the other two diffusion models. The Tonic model was developed within the IST-TONIC project and provided reasonably accurate fitting over historical data related to hightechnology products. Three curves on demand were calculated, a pessimistic one based on countries with low NGA demand, an optimistic one based on countries with high NGA demand and a baseline of medium demand based on countries close to the EU average NGA demand in rural areas (Fig. 1). The results presented in this paper involve only on the medium demand case. The network penetration is expected to reach 30% of the population in 2027. 2.2. Geographical classification The assessed areas include only rural areas according to Eurostat's definition of territorial typologies (EUROSTAT, 2018d). In this methodology, typologies of territory are based on clusters using 1 km2 continuous grid cells with similarities in terms of population and density. Areas of local administrative units at level 2 (LAU level 2) can then be classified into three degrees of urbanization based on population share in different types of clusters. Areas, where more than 50% of the population lives in rural grid cells, are classified as rural areas. This classification is also defined at higher geographical scale like NUTS (Nomenclature of Territorial Units for Statistics) level 3 regions. For the purpose of this study, three groups of European countries are considered with different population density in rural areas based on the Eurostat data for LAU level 2 share of population and land area. These groups were formed for low, medium and high population density and referred to countries with an average rural population density of 30, 40 and 55 people per km2 (in LAU2 areas), respectively. Countries like Bulgaria, Cyprus, Greece, Croatia, Ireland and Portugal belong to the low density group while Romania, UK, Hungary, France and Austria belong to the medium density group. The high density group consists of countries like Poland, Czechia, Slovakia, Italy, Slovenia and Germany. Additionally, Eurostat's data on housing (EUROSTAT, 2018b) and land use (EUROSTAT, 2018c) in rural areas were taken into consideration for calculating the average number of households per km2. The geotype parameters of the model which were defined by the aforementioned data are presented in Table 1. Using the above data the household density is calculated according to equation:

HouseholdDensity =

PopulationDensity 1 ∗ PeoplePerHousehold 1 − ShareOfNonResidentialArea

(1)

This methodology aims to exclude the non-residential areas (land used for agriculture, forestry, etc.), a process which is essential 4

Telecommunications Policy xxx (xxxx) xxxx

N. Ioannou, et al.

Table 1 Geotype input parameters. Parameter 30/40/55 pop. per km2 2.1 85% 15% 70%

Average population density Low/Medium/High Average number of people per household Percentage of Single Dwelling Units Percentage of Multiple Dwelling Units Average share of non-residential area

for network dimensioning, especially for fixed access networks. 2.3. Technology driven scenarios In this section, the main characteristics and architectures as well as technical modelling and dimensioning methodology of the three different technologies will be presented. Some important technical modelling assumptions are also included in the discussion. 2.3.1. LTE fixed wireless access The LTE FWA network architecture consists of four (4) Aggregation Nodes (AN) linked through Aggregation Links (AL) which use only microwave links in a ring topology. The AN1 is the radio access network and is made up of four eNodeB. AN2 and AN3 constitute the transport network and consist of eight (8) Ethernet Switches and eight (8) Access Routers, respectively. AN4 is made up by the core network elements of the LTE network which include the MME, PDN, etc. The dimensioning of AN1 is made by using hexagonshaped cells while the dimensioning of the other ANs uses square-shaped areas. Finally, each household is connected to eNodeB using an outdoor directional antenna which is connected through a cable to the indoor Customer Premises Equipment (CPE). This is a key aspect of network planning because it increases the received signal quality in order to provide the necessary throughput for 30 Mbps speeds over the longer distances in rural areas. It is worth mentioning that ring topology is chosen in accordance with industry best practices used for backhauling and core transmission (usually mesh and ring topology). An overview of the full network architecture is presented in Fig. 2. The technical model starts with a link budget, followed by coverage dimensioning and finally capacity dimensioning. The link budget is used to calculate the cell radius which is used along with geotype data as input to coverage dimensioning to calculate the minimum number of cells and network elements required for the specified network coverage. The capacity dimensioning is used to calculate the minimum number of cells and network elements to meet traffic requirements based on service demand. The final size of the network is the maximum number of cells and network elements between coverage and traffic dimensioning. For the assessment purposes, LTE 2 × 20 MHz FDD carrier in 800 MHz is considered. In other studies (Ovando et al., 2015) (Jha & Saha, 2016), 2 × 10 MHz FDD carriers in the 800 MHz have been used which is reasonable considering the fact that the 800 MHz band consists of 2 × 30 MHz only. On the other hand, such low bandwidth greatly reduces the capacity capabilities of the network. As a result, a denser radio access network is needed in order to provide 30 Mbps speeds, increasing the overall cost of the investment. Thus, in order to simplify the technical model, we consider that 20 MHz of carrier bandwidth should be available for the investment most probably through Carrier Aggregation (CA) with higher frequency bands. Inter-band Carrier Aggregation is already supported by LTE-advanced (LTE-A) since release 11 which allows aggregating of up to 5 carriers of up to 20 MHz each into one “virtual” carrier with a theoretical downlink peak rate of up to 1 Gbps (Jeanette Wannstrom for 3GPP, 2013). The CA configurations which include 800 MHz (B20) band along with higher frequency bands like 1800 MHz (B3) and 2600 MHz (B7) are already in the top 4 mostly used CA configurations among global LTE networks (Halberd Bastion, 2019). Our technoeconomic evaluation results have shown that CA does not have any significant negative effect on the investment even when using higher frequency bands since capacity has a much greater impact on network planning (capacity driven) of the FWA network than coverage. The use of directional outdoor antennas is another factor, which can facilitate CA with higher frequency bands. However, results of 2 × 10 MHz FWA have also been included in the ARPU results (Fig. 9 and Fig. 10) presented in section 3.1 in order to illustrate the impact of carrier bandwidth to the feasibility of the investment. Additionally, we assume the deployment of 3-sector cell sites and frequency reuse factor 1 by assigning all the available

Fig. 2. LTE FWA network architecture. 5

Telecommunications Policy xxx (xxxx) xxxx

N. Ioannou, et al.

Table 2 Technical input parameters. Parameter Average Downlink throughput Average Uplink throughput (to downlink) Frequency Band Bandwidth Number of sector per site Frequency Reuse Factor Propagation Model Contention Ratio Control Channel Overhead BS transmitter power BS antenna gain plus cable and connection losses Outdoor CPE antenna gain minus cable and connection losses Outdoor CPE antenna height

30 Mbps 10% 800 MHz 2 × 20MHz FDD 3 1 Okumura-Hata 10:1 20% 46 dBm 15 dBm 8 dBm 5m

bandwidth on each sector. This is possible due to the advanced scheduling and interference management capabilities of LTE technology. Among other technical parameters (Table 2) a contention ratio of up to 10:1 was used, in line with existing cable networks. This contention ratio can assure an acceptable user experience taking into consideration Cisco's estimations (Cisco, 2018) of 140 GB monthly average internet traffic per household in Western Europe in 2021 which results in less than 4 Mbps per household in the busy hour. Although a 20:1 contention ratio could theoretically allow us to offer 30 Mbps broadband services with 2 × 10 MHz FDD carriers, it would greatly reduce the quality of service in the following years. Although TDD seems more appropriate considering the 10% uplink traffic proportion, FDD systems are more popular and have already been widely implemented in the 800 MHz band in Europe. TDD systems in Europe are usually deployed in higher frequency bands (higher than 2 GHz). In various countries, there are commercial implementations of LTE-A systems which use Carrier Aggregation to combine TDD with low-band FDD spectrum in order to improve the high-band coverage area and the overall downlink throughput. The average uplink proportion is mostly used for the dimensioning of the backhaul transmission and the core network. 2.3.2. VDSL FTTC For the dimensioning of the fixed networks, we use a geometric model according to which the NGA network consists of flexibility points (FP). These flexibility points are FP1 the floor (in case of MDUs), FP2 the building, FP3 the street cabinet, FP4 the branch box and FP5 the Local Exchange (LEX or Central Office). Each FP has a distribution rate based on the maximum connections from one FP to the next (Fig. 3). These distribution rates are not constant and can change according to the input parameters for the capacity of each FP (available space, ports, etc.). Each FP layer is dimensioned based on a star-mesh topology in a rectangular area of L x L’ size (in km) which is divided equally among the units of the FP. These units are placed at the center of each part while the next level FP is placed in the center of the whole layer (Fig. 4). For FTTC VDSL network we have assumed a maximum 1 km distance from the buildings in order to provide speeds higher than 30 Mbps for all households covered. The VDSL cabinets consist mainly of up to four VDSL cards with 48 ports each. The model makes the FTTC network dense enough in order to reduce the maximum distance to 1 km distance while adjusting the number of installed VDSL cards accordingly. Both the power consumption of the active equipment of VDSL cabinets and the utilization based on average network load were included in the calculation of OPEX. 2.3.3. G.fast FTTdp For the FTTdp network architecture, it is assumed that a Distribution Point Unit (DPU) cabinet will be used which will be installed on the street between the distribution and the drop segment, with a maximum distance to the end user's premises up to 300 m. FTTdpstreet is preferred over FTTdp building due to higher cost savings especially in rural areas (Schneir & Xiong, 2016b; 2016a). The DPU

Fig. 3. NGA network dimensioning model. 6

Telecommunications Policy xxx (xxxx) xxxx

N. Ioannou, et al.

Fig. 4. Example of Geometric model for FP3, layer 3.

cabinet contains DPUs with up to eight ports, one for each subscriber connected, and an internal main distribution frame (MDF). A Reverse Power Feeder (RPF) will be installed in each user's household for supplying power to the DPU. For the drop segment of the FTTdp-street, the existing copper infrastructure will be reused and for the fiber segments, a Gigabit Passive Optical Network (GPON) architecture will be used. There will be two splitting levels in total – up to 1:32 located in the street cabinet and up to 1:8 in the DPU cabinet (Fig. 5). For the dimensioning, we have used the same geometric model as with the FTTC network by extending the optical fiber in the distribution segment up to the DPU cabinet. In the case of FTTdp-street, the DPU cabinets are placed according to the maximum copper cable distance from the premises in the FP3 layer right after the calculation of trenches from the street cabinet to the buildings. This method has been used in order to reuse efficiently the already installed copper in the drop segment. Furthermore, different distribution rates were used because FTTdp does not require as many street cabinets as the FTTC network does. Because of 10 Gigabit PON deployment, the G.fast FTTdp network is able to provide speeds up to 150 Mbps, much higher than the speeds that can be supported by the LTE FWA network and the average speeds of FTTC VDSL network. Nonetheless, we have not included services of higher internet speeds in the techno-economic model.

2.4. Regulation and policy-driven scenarios The regulation and policy-driven scenarios consist of various policy options that could contribute to the feasibility of the investment depending on the network technology deployed (technical scenario) and the market condition in the area. These policy choices have been added as separate inputs to the technoeconomic model and have a great impact on the costs and/or the revenues of the investment as shown in the results in section 3. Two main scenarios were considered for each network technology. In the first scenario, we have an infrastructure-based competition where more than one NGA network operators can provide services of at least 30 Mbps in the same rural areas without having access to the network infrastructure of any other operator (grey or black NGA area). However, the low demand and high costs can lead to market failure situations where no operator is willing to invest in NGA network deployment in rural areas. Therefore, in the

Fig. 5. G.fast FTTdp-street network architecture. 7

Telecommunications Policy xxx (xxxx) xxxx

N. Ioannou, et al.

second scenario, there is only one infrastructure operator allowed to deploy the NGA network in rural areas (white NGA area) but provides wholesale (bitstream virtual) access to other competitors. The wholesale price is defined as a percentage of the retail price (up to 50%). Those two scenarios refer to future market scenarios that could occur in current grey and white NGA areas respectively. For LTE FWA, we also present results having considered the cases of spectrum subsidization (up to 90%) and passive site sharing between different network operators (up to two operators). Site sharing is not possible in the second main scenario where there is only one infrastructure operator. It is worth mentioning that subsidization of spectrum has not been extensively discussed in EU, but our results have shown that spectrum is key factor for the feasibility of the FWA investment. The cost of the spectrum for rural areas could be considered as a “sunk” cost covered from the urban areas for a telecom actor operating in both areas. As an alternative no real cost should occur for an investor in such areas since the State could subside this cost by offering free bands to the providers (investors). Analogous approach has been considered in Greece for access spectrum where it is provided for free by the Greek State to the PPP scheme. However, similar results in terms of cost can be achieved through other policy choices like spectrum sharing (Teng, Guo, & Honig, 2017; Ye, Wu, Shu, & Qian, 2016), but result in more complicated regulation schemes and modelling methodologies. For FTTdp G.fast and FTTC VDSL networks, we have included results for network deployment using aerial cables and reusing the already available utility poles (of power or telephone lines) in rural areas. In this case, we assume 80% of the distribution segment and 15% of the feeder segment of the network to be covered with aerial cables. The above scenarios were selected based on the most influential factors for the CAPEX and OPEX of the investment as shown by the results of the model which will be presented in the next sections. In the case of a single infrastructure operator which could occur in current white NGA areas, it is likely that part of the investment could be subsidized through State Aid or EU structural funds. As previously reported, State Aid and PPP initiatives have already been used in EU member States and have contributed to the deployment of NGA networks. The amount of subsidization should also be linked to the uncertainties of demand and the purchase power (desirable tariff) of rural areas which differs between countries. In that context, a risk assessment has been performed in order to calculate the amount of subsidization per amount of desirable tariff for ensuring investment profitability. The respective results are shown in section 3.3 of this paper. 2.5. Techno-economic methodology The techno-economic methodology is based on a bottom-up analysis of Discounted Cash Flows (DCF) for network deployment, operation, and maintenance. The techno-economic tool is based on the tool developed within a series of techno-economic projects, namely TONIC, ECOSYS, TITAN and it has been used in several related network evaluation studies for mobile (Varoutas, Katsianis, Sphicopoulos, Stordahl, & Welling, 2006) and fixed networks (Katsianis et al., 2012; Monath, Elnegaard, Cadro, Katsianis, & Varoutas, 2003; Olsen et al., 2006; T.; Rokkas, Katsianis, & Varoutas, 2010) (Katsianis et al., 2012) (Theodore Rokkas, Neokosmidis, Katsianis, & Varoutas, 2012). We assume an eight-year study period is a reasonable period for fixed access network deployments, considering the time it usually takes to reach market maturity. The market penetration of broadband services and the tariffs for these services as well as their market share have been defined. Concerning the analysis, demand and price forecasts have been included in order to calculate the network components needed as well as the revenues generated by network services. The methodology incorporates the geometric model described in the previous sections for calculating lengths for cables and ducting for the FTTdp G.fast and FTTC VDSL networks. As far as the LTE FWA network is concerned, a radio propagation model is incorporated for dimensioning assuming the use of 800 MHz frequency band. In all cases, the network is designed to provide at least 30 Mbps throughput per household. The result of the architecture scenario definition is the “shopping list” for each year of the study period that shows the volumes of all network cost elements and the distribution of these network components over different flexibility points and link levels. The costs of the network components are calculated using an integrated cost database. Architecture scenarios are used together with the cost database to calculate investments for each year (CAPEX and OPEX). Power consumption costs for each active element of the network adjusted to the average network utilization is included in the OPEX of the investment. 2.5.1. Model assumptions For the purposes of this study, the internet service of 30 Mbps is considered without data caps or limits. No other services were included in the current study. We have assumed a network coverage of up to 95% of rural areas within the first two years (2019 and 2020) of the investment. The 5% of the rural areas were not included in the assessment because they are expected to be very sparsely populated or almost open areas, which will be covered through other wireless technologies (e.g. satellite). These areas could greatly increase the cost of network deployment without a significant increase in population coverage. Furthermore, a fixed 50% market share is used because it is expected that such risky investments will most probably be implemented by the incumbent operator or an operator that can ensure a significantly high enough market share for a large scale rollout of NGA networks in rural areas. As shown in the results of sensitivity analysis which will be presented in the next section, market share is a key factor for the profitability of the project. Additionally, we have assumed an annual churn rate of 2% and an annual tariff degression of 3%. The year 2019 is considered to be the initial year of investment, while the network is expected to be completed within 2020. Network deployment will be conducted in a greenfield scenario. Therefore, no existing infrastructure is reused except for the copper network (trenches, ducts, cables) between the DPU or the FTTC cabinet and the households. However, in the case of G.fast with aerial cables, we assume that 95% of poles will be reusable. It is worth mentioning that the weighted average cost of capital (WACC) is set to 12%, a relatively high value, due to the risk profile of the investment of NGA networks in rural areas. Therefore it is expected that a risk premium (European Commission, 2010a) (Katsianis et al., 2012) could be incorporated into the WACC calculation in order to reflect the high uncertainties mainly in terms of 8

Telecommunications Policy xxx (xxxx) xxxx

N. Ioannou, et al.

Table 3 Economic and demand input parameters. Parameter WACC Tax Rate Annual Tariff Degression Churn Rate Network Take-up (service demand in 2027) Market Share Network Coverage

12% 20% 3% 2% 30% 50% 95%

demand and revenues. The most important economic and demand input parameters of the model are presented in Table 3. For LTE FWA network, the spectrum price is set to the average price in European countries and is measured in euro per MHz per population. The data is based on GSMA's effective spectrum pricing report (GSMA, 2017). It is important to note that only the share of the population of rural areas is used for the calculation of the total cost of the spectrum, although this cost can be omitted in a subsidization approach by the regulator. As already mentioned, power consumption for each active element of the network is included in the OPEX of the investment. The average electricity price in European countries was used according to the available Eurostat data (EUROSTAT, 2018a). In addition, a straight-line depreciation was used based on the lifetime of assets. The minimum monthly Average Revenue Per User (ARPU) is calculated for the NPV to become zero in 2027 (last year of study). The subsidization of CAPEX for the first two years is calculated as the ARPU is reduced close to the average minimum EU prices (adjusted by means of purchasing power parities) for services of speeds 30 Mbps to 100 Mbps (European Commission, 2017a). In conclusion, unit costs for the technology scenarios are listed in Table 4 and Table 5. 3. Results and discussion The results of the technoeconomic model shown below are divided into three subsections. First, we present the results of the Discounted Cash Flows (DCF) analysis based on standard financial indices like ARPU, Net Present Value (NPV) and Cash Balance along with cost break down per cost category (CAPEX and OPEX) and asset category. Moreover, the sensitivity results shown in section 3.2 aim to indicate the most influential factors of the investment for each NGA technology. In conclusion, results for subsidization given a range of desirable tariffs are discussed comparatively for all the technologies for the single infrastructure scenario which is considered the most eligible case for State aid. 3.1. Discounted cash flows and costs analysis The results presented in this study will focus mostly on ARPU, Net Present Value and Cash Balance. The Net Present Value (NPV) describes today's value of the sum of resultant discounted cash flows (annual investments, running costs, revenues, etc.), or equivalently, the volume of money, which is expected to yield over a given time period. If the NPV is positive, the project is acceptable and it is a good indication of the profitability of an investment project, taking into consideration the time value or opportunity cost of money, which is expressed by the discount rate. NPV is given by: n

NPV =

CF

∑ (1 + ti)t

(2)

t=0

Table 4 Unit costs for FTTC VDSL and FTTdp G.fast (€). Parameter

FTTC VDSL

Trenching Cost (average per km) Cable Cost (average per fiber per km) Fiber Cost (per km) OLT (per port) Optical Splitter Branching Box Customer Equipment & Installation Average Energy Price (per KWh) VDSL card Cabinet DPU Passive Cabinet Aerial Cable Cost (average per fiber per km) Aerial Fiber Cable Installation (per km) Pole Cost

8,500 70 90 100 20 300 40 0.17 300 2,500

FTTdp G.fast

300 1,000 90 1,000 1,500

9

Telecommunications Policy xxx (xxxx) xxxx

N. Ioannou, et al.

Table 5 Unit costs for FWA LTE (€). Parameter Spectrum Fees (per MHz per Pop) LTE eNodeB Site Rent eNodeB (per year) Ethernet Switch Access Router MME PDN SGW Microwave Link 1 Gbps (per 10 km hop) CE Outdoor Antenna Cost CE Modem-Router cost CE Installation Cost Average Energy Price (per KWh)

0.30 30,000 5,000 20,000 70,000 5,000,000 1,000,000 1,000,000 4,000 40 50 60 0.17

where t is the number of the time period, CF the net cash flows at year t and i the discount rate. The Cash Balance (accumulated Cash Flow) curve generally goes negative in the early part of the investment project because of initial capital expenditures. Once revenues are generated, the cash flow turns positive and the Cash Balance curve starts to rise. The lowest point in the Cash Balance curve gives the maximum amount of funding required for the project. The point in time when the Cash Balance turns positive represents the Payback Period for the project. Fig. 6, shows the share of accumulated CAPEX and OPEX in the costs of investment during the study period for a medium population density country. The results are almost the same (within 1%–2%) for the low and high population density scenarios. For the LTE network, eNodeB site acquisition and spectrum fees are the largest part of CAPEX costs having a share of 35.23% and 22.58% of total accumulated CAPEX, respectively. Moreover, eNodeB site rent is 60.57% of total accumulated OPEX. In the case of fixed NGA networks, the cost of trenches is 85.37% of total accumulated CAPEX for the G.fast network and 64.76% for the VDSL network. It is also worth mentioning that power consumption is more that one-quarter of the total OPEX of the VDSL network. The total unit costs as a percentage of CAPEX are presented in Fig. 7. FTTdp G.fast has the highest cost when comparing the total project costs of the three technologies. In Fig. 8, a comparison of total project costs of VDSL and LTE networks to the total project costs of G.fast network is presented for the three population density scenarios. It is important to note that with higher population density the total cost of the VDSL network seems to improve significantly compared to G.fast network while the LTE network cost does not seem to improve as much as the cost of fixed networks. However, the reason for this phenomenon is that service costs per user for VDSL and G.fast improve more rapidly with higher population densities than the service costs of the LTE network. For each technology in the three population density cases, we calculated the minimum monthly ARPU required for NPV to be

Fig. 6. Percentage of CAPEX and OPEX in total investment. 10

Telecommunications Policy xxx (xxxx) xxxx

Fig. 7. CAPEX and OPEX per asset category.

N. Ioannou, et al.

11

Fig. 8. Total project costs compared to total project costs of FTTdp G.fast (percentage and absolute values).

N. Ioannou, et al.

Telecommunications Policy xxx (xxxx) xxxx

12

Telecommunications Policy xxx (xxxx) xxxx

N. Ioannou, et al.

Fig. 9. Minimum viable monthly ARPU, in infrastructure-based competition scenario.

equal to zero at the end of the study period (“minimum viable ARPU”) (Figs. 9 and 10). These results can be used as an indicator of the cost of service for each subscriber because “minimum viable ARPU” means that the cost of the investment is equal to revenues at the end of the study period. In all cases, the FTTdp G.fast cost per user is much higher than the other two technologies even when using aerial cables. The ARPU for the LTE FWA network is the lowest of the three technologies and improved further with spectrum subsidization and passive site sharing especially in the case of the default carrier bandwidth (20 MHz). Again, we observe that FTTC VDSL seems to be greatly influenced by population density showing the biggest improvements as the population density rises while LTE FWA does not show any notable improvements. In the case of 10 MHz LTE FWA, the ARPU is quite higher even surpassing the

Fig. 10. Minimum viable monthly ARPU, in single infrastructure operator scenario. 13

Fig. 11. Revenues, investments and cash balance for ARPU equal to 35 euro.

N. Ioannou, et al.

Telecommunications Policy xxx (xxxx) xxxx

14

Telecommunications Policy xxx (xxxx) xxxx

N. Ioannou, et al.

Fig. 12. Sensitivity analysis results for LTE FWA.

ARPU of FTTC VDSL in the high population density scenario. Passive sharing seems to be a key factor for the 10 MHz LTE investment since it greatly reduces the ARPU but can only be implemented in the infrastructure competition scenario. Lastly, significant reductions in ARPU can be seen for all technologies in the single infrastructure operator scenario. Fig. 11 illustrates the revenues, investments and cash balance of the most promising scenarios for each technology assuming ARPU equal to 35 euros, which is close to the average pricing of NGA services in Europe (European Commission, 2018). In particular, the chosen scenarios which showed the best results are: spectrum subsidization of LTE FWA, using aerial cables for FTTC VDSL and FTTdp G.fast and one infrastructure operator scenario. We also included results that derive from the comparison of the cash balance curves of the three technologies for medium population density. In all cases, the gradient of the cash balance curves at the end of the study period indicates the future earning potential. The case of the high population density country has much more potential revenues for LTE and VDSL than the other two cases due to the greater potential of market size. Still, these potential revenues are not enough to cover the high amounts of CAPEX investment of the G.fast network deployment. The lowest point of the cash balance indicates the amount of investment funding required by the operator. Significant investments are needed during the first two years of the project in order to deploy the NGA networks in rural areas. Of course, the investments and operational costs show crucial differences between the three technologies. The maximum need for funding is quite significant, especially for FTTdp where we see negative cash balance in all cases compared to FWA where the discounted cash balance turns to steady growth after 7 years. Finally, large subsidization schemes are needed due to the negative cash balance in the first two years primarily for the fixed NGA networks. 3.2. Sensitivity analysis Sensitivity analysis has been carried out in order to rank a number of selected uncertainty assumption variables according to their impact on the NPV (Figs. 12–14). In a “traditional” approach, each of the selected parameters is changed on a one-by-one basis by the same percentage up and down. This is basically wrong, as some variables are inherently more uncertain than others. Instead, 5% and 95% percentiles have been chosen as lower and upper limits respectively for each variable and probability density functions have been applied to all parameters. For the majority of demand and economic input parameters of the model, these limits are based on the available Eurostat data for the European countries. The results are shown in Figs. 12–14 are based on the infrastructure competition scenario for the medium rural population density countriess. As was expected, all the networks are sensitive to changes in ARPU and demand related variables like market share and take-up

Fig. 13. Sensitivity analysis results for FTTC VDSL. 15

Telecommunications Policy xxx (xxxx) xxxx

N. Ioannou, et al.

Fig. 14. Sensitivity analysis results for FTTdp G.fast.

(service demand). Also, costs which make up the biggest part of CAPEX (spectrum fees, eNodeB site cost, trenching cost) and OPEX (site rent) have a serious impact on the feasibility of the investment. The influence of these factors in the profitability of the investment is another indicator that shows the importance of investment options like passive site sharing or the use of aerial fibers and policies like spectrum subsidization or regulation of infrastructure competition.

3.3. Investment risk and subsidization Demand is expected to be very sensitive to the price of offered services especially in rural areas. Therefore, it is important to specify a desirable tariff as well as the subsidization required to reduce the risks and ensure the profitability of the investment given the specified tariff. However, the differences in purchasing power, current broadband fares and regulation policies between the European countries do not allow the estimation of a single desirable tariff for all countries. Thus, an extensive risk assessment has been performed on different tariff scenarios and subsidization values. Τhe result is the minimum subsidization required (as a percentage of CAPEX) for the first two years of investment given an ARPU/tariff value in order to ensure (with the certainty of 95%) a

Fig. 15. Subsidization (percentage of CAPEX of each technology) required for given tariff value for NPV > 0 with 95% certainty. 16

Telecommunications Policy xxx (xxxx) xxxx

N. Ioannou, et al.

Fig. 16. Subsidization (percentage of CAPEX) required for given tariff value for NPV > 0 with 95% certainty, scaled to FTTdp G.fast CAPEX.

positive NPV at the end of the study period. The results in Fig. 15 show that in all the scenarios, the FTTdp network requires huge subsidization to ensure profitability compared to the FWA and FTTC networks, with FWA being the most cost-efficient option. In view of the big difference in the total CAPEX between the three networks, it is more useful to rescale (normalize) the results (Fig. 16) to the CAPEX of FTTdp G.fast (100%) in order to realize the actual difference in the amount of subsidization between FTTdp G.fast and FTTC VDSL or LTE FWA. The subsidization needed for LTE FWA is two times less that VDSL and nine times less than G.fast. In contrast to the results in Fig. 8 where total costs for LTE and VDSL are comparable, less subsidization is needed for LTE since OPEX is a big part of the total cost and is spread among the years of the project. It is important to note that in Figs. 15 and 16, we present results for the single infrastructure operator scenario as it represents a white NGA area case, which is in principle eligible for State Aid. In the infrastructure competition scenario, LTE requires 35–45% more subsidization (or 5–10% normalized to FTTdp CAPEX), VDSL requires 9–25% more (or 4–9% normalized to FTTdp CAPEX) and G.fast needs 5–10% more. However, as we have already mentioned, State Aid is less likely in grey NGA areas where there is an existing NGA network infrastructure or is already being deployed in the coming three years and specific criteria have to be met (“EU Guidelines for the application of State aid rules in relation to the rapid deployment of broadband networks,” 2013).

4. Conclusions In this paper, the feasibility of providing 30 Mbps internet services in rural areas of Europe through LTE FWA or FTTC VDSL or FTTdp G.fast is assessed. FTTdp G.fast is a promising technology that, in theory, can deliver up to gigabits of throughput per subscriber with less cost than FTTH. The FTTdp architecture enables an easy bandwidth upgrade and therefore it can be considered as a futureproof solution. However, the cost of FTTdp network rollout is too high for rural areas where population density is low and demand is not high enough to ensure the profitability of the investment even with subsidization. In case of fixed broadband network, other more cost-efficient solutions like FTTC VDSL networks offer a more realistic option, especially in denser rural areas. The FTTdp G.fast technology could provide FTTC VDSL a future upgrade to gigabit internet speeds in rural areas with high demand at a reasonable cost. Additionally, other methods of reusing the existing copper infrastructure like aerial cables should be used to reduce the trenching costs which is the highest portion of investment CAPEX for fixed NGA networks. The results showed that LTE Fixed Wireless Access networks could be a viable, cost-efficient solution for delivering 30 Mbps internet services in sparsely populated rural areas even in greenfield scenarios. Thus, considering the already high LTE radio access network (RAN) coverage and the percentage of site acquisition costs in investment CAPEX, substantial cost savings can be expected from passive and active RAN sharing. The existence of the current extensive LTE network infrastructure in rural areas of most 17

Telecommunications Policy xxx (xxxx) xxxx

N. Ioannou, et al.

European countries could also allow a very fast delivery of 30 Mbps services in an attempt to meet DAE objectives even as a “last minute” solution. In the case of only 10 MHz available spectrum, the results show a clearly worse prospect for the FWA investment but significant improvement can be expected through RAN sharing which should be used as a guideline for the operators. Furthermore, since spectrum is another large part of investment CAPEX, spectrum subsidization or new spectrum policies like spectrum sharing between operators should be discussed and encouraged by NRAs. In terms of the regulation policy, the low expected service penetration and the high investment costs greatly reduce the profitability aspects of NGA network investments in infrastructure-based competition scenarios leading to a market failure. We have seen that significantly lower ARPU is needed to ensure the profitability of the investment for all technologies in the scenario of a single infrastructure operator who provides virtual access to other competitors. Finally, new mobile network technologies like 5G networks could be a solution for upgrading LTE fixed wireless access networks in the future. More studies should be elaborated to evaluate the cost of such investments and to determine whether fixed wireless networks can be a futureproof solution for providing speeds comparable to fixed NGA networks in rural areas. Appendix A. Supplementary data Supplementary data to this article can be found online at https://doi.org/10.1016/j.telpol.2019.101875. References Bastion, H. (2019). LTE carrier aggregation. Retrieved July 30, 2019, from https://halberdbastion.com/technology/cellular/4g-lte/lte-carrier-aggregation. Cisco (2018). Visual networking index (VNI) forecast highlights. Retrieved October 31, 2018, from https://www.cisco.com/c/m/en_us/solutions/service-provider/vniforecast-highlights.html. Council of EU (2016). Press release 274/16, Freeing up 700 MHz band for mobile: Council agrees its position. Retrieved from https://www.consilium.europa.eu/en/press/ press-releases/2016/05/26/freeing-up-700-mhz-band-for-mobile/pdf. European Commission Communication. (2013). EU Guidelines for the application of State aid rules in relation to the rapid deployment of broadband networks. OJ C, 25, 1–26. Retrieved from: 26.1.2013 https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=uriserv:OJ.C_.2013.025.01.0001.01.ENG. European Commission (2010a). COMMISSION RECOMMENDATION of 20 September 2010 on regulated access to next generation access networks (NGA). Official Journal of the European Union §. European Commission (2010b). Press release IP/10/540, radio spectrum: Harmonized EU rules to foster high-speed wireless internet services and avoid harmful interference. Retrieved October 31, 2018, from http://europa.eu/rapid/press-release_IP-10-540_en.htm?locale=en. European Commission (2013). The socio-economic impact of bandwidth. Retrieved from https://ec.europa.eu/digital-single-market/en/news/study-socio-economicimpact-bandwidth-smart-20100033. European Commission (2017a). Digital single market, Europe's digital progress report 2017. Retrieved October 31, 2018, from https://ec.europa.eu/digital-single-market/ en/news/europes-digital-progress-report-2017. European Commission. (2017). Digital single market, study on broadband coverage in Europe 2017. Retrieved October 31, 2018, from https://ec.europa.eu/digital-singlemarket/en/news/study-broadband-coverage-europe-2017. European Commission. (2018). Digital single market, digital agenda scoreboard key indicators. Retrieved October 31, 2018, from https://digital-agenda-data.eu/datasets/ digital_agenda_scoreboard_key_indicators. European Court of Auditors. (2018). Broadband in the EU member states: Despite progress, not all the Europe 2020 targets will be met. Retrieved from http://publications. europa.eu/webpub/eca/special-reports/broadband-12-2018/en/. European Parliament and Council. (2012). Decision No 243/2012/EU of the European Parliament and of the Council of 14 March 2012 establishing a multiannual radio spectrum policy programme Text with EEA relevance. EUROSTAT (2018a). Statistics explained, electricity price statistics. Retrieved October 31, 2018, from https://ec.europa.eu/eurostat/statistics-explained/index.php/ Electricity_price_statistics. EUROSTAT (2018b). Statistics explained, housing statistics. Retrieved October 31, 2018, from https://ec.europa.eu/eurostat/statistics-explained/index.php/Housing_ statistics. EUROSTAT (2018c). Statistics Explained, Land cover and land use statistics at regional level. Retrieved October 31, 2018, from https://ec.europa.eu/eurostat/statisticsexplained/index.php/Archive:Land_cover_and_land_use_statistics_at_regional_level. EUROSTAT (2018d). Statistics explained, territorial typologies. Retrieved October 31, 2018, from https://ec.europa.eu/eurostat/statistics-explained/index.php?title= Territorial_typologies. Falch, M., & Henten, A. (2010). Public private partnerships as a tool for stimulating investments in broadband. Telecommunications Policy. https://doi.org/10.1016/j. telpol.2010.07.010. Feijóo, C., Ramos, S., Armuña, C., Arenal, A., & Gómez-Barroso, J. L. (2018). A study on the deployment of high-speed broadband networks in NUTS3 regions within the framework of digital agenda for Europe. Telecommunications Policy. https://doi.org/10.1016/j.telpol.2017.11.001. Frías, Z., González-Valderrama, C., & Martínez, J. P. K. (2015). Eys and challenges to close the rural broadband gap. The role of LTE networks in Spain. 26th European regional conference of the international telecommunications society (ITS), Madrid, Spain, 24-27 June 2015. Furuskär, A., Rao, J., Blomgren, M., & Skillermark, P. (2011). LTE and HSPA for fixed wireless broadband: Datarates, coverage, and capacity in an Indian rural scenario. 2011 2nd international conference on wireless communication, vehicular technology, information theory and aerospace and electronic systems technology, wireless VITAE 2011, 1–5https://doi.org/10.1109/WIRELESSVITAE.2011.5940858. GSMA (2017). GSMA, full report February 2017, effective spectrum pricing: Supporting better quality and more affordable mobile services. Retrieved October 31, 2018, from https://www.gsma.com/spectrum/wp-content/uploads/2017/02/Effective-Spectrum-Pricing-Full-Web.pdf. Jeanette Wannstrom for 3GPP. (2013). Carrier aggregation explained. Retrieved from 3GPP website https://www.3gpp.org/technologies/keywords-acronyms/101carrier-aggregation-explained. Jha, A., & Saha, D. (2016). Techno-economic assessment of the potential for LTE based 4G mobile services in rural India. International symposium on advanced networks and telecommunication systems, ANTS, 2016-Februhttps://doi.org/10.1109/ANTS.2015.7413612. Katsianis, D., Rokkas, T., Neokosmidis, I., Tselekounis, M., Varoutas, D., Zacharopoulos, I., et al. (2012). Risks associated with next generation access networks investment scenarios. IEEE Network, 26(4), 11–17. https://doi.org/10.1109/MNET.2012.6246747. Monath, T., Elnegaard, N. K., Cadro, P., Katsianis, D., & Varoutas, D. (2003). Economics of fixed broadband access network strategies. IEEE Communications Magazine, 41(9), 132–139. https://doi.org/10.1109/MCOM.2003.1232248. Nucciarelli, A., Castaldo, A., Conte, E., & Sadowski, B. (2013). Unlocking the potential of Italian broadband: Case studies and policy lessons. Telecommunications Policyhttps://doi.org/10.1016/j.telpol.2013.05.008. Olsen, B. T., Katsianis, D., Varoutas, D., Stordahl, K., Harno, J., Elnegaard, N. K., et al. (2006). Technoeconomic evaluation of the major telecommunication investment options for European players. IEEE Network, 20(4), 6–15. https://doi.org/10.1109/MNET.2006.1668398.

18

Telecommunications Policy xxx (xxxx) xxxx

N. Ioannou, et al.

Ovando, C., Pérez, J., & Moral, A. (2015). LTE techno-economic assessment: The case of rural areas in Spain. Telecommunications Policy, 39(3–4), 269–283. https://doi. org/10.1016/j.telpol.2014.11.004. Rokkas, T., Katsianis, D., & Varoutas, D. (2010). Techno-economic evaluation of FTTC/VDSL and FTTH roll-out scenarios: Discounted cash flows and real option valuation. IEEE/OSA Journal of Optical Communications and Networking, 2(9), 760–772. https://doi.org/10.1364/JOCN.2.000760. Rokkas, T., Neokosmidis, I., Katsianis, D., & Varoutas, D. (2012). Cost analysis of WDM and TDM fiber-to-the-home (FTTH) networks: A system-of-systems approach. IEEE Transactions on Systems, Man, and Cybernetics - Part C: Applications and Reviews, 42(6), 1842–1853. https://doi.org/10.1109/TSMCC.2012.2227999. Schneir, J. R., & Xiong, Y. (2016a). A cost study of fixed broadband access networks for rural areas. Telecommunications Policy, 40(8), 755–773. https://doi.org/10. 1016/j.telpol.2016.04.002. Schneir, J. R., & Xiong, Y. (2016b). Cost assessment of FTTdp networks with G.fast. IEEE Communications Magazine, 54(8), 144–152. https://doi.org/10.1109/MCOM. 2016.7537189. Teng, F., Guo, D., & Honig, M. L. (2017). Sharing of unlicensed spectrum by strategic operators. IEEE Journal on Selected Areas in Communications. https://doi.org/10. 1109/JSAC.2017.2672220. Varoutas, D., Katsianis, D., Sphicopoulos, T., Stordahl, K., & Welling, I. (2006). On the economics of 3G mobile virtual network operators (MVNOs). Wireless Personal Communications, 36(2), 129–142. https://doi.org/10.1007/s11277-006-0027-5. Ye, Y., Wu, D., Shu, Z., & Qian, Y. (2016). Overview of LTE spectrum sharing technologies. IEEE Accesshttps://doi.org/10.1109/ACCESS.2016.2626719.

19