Analyzing digital divide within and between member and candidate countries o f European Union

Abstract

The main purpose of this study is to analyze whether a digital divide exists among European Union (EU) members, new members, and candidate countries. Beyond this, the second goal is to find out if a digital divide has a significant association with the process of becoming an EU member. First, member, new member, and candidate countries are classified into three groups, and MANOVA (Multivariate Analysis of Variance) is applied to determine differences among these groups in terms of Information Society levels. Then, Information Society variables are analyzed using Discriminant Analysis. According to the results of the research, there is a significant level of digital divide in the EU and a certain information society level currently is not associated with EU membership. The EU must address the digital divide among member countries if they are to become a close-knit community, and to continue to be one of the most competitive economic powers in the world. Finaly, the EU might consider using information society level as an objective criteria along with other objective and subjective criteria currently being used as EU membership criteria.

Full text

Analyzing digital divide within and between member and candidate
countries o f European Union
Çiğdem Arıcıgil Çilan
a
, Bilge Acar Bolat
a
, Erman Coşkun
b,c,
⁎
a
Faculty of Business Administration, Istanbul University, Avcilar-Istanbul, Turkey
b
Sakarya University Business Department, Sakarya Turkey
c
LeMoyne College Business Department and Information Systems Program Syracuse, NY 13214, USA
abstractarticle info
Available online 7 September 2008
Keywords:
Digital divide
Information society
MANOVA
Discriminant analysis
European Union
The main purpose of this study is to analyze whether a digital divide exists among European Union (EU)
members, new members, and candidate countries. Beyond this, the second goal is to find out if a digital
divide has a significant association with the process of becoming an EU member. First, member, new member,
and candidate countries are classified into three groups, and MANOVA (Multivariate Analysis of Variance) is
applied to determine differences among these groups in terms of Information Society levels. Then,
Information Society variables are analyzed using Discriminant Analysis. According to the results of the
research, there is a significant level of digital divide in the EU and a certain information society level currently
is not associated with EU membership. The EU must address the digital divide among member countries if
they are to become a close-knit community, and to continue to be one of the most competitive economic
powers in the world. Finaly, the EU might consider using information society level as an objective criteria
along with other objective and subjective criteria currently being used as EU membership criteria.
© 2008 Elsevier Inc. All rights reserved.
1. Introduction
Information and communication technology (ICT) utilization,
especially in developed and developing countries, has been prolifer-
ating greatly in the last decade. In recent years, new developments in
hardware, software, and computing fields have resulted in affordable
technologies and new application areas for ICT, along with the
widespread availability of internet access. These developments have
played a role in every aspect of society, including business transactions
and communications, daily routines and lifestyles, politics, and the
economy. e-government, e-health, e-democracy, e-learning, e-com-
merce, e-business, e-municipality, e-banking, e-finance, and e-tax are
part of daily life in many countries at many different levels, while e-
mail, web-surfing, blogs, Youtube, internet news, online shopping,
online services and online dictionaries such as Wikipedia, are a major
component of life for many citizens.
As a result, in recent years, governments of developed countries
and developing countries have emphasized the need for advances in
information and communications technologies (ICT) when they form
their policies. Many countries have launched their national e-strategy,
E-Government, e-trade, and e-health implementations. The World
Bank, by taking into account the declaration prepared by the United
Nations in 2000, has determined many specific targets related to ICT in
the Millennium Development Targets (World Bank, 2006). Further-
more, the World Bank has presented ICT as a way of the accelerating
economic development and reducing the poverty.
In the 1960's and 70's, telecommunications began to play a signi-
ficant role in production, public service, and in management. In the
1980's, information became an accepted production factor, together
with labor and capital. In the 1990's, the effect of globalization, the
increasing significance of information in production processes, rapid
changes in technology, and increases in demand have proved the
importance of ICT for competition and economic growth. In recent
years, it has been accepted that ICT are significant inputs to eco-
nomic growth. Moreover, the efficiency of ICT in the development
of international competitive power, health and education, and the
power of ICT in creating new job possibilities have been assumed as a
significant component determining the socio-economical structure of
countries and as a way of decreasing poverty in developing and
underdeveloped countries (World Bank, 2006).
Today, both developed and developing countries seek a society in
which all citizens can reach and share information. However, this is
not currently the case, and there are significant differences between
individuals, groups, regions, and countries in terms of reaching and
sharing the information. Many countries are trying to form suppor-
tive policies in order to achieve this goal. Forming these policies
successfully is only possible by determining the differences in the use
of ICT between individuals, regions, or countries. At that point, the
concept of a digital divide becomes important.
Government Information Quarterly 26 (2009) 98–105
⁎Corresponding author. LeMoyne College Business Department and Information
Systems Program Syracuse, NY 13214, USA.
E-mail addresses: ccilan@istanbul.edu.tr (Ç.A. Çilan), bacar@istanbul.edu.tr
(B.A. Bolat), coskune@lemoyne.edu (E. Coşkun).
0740-624X/$ –see front matter © 2008 Elsevier Inc. All rights reserved.
doi:10.1016/j.giq.2007.11.002
Contents lists available at ScienceDirect
Government Information Quarterly
journal homepage: www.elsevier.com/locate/govinf

As defined by the OECD (the Organization for Economic Coopera-
tion and Development), the term digital divide refers to “the gap
between individuals, households, businesses and geographic areas at
different socio-economical levels with regard both to their opportu-
nities to access information and communication technologies and to
their use of Internet for a wide variety of activities”(OECD, 2001).
Our study has two goals. The first is to analyze whether a digital
divide exists between EU members, new members, and candidates,
and if so, to determine what is the level of the digital divide. The
second is to find out if a digital divide has a significant association with
becoming an EU member. In order to achieve these goals, member and
candidate countries are classified into three groups. The three groups
are (1) the original EU15 countries, (2) the countries which became
members in January 2004, and (3) the candidate countries in 2004,
respectively (Table 1).
Section two summarizes definitions of a digital divide, and the
main reasons explaining a digital divide, as expressedin the literature.
Section three looks at different metrics from previous studies that
have been used to measure digital divide. Section four explains our
methodology and techniques. Section five explains our findings and
results. Section six concludes the paper by analyzing results and
developing suggestions.
2. Reasons for digital divide
The term ‘digital divide’was first used (in the mid 90s) by former
Assistant Secretary for Communications and Information of the U.S.
Department of Commerce, and director of the National Telecommu-
nications and Information Administration (NTIA), Larry Irving, Jr. His
purpose was to focus public attention on the existing gap in access to
information services between those who can afford to purchase the
computer hardware and software necessary to participate in the
global information network, and low-income families and commu-
nities that cannot (Dragulanescu, 2002). Formerly, the term digital
divide had been perceived as the differences for having or not having,
using or not using, and knowing or not knowing how touse computers
and the internet (Tapscott, 1998). Recently, the words ‘computers and
the internet’have been replaced with ‘new forms of information
technology,’(Van Dijk, 2006). The digital divide has also been
generally defined as the socio-economical difference in the use of
ICT (Vehovar et al., 2006). In this paper, we use the OECD's definition
‘the gap between individuals, households, businesses, and geographic
areas at different socio-economical levels with regard both to their
opportunities to access information and communication technologies
and to their use of internet for wide variety of activities’(OECD, 2001).
There are two primary dimensions of digital divide: domestic and
international. Domestic digital divide refers to a digital divide in a
certain country or region, while the international digital divide refers
to agap between regions, countries, or continents. Although indicators
used for determining international and domestic digital divide may
vary, many common indicators are also used. Other than regions or
geographic locations, a digital divide can also occur between genders,
ages, education groups, income groups, racial, and ethnic groups (Ono
& Zavodny, 2006).
Studies indicate that an international digital divide originates from
the difference in social development and economical growth in
developing and developed countries and regions, in the wide use of
English (since 3/4 of websites in the world are in English and 1/2 of the
Internet Users' mother tongue is English), and in differences between
the demographic quality of the citizens (gender, race, lifestyle, family
structure, and size of family) (Chen & Wellman, 2004; Cuervo &
Menendez, 2006; Cullen, 2001; Ono & Zavodny, 2006; Primo et al.,
2000). In many studies, it is seen that the digital divide negatively
affects women, old people, people with low education and with low
income, large size families, people living in rural areas, low-skilled
persons, and minorities (Chinn & Fairlie, 2004) at the domestic level,
and poor or low-income countries at the international level.
The international digital divide is very significant between under-
developed, emerging, and developed countries. According to Fuchs and
Horak (2008), while Africa has 14.1% of the world population (57
countries), it has only 2.3% of internet users (while the world average is
15.7% ). Cuervo and Menendez (2006) studied15EUcountriesandfound
that a digital divide occurs even between developed countries. They
found that multiple dimensions of digital divide can be summarized into
two factors: ICT infrastructure and use, along with the costs and
availability of online public services. Their study grouped 15 Eurupean
Union countries into four main categories by cluster analysis.
3. Measuring the Digital Divide
As a first step, it is necessary to determine the appropriate
variables to measure the digital divide. At this stage, the lack of a
theoretical framework describing the information society and digital
divide is problematic. Initial studies on the digital divide have been
made in the USA. The United States Department of Commerce
conducted one of the first studies in order to determine differences
in the usage of digital technologies in business and in public
administration (US Department of Commerce, 1999–2000). One of
the pioneer methodological studies on measuring the digital divide
was made by Ricci, who developed an “adoption scale”for digital
technology by collecting elementary indicators (Ricci, 2000).
Over time, many national and international institutions showed
interest in this area and have based their studies on different variables
in order to measure the digital divide. In particular, the study variables
have been taken from international institutions such as EUROSTAT, the
OECD, the World Bank, the UNDP (United Nations Development
Program), the IDC (International Data Corporation), and the ITU
(International Telecommunication Union). Questionnaires have been
often used in data gathering, and the multi-dimensional aspects of the
information society have led to the development of various index
measures to compare levels of the information society, such as the
information society index (IDC, 2007), digital access index (ITU, 2003),
and the technology achievement index (UNDP, 2001).
Table 1
EU membership status for countries
Countries EU Membership Status Category
Belgium Old member 1
Czech Rep. 2004 member 2
Denmark Old member 1
Germany Old member 1
Estonia 2004 member 2
Greece Old member 1
Spain Old member 1
France Old member 1
Ireland Old member 1
Italy Old member 1
Cyprus 2004 member 2
Latvia 2004 member 2
Lithuania 20 04 member 2
Luxembourg Old member 1
Hungary 2004 member 2
Netherlands Old member 1
Austria Old member 1
Poland 2004 member 2
Portugal Old member 1
Slovenia 2004 member 2
Slovakia 2004 member 2
Finland Old member 1
Sweden Old member 1
United K Old member 1
Bulgaria⁎⁎⁎ 2007 Member 3
Romania⁎⁎⁎ 2007 Member 3
Turkey Candidate 3
⁎⁎⁎Bulgaria and Romania became EU members as of January 2007. Since this study uses
data belong to 2004, these two countries considered as candidates.
99Ç.A. Çilan et al. / Government Information Quarterly 26 (2009) 98–105

Various components of the information society have been
measured with these indices. For instance, Corrochner and Ordaninin
(2002) identified 6 components of a ‘synthetic index of digitalization’:
markets, diffusion, infrastructure, Human Resources, competitiveness,
and competition. Other components determining international digital
divide identified in other studies include ICT sector; ICT market and
external trade; ICT penetration; ICT usage in households; ICT usage in
enterprises; ICT education, training, and skills; and ICT in government
and health. In addition, there are many variables measuring each
component. ICT penetration; ICT usage in households; and ICT
education, training, and skills components have gained importance
in measuring the digital divide between individuals. At the interna-
tional level, the most basic and most important indicators are “the
number of access lines per 100 habitants,”along with “Internet hosts
per 1000 inhabitants,”which includes “fixed plus telecommunication
access paths per 100 inhabitants”(OECD, 2001).
Despite the numerous variables just discussed, additional and more
descriptive variables are needed. For instance, in the report prepared by
UNESCO (2003) under the heading of “what might be useful to collect in
the future;”indicators related to ICT, ICT-education, and ICT-culture were
identified. In that report, currently used indicators include the number of
telephone lines, the number of cellular subscribers, the number of
personal computers, and the number of internet users, all expressed
either per 100 or 1000 inhabitants. The same report also lists the following
as variables which must be included in the near future: indicators
showing how frequently ICT is used; education in relation to e-learning
and distant education (to what extent radio, television, and the internet
are used for this purpose); indicators related to internet connectivity in
schools; available speed, bandwidth, systems, and hardware; ICT
curriculum, teachers, training, learning, and outcomes; and ICT-culture
relations showing number of online newspapers, online radio, and
television stations, access statistics, internet presence of libraries and
archives, artists' personal websites, and virtual galleries (UNESCO, 2003).
Cuervo and Menendez (2006) used several variables to determine
digital divide between 15 European countries. These variables included:
the number of computers per 100 inhabitants, the number of main
telephone lines per 100 inhabitants, the number of broadband connec-
tions per 100 inhabitants, the number of secure servers per million
inhabitants, the percentage of businesses with a website, the percent-
age of businesses buying online, internet dial-up access costs for a
residential users (off peak USD PPP), the percentage of households
connected to the internet, the percentage of public services online, and
the percentage of the active population using a computer for profes-
sional purposes.
4. Previous and current digital divide and information society
level assessment studies for European countries
Several international organizations are conducting yearly studies to
determine different aspects related to the digital divide. One of them is
the Economist Intelligence Unit (EIU). This organization has calculated
and published annual e-readiness scores for the world's largest
economies since 2000 using a model developed together with the
IBM Institute for Business Value (EIU, 2005). They currently calculate
and publish scores for countries throughout the world. The rankings,
according to this model, are a measure of a counry's e-business
environment, which is measured by a number of factors that indicate
how amenable a market is to internet-based opportunities. For the e-
readiness index scores, different indicators are used with different
weights. They are Connectivity and Technology Infrastructure (25%),
Business Environment (20%), Consumer and Business Adoption (20%),
Legal and Policy Environment (15%), Social and Cultural Environment
(15%), and Supporting E-services (5%). Table 2 uses data from the EIU
study to rank EU member and candidate nations.
This table shows that 14 of our first group (first 15 EU member
countries) are leading the EU in terms of e-readiness scores. Greece,
which is one of the 15 members, has slightly lower e-readiness score
than two new member countries (Estonia and Slovenia). Bulgaria and
Romania (which became members in January 2007) have the lowest
Table 2
Economist intelligence unit e-readiness scores and rankings for EU member and
candidate countries
Country 2006 e-rank
(of 68)
2005 rank 2006 e-readiness score
(out of 10)
2005 score
Denmark 1 1 9.00 8.74
Sweden 4 3 8.74 8.64
UK 5 5 8.64 8.54
Netherlands 6 8 8.60 8.28
Finland 7 6 8.55 8.32
Germany 12 12 8.34 8.03
Austria 14 14 8.19 8.01
Ireland 16 15 8.09 7.98
Belgium 17 17 7.99 7.71
France 19 19 7.86 7.61
Spain 24 23 7.34 7.08
Italy 25 24 7.14 6.95
Portugal 26 25 7.07 6.90
Estonia 27 26 6.71 6.32
Slovenia 28 27 6.43 6.22
Greece 29 28 6.42 6.19
Czech Rep. 32 29 6.14 6.09
Hungary 32 30 6.14 6.07
Poland 34 32 5.76 5.53
Slovakia 36 34 5.65 5.51
Lithuania 38 40 5.45 5.04
Latvia 39 37 5.30 5.11
Bulgaria 44 42 4.86 4.68
Turkey 45 43 4.77 4.58
Romania 49 47 4.44 4.19
The full table with 68 countries can be found at http://graphics.eiu.com/files/ad_pdfs/
2005Ereadiness_Ranking_WP.pdf (Economic Intelligence Unit (2005 and 2006).
Table 3
2006 UN ICT diffusion rankings table for EU and Other European Countries
Rank Country Access index Connectivity index Diffusion index
1 Luxembourg 0.928 0.703 0.815
4 Sweden 0.836 0.7 0.768
5 Denmark 0.828 0.667 0.748
6 Netherlands 0.803 0.642 0.723
7 Switzerland 0.764 0.645 0.705
10 United Kingdom 0.804 0.557 0.68
11 Finland 0.799 0.546 0.672
14 Norway 0.758 0.558 0.658
18 Germany 0.753 0.538 0.646
20 Estonia 0.704 0.567 0.635
21 Austria 0.76 0.51 0.635
23 Ireland 0.727 0.496 0.611
24 Italy 0.753 0.452 0.602
25 France 0.73 0.464 0.597
27 Malta 0.764 0.394 0.579
28 Belgium 0.735 0.421 0.578
29 Slovenia 0.719 0.406 0.562
30 Czech R 0.712 0.397 0.555
31 Spain 0.697 0.402 0.549
32 Cyprus 0.685 0.407 0.546
34 Portugal 0.659 0.393 0.526
37 Slovak 0.678 0.321 0.499
39 Hungary 0.64 0.349 0.494
40 Greece 0.607 0.38 0.493
43 Lithuania 0.63 0.329 0.479
44 Latvia 0.649 0.289 0.469
45 Croatia 0.627 0.299 0.463
49 Poland 0.616 0.272 0.444
52 Bulgaria 0.607 0.248 0.428
60 Serbia 0.699 0.182 0.403
66 Romania 0.582 0.184 0.383
73 Turkey 0.535 0.193 0.364
Source: http://209.85.165.104/search?q=cache:BACvgeTx0h0J:www.unctad.org/en/docs/
iteipc20065_en.pdf+ict+diffusion+rankings&hl=en&ct=clnk&cd=1&gl=us&client=firefox-a.
100 Ç.A. Çilan et al. / Government Information Quarterly 26 (2009) 98–105

e-readiness scores in the EU. A long-time candidate, Turkey, has a
better e-readiness score than Romania.
When we look e-readioness scores, we see that there are significant
differences among EU countries, as well as between the original EU15
and new members, or EU15 and candidate countries. The highest e-
readiness score is more than twice the lowest score.
The United Nations also publishes reports related to the digital
divide and the information society levels. The UN publishes ICT
Diffusion rankings for 180 countries throughout the world. Table 3
includes only EU member and candidate country scores and it is
created from the UN data for the 180 countries.
In Table 3,Connectivity is defined as “the physical infrastructure
available to a country, as distinct from broader factors determining
access (e.g. literacy, cost).”Connectivity represents the basic “limiting
factor”regarding access to and use of ICTs —for without the essential
physical hardware, ICT use is not possible. Internet hosts per capita,
PCs per capita, telephone mainlines per capita, mobile subscribers per
capita are all metrics to determine connectivity.
The Access score in Table 3 is defined as the “availability and
affordability of access equipment …and pervasiveness of telematics (a
mix of hard/software with human/organizational skills and knowl-
edge transfer).”This introduces a broader definition of access and the
factors determining use of ICTs, that is, factors beyond narrowly-
defined connectivity. The number of internet users, literacy, costs of a
local call, and GDP per capita, often determines access.
In Table 3,theDiffusion index is an average of access and connect-
ivity. Table 4 shows diffusion index score rankings for EU countries, new
members, and candidates, for the time period between 1997 and 2004.
Table 4 shows that most of the EU15 countries have high rankings
and can be called advance information societies. New members and
candidates form a second group and there are some differences bet-
ween these two groups.
5. Research
Previous studies in this area focused primarily on measuring the
level of Information Technology / Information Systems (IT/IS)
utilization in different countries, and very few focused on the digital
divide itself. Furthermore, none of the previous studies looked at the
correlation between EU membership and information society levels.
Thus, this study contributes in two areas. First, it analyzes existence of
digital divide in the EU. Second, this is one of the first studies to
analyze existence of a relationship between EU membership and a
country's information society level. The study uses MANOVA to
determine existence of a digital divide, and Discriminant Analysis to
determine which factors are significant in creating different informa-
tion society levels and a digital divide among the original EU15, new
members, and candidate countries. Based on our knowledge, this is
the first study to include new members and candidate countries in
digital divide analysis.
5.1. The purpose and scope of the research
Our main research question is whether there is a significant
digital divide between EU members, new members, and candidate
countries. The second research question is whether information
society level would be an indicator for becoming an EU member. In
this study, the indicators for member and candidate countries of the
EU have been analyzed by Multivariate Analysis of Variance
(MANOVA) in terms of information society indicators and the
question, “which information society indicators within the coverage
of the study have affected the membership”has been evaluated by
Discriminant Analysis. Croatia and Malta are not included in this
analysis because there is a significant amount of missing data re-
garding information society variables for them in the EUROSTAT
database (Eurostat, 2003).
Since the candidate and member countries of the EU will be
compared in terms of the digital divide concept, the indicators from
EUROSTAT regarding information society have been taken into
consideration. EUROSTAT has 6 main groups of variables concerning
information society: (1) Policy indicators, (2) Structural indicators, (3)
Telecommunication services, (4) Computers and the internet in
households and enterprises, (5) E-commerce, and (6) Individuals e-
skills. In each main group there are variables representing this
component. The variables in EUROSTAT, particularly the variables
related to the candidate countries, are missing for the years 2005 and
2006, thus the year of 2004 has been taken as base year since the
variables of the candidate countries are available for this year. The
analysis is limited to 10 variables because of incomplete data in some
categories (Table 5).
Table 4
ICT diffusion rankings 1997–2004
Country 1997 1998 1999 2000 2001 2002 2003 20 04
Luxembourg 8 7 8 5 3 2 2 1
Sweden 5 5 5 4 4 4 4 4
Denmark 7 6 7 8 6 5 5 5
Netherlands 17 11 9 7 7 8 9 6
Switzerland 14 14 13 9 5 6 6 7
UK 19 19 20 20 15 14 12 10
Finland 3 4 6 10 8 7 10 11
Germany 21 21 21 18 20 18 17 18
Estonia 35 32 33 31 33 33 28 20
Austria 20 18 17 17 14 17 18 21
Italy 26 24 25 24 26 23 24 24
France 22 23 24 26 25 24 23 25
Malta 38 37 38 35 29 29 30 27
Belgium 25 26 26 25 23 25 25 28
Slovenia 28 30 29 27 27 27 27 29
Czech 41 39 41 38 32 31 31 30
Spain 31 31 30 29 28 30 32 31
Cyprus 23 25 27 28 30 28 29 32
Portugal 29 28 28 30 31 32 33 34
Slovak 47 47 47 47 45 45 45 37
Hungary 45 43 44 42 39 39 37 39
Greece 34 34 32 32 34 35 39 40
Lithuania 49 48 53 55 54 52 49 43
Latvia 54 49 55 56 51 51 41 44
Poland 57 53 51 52 53 48 48 49
Bulgaria 55 59 60 60 58 57 56 52
Serbia 56 58 54 50 57 58 61 60
Romania 77 84 87 89 77 70 68 66
Turkey 78 74 71 70 68 71 75 73
Table 5
Selected Information Society Indicators
Label Variable
IS1 Percentage of individuals who haveused a mouse to launch programs such as an
internet browser or word processor
IS2 Percentage of individuals regularly using internet
IS3 Percentage of households having access to internet
IS4 Percentage of individuals used the internet in relation to training and
educational purposes
IS5 Percentage of individuals having ordered /bought goods or services for private
use over the internet in the last three months
IS6 Percentage of households with broadband access
IS7 Percentage of individuals using internet for interacting with public authorities
broken down by purpose : obtaining information, obtaining forms, returning
filled forms
IS8 Percentage of individuals having taken ICT security precaution within the last
three months.
IS9 Access to networks (per 100 inhabitants)
IS10 Percent of individuals who have used the internet in the last 3 months for
interaction with public authorities
Source: http://epp.eurostat.ec.europa.eu/portal/page?_pageid=1996,45323734&dad=portal&
schema=PORTAL&screen=welcomeref&open=/science/isoc&language=en&product=
EUMAIN_TREE&root=EU_MAIN_TREE&scrollto=217.
101Ç.A. Çilan et al. / Government Information Quarterly 26 (2009) 98–10 5

5.2. Statistical methods
5.2.1. MANOVA and discriminant analysis
The purpose of MANOVA (Multivariate Analysis of Variance) is to
test whether there is difference in mean vectors of k groups. We have
selected for this study One-Way MANOVA, which is used for cases in
which more than one dependent variable affects one independent
variable. Our independent variable is qualitative (shows which group
a country belongs to: 1 for EU15, 2 for new members, and 3 for
candidate countries) and the dependent variables are quantitative
(Table 5).
Discriminant Analysis is the appropriate statistical technique when
the dependent variable is qualitative and the independent variables are
metric. The single dependent qualitative variable in Discriminant Analysis
turns into an independent variable in MANOVA (Hair et al., 1998).
The study also uses Multiple Group Discriminant Analysis. There
are three groups: the original EU15, the countries which becomes
member of the EU in 2004, and the candidate countries of the EU. The
objective of the analysis is to determine the variables which
significantly define the groups and to describe the functions of these
discriminating variables. In the analysis, the ability of the variables to
discriminate between the groups has been determined through the
tests and by considering the question of “whether the predicted
discriminant function classify properly the cases to their own groups”
(Tacq, 1999). Function calculations of Discriminant Analysis have been
based on Fisher approach.
In order to apply the analysis, the common assumptions (homo-
geneity of covariance matrices and multivariate normality) of the two
analyses should be realized (Klecka, 1980). MANOVA can construct a
linear relationship only between the dependent variables. On the
other hand, the problem of multi-collinearity between independent
variables should not exist in Discriminant Analysis. This study follows
the practice and uses the method of “Stepwise Discriminant Analysis”
and the results are adopted accordingly.
5.3. Research Results
5.3.1. Achieving the assumptions
5.3.1.1. Normality. There are many methods to assess the multiva-
riate normality assumption. One of the frequently used methods is
constructing a chi-square plot (Johnson & Wichern,1998). The plot can
be graphed with the Mahalinobis distances (d
i
2
) and chi-square values.
The plot should resemble a straight line through the origin. In other
words, the Pearson correlation coefficient of chi-square values and
Mahalinobis distances should be near one. The plot drawn for the
examined data group is presented in Fig. 1.
The value of the Pearson correlation coefficient is 0.9769. This
value is very close to one. Furthermore, through this graph, it can be
accepted that the relation is linear. The data set which has been
examined shows multivariate normal distribution.
5.3.1.2. Homogeneity of covariance matrices. The tests developed to
investigate the equality of the variance–covariance matrix are very
sensitive to multivariate normality (Johnson & Wichern, 1998). In
practice, the frequently used test is Box-M. This test is also very
sensitive to deviations from normalcy.
In this study, we chose to apply MANOVA to determine whether
there is a significant difference between candidate and member
countries of the EU in terms of information society levels and to
analyze whether the membership process of candidate countries
has been affected by the information society level. For this purpose,
the countries have been classified into three groups: the EU15, the
countries which became members of the EU in 2004, and the EU
candidate countries. With Discriminant analysis, IT indicators
properly discriminating these three groups are determined. Accord-
ing to this, the significance level of Box-M test statistics obtained
for the three groups is (0.055) and the assumption has been
proved.
5.3.2. Results of MANOVA
In order to examine the MANOVA test results, four statistics are
considered and the scores for this study are presented in Table 6.
Wilks' Lambda is the one most frequently used tests and it is used for
this study. Significance level of Wilks’lambda is 0.023. In other words,
for the EU15, the countries which became members in 2004, and the
EU candidates, the information society indicators are significantly
different and a significant digital divide exists.
Fig. 1. Chi-Square Plot.
Table 6
Results of Multivariate Tests
Multivariate tests Sig
Pillai's trace 0.016
Wilks' lambda 0.023
Hotelling's trace 0.035
Roy's largest root 0.027
Table 7
Mean Values for Each Indicator
Label Countries Mean Values
IS1 EU15 57.60
New Member 44.26
Candidate 16.00
IS2 EU15 45.98
New Member 29.67
Candidate 11.33
IS3 EU15 48.40
New Member 26.67
Candidate 7.67
IS4 EU15 10.63
New Member 10.15
Candidate 3.10
IS5 EU15 16.42
New Member 3.11
Candidate 0.33
IS6 EU15 20.73
New Member 7.00
Candidate 3.37
IS7 EU15 12.66
New Member 6.92
Candidate 2.26
IS8 EU15 15.59
New Member 10.39
Candidate 5.00
IS9 EU15 49.47
New Member 35.11
Candidate 28.33
IS10 EU15 25.68
New Member 14.22
Candidate 7.92
102 Ç.A. Çilan et al. / Government Information Quarterly 26 (2009) 98–10 5

When arithmetic means of these three groups are examined for
each variable in Table 7, it can be seen that the EU15 has the maximum
mean values for all variables, the countries which became members in
2004 follow EU15 in terms of mean values, and lastly the candidate
countries have the minimum mean values. Taking into consideration
that the EU15 countries are socio-economically well-developed, and
the IT indicators are closely related to socio-economical development,
it can be inferred that the EU candidacy status mirrors or echoes
information society levels.
In order to determine difference between the groups for each
variable, the Tukey HSD procedure is applied and the results are
presented in Table 8. When, Table 8 is analyzed, it is seen that the
states which became member countries in 2004 and the candidate
countries of the same period are not significantly differentiated in
terms of the variables except IS1 (percentage of individuals who
have used a mouse to launch programs such as an internet
browser or word processor). However, the EU15 and the candidate
states significantly differentiate in terms of all variables except IS4
(percentage of individuals who used the internet in relation to
training and educational purposes) and IS6 (percentage of house-
holds with broadband access). In addition, EU15 and the countries
which became members in 2004 significantly differentiate in terms
of IS3 (percentage of households having access to internet), IS5
(percentage of individuals having ordered /bought goods or
services for private use over the internet in the last 3 months),
IS6 (percentage of households with broadband access), and IS9
(access to networks (per 100 inhabitants).
5.3.3. Results of multiple group discriminant analysis
The dependent variable in Discriminant Analysis is formed by
three groups (the first group is EU15, the second group is the states
which became members in 2004, and the third group is the
candidate countries). For analysis, Stepwise Discriminant analysis is
applied by taking the multi-collinearity problem into consideration.
As a result of the analysis, only IS3 (Percentage of households
having access to internet), which discriminates the groups sig-
nificantly (0.000), takes place in the Discriminant function as a
discriminating variable. The Eigenvalue of the discriminant function
shows to what extent the function discriminates between the
groups. It can be claimed that if the Eigenvalue of the function is
higher than the rate of 0.40, this function has the significant power
of discrimination. However, the rate of 0.40 is not certain (Albayrak
et al., 2005). Thus, the function has significant discrimination power
(eigenvalue =0.895). When the Eigenvalue of the discriminant
function is tested in terms of significance, it is seen that it is
significant at the 0.000 level. When correlation coefficients between
the discriminating function and the variables of the Structural
Matrix are analyzed, it can be seen that IS3 (percentage of
households having access to internet) is the variable which has
the maximum (0.98) ability of discriminating.
As a result of Stepwise Discriminant Analysis, IS3 (percentage of
households having access to internet) is discriminating the members
of EU (EU15, the countries which became members in 2004) and the
candidate countries. This variable is classified as a citizen indicator
within Information Society indicators of EUROSTAT.
The correct classification rate of Discriminant Analysis is 74.1%
(Table 9). Maximum chance criterion has been calculated in order to
determine whether the proper classification rate is valid or not. The
maximum chance criterion is calculated by means of the group which
Table 8
Multiple comparisons of groups (Tukey HSD)
Label Multiple Comparisions Sig.
IS1 EU 15–New Member 0.147
EU15–Candidate 0.001⁎
New Member–Candidate 0.039⁎
IS2 EU 15–New Member 0.055
EU15–Candidate 0.005⁎
New Member–Candidate 0.211
IS3 EU 15–New Member 0.010⁎
EU15–Candidate 0.001⁎
New member–Candidate 0.197
IS4 EU 15–New Member 0.975
EU15–Candidate 0.082
New member–Candidate 0.133
IS5 EU15–New Member 0.006⁎
EU 15–Candidate 0.027⁎
New member–Candidate 0.892
IS6 EU 15–New Member 0.032⁎
EU15–Candidate 0.078
New member–Candidate 0.894
IS7 EU 15–New Member 0.065
EU 15–Candidate 0.022⁎
New member–Candidate 0.455
IS8 EU 15–New Member 0.073
EU 15–Candidate 0.012⁎
New Member–Candidate 0.300
IS9 EU 15–New Member 0.011⁎
EU 15–Candidate 0.013⁎
New member–Candidate 0.616
IS10 EU 15–New Member 0.069
EU 15–Candidate 0.059
Newmember–Candidate 0.698
EU 15–New Member 0.006⁎
⁎The mean difference is significant at the .05 level.
Table 9
Summary Table of Classification
Original Predicted Group Membership Total
EU15 New Member Candidate
Count EU15 13 2 0 15
New Member 2 7 0 9
Candidate 0 3 0 3
Percent (%) EU15 86.67 13.33 0 100
New Member 22.22 77.78 0 100
Candidate 0 100 0 100
74.1% of original grouped cases correctly classified.
Table 10
Actual and Predicted Groups
Countries Actual Group Predicted Group
Belgium 0 0
Czech Rep. 1 1
Denmark 0 0
Germany 0 0
Estonia 1 1
Greece 0 1
Spain 0 0
France 0 0
Ireland 0 0
Italy 0 0
Cyprus 1 0
Latvia 1 1
Lithuania 1 1
Luxembourg 0 0
Hungary 1 1
Netherlands 0 0
Austria 0 0
Poland 1 1
Portugal 0 1
Slovenia 1 0
Slovakia 1 1
Finland 0 0
Sweden 0 0
United K 0 0
Bulgaria 2 1
Romania 2 1
Turkey 2 1
103Ç.A. Çilan et al. / Government Information Quarterly 26 (2009) 98–105

has maximum number of cases. In the study, the numberof cases is 15
for the first group (EU15), 9 for the second group (the countrieswhich
became members in 2004) and 3 for the third group (candidate
countries). As mentioned at the beginning of the study, Croatia
(candidate country) and Malta (which became a member in 2004) are
not included in this study because of missing data. The group which
has maximum number of cases is the first group. Thus, maximum
change criteria is 15/(15+9 + 3) =0.55. In order to have valid and
effective proper classification rate, the correct classification rate
should be higher according to maximum change criteria (Hair et al.,
1998). In Table 9, correct classification rate is 74.1% and this rate is
more than 55%; thus it can be said that the classification is successful.
As known, Bulgaria and Romania have been member countries since 1
January 2007. However, since the analysis is realized through the data
from 2004, Romania and Bulgaria are included as candidate countries
of EU. In this analysis, EU15 countries are coded as 0, the countries
which became member countries of EU in 2004 are coded as 1, and the
candidate countries of EU are coded as 2. When Table 10, including
casewise statistics are examined, it is seen that Portugal and Greece,
which are EU15 countries, have the same level with the countries
which became member countries of EU in 2004 in terms of IT
indicators. Cyprus and Slovenia, the countries which became member
countries of EU in 2004, have the similar IT level with EU15 countries.
Romania, Bulgaria, and Turkey have similar IT level with the countries
which became member countries of EU in 2004.
6. Conclusions
As a result of MANOVA Analysis, it has been seen that there has
been a digital divide between the EU15 countries and the countries
which are candidate of EU in 2004 (Romania, Bulgaria, and Turkey).
However, there is no digital divide between the countries which
became member countries in 2004 and the countries which were
candidates in 2004. On the other hand, Romania and Bulgaria, which
are candidate states in 2004, became member countries in January
2007. Furthermore, as a result of MANOVA Analysis, it has been seen
that the countries which have maximum Information Society level are
EU15 countries.
It is known that EU membership is not only related to socio-
economical development, but also to politics. In the Copenhagen
Summit (21–22 June 1993) the criteria to be a member of EU has been
determined for the candidate countries of EU (Copenhagen Criteria).
According to the EU, Romania and Bulgaria have met the criteria and
they became members of the EU in January 2007. On the other hand,
Turkey is still trying to fulfill the criteria of membership. By taking into
consideration that the membership is also related to socio-economical
structure, it has been seen in discriminant function that Romania,
Bulgaria, and Turkey have similar characteristics in terms of economic
and social context and, according to 2004 data, Bulgaria and Romania
occur in the Lower-Middle Income Group. Turkey is in the Upper–level
income group (World Bank, Information and Communication for
Development 2006 report, 167–280). Turkey has similar character-
istics with Bulgaria and Romania in terms of socio-economical
structure. Furthermore, all three countries are at the similar level in
terms of Information Society.
The estimates made as a result of discriminant function show that
Portugal and Greece, which are in the EU15, have the same Information
Society level with the countries which become member state of EU in
2004. Cyprus and Slovenia, which became member state in 2004, have
similar Information Society levels with the EU15. Romania, Bulgaria,
and Turkey (candidates in 2004) have similar Information Society
structure with the states which become members in 2004.
As a result of Discriminant Analysis, it has been determined that
IS3 variable (Percentage of households having access to internet),
which discriminates the groups significantly, takes place in the
discriminant function as a discriminating variable for the EU15
countries, those which became member countries in 2004, and the
countries which are candidate countries in 2004. The significant
difference has been determined between the countries in terms of this
variable. As a result we can say that socio-economical development is
closely related with Information Society indicators and we propose
that these indicators should take place in basic socio-economical
indicators in the short-run.
Finally, we can conclude that EU and member countries must take
action to develop and implement policies reducing the current digital
divide among member states. This will help to achieve the mission
and goals of the European Union.
7. Limitations of the study
As with other research, this study has some limitations. The most
important limitation is the availability of data. This research
conducted with the data from EUROSTAT. However, EUROSTAT data
is missing for some new member countries. Secondly, the study
conducted with 2004 data because later data was not available.
Another related limitation might be using more metrics to
measure digital divide. As mentioned, the literature review suggests
many metrics to measure existence of digital divide. However this
kind of data is unavailable for many countries.
Nevertheless, this study uses a unique approach and unique set of
metrics to determine the existence of a digital divide in the EU. We
strongly believe that our findings might be helpful for policy makers
and decision makers of the EU and other countries.
References
Albayrak, A. S., Eroglu, A., Kalaycı,Ş., Kayış, A., Öztürk, E., Küçüksille, E., et al. (2005).
SPSS UygulamalıÇok Değişkenli İstatistik Teknikleri. Ankara, Asil Yayın.
Chen, W., & Wellman, B. (2004, spring/summer). The global digital divide –Within and
between countries. IT&Society,1(7), 39−45 (Spring/Summer).
Chinn, M. D., & Fairlie, R. W. (2004). The determinants of the global digital divide: A
cross country analysis of computer and internet penetration. Yale University
Economic Growth Center Discussion Paper No:881.
Corrochner, N., & Ordaninin, A. (2002). Measuring the digital divide : A framework for
the analysis of cross-country differences. Journal of Information Technology,17,9−19 .
Cuervo, M. R. V., & Menendez, A. J. L. (2006). A multivariate framework for the analysis
of the digital divide evidence for the European Union -15. Information and
Management,43,56−766.
Cullen, R. (2001). Adressing the Digital Divide. Online Information Review,5,311−320.
Eurostat (2003). Information society statistics pocketbook. Brussels: Eurostat.
Dragulanescu, N. G. (2002). Social Impact ofthe “Digital Divide”in a Central-Eastern
European Country”.International Information & Library Review,34,139−151.
Economic Intelligence Unit (2005 and 2006). The 2005 e-readiness rankings.A white
paper from Economic Intelligence Unit. Available online http://graphics.eiu.com/
files/ad_pdfs/2005Ereadiness_Ranking_WP.pdf
Fuchs, C., & Horak, E. (2008). Africa and digital; divide. Telematics and Informatics,25,
99−116 .
Hair, J., Anderson, R., Tatham, R., & Black, W. (1998). Multivariate data analysis. (5th ed.).
New Jersey: Prentice Hall Inc.
International Data Corporation (IDC). (2007, January). Information Society Index. http://
www.idc.com/getdoc.jsp?containerId=204122
International Telecommunication Union. (2003). World Telecommunication Development
Report, Geneva.
Johnson, R., & Wichern, D. (1998). Applied multivariate statistical analysis. (4 edition).
U.S.A.: Prentice Hall Inc.
Klecka, W. R. (1980). Discriminant analysis. California: Sage Publications.
Ono, H., & Zavodny, M. (2006). Digital inequality: A five country comoparison using
Microdata. Scoaila Sceince Research, doi:10.1016/j.ssresearch.2006.09.001.
Organization for Economic Cooperation and Development (OECD). (2001). Under-
standing The Digital Divide.
Primo, Braga, C. A., Kenny, C., Qiang, C., Crisafulli, D., Di Martino, D., et al. (2000). The
networking revolution opportunities and challenges for developing. The World
Bank Working Paper.
Ricci, A. (2000). Measuring information society dynamics of european data on useage of
information and communication technologies in Europe since 1995. Telematics and
Informatics,17,141−167.
Tacq, J. (1999). Multivarate technique in social sciences. Great Britain: Sage Publications.
Tapscott, D. (1998). Growing up Digital: The rise of the net generation. New York: Mc-G raw-Hill.
United Nations Development Program. (2001). The Technology Achievement Index.
http://hdr.undp.org/reports/global/2001/en/pdf/techindex.pdf
UNESCO. (2003). Measuring and Monitoring The Information And Knowledge Societies: A
Statistical Challenge.
104 Ç.A. Çilan et al. / Government Information Quarterly 26 (2009) 98–10 5

US Department of Commerce. (1999–2000). Falling through the net: toward digital
inclusion. Washington D.C.: US Census Bureau.
Van Dijk, J. A. G. M. (2006). Digital divide research, achievements and shortcomings.
Poetics,34,21−235.
Vehovar, V., Sicherl, P., Hüsing, T., & Dolnicar, V. (2006). Methodological challenges of
digital divide measurements. The Information Society,22,279−290.
World Bank. (2006). Information and communications for development global trends and
policies. Washington, D.C.: The World Bank.
Çiğdem Aricigil Çilan is an assistant professor of Statistics in the Quantitative Methods
Department of Istanbul University Business School. She holds a Ph.D. degree from
Istanbul University, a Masters from Marmara University, and Bachelor’s degree from
Yildiz Technical University.
Bilge Acar Bolat is a Ph.D. student and research assistant in the Quantitative Methods
Department of Istanbul University Business School. Her research area is business
statistics.
Erman Coşkun is an associate professor of Quantitative Methods and Information
Systems at Business and Administrative Sciences Faculty of Sakarya University in
Turkey and Associate Professor of Information Systems at LeMoyne College,
Syracuse NY. He also teaches at Business School of Istanbul University. He holds a
Ph.D. and a Master of Engineering degree from Rensselaer Polytechnic Institute, an
MBA from Pace University, and BBA from Istanbul University. His research interests
are information systems, evaluation of IS, IS education, impacts and complexity of
IS, and IS applications in different industries, intelligent systems, risk management,
safety-critical large scale systems, and human-computer interaction. He published
several articles in the Journal of Systems and Software, Journal of Information and
Software Technology, and Journal of Homeland Security and Emergency Manage-
ment and conference papers in International Conference on Information Systems
(ICIS), Americas Conference on Information Systems (AMCIS), European Conference
on Information Systems (ECIS), Information Systems for Crisis Response and
Emergency Management. Dr. Coşkun conducts several research projects in the US
and in Turkey.
105Ç.A. Çilan et al. / Government Information Quarterly 26 (2009) 98–105