Chapter XLV Diffusion Forecasting and Price Evolution of ...cgi.di.uoa.gr/~michalak/publications/Bookchapters/Chapter45_cm.pdfa product is offered is a substantially influential factor,

730

Chapter XLVDiffusion Forecasting and Price

Evolution of BroadbandTelecommunicationServices in Europe

Dimitris VaroutasUniversity of Athens, Greece

Christos MichalakelisUniversity of Athens, Greece

Alexander VavoulasUniversity of Athens, Greece

Konstantina DeligiorgiUniversity of Athens, Greece

Copyright © 2008, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited.

ABSTRACT

This chapter is concerned with the methodologies for the study of the diffusion patterns and demand estimation, as well the pricing schemas for broadband telecommunication services in Europe. Along with the introduction of diffusion models and price indexes which can represent broadband convergence and diversity, a description of the theoretical models and methodologies are given and application of these models in European telecommunication market is performed. Evidence from Europe outlines telecom market behavior and contributes to better understanding of broadband diffusion worldwide. To this direction, a price index is constructed regarding the ADSL technology.

731

Diffusion and Price of Broadband in Europe

InTRoduCTIon

The worldwide expansion of the Internet and other network-based services forced the partici-pant agents of the area of telecommunications to reconsider their strategic plans, in order to be able to meet the continuously increasing demand for required bandwidth as well as the continuously increasing numbers of users.

On the other hand, as network infrastructures are being constructed to support the increasing demand, many multimedia applications will be realized. Nowadays, customers’ desires consist of high-speed Internet access, as much as in-teractive, bandwidth—consuming multimedia applications.

Following these considerations, the area of telecommunications merits a continuous improve-ment and development towards the direction of the quantity and quality of the offered services. Con-temporary technology allows extended network capabilities and the development of new products, which in turn increase the quality of services offered to customers. However, convergence of telecommunication services often disorientates customers and regulators; the former concerning their potential selections among the offered prod-ucts and the latter regarding market monitoring and regulation. The price at which a service or a product is offered is a substantially influential factor, as it is related to its future diffusion among the potential adopters. In addition, from an eco-nomics point of view, telecommunication services obviously reveal the characteristics which strongly relate them with the network externalities effects, the phenomenon that a good becomes more valu-able to each user the more other consumers use the same or a compatible product.

Summarizing the above considerations, re-garding demand evolution and pricing shapes of products and the related influential parameters, this chapter is devoted to the provision of answers/out-lines concerning the following questions:

• How can we model the diffusion of broadband networks and services?

• How can we model the impact of countries/regions with higher broadband penetration to those with lower penetration levels?

• How can estimate prices for broadband services and products that enter the market for the first time or have been modified/con-verged?

• Can we determine a unified price index for these products in a specific period and what do these prices tend to become over time?

The rest of the chapter is organized as follows: a short overview of diffusion theory and diffusion models followed by a study of the evaluation of diffusion models over selected cases. The factors affecting broadband access development in whole Europe are then presented. An introduction of telecommunication services and ADSL high-speed Internet connections is given in the next section and then theoretical background to econometric methods and empirical models are given. The final section proceeds with a concluding discussion and suggestions for future development in the telecom-munications area.

BACkgRound

diffusion Theory and diffusionModels

As it is evident that innovations in the area of telecommunications and high technology, in gen-eral, are facing a significant and quite promising expansion, corresponding research activities focus on the study of the diffusion process dynamics. Their main targets aim towards the direction of developing and applying methods to provide comprehensive analyses concerning the demand for telecommunications products, services, and technology (Fildes & Kumar, 2002). The main The main areas of research interest include:

• The study and development of demand mod-els

• The development of mathematical methods for the estimation of the models’ parameters

732


• The evaluation of the results related to the models’ performance over cases of interest

The above cases are not exhaustive, as in a more detailed analysis, the study of the diffusion of multigenerational products (Rai, 1999), the influence on demand of cross-national diffusion (Kumar & Trichy, 2002), as well as the development of methods for estimating diffusion without or with limited prior data should be included (Venkatesan & Kumar, 2002). The latter cases constitute quite common characteristics in the area of telecom-munication innovations and in the wider area of high technology products.

Diffusion theory is a methodological approach used for estimating the adoption of technological innovations or other products or services. DiffusionDiffusion models are, consequently, mathematical functions of time which are used to estimate the parameters of the diffusion process of a product’s life cycle at an aggregate level, without taking in consideration the underlying specific parameters that drive the process (Bass, 1969). They are constructed such that they have a main target of capturing the general trends of the market reactions to an innovation’s introduction. These modelsareofma�or importance are of ma�or importance in determining the product’s expected life cycle and the associated parameters such as maximum penetration. The cumulative diffusion shapes of innovations are often described by sigmoid pat-terns, the so called S-shaped growth patterns, which originate from a number of early buyers (innovators). These initial adopters of the services and the size of the “critical mass” needed are of significant importance for the rest of the diffusion process, especially for the saturation level and the saturation time. Innovators’ decisions to adopt the service are independent from the decisions of the rest of the social system population. Apart from the innovators, there is another category of adopters, the imitators, who proceed to the adop-tion of the service, influenced by the interaction with innovators (word-of-mouth) and by external influence such as mass media communication and other communication channels. Finally, the market reaches maturity, when the maximum number of adopters among the considered population is met (market saturation).

Based on the concepts of diffusion theory presented above, the present section attempts to provide an insight concerning estimation and forecasting of broadband technology diffusion underlying mechanics. This purpose is the main reason for employing a number of existing diffu-sion models, the results of which can be considered as an intuitive confidence interval describing the expected values of the specific market’s ultimate potential. These aspects constitute the main func-tional utility of the present work.

Diffusion Models

The most widely used representatives of the ag-gregate models developed for diffusion estimation are the Bass model (Bass, 1969), the Fisher-Pry model (Fisher & Pry, 1971), logistic family models (Bewley & Fiebig, 1988), and the Gompertz model (Rai, 1999). Logistic models and variations of the Gompertz model provide “S-shaped” curves which are used in common in forecasting diffusion of products or services. S-shaped patterns derive from the following differential equation:

dY(t)=

dt⋅ ⋅

(1)

In (1), Y(t) represents total penetration at time t, S the saturation level of the specific technology, and δ is a constant of proportionality, the so-called coefficient of diffusion. Penetration is defined as the proportion of the population that uses the product or service being examined. In that sense, the diffusion rate of a product is proportional to the already recorded penetration as well as to the remaining potential of the market’s users.

As observed in (1), diffusion speed, as a func-tion of time, is proportional to a product of two factors: the population that has already adopted the service, denoted by Y(t), as well as to the remain-ing market potential represented by S-Y(t). It is easily derived, given t>0, that as the time variable increases Y(t) also increases towards S, and this can be interpreted as an increase of the “pressure” of the adopters over the nonadopters.

michalak

Note

the choice of the most appropriate model to describe diffusion (Taneda, Takada & Araki, 2001)

733


Based on the above assumption, a number of diffusion models have been developed for diffusion estimation and forecasting purposes. The most widely used ones include the Bass model (Bass, 1969), the Gompertz model, and models of the lo-models of the lo-gistic family (Bewley & Fiebig, 1988). The linear. The linear The linear logistic model, a representative diffusion model which belongs to the logistic family models, was used for evaluating the cases presented in this work. The linear logistic model is also known as Fisher-Pry model (Fisher & Pry, 1971). It is described by(Fisher & Pry, 1971). It is described by the following equation:

SY(t) = -a-b t1+e ⋅ (2)

Cross-National Diffusion

In many cases, the same product is introduced either simultaneously, or after a time lag, into a number of different markets (Kumar, Ganesh, & Echam-badi, 1998). This is quite frequent especially in the markets of high technology products because they usually target international markets. In such cases, it is quite important to study the diffusion proce-dure in well defined groups of markets, according to their characteristics, and the interaction among them. These market groups can be defined either as a group of neighboring countries, as a number of areas within the boundaries of the same country, or any other kind of geographical segmentation in general, according to various geographical or social characteristics imposed (Gatignon, Eliashberg, & Robertson, 1989; Michalakelis, Dede, Varoutas, & Sphicopoulos, 2005).

The effects of simultaneous introduction are related to the influential behavior between users of the corresponding markets as a result of people interaction (Bass, 1969). This fact is usually not taken into account when estimating the diffusion process of the product and the penetration among studied population. Thus, the effect of market interaction and the consequent co-influence in the diffusion rates is overlooked, although it can be able enough to modify the initially estimated diffusion process.

EvAluATIon of dIffuSIonESTIMATIon And foRECASTIng

In this section, evaluation results of the diffusion theory introduced in the preceding sections are presented, with respect to the ADSL technology. The datasets used for this evaluation describe penetration of ADSL connectivity across the Eu-ropean area. The corresponding data sources were Eurostat’s and OECD’s Web sites, as indicated while presenting the datasets. The cases consid-ered are the average values of penetration in the European Union (15 countries) and in the OECD countries. They were selected in order to depict the worldwide dynamics in ADSL adoption. In addition, two specific European countries are also included, Portugal and the United Kingdom. This particular choice was made in order to examine the diffusion process between two diverging countries; one considered developing (Portugal) and one considered developed (UK), together with their relative position as compared to EU and OECD mean values of ADSL technology penetration.

The historical data are semiannual values, as indicated by the values next to the year (e.g., 2002m10 refers to October of 2002) and they refer to penetration of ADSL technology for residential access only. It is obvious that the same evaluation can be performed over data regarding business con-nections. In the presented diagrams the horizontal axis (Years) depicts time, whereas the vertical (S) refers to the penetration value, over the referenced population. At this point, it should be noted that the term “population” has a meaning that varies, depending on each examined case, which must be clarified each time. Thus, in this particular case, the term population refers to individuals rather than households. A quite accurate approximation for performing a transformation between households and people, as this derives from demographic data (source: Eurostat), is the ratio 1:3. In this way, the results can be easily converted to depict penetra-tion among households.

The estimation and forecasting evaluation process was performed by using the linear logistic model (Fisher & Pry, 1971) while the estimation of the parameters was performed by employing

michalak

Note

;Takada & Jain, 1991

734


nonlinear regression (NLS). Moreover, the ac-curacy of the results provided and the consequent effectiveness of the model was based on appropriate statistical measures, such as the mean square error (MSE), the mean absolute error (MAE) and the mean absolute percentage error (MAPE). Another practice adopted for determining the model’s data fitting ability was to hold back some years of data and use them in order to determine the model’s forecasting power, based on the earlier portion of

the dataset, the training data. More specifically, four simulations have been performed by hold-ing back one, two, three, and four points of data, respectively.

The evaluation results are presented in the following tables (Table 1 through Table 4) and in the corresponding figures (Figure 1 through Figure 4).

Year EU 15 – Actual EU 15 - Est.

2002m07 2,3 2,4

2002m10 2,7 2,9

2003m01 3,4 3,5

2003m07 4,5 4,2

2003m10 5,1 5,0

2004m01 6,0 6,0

2004m07 7,6 7,2

2004m10 8,4 8,6

2005m01 9,9 10,2

2005m07 12,0 12,0

2005m12 14,2 14,1

2006m01 16,5

2006m10 19,1

2006m12 22,0

2007m04 25,2

AdSl diffusion - Eu 15

0,0

5,0

10,0

15,0

20,0

25,0

30,0

2002

m07

2002

m10

2003

m01

2003

m07

2003

m10

2004

m01

2004

m07

2004

m10

2005

m01

2005

m07

2005

m12

2006

m01

2006

m10

2006

m12

2007

m04

Years

S

EU 15 - Actual EU 15 - Est.

Figure 1. ADSL diffusion: European Union (15 countries)

Table 1. Actual, estimated, and forecasted values of ADSL penetration over EU (15 Countries)1

735


Year OECD - Actual OECD - Est.

2001 2,9 2,9

2002 4,9 4,8

2003 7,3 7,3

2004 10,2 10,2

2005 13,6 13,6

2006 17,2

2007 20,8

2008 24,4

2009 27,9

Table 2. Actual, estimated, and forecasted values of ADSL penetration for OECD countries2

AdSl diffusion - oECd Countries

0,0

5,0

10,0

15,0

20,0

25,0

30,0

2001 2002 2003 2004 2005 2006 2007 2008 2009

Years

S

OECD - Actual OECD - Est.

Figure 2. ADSL diffusion: OECD countries

736


Year Portugal - Actual Portugal - Est.

2002m07 1,5 1,6

2002m10 1,8 2,0

2003m01 2,5 2,6

2003m07 3,6 3,2

2003m10 4,1 4,0

2004m01 4,8 5,0

2004m07 6,4 6,0

2004m10 7,2 7,3

2005m01 8,2 8,6

2005m07 10,1 10,0

2005m12 11,5 11,4

2006m01 12,9

2006m10 14,2

2006m12 15,5

2007m04 16,6

Table 3. Actual, estimated, and forecasted values of ADSL penetration in Portugal3

AdSl diffusion - portugal

0,0

2,0

4,0

6,0

8,0

10,0

12,0

14,0

16,0

18,0

2002

m07

2002

m10

2003

m01

2003

m07

2003

m10

2004

m01

2004

m07

2004

m10

2005

m01

2005

m07

2005

m12

2006

m01

2006

m10

2006

m12

2007

m04

Years

S

Portugal - Actual Portugal - Est.

Figure 3. ADSL diffusion: Portugal

737


Year UK - Actual UK - Est.

2002m07 1,6 1,7

2002m10 2,1 2,1

2003m01 2,6 2,7

2003m07 3,7 3,5

2003m10 4,4 4,4

2004m01 5,3 5,6

2004m07 7,4 7,0

2004m10 8,8 8,7

2005m01 10,3 10,8

2005m07 13,5 13,2

2005m12 15,9 15,9

2006m01 19,0

2006m10 22,3

2006m12 25,8

2007m04 29,3

Table 4. Actual, estimated, and forecasted values of ADSL penetration in UK4

AdSl diffusion - uk

0,0

5,0

10,0

15,0

20,0

25,0

30,0

35,0

2002

m07

2002

m10

2003

m01

2003

m07

2003

m10

2004

m01

2004

m07

2004

m10

2005

m01

2005

m07

2005

m12

2006

m01

2006

m10

2006

m12

2007

m04

Years

S

UK - Actual UK - Est.

Figure 4. ADSL diffusion: UKADSL diffusion: UK

738


Table 5 summarizes the saturation levels, as these were estimated by the model, together with the expected year of reaching them, in each case. In addition, Figure 5 provides a comparison view among the evaluated cases. As it can be observed, the diffusion process in the European area is mov-ing towards a rapid growth, and it is expected to have covered a percentage of approximately 76% until year 2012. A somewhat slower but also steady growth is the process of ADSL technology adoption among the OECD countries. The mean European Union diffusion is expected to reach a higher level of saturation than the OECD mean value earlier in time than the latter.

Regarding the evaluation results concerning the two participating countries, Portugal and UK, although diffusion was at the same levels at the initial stages, the results indicate that in UK the diffusion rate is expected to be higher and be met earlier. Consequently, the saturation level is expected to reach a quite higher level in UK than

in Portugal. This is strongly related to the rate that high technology products penetrate in each country, together with other reasons such as demographic, cultural, and economic ones.

By inspecting Figure 5, the dynamics of ADSL technology penetration can be derived. More specifically, it can be observed that in the OECD countries ADSL has met a higher level of penetra-tion than in EU, from the year 2002 to the year 2006. However, the diffusion rate was higher in EU (the slope of the corresponding curve), which resulted in EU mean penetration becoming higher than the corresponding one of the OECD countries. According to this estimation, a higher saturation level is expected to be recorded in EU than in OECD, almost two times higher.

In addition, in the United Kingdom penetration was steadily above the European mean value and after the year 2005 revealed a substantial diver-sion above it. Exactly the opposite is the case of Portugal, which was below the European mean

EU OECD Portugal UK

Saturation Level 76,02 35,46 22,01 54,61

Year 2012 2014 2019 2018

Table 5. Saturation levels of ADSL penetration and expected time to saturation

Figure 5. ADSL diffusion: Comparison among EU, OECD, and selected European countries

AdSl diffusion - comparison

0,0

5,0

10,0

15,0

20,0

25,0

30,0

35,0

2002

m07

2002

m10

2003

m01

2003

m07

2003

m10

2004

m01

2004

m07

2004

m10

2005

m01

2005

m07

2005

m12

2006

m01

2006

m10

2006

m12

2007

m04

Years

S

Eu-15 oECd portugal uk

739


diffusion during all the time of study. In addition, after the year 2005 diverged negatively from the European mean, although it retains a steadily incremental rate.

Concluding, the performed evaluation provided an estimation of the so-far recorded diffusion of the ADSL technology among the studied markets and a forecasting for the expected future process. These results, especially for the part related to forecasting, were derived based on the current dynamics of the diffusion process. This means that they can be considered more as a worst case scenario of the expectations for ADSL adoption than as absolute values of penetration. To be more specific, if governmental and other public or private initiatives are supported towards this direction, the diffusion of the ADSL technology should be expected to grow faster and saturate at a much higher level. In any case, the recorded penetration among the referenced market must be fed into the diffusion model and the estimation should be recalculated in order to provide refined forecasting estimations. As mentioned above, the drivers affecting diffusion are related to a number of demographic, social, and financial aspects, prob-ably one the most important of them being pricing

policies. This is the main sub�ect of consideration of the following section.

fACToRS AffECTIng BRoAdBAnd ACCESS dEvElopMEnT In ThE WholE of EuRopE

Many factors influence the diffusion process in the telecommunication sector and, consequently, broadband access development in all of Europe. The amount of investments in telecommunications networks is one of the most important factors. In addition, liberalization and privatization in Europe is of great interest while the increased number of operators and carriers has a great impact to the whole market. Furthermore, the globalization of the economy in Europe provokes the necessity of a regulatory framework by the European Commis-sion. Finally, the subscribers’ profile is examined while the governmental policies and guidelines for the telecom market are also of considerable importance.

0

5

10

15

20

25

30

D e nmark

N e th e rland sIce

la n d

S w itzer la

ndF inla nd

N o rwa y

Sweden

U nited K i ng

dom Belgiu m

Lu xembou rgAustri

aF ra nce

German y

Spai n Ita l y

Portug a l

Czec h Re p ubl i c Irelan d

Hunga r yPo la n d

S lova k Re publ ic Gre ec e

0

10.000

20.000

30.000

40.000

50.000

60.000

70.000Broadband penetration (subscribers per 100 inhabitants, June 2006)GDP per capita (USD PPP, 2004)

Broadband penetration GDP per capita

Source : OECD

Figure 6. Broadband penetration and GDP per capita5

740


gdp per Capita and Investments in Whole Europe

A review in the European status shows a variation in telecom sector investments and Eastern Europe countries are trying to balance Western Europe countries. Overall, the level of existing investment in Eastern European countries is quite less than that of Western countries. However, the rate at which Eastern countries are investing in telecommunica-tions remains amazing. The strongest per capita subscriber growth comes from Denmark, Austria, Norway, the Netherlands, Finland, Luxembourg, Sweden, and the United Kingdom. In each of these countries more than six subscribers per 100 inhabitants were added, during the past year.

Generally, investments expected to be made during each year, or a period of time, depend on the required equipments and network infrastruc-ture. However, a crucial role to these investments is the fact that the equipment prices are decreasing over time due to the mass production (Broadwan, 2004). Other significant parameters in the invest-ment plans construction are the expected costs that will probably appear at the studied period, such as maintenance and repair costs of the existing network infrastructure, licenses costs (software and hardware), marketing costs for promoting a product, and customers’ installation. Many invest-ments plans have been proposed in order to estimate these parameters. However, the most difficult part to estimate are the running costs, since new services appear and in replacement of others.

In addition, it has been noticed that higher-in-come households are using Internet more often, for a longer period of time (Dickenson & Ellison, 1999) and for a variety of services (e.g., goods purchase, banking services, services under credit-card, video, audio, etc.) while people who live in urban areas are more likely to use Internet more often than people who live in suburban and rural areas. Thus, there is a higher investment risk in the latter areas with a long period of pay back, which in turn leads in a growing inequality of the Internet access capability among different geographical regions (Monath, Elnegaard, Cadro, Katsianis, & Varoutas, 2003).

from a Monopoly Situation to the liberalization

Telecommunications operated under governmen-tal provision and monopolistic status quo for a long period of time. The main reason, besides the governmental strategic perception, was military security. However, in the last two decades of the 20th century, governments reconsidered this op-tion which resulted in the start of liberalization in telecommunications. In Table 6, the year of liberalization for several countries across Europe is provided.

Legislation alone does not seem to be enough for the liberalization process to begin. The change from a central planned economy into a free market is difficult and customers have to learn, react, and interact in order to take advantage. When liberal-ization started in the telecommunications market, a fierce competition emerged among companies as they tried to attract as many customers as possible by offering packages in quite competitive prices. As a consequence, tariffs started to decrease and be-ing more accessible to potential customers. Within West Europe we can distinguish a difference in behavior between the North and the South. In some European countries, telecommunication sector followed a great development, but as it is believed they were “forced” by European Community to change to liberalization and privatization facts in order to have investments in their countries and become more competitive. Sweden can be consid-ered as a characteristic example, the leader of all European countries, where no monopolies exist. However, privatization raised other problems such as staff reduction in telecom companies, which is an accustomed way to increase profit. On the other side, the growing competition led to advantages for consumers who en�oyed lower tariffs. Coun-tries of Eastern Europe have to overcome ma�or economic problems which are preventing them to adopt European legislation more quickly.

Rapid Technological Change

The telecom sector is changing rapidly in the last decades, due not only to the fast technological

741


improvement in software and hardware but also in network infrastructure. The cooperation devel-opment among telecom industries across Europe inducted to the knowledge diffusion and the fast information trading. As a result, various changes in countries’ and companies’ profile took place. Companies needed resources and this translated into people, engineers, and investments. Another reason for this international cooperation derived from the mass production of several, continuously smaller, components. These parts are constructed in countries which have cheap “labor hands” and lower operational costs. In Eastern Europe the qual-ity of networks is generally low, with digitalization in low levels in Ukraine, Moldova, Belarus, and Bulgaria (Broadwan, 2004).

Telecom operators are expected to provide broadband communication services in all geo-graphical areas, due to subscribers’ preferences and the growing political interest (e.g., e-Europe 2005 guideline), necessitating the technological upgrad-ing as inevitable. As a result, network operators are enforced to increase the available bandwidth in order to give more services to customers. This is achieved by replacing parts of existing networks network with fiber infrastructure, especially in areas with specific characteristics such as high density of population—potential subscribers. In addition, they have to identify these characteristics and make corresponding plans in order to ensure a secure profit after upgrading their network. More specifically, they have to examine the existence of these characteristics in urban, suburban, and rural areas. These include the necessary cable length for the expected connections, the number of subscrib-ers per square km, the number of buildings in each specific area, as well as the number of potential subscribers per building (Monath et al., 2003). As a consequence, a detailed study and identification of the necessary network infrastructure in order to satisfy Internet services and traffic characteristics is extracted. It is worth mentioning that in North Europe, the differences between rural and urban areas concerning the use of Internet are quite small, while the differences in south Europe are observable (Kalhagen & Olsen, 2002). Rural areas of Northern Europe have low population density to

Country Liberalization

United Kingdom 1991

Sweden 1996

Nertherlands 1997

Italy 1997

Belgium 1998

Austria 1998

France 1998

Germany 1998

Spain 1998

Luxenbourg 2000

Ireland 2000

Portugal 2000

Greece 2001

Czech Republic 2001

Estonia 2001

Poland 2001

Slovenia 2001

Hungary 2002

Latvia 2003

F.Y.R.O.M. (Former Yugoslavian Republic of

Macedonia)2003

Bulgaria 2006

Lithuania 2006

Romania 2006

Table 6. Year of liberalization for countries of Europe6

742


present which leads to very high costs of network infrastructures. On the contrary, in South Europe the situation is quite different. In rural areas, with high population density, which is an advantage for the network infrastructure and equipment cost, the demand for broadband services is low as a result of social and economic differences between them and the urban areas.

Generally, as observed in Figure 7, Iceland, Sweden, and Norway (Northern Europe) met a high level of broadband penetration, even if in these areas population density is low, where in countries like Greece, Poland, and Germany (Central and Southern Europe) where population density is higher, the corresponding broadband penetration is low.

operators, Carriers, and Competition

Network operators have the responsibility to install, manage, and operate telecommunication transmission network so as to offer public tele-phone or network services. They are distinguished

in local operators, which offer services to users who live in specific areas, and national operators, which offer services no matter where the user lives. They were “forced” by the EU Commission to provide a minimum set of telecommunication services according to specific technical standards. This means that EU Commission applies similar conditions in similar circumstances to companies which provide similar services by the meaning of the same quality.

Different number of operators that have license or authorization to offer network services with different entrance times in the market can be found across member states. Thus, operators are developing new tariff policies with different cus-tomers paying different prices for the same service, depending on the time they signed a contract.

In Eastern Europe, only a few operators are making much profit because even if there is a high Internet penetration rate; the usage of Internet is far away from Southern European countries (Gunther, 2001). Low Internet penetration constitutes a brake on their market. In addition, according to Table 7,

0

5

10

15

20

25

30

Icel a n d

S w itz e r l a n dF in la n d

B elg i umS p a i n

0

50

100

150

200

250

300

350

400

450Broadband penetration (subscribers per 100 inhabitants, June 2006)

Population density (inhab/km2, 2004)

Broadband penetration Population density

Source : OECD

United Kingdo

mDenmark

Netherlands SwedenNorway

Luxembour

gAustria France Germany Italy Portug

al

Czech Repub

licIrela

ndHunga

ryPoland

Slovak Repub

lic Greece

Figure 7. Broadband penetration and population densities7

743


it is obvious that in EU candidate countries which are Bulgaria, Croatia, F.Y.R.O.M., Romania, Tur-key, and the rest of European countries there is a significant usage growth across years, where as the Internet penetration compared to the population is quite less yet.

In addition to the information provided in Table 7, carriers are promoted in all of Europe. Carriers are companies that buy lines in wholesale and are independent of technical changes as they can act quicker. They buy big capacities of lines at lower prices and they rent them. This will lead to a continuing reduction of prices since customers can choose among several carriers.

In order to ensure high quality and reliability of Internet services, telecom operators offer a number of choices according to the customer de-mands and paying abilities. As a result, there is a fierce competition among them to attract more subscribers (Goff, 2002). Some of them offer con-tracts with free of charge extra time to an Internet connection, the duration of which depends on the offered bandwidth. There is also a great differ-ence between residential and business packages offered. Currently, operators offer ADSL packages with varying data rates. There are low prices for Internet connections that are designed to attract residential customers, while the most expensive ones, providing higher speed connections, are designed for business customers. On the other hand, some operators offer free unlimited access to Internet, provided that telephone companies agree to pay these providers (Haan, 2001). It is obvious that as the competition among firms is growing, customers will en�oy lower prices with better services.

delivery and Repair Times ofTelecommunication Servicesby Companies

The time period from the date when the user makes a request for a specific type of a telecom-munication service until the request is fulfilled is called delivery time. The determination for each country is difficult, mainly due to the large num-ber of operators. However, it seems that serious delays still remain in some countries, even if there are improvements. This is a ma�or problem for the telecommunication sector and especially the broadband expansion, since the rapid accommoda-tion is a key component for the ongoing diffusion of broadband services.

On the other hand, the intermediate period of time from when a failure message has been given to the responsible unit within the organization, which provided the specific type of telecom-munication service, until its connection has been re-established and notified back in operation to the user is called repair time. There are remarkable differences across member states at repair times too. This points out an important impact on the market again because the quality of service and the response time for repairing it, if any damage happens, shows a continuing competition among companies, so that customers are satisfied and, as a result, prices decline.

Subscribers’ Profile

It is evident that people across Europe are accessing Internet at a growing rate. Northern European coun-tries such as Denmark, the Netherlands, Iceland,

EuropePopulation (2006

estimation)% population

of worldInternet users,

Latest DataPenetration (%

population)% Usage of

worldUse growth

(2000 – 2006)

European Union 462.371.237 7.1 % 239.881.917 51,9 % 22,1 % 157,5 %

EU candidate countries

110.206.019 1.7 % 24.983.771 22,7 % 2,3 % 622,1 %

Rest of Europe 234.711.764 3.6 % 43.847.215 18,7 % 4,0 % 417,5 %

Table 7. Internet usage in Europe8

744


Switzerland, and Finland continued their advance with high broadband penetration rates (Table 7). Internet penetration imposes significant diversions across European countries, being influenced by several characteristics such as the age, the educa-tional level, the level of the income, and the type of the family (Dickenson & Ellison, 1999). It is well known that different needs are satisfied in different ages. For example, young people are interested in Internet for playing games, listening to the radio, or en�oying educational services. On the contrary, an adult has other interests like e-mail accessing, health and tourism information, electronic banking, and purchase of goods. Moreover, the educational level of the subscriber is an important factor, since higher-educated people are more familiar and interested in new technologies (Lu, 2003) and are

more willing to use Internet for their needs. In addition, there is a strong relationship between income and Internet use (Madden & Coble-Neal, 2005). Households with a higher income have a stronger effect on technology adoption, as they are expected to have already used some kind of such technology in the past, obtaining the neces-sary familiarity. It is obvious that in developing countries possible customers have higher income, which means greater payment ability, and they can afford to pay higher prices for technology equipment and services. A last characteristic of the subscribers’ profile is the type of family. The demand in Internet usage seems to be higher in families with children, instead of families with no children, or one person households. This is because all members of a family are accessing

2001 2002 2003 2004 2005

Austria 3,6 5,6 7,6 10,1 14,1

Belgium 4,4 8,7 11,7 15,5 18,3

Czech Republic 0,1 0,2 0,5 2,5 6,4

Denmark 4,4 8,2 13,0 19,0 25,0

Finland 1,3 5,5 9,5 14,9 22,5

France 1,0 2,8 5,9 10,5 15,2

Germany 2,3 4,1 5,6 8,4 13,0

Greece 0 0 0,1 0,4 1,4

Hungary 0,3 0,6 2,0 3,6 6,3

Iceland 3,7 8,4 14,3 18,2 26,7

Ireland 0 0,3 0,8 3,3 6,7

Italy 0,7 1,7 4,1 8,1 11,9

Luxembourg 0,3 1,5 3,5 9,8 14,9

Netherlands 3,8 7,0 11,8 19,0 25,3

Norway 1,9 4,2 8,0 14,8 21,9

Poland 0,1 0,3 0,8 2,1 2,4

Portugal 1,0 2,5 4,8 8,2 11,5

Slovak Republic 0 0 0,3 1,0 2,5

Spain 1,2 3,0 5,4 8,1 11,7

Sweden 5,4 8,1 10,7 14,5 20,3

Switzerland 2,0 5,6 10,1 17,5 23,1

United Kingdom 0,6 2,3 5,4 10,5 15,9

Table 8. Broadband subscribers per 100 inhabitants, 2001-20059

745


Internet for different reasons and preferences. Northern European countries such as Denmark, the Netherlands, Iceland, Switzerland, and Finland have continued their advance with high broadband penetration rates (Table 8).

globalization of the Economy

In the decade of acceleration, globalization brings chained reactions to legislation, economies, and social movements, which are formed at an in-ternational level. As the competition is growing, company merging becomes the solution for the dominant market share acquisition. This led to a continuing reduction of service prices and there is a trend in all companies to charge the same prices for the same services. The global economic slowdown has created a negative balance for the telecom companies and the telecom market is quite fragile. As a result, a unified regulatory framework for telecom companies is derived by the European Commission in order to protect consumers and to promote a healthy competition among them.

governments’ policy and Instructions

Governments are trying to promote broadband communications since the information being traded through Internet includes health, tourism, entertainment, and commerce-oriented sub�ects (Kalhagen & Olsen, 2002). In general, the promo-

tion of broadband services has a positive effect to economic activity and speeds up the economic development (Kim & Galliers, 2004) while such policies will improve the quality of life and mod-ernize the less developed areas by offering these services. Consequently, as the governments rec-ognizes the need for special attention, treatment, and confrontation in rural areas, they proceed to adopt active measures by creating network in-frastructures, which leads to high speed Internet access and so all citizens can participate in an information society.

TElECoMMunICATIon SERvICES: AdSl hIgh-SpEEd InTERnETConnECTIonS

Broadband services and applications are classified according to the offered data rate. The domination of ADSL technology for broadband access across Europe during the last years demonstrated the high-speed Internet access and IP-telephone as the most common broadband services. However, a wide range of new services is now present in the mass market with an increased penetration of new technologies (e.g., FTTx, ADSL2++, WiMAX) (Ecosys, 2004). Table 9 summarizes typical. Table 9 summarizes typical broadband services that are expected to dominate in the near future, together with the corresponding demand in bandwidth.

Figure 8. Typical ADSL service basket

746


In order to specify the basic basket of broadband services an extended survey of providers across European countries took place, focused on the services offered, the pricing policy as well as the development in broadband market. As a result, the typical ADSL service basket was determined as a combination of main and additional services ac-cording to Figure8.MainservicesaredistinguishedFigure 8.Mainservicesaredistinguished. Main services are distinguished into “horizontal” and “vertical” according to the number of fixed variables among supported data rate (DR), maximum consumed data volume (V) and maximum allowed minutes on line (T).

Additional services include a number of e-mail addresses, Web space for Web hosting and/or file storage, and optional free local and nationwide phone calls and static IP addresses. The choice of the appropriate combination for each operator is depending on the specific business plan as well as the techno-economic model parameters and assumptions.

The evolution of broadband technology offers new and challenging options. The EU Commis-sion’s “Broadband for all” policy is expected to grow the interest for broadband in the next years and to enforce the infrastructure competition among providers. As a result of this competitive environment, the provision of enhanced broad-band services with reduced tariffs is expected to increase significantly the number of broadband subscribers.

ThEoRETICAl ModEl

Econometric Methods

Econometric methods have been used for price index calculation for a long period of time: cars (Griliches, 1961), refrigerators (Triplett & Mc Don-ald, 1977), and computers (Cole, 1986) are some examples. Furthermore, indices about information technology can be found in Cartwright and Smith (1988) and Moreau (1991). In addition, statisticians use econometric methods in the U.S., but the root of the hedonic approach, which is a part of economic research, goes back to Waugh (1928), Court (1939), and Stone (1954,1956).

There are two types of econometric methods, hedonic methods and matched model methods, each of which have both advantages and disadvantages. One choice is to apply the “hedonic methods,” such as two-period method, single-period method, two-period method with an indicator for new models or single-regression method. Such indices are commonly used for products which undergo rapid technological changes.

Hedonic methods refer to regression models in which product prices are related to product charac-teristics and the observed price of a product (service) is considered to be a function of its characteristics. Generally, hedonic methods are based on the idea that a service (product) is a bundle of characteristics and that consumers �ust buy bundles of product characteristics instead of the product itself. These methods can be used to construct a quality-ad�usted

Service Bandwidth (Mbps)

High definition TV (HDTV) 16 – 20

Telemedicine 6

Video on demand (VoD) 6 – 18

Internet access 1,5

Video conferencing 1,5

Telecommuting 3

Multiple digital TV 6 – 24

Table 9. Required bandwidth for typical broadband services

747


price index of a service. Berndt (1991) and Triplett (2000) described an overview on hedonic price equations. Rosen (1974) states that from a large amount of product varieties, a consumer chooses without influencing prices. Therefore, consumers maximize utility and producers maximize profits. In hedonic studies, it is possible to ad�ust the price of a service for its quality not quantity. All of them are based on some estimated coefficients that are inflicted on the characteristics of the products in both periods: m and m+1. Someone can estimate the coefficients for every year separately or can have observations of two or all years together and estimate a common set of coefficients. The advan-tage of this method is that calculations are easy and fast. Indeed, hedonic methods are very fast to apply but the disadvantage is that index price can change even if no new products are existed or all prices remain the same.

Another choice is to apply a matched model method such as chained Laspeyres (LCPI, LPI) or chained Paashe (PCPI, PPI) or chained Fisher or chained Tornqvist or chained geometric-mean (Okamoto & Sato, 2001). A classic LPI cannot deal with such complexity due to rapid technologi-cal changes or the introduction of new products (services). With LPI, an index shows how much the product would cost in period m+1 relatively to what it cost in period m. Other price indices function in the same way with slight differences.

The hedonic price indices are commonly used as approximations to the true cost-of-living in-dices (COLI), which indicate how much money a consumer would need in period m+1 relatively to the amount of money he needed in period m so as to keep the same level of utility in period t as in period t0 (Jonker, 2001). The solution is to determine consumer’s profile so as to react to a varied and fast-changing supply of products. But how can this profile be determined when everyone has different needs and requirements? No matter what profile is decided, it will be a hypothesis and an assumption that will respond at a specific model. In addition to the above, someone can see that consumer’s desire is not stable and this is not unreasonable because there is a great offer as the “goods” of technology become more and more at-

tractive. However, according to this approach the price index is constructed only using the prices of products, which are available in two ad�acent periods.

Matched Model Method (laspeyres Method)

According to Laspeyres, in order to create a price index at a time, someone observes the number of units sold in a period m (for example a month) and the average unit price in the period m and m+1. These data are used in the following formula:

(3)1

11/

1

n

im imi

m m n

im imi

p qI

p q

+=

+

=

=∑

∑

Price indices are measured, as it is mentioned above, by the matched model method of Laspeyres with chaining average unit prices, which are re-ferred to a previous period among units sold in the same period.

The problem with such a price index is that, as it is known, a basket of products does not remain the same over time. Furthermore, it is noticeable that some products disappear from the market (especially in the telecom market) and some others are modified, so someone has to introduce new products in order to keep the basket indicative of customers’ preferences. If the quality changes are ignored, the resulting price index will be biased.

hedonic Method

The term “hedonic methods” refers to a “hedonic function” f(X), which is used in economic mea-surement,

Pi = f(Xi) (4)

where Pi is the price of a variety (or a model), i of a product, and Xi is a vector of characteristics associ-ated with the variety. The hedonic function is then used, for different characteristics among varieties of the product, in calculating the price index.

748


As soon as it is determined which characteris-tics have to be considered, then the equations (13) and (14) are estimated for N telecommunication products in period m and m+1:

ln(pim) = b0+ b1 ⋅ X1i + b2 ⋅ X2i + uim , i = 1,…,N (5)

ln(pim+1) = b0+ b1 ⋅ X1i + b2 ⋅ X2i + b3+ uim+1 (6)

where bi are some coefficients that have to be estimated.

True and Candidate Model by using the hedonic function

In this approach, there is a set of consumers who have preferences over some characteristics of a service. The construction of a price index is com-plicated by product-pricing limits such as different charges for various characteristics, so the first thing someone has to do is to define a basket of services (products). The model which describes the attributes (characteristics) of a product and their prices is given by the function:

Pi = f(Xb)+u (7)

where X =(x1,x2,……xn) is an n×p matrix of random regressors values, xi and b are p×1 vectors and f(Xb) is an n×1 vector with i-th component f(xi΄�b), (i=1,2,3,….n), u for given X=x is distributed as N(0,��2 I n×n) and � is an unknown scalar. Fur-� is an unknown scalar. Fur- is an unknown scalar. Fur-thermore, we assume that f is an unknown func-tion, which means that its shape is unknown and b estimators have a unit norm. In a similar way to the true model, we construct the candidate models by the definition of characteristics that “play” a significant role to the construction of the price index. In order to estimate the distance between the true and candidate models, is necessary the single-index model to be described.

Single-Index Model

Single-index models can be estimated by using iterative or direct methods. Whatever method will be followed, the result will be the same. In the

iterative method nonparametric regression is ap-plied in order to be calculated the mean regressor b. By using the iterative method the computation is difficult because it is required an estimate of nonparametric mean regression at each data point in order to compute a representative function f. In contrast, by using the direct method the com-putation is easier because the b relative weights are estimated by sliced inverse regression (SIR) (Li, 1991) and it is not necessary the estimation of function f.

Once the relative weights bi are estimated, an index SIRz Xb= is constructed and the nonlinear link ˆ ( )price f z= is estimated by applying local is estimated by applying local polynomial regression (LPR) (Simonoff, 1996).

As soon as ˆ ˆfandb are estimated by iterative or direct method we compute:

(8)2ˆ ˆ ˆ ˆ{ ( )} { ( )}ˆ f Xb Y f Xb

n′Υ − −

=

So, the suitable model from a variety of candi-date models via the AICc by using 2ˆ ˆ ˆ( , , )f b .

Assumptions for AICc Criterion

In order to find out which is the best model, from a variety of candidate models that describe a product with a set of characteristics, the following equa-tions are used:

d( 2ˆ ˆ ˆ( , , )f b = E0{-2log f(Y)} (9)

where f(Y) shows the possibility for the candidate model and E0 shows expectation under the true model. So,

2ˆ ˆ ˆ ˆ1 ( ) /

ˆlog ˆ ˆ ˆ ˆ1 { ( ) 2} /p np p np

cp np p np

tr � � � � nAIC

tr � � � � n+ + −

= +− + − +

(10)

where 1ˆ ˆ ˆ ˆ ˆ ˆ( ) ,p� V V V V V−′ ′= is obtained by replac-ing b* and .f inV

with estimators ˆ ˆbandf , and ˆ

np np� is� evaluated at ˆXb Xb= . Because of the difficult computation, since

there is an unknown function, without having great

749


inclination, the following assumptions facilitate the computation of the AICc value:

Assumption 1: ( *) npf Xb � Y≅

Assumption 2: Similarly, 0 0{ ( *)} ( *)E f Xb f Xb≅

Assumption 3: In addition to the above,ˆ ˆ ˆ ˆ( ) ( *) ( *) { ( *)}pf Xb f Xb V b b � Y f Xb− ≅ − ≅ −

where1

*( ) , ( ) / | .( *)p b b� V V V V V f Xb b f Xb X−=′ ′= = ∂ ∂ =

A problem which is posed is the selection of the best model, so in order to estimate prices, a sliced inverse regression (SIR) is performed, without knowing the shape of the function (Li, 1991). Then a local polynomial regression (LPR) is applied and a possible shape of the hedonic function is extracted. Among several candidate hedonic models the best one is derived by applying the Akaike information criterion (AIC) (Naik & Tsai, 2001).

EMpIRICAl ExAMplE

Models with nonlinear functions

Let’s have an example in which the characteristics of the product are determined, so as to show the

quality of this product. As soon as the character-istics of a product are determined, by applying the hedonic method, someone estimates its price. Otherwise the price of a product can be estimated “manually” by comparing the new product with the most similar old one.

The main assumption is that ADSL connections have four main physical characteristics: supported DR (up and down), maximum V, and maximum T. These four characteristics are widely used from telecom operators for valuating and selling leased lines across Europe.

Regarding the ADSL connections data, which are collected from year 2003 to 2005 for both residential and business connections, it is clear that there is a large variation of values in the variables and even more the price level is also varying from country to country, for the same service considered.

First of all, the hedonic model is estimated, de-scribed by (17). The data are sorted by Pi and then are divided into three slices. Without specifying the unknown link function we derive:

(0.581899, 0.78326, 0.00011, 0.21886)SIRb = − − −

The above results imply that the price is strongly related only with the downlink DR (Triplett, 2004). As it can be seen after inspecting the diagram

-100000

-80000

-60000

-40000

-20000

0

20000

0 20 40 60 80 100 120 140 160 180 200

Figure 9. Plot of Pi against SIR directions

750


1 2 3 4 5C3

-100000

-80000

-60000

-40000

-20000

0

20000

C2

Figure 10. Local polynomial regression with kernel smoothing

-60

-40

-20

0

20

40

60

80

100

1 2 3 4 5

Residuals of linear modelon price

Residuals of linear modelon ln-ned price

Figure 11. Residuals

(Figure 14), the described relationship can be easily expressed as a linear model.

By having four characteristics, we take under consideration 42-1 nested candidate models. For each of the nested models SIR estimates are ob-tained (Figure 11) and then by applying the LPR (Figure 12) the link function ( 1, 2,3,..15)kf k =

. Figures 11 and 12 also show that some individual

tariffs are existing which decline significantly from the main cluster.

In order to examine the relationship between the price of an ADSL connection and their main characteristics, such as the supported data rate and the maximum consumed data, several candidate link functions are applied. Across the candidate models in several shapes of link function, the

751


one which gives the smallest AICc (0,2125746 vs. 7,846617) value from (20), is described by:

ln(P) = b0+b1•DR(down)+b2•DR(up)+b3•(V)+b4•(T) (11)

Even link functions such as hyperbolic sine or hyperbolic cosine give almost the same results with the non linear functions. Although linear model on logged price is not comparable with all the other models, because of the AICc value, linear model on logged scale has an advantage.

Working in the logarithmic scale using a linear model results shows a better fit than all the other models because the residuals from the log-linear model are all around zero (Figure 13) (Triplett, 2004) and have less standard error (0,5845 vs.26,58).

Constructing the price Index

Finally, when there are N telecommunication prod-ucts in period m and m+1, the proposed hedonic price index can be calculated by the following equation:

1/ 1( * ) ( * )m m m i i m i iI f b X f b X+ += −

(12)

Using data such as those presented in Table 4 for the case of Europe, it can be observed in Figure 14 that it is not easy to compare prices for different data such as data rate, consumed volume, allowed time on line, and so forth, but there are similarities

and patterns that must be evaluated. It is obvious that the more a consumer demands for a product with upper services, the more the prices of this product are increased, even if there is a remarkable trend to be kept prices in low levels. Because of this consumers’ “behavior” there are no implicit prices for all characteristics.

This behavior fits to the hedonic approach and it can be observed by calculating the hedonic price index from (22). In Figure 15, this index and its evolution are presented for the case of European countries broadband connections market, even if market, even ifmarket, even if there isn’t historical data.

ASpECTS And ConCluSIonS

As observed in the preceding sections, diffusion methodology was quite capable of describing the diffusion process of ADSL technology over the studied cases. According to the results, there is obviously an increasing rate in the adoption of ADSL services, as was probably expected. As indicated, the used actual data refer to the penetra-tion of ADSL technology over 100 inhabitants that access this kind of broadband networks. If these numbers were expressed in terms of households, considering an average number of three people in a household, it is concluded that the penetration has already reached a substantially high level and it is expected to grow even more.

In the context of the newcomer countries in the European Union, there is an increased interest in considering the application of such methodologies

Subscription PriceData rate(Down)

(Kbps)Data Rate(Up)

(Kbps)Volume(Mbps)

59,95 10000 512 15.360

14,95 256 64 307,2

21,95 1536 256 1.536

29,95 800 256 -

74,95 12000 1024 8.192

Table 10. Data for ADSL connections

752


0

500

1000

1500

2000

2500

3000

3500

4000

4500

2003 2004 2005

Years

DownUpSubscription pricesLinear (Down)

y = 0,5339x + 0,0221R2 = 1

0

0,2

0,4

0,6

0,8

1

1,2

2004/2003 2005/2004

years

inde

xes

Figure 12. Evolution data rate (down) in EuropeEvolution data rate (down) in Europeate (down) in Europe

Figure 13. �edonic price index evolution for the case of ADSL market in Europe�edonic price index evolution for the case of ADSL market in Europe for the case of ADSL market in Europe

in order to estimate the expected process of adop-tion of broadband technologies. In this case, the application of the corresponding cross-national methodology is also expected to provide interesting and useful results, to be used as indicative drivers to the performance of the pertinent technoeconomic analysis. More specifically, in the context of ADSL networks there is always the case of the need of a precise estimation of the network that should be developed in order to support the increasing future demand of the offered services. This example re-veals that the accuracy of the forecasting results

is an extremely critical factor, mainly because of the strong relationship with future investments that will be required.

An even more detailed methodology, which is one of the ma�or ob�ectives in the corresponding area of research, should include the economical and social indicators in the areas of interest. Although the methodology is quite robust, demand can be expected to rise even more, depending on the levels of some important economical and social indicators in the markets of reference. The most representative of these indicators include the GDP per capita, the

753


industrialization level of the country, the unemploy-ment level, the potential user’s perceived risk of choice, the time of the product’s introduction, the education level of population, and the competition level of telecommunications’ market. These fac-tors, among others, are major drivers for defining the demand function of the product.

In addition, a study of factors that affect the broadband expansion and development, the domi-nant characteristics that influence the tariff policy of ADSL services, and finally the construction of price index for ADSL connections is presented. It has been shown that the general trend is the decrease of subscription prices, as expected. It is evidenced that customers are taking advantage from opera-tors’ competition. This is extracted from the fact that they are interested in the offered downstream data rates, since these parameters designate the quality of the subscribed services.

In addition, this analysis is to foresee the trend of prices of telecommunication services (products) and especially ADSL connections over time across all countries of Europe, applying a hedonic method for some defined characteristics. This method works better when there is a variety of important characteristics but less satisfactorily when these change rapidly over time. ADSL connections have important and specific characteristics indeed and their prices vary slowly over time. The results give a view of telecommunication prices over time and show how the prices will fluctuate the next year.

The application of these econometric methods, following the definition of products’ characteristics, provides a reliable and accurate method able to produce an exact estimate of prices both for new products and over next years. The validity of the model and the appropriate selection of the func-tional form that has been chosen to relate price to characteristics can be validated over the next years and more observations.

However, as the factors affecting broadband tariffs can be numerous, a more detailed study that takes into account additional parameters is needed.

EndnoTES

1 Source for actual data: Eurostat (available online at: http://ec.europa.eu/eurostat/).

2 Source for actual data: OECD (available online at: http://www.oecd.org/).

3 Source for actual data: Eurostat4 Source for actual data: Eurostat.5 Source: OECD6 Source: Yankee Group Europe (1997); Regu-

latory Developments (2000).7 Source: OECD8 Source : Internet World Stats9 Source: OECD broadband statistics to June

2006.

REfEREnCES

Baltas, G., & Freeman, J. (2001). Hedonic price methods and the structure of high-technology in-dustrial markets, an empirical analysis. Industrial Marketing Management, 30, 599-607.

Bass, F.M. (1969). A new product growth model for consumer durables. Management Science, 15(5), 215-227.

Berndt, E.R. (1996). The practice of economet-rics: Classic and contemporary. Addison-Wesley Publishing.

Bewley, R., & Fiebig, D.G. (1988). A flexible logistic growth-model with applications in telecommu-nications. International Journal of Forecasting, 4(2), 177-192.

Bonnetain, P. (2003). A hedonic price model for islands. Journal of Urban Econ, 54, 368-377.

Broadwan, (2004). IST project broadband access roadmap based on market assessment and techni-cal-economic analysis. Deliverable 15.

Budde, P. (2004). Eastern European infrastructure, fixed voice and data market report. Communica-tion Pty Ltd.

754


Buiter, (2004). To purgatory and beyond. In Con-ference on challenges for central banks in an enlarged EMU. Vienna, Austria.

Cartwright, D.W., & Smith, S.D. (1988, November). Deflators for purchases of computers in GNP: Revised and extended estimates 1983-88. Survey of Current Business.

Central and Eastern European Countries Synthesis of Master Reports. (2000). Regulatory Develop-ments.

Cole, R. (1986). Quality-ad�usted price indexes for computer processors and selected peripheral equipment. Survey of Current Business. Bureau of Economic Analysis, U.S. Department of Com-merce.

Communications Committee of European Com-mission. (2001a). Final 2001 report on performance in the supply of leased lines pursuant to Directive 92/44/EC (Working Document).

Communication from the Commission to the Coun-cil, the European Parliament. (2001b). The econom-ic and social committee and the committee of the regions. In Seventh Report on the Implementation of the Telecommunications Regulatory Package. Commission of the European Communities.

Communication from the Commission to the Council, the European Parliament. (2001c). Status of licensing and fees for fixed networks and ser-vices. In Seventh Report on the Implementation of the Telecommunications Regulatory Package. Commission of the European Communities.

Communication from the Commission to the Coun-cil, the European Parliament (2002). The economic and social committee and the committee of the regions. In Eighth Report on the Implementation of the Telecommunications Regulatory Package. Commission of the European Communities.

Court, A.T. (1939). Hedonic price indexes with automotive examples. In The dynamics of automo-bile demand (pp. 99-117). New York: The General Motors Corporation.

Dewenter, R., Haucap, J., Luther, R., & Rotzel, P. (2004). �edonic prices in the German market for mobile phones (Discussion Paper No.29). Hamburg, Germany: University of the Federal Armed Forces.

Dickenson, P., &Ellison, J. (2001).&Ellison, J. (2001). Ellison, J. (2001). Plugging in: The increase of household Internet use continuous into 1999. Connectedness Series, Statistics Canada.

Diewert, (2003). Services and the new economy: Data needs and challenges. Vancouver, Canada: University of British Columbia.

Ecosys, (2004). Overview of demand forecasts for the fixed and mo�ile networks and services in Europe. Deliverable 2.

Fildes, R., & Kumar, V. (2002). Telecommunica-tions demand forecasting - a review. International Journal of Forecasting, 18(4), 489-522.

Fisher, J.C., & Pry, R.H. (1971). Simple substitu-tion model of technological change. Technological Forecasting and Social Change, 3(1), 75-88.

Fixler, D., Greenless, J., & Lane, W. (2001). Tele-communications indexes in the U.S. Consumer price index. In Sixth Meeting of the International Working Group on Price Indices. Canberrra, Australia.

Gatignon, H., Eliashberg, J., & Robertson, T.S. (1989). Modeling multinational diffusion pat-terns—an efficient methodology. Marketing Sci-ence, 8(3), 231-247.

Goff, D.H. (2002). An assessment of the broadband media strategies of Western European telecoms. In Proceedings of the Fifth World Media Economics Conference. Turku, Finland.

Griliches, Z. (1961). Hedonic price indexes for automobiles: An econometric analysis of quality change. In The price of the federal government (General Series No. 73) (pp. 137-196). New York: Columbia University Press for NBER.

Gunther, J. (2001). Regulations of telecommu-nications in Europe. Krems, Austria: Danubue University.

755


Haan, M. (2001). The economics of free Internet access. Journal of Institutional and Theoretical Economics, 157(3), 359-379.

Harald, E. (2005). Do hedonic price indexes change history? The case of electrification. Stockholm School of Economics.

Harhoff, D., & Moch, D. (1997). Price indexes for PC database software and the value of code compatibility. Research Policy, 26, 509-520.

Jonker, N. (2001). Constructing quality adjusted price indexes: A comparison of hedonic and dis-crete choice methods. De Nederlandsche Bank, Econometric Research and Special Studies De-partment.

Kalhagen, K.O., & Olsen, B.T. (2002). Provision of broadband services in non-competitive areas in Western Europe countries. In Proc. ISSLS 2002, Seoul, Korea.

Kam, Y.U. (2001). An elementary price index for Internet service providers in Canada: A hedonic study. In Sixth Meeting of International Working Group on Price Indices. Canberrra, Australia.

Kim, C., & Galliers, R.D. (2004). Toward a diffu-sion model for Internet systems. Internet Research Journal, 14(2), 155-166.

Kumar, V., Ganesh, J., & Echambadi, R. (1998). Cross-national diffusion research: What do we know and how certain are we? Journal of Product Innovation Management, 15(3), 255-268.

Kumar, V., & Trichy, K. (2002). Research note multinational diffusion models: An alternative framework. Marketing Science, 21(3), 330.

Lee, H.S., Park, K., & Kim, S.Y. (2003). Estimation of information value on the Internet: Application of hedonic price model. Electronic Commerce Research and Applications, 2, 73-80.

Li, K.C. (1991). Sliced inverse regression for dimen-sion reduction (with Discussion). J. Am. Statist. Assoc., 86, 316-342.

Lu, J. et al. (2003). Technology acceptance for wireless Internet. Internet Research Journal, 13(3), 206-222.

Madden, G., & Coble-Neal, G. (2005). Australian residential telecommunications consumption and substitution patterns. Review of Industrial Orga-nization, 26, 325-347.

Michalakelis, C., Dede, G., Varoutas, D., & Sphi-copoulos, T. (2005). Impact of cross-national diffusion process in telecommunications demand forecasting. In Proceedings of NAEC 2005, Garda, Italy.

Monath, T., Elnegaard, N.K., Cadro, P.H., Kat-sianis, D., & Varoutas, D. (2003). Economics of fixed broadband access network strategies. IEEE Commun. Mag., 41, 132-139.

Montella, M., Mostacci, F., & Zanolini, G. (2001). Consumer price indexes for telecommunication services in Italy: Work in progress. In Sixth Meet-ing of the International Working Group on Price Indices. Canberrra, Australia.

Moreau, A. (1991). A price index for microcompu-ters in France. Document de Travail de la Direction des Statistiques Economiques, 9109.

Naik, P.A., & Tsai, C.L. (2001). Single-index model. Biometrica Trust, 88(3), 821-832.

Nerlove, M. (1995). Hedonic price functions and the measurement of preferences: The case of Swedish wine consumers. Eur. Econ. Rev., 39, 1697-1716.

Nicholson, J.L. (1967). The measurement of quality changes. The Economic Journal, 77, 512-530.

Okamoto, M., & Sato, T. (2001). Comparison of hedonic method and matched models method using scanner data: The case of PCs, TVs, and digital cameras. In Sixth Meeting of International Working Group on Price Indices. Canberra, Australia.

Rai, L.P. (1999). Appropriate models for technol-ogy substitution. Journal of Scientific & Industrial Research, 58(1), 14-18.

Rosen, S. (1974). Hedonic prices and implicit mar-kets: Product differentiation in pure competition. Journal of Political Economy, 92, 34-55.

Silver, M. (2000). Hedonic regressions: An ap-plication to VCRs using scanner data. Omega, 28, 399-408.

756


Simonoff, J.S. (1996). Smoothing methods in sta-tistic. New York: Springer.

Simonoff, J.S., & Tsai, C.L. (1999). Semiparametric and additive model selection using an improved Akaike information criterion. J. Comp. Graph. Statist. 8, 22-40.

Stiroh (2001). The economic impact of information technology. Federal Reserve Bank of New York, Academic Press.

Stone, R. (1954). The measurement of consumer behavior and expenditure in the United Kingdom, 1920-1938. Studies in the National Income and Expenditure of the United Kingdom.

Stone, R. (1956). Quantity and price indexes in national accounts. Organization for European Economic Cooperation.

Takada, H., & Jain, D. (1991). Cross-national analysis of diffusion of consumer durable goods in Pacific Rim countries. Journal of Marketing, 55(2), 48-54.

Taneda, M.A., Takada, J., & Araki, K. (2001). The problem of the fading model selection. IE-ICE Transactions on Communications, E84b(3), 660-666.

Triplett, J.E. (2004). �andbook on hedonic indexes and quality adjustments in price indexes (OECD Science, Technology and Industry Working Papers). OECD Publishing.

Triplett, J.E. (2000). Draft copy handbook on quality adjustment of price indexes for informa-tion and communication technology products. Paris: OECD.

Triplett, J.E., & McDonald, R.J. (1977). Assessing the quality error in output measures: The case of refrigerators. Review of Income and Wealth, 137-176.

Waugh, F.V. (1928). Quality factors influencing vegetable prices. Journal of Farm Economics, 10(2), 185-196.

michalak

Note

Venkatesan, R. & Kumar, V. (2002). A genetic algorithms approach to growth face forecasting of wireless subscribers. International Journal of Forecasting, vol. 18, pp. 625-646

Documents

Chapter XLV Diffusion Forecasting and Price Evolution of ...cgi.di.uoa.gr/~michalak/publications/Bookchapters/Chapter45_cm.pdfa product is offered is a substantially influential factor,