Вы находитесь на странице: 1из 14

This article was downloaded by: [University of Huddersfield]

On: 30 January 2011


Access details: Access Details: [subscription number 773557273]
Publisher Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-
41 Mortimer Street, London W1T 3JH, UK

The Information Society


Publication details, including instructions for authors and subscription information:
http://www.informaworld.com/smpp/title~content=t713669588

A Critical Analysis of Current Indexes for Digital Divide Measurement


Giuseppe Brunoa; Emilio Espositoa; Andrea Genovesea; Kholekile L. Gwebub
a
Department of Business and Management Engineering, University of Naples “Federico II”, Naples,
Italy b Whittemore School of Business and Economics, University of New Hampshire, Durham, New
Hampshire, USA

Online publication date: 06 January 2011

To cite this Article Bruno, Giuseppe , Esposito, Emilio , Genovese, Andrea and Gwebu, Kholekile L.(2011) 'A Critical
Analysis of Current Indexes for Digital Divide Measurement', The Information Society, 27: 1, 16 — 28
To link to this Article: DOI: 10.1080/01972243.2010.534364
URL: http://dx.doi.org/10.1080/01972243.2010.534364

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf

This article may be used for research, teaching and private study purposes. Any substantial or
systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or
distribution in any form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses
should be independently verified with primary sources. The publisher shall not be liable for any loss,
actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly
or indirectly in connection with or arising out of the use of this material.
The Information Society, 27: 16–28, 2011
Copyright 
c Taylor & Francis Group, LLC
ISSN: 0197-2243 print / 1087-6537 online
DOI: 10.1080/01972243.2010.534364

A Critical Analysis of Current Indexes for Digital


Divide Measurement

Giuseppe Bruno, Emilio Esposito, and Andrea Genovese


Department of Business and Management Engineering, University of Naples “Federico II,”
Naples, Italy

Kholekile L. Gwebu
Whittemore School of Business and Economics, University of New Hampshire, Durham,
New Hampshire, USA
Downloaded By: [University of Huddersfield] At: 09:44 30 January 2011

information society (readers interested in the origins of


Several composite indexes grouping several variables into a sin- the term “digital divide” are encouraged to read Gunkel
gle figure have been proposed for measuring the digital divide. In 2003). In this context, different questions have surfaced
this article, the authors analyze shortcomings of extant indexes. about the dynamic evolution of the digital divide and the
Using multivariate analysis tools, they show that in the definition identification of its main determining factors. We focus on
of composite indexes it is possible to individuate a limited set of the emerging need for performance evaluation and bench-
more influencing and significant variables. This finding suggests marking tools to assess the magnitude and the evolution
the possibility of replacing some redundant variables with other of digital divide.
dimensions that include other aspects of the digital divide that are Institutions and organizations such as the International
not considered in extant measures.
Telecommunication Union (ITU) have proposed various
composite indexes that aggregate several indicators into
Keywords composite indexes, digital divide, principal component a single number to capture the complexity of the digital
analysis divide (ITU 2009). The first significant index released by
the ITU was the “Digital Access Index” (DAI), which was
developed to measure the overall ability of individuals in
The rapid development of information and communi- a country to access and use ICT (ITU 2003). At about the
cation technologies (ICTs) and their pervasive role in eco- same time, a network of hundreds of communication lead-
nomic activities can be considered one of the fundamental ers from academia, the media, business, and government
factors fueling growth and transforming the way business circles (Orbicom) developed a conceptual framework for
is conducted in many countries (Mansell and Wehn 1998; digital divide measurement that formed the basis of the
Mansell 1999; Cette et al. 2000; Jorgenson 2001; Rao “Infostate Index” (Orbicom 2003). In 2005 ITU and Or-
2001; Colecchia and Schreyer 2002). However, there is bicom decided to merge the DAI and the Infostate Index to
a striking and alarming disparity in ICTs access and us- create the “ICT Opportunity Index” (ICT-OI) (ITU 2005)
age among economies. The term “digital divide” has been to avoid a duplication of indexes with similar characteris-
coined to describe this phenomenon and to draw atten- tics. In the same year the ITU also proposed the “Digital
tion to the existing gap in ICT access and in the ability Opportunity Index” (DOI), later updated in 2007, to mea-
of individuals and economies to participate in the global sure the potential opportunities of countries to benefit from
access to ICTs (ITU 2007). Although the two indexes (DOI
and ICT-OI) differ in terms of indicators and the method-
Received 15 September 2008; accepted 10 August 2010. ology used, an effort has been made to develop a single
Address correspondence to Andrea Genovese, University of index because statistical analysis has demonstrated that
Naples “Federico II,” Business and Management Engineering De- the two indexes were closely correlated (Abuqayyas and
partment, Piazzale Tecchio 80, 80125 Napoli (NA), Italy. E-mail: Audin 2008). The need for a single index has also been
giuseppe.bruno@unina.it affirmed during various international summits on ICTs
16
DIGITAL DIVIDE INDEXES 17

development (ITU 2003, 2005, 2009). Accordingly, ITU the phenomenon as a whole is the focus of this analysis.
recently released a new index, the “ICT Development In- In particular, this analysis seeks to verify the potential
dex” (IDI), which represents a synthesis of the previous presence of a limited set of more influencing and signif-
ITU indexes (2009). icant indicators. If such indicators are indeed present, it
The development of an assortment of indexes in such could open the possibility of replacing some redundant
a short period suggests that the digital divide index is indicators with other dimensions that allow inclusion of
still a work in progress. Moreover, it highlights the dif- other aspects of the digital divide, not considered in extant
ficulty in synthesizing a composite index for a complex measures.
concept that relies on aggregation of indicators with dif- The remainder of the article is organized as follows.
ferent characteristics. From a methodological viewpoint, The next section provides a brief overview on the defi-
the indexes are calculated by a linear or nonlinear com- nitions proposed for the digital divide. Subsequently, the
bination of the indicators using a set of weights. Even if methodologies used to build the proposed indexes are de-
composite indexes are widely exploited as useful tools in scribed with particular reference to the ICT-OI and the IDI,
policy analysis, caution should be exercised in their design along with the main comments and criticisms in the liter-
and implementation to avoid possible misrepresentation or ature. Thereafter the methodology used for the analysis is
misinterpretation, as highlighted in Organization for Eco- illustrated and the results are presented. Finally, findings
nomic Cooperation and Development (OECD; 2008). For are discussed and possible avenues for future research are
Downloaded By: [University of Huddersfield] At: 09:44 30 January 2011

this reason, a debate has arisen about the reliability of the outlined.
digital divide indexes.
In this article, we discuss digital divide measurement
THE DIGITAL DIVIDE CONCEPT AND
through an assessment of two existing indexes (ICT-OI and
DEFINITIONS
IDI), based on principal component analysis, and high-
light their shortcomings. The analysis focuses on these The digital divide is a complex, dynamic, multifaceted
two indexes because they are likely to be used exten- concept. Consequently, different streams of research of-
sively in years to come by a wide array of stakeholders— ten define it in different ways. For example, in one stream
policymakers, academics, and the public—to measure dig- of research, which focuses on the crucial role of techno-
ital divide. Here are some important points to consider: logical resources, Mehra et al. (2004) simply define digital
r The IDI is the most recent index and represents divide as the gap between individuals who use comput-
ers and the Internet and individuals who do not. Other
the evolution and the synthesis of the previous
authors in this stream suggest that it refers to the distinc-
r indexes suggested by the ITU.
The ICT-OI is the result of the merging of two
tion between the information haves and have-nots (Dewan
and Riggins 2005; Ida and Horiguchi 2008; Bélanger and
indexes, the DAI proposed by the ITU and the
Carter 2009). Along similar lines, Dewan and Riggins
r Infostate developed by the Orbicom.
The ICT-OI is also representative of the DOI
(2005) suggest that the digital divide refers to the division
among persons who have access to digital ICTs and those
since the two indexes are strongly correlated
who do not. Moreover, they consider three different levels
r (Abuqayyas and Audin 2008).
The ICT-OI (employing the geometric average)
of digital divide: individual, organizational, and global.
Campbell (2001) and James (2008) describe the digital
and the IDI (utilizing a weighted sum) are repre-
divide as the division between those who have access to
sentative of two different aggregation methodolo-
ICT and are using it effectively, and those who do not.
gies.
A second stream of research highlights the determi-
This analysis is relevant since an index, whose aim is to nants of the digital divide, emphasizing that both techno-
measure digital divide among different countries, should logical and nontechnological factors are important. Norris
be an operative tool for international institutions and local (2001) believes that the digital divide should be viewed
governments. From this standpoint, a good index should as consisting of three distinct dimensions—global divide,
be both efficient and effective (Jollands et al. 2004). social divide, and democratic divide. The global divide
Efficiency refers to the ability to represent the phe- is the difference in Internet access among industrialized
nomenon without using redundant variables, which can and nonindustrialized nations; the social divide is the gap
confuse policymakers. Effectiveness refers to the ability between those with and those without information access
to interpret the different ways in which the phenomenon within a country; and the democratic divide represents
occurs and thereby provide policymakers with an opera- the gap between individuals who do and individuals who
tive and concrete opportunity to intervene. The trade-off do not utilize digital resources to engage and mobilize
between the need to reduce the number of used variables in the public life. Other authors also highlight that fac-
and the necessity to include variables representative of tors such as the availability of digital resources (relevant
18 G. BRUNO ET AL.

content in diverse languages), human resources (literacy information society, and of participation in institutions
and education), and social resources (institutional struc- governing ICTs and society.”
tures) contribute to the digital divide (Warschauer 2003; The preceding discussion highlights two main points:
Bertot 2003; Cuervo and Menendez 2006). Mossberger r There exists a heated ongoing debate among prac-
et al. (2003) see the digital divide as the disparity in in-
titioners and academic circles about the definition
formation technology due to various factors such as race,
ethnicity, gender, and income. Chinn and Fairlie (2006), r of the digital divide.
The digital divide is an evolving concept, which
through a statistical analysis based on a sample encom-
is shifting from a preoccupation with mere access
passing 161 developed and developing countries, highlight
to technological resources to a multidimensional
the importance of per capita income, of the telecommu-
understanding of inequality (e.g. global, social,
nications infrastructure, and of the quality of regulation
and democratic divides).
in explaining the gap in computer and Internet use. They
also affirm that education has a low impact on Internet ac- The liveliness of the discussion has also influenced the
cess. Using data on 118 countries during the 1997–2001 measurement methodologies, as they reflect the various
period, Guillen and Suarez (2005) argue that the global conceptual frameworks proposed for underlying the digi-
digital divide is the result of the economic, regulatory, and tal divide concept. The body of literature on theoretical and
sociopolitical characteristics of countries and their evolu- empirical studies in this area is significant (e.g., Corrocher
Downloaded By: [University of Huddersfield] At: 09:44 30 January 2011

tion over time. and Ordanini 2002; Sciadas 2002; Cuervo and Menendez
Examining a panel of 40 countries from 1985–2001, 2006; ITU 2005, 2006, 2007). In any case, even if the par-
based on data from three distinct generations of informa- ticipation in the debate on the definition of digital divide
tion technology (IT; mainframe, personal computer, and definition is beyond the scope of this article, the variety of
Internet), Dewan et al. (2005) find that information tech- positions seems to confirm, as noted by Fuchs (2008), that
nology (IT) penetration is positively associated with per its measurement it cannot be limited to economic factors
capita income, years of schooling, and size of the trade but should be more comprehensive in scope.
sector, while it is negatively related to telecommunica-
tion costs and size of urban population. In fact, according
DIGITAL DIVIDE MEASUREMENT
to Van Dijk (2006, 223), the urgent question is: “What
is exactly new about the inequality of access to and use Comparing country performances to identify evolutionary
of Information and Communication Technology as com- trends and establish benchmarks is a common practice in
pared to other scarce material and immaterial resources a wide range of fields (e.g. environment, economics, and
in society?” Finally, Fuchs (2008) criticizes approaches technological development). Such comparisons are often
that attempt to explain the digital divide primarily based performed by introducing composite indexes calculated
on economic factors. Accordingly, he suggests that other through the aggregation of individual indicators, repro-
aspects should be taken into account, such as motivational ducing quantitative or qualitative measures of factors with
access, skills access, usage access, the degree to which the aim of representing the relative position of a country
users benefit from usage, and the degree to which tech- on a conceptual space (OECD 2008). Indicators can be
nologies enable political participation. grouped into categories in accordance with their meaning.
A third stream of research tries to deal with the criti- The aggregation of indicators and/or categories is per-
cisms that have emerged from the second research stream formed according to an underlying mathematical model.
and suggests more comprehensive definitions. In line with Hence composite indexes mainly differ on the basis of
requests for a sophisticated depiction of the phenomenon, the selected indicators and the aggregation methodology.
ITU and Orbicom propose a more comprehensive def- The proposed digital divide indexes are characterized by
inition that sees the digital divide as “the relative dif- different numbers and types of indicators and methodolo-
ference in Infostate among economies” (Sciadas 2005) gies. The choice of the indicators to be considered depends
based on the concept of Infostate, an aggregation of info- on the conceptual framework assumed to describe and to
density (including ICTs infrastructure and skills) and info- represent the process of ICTs penetration and diffusion
use (including ICTs uptake and intensity of use). Fuchs among countries.
(2009, 46) proposes the following comprehensive defini- The first index considered here, the DAI (“Digital Ac-
tion, which strives to address the already-mentioned crit- cess Index”), uses eight indicators, which are grouped into
icisms, stating that the digital divide represents “unequal five categories (infrastructure, affordability, knowledge,
patterns of material access to, usage capabilities of, and quality, and actual usage of ICTs). Indicator values are
benefits from computer-based information and communi- normalized with reference to upper values limits and then
cation technologies that are caused by certain stratification averaged to obtain category scores, and, finally, averaged
processes that produce classes of winner and losers of the to calculate the overall index value. The next index, the
DIGITAL DIVIDE INDEXES 19

TABLE 1
Characteristics of the released indexes for the digital divide measurement

Index Year Institution No. categories No. indicators Methodology

DAI 2003 ITU 5 8 Average of category scores which are averages of the
respective indicators
Infostate 2003 Orbicom 2 21 Geometrical mean of category scores which are geometrical
means of the respective indicators
DOI 2005 ITU 3 11 As for the DAI
ICT-OI 2005 ITU-Orbicom 2 10 As for the Infostate
IDI 2009 ITU 3 11 As for the DAI
Notes. DAI = Digital Access Index; DOI = Digital Opportunity Index; ICT-OI = Information and Communication Technologies–Opportunity
Index; IDI = Information and Communication Technologies Development Index; ITU = International Telecommunication Union.

Orbicom Infostate index, introduces two categories (info- shows the ICT-OI indicators and their groupings into cat-
Downloaded By: [University of Huddersfield] At: 09:44 30 January 2011

density and info-use) with a total of twenty-one indicators. egories and the aggregation schema used to calculate the
Although many indicators are aggregated, others are pre- index.
liminarily combined to form intermediate-level composite As mentioned earlier, the ICT-OI adopts most of the
indicators. After the normalization, the aggregation is per- indicators from the DAI, reducing them from seventeen to
formed through the geometric mean (for a more detailed
explanation, see Sciadas 2005).
The DOI (“digital opportunity index”) includes eleven TABLE 2
indicators grouped into three categories (opportunity, in- Indicators used for the calculation of Information and
frastructure, use). The category scores are calculated as Communication Technologies–Opportunity Index (ICT-OI)
arithmetic averages of the respective indicators, while the
DOI is the arithmetic average of the three category scores. Indicator
The ICT-OI (“ICT opportunity index”) assumes the same Category Subcategory symbol Indicator
framework, main categories (info-density and info-use),
and mathematical model used in the definition of the In- Info-density Networks DN1 Main telephone lines per
fostate. The total number of adopted indicators is reduced 100 inhabitants
to ten, with many overlapping with the ones introduced DN2 Mobile cellular subscribers
for the definition of the DAI. per 100 inhabitants
The last and most recently developed index, the IDI DN3 International internet
(“ICT development index”), considers eleven indicators bandwidth
organized into three categories (access, use, and skills) (kbps/inhabitant)
and the calculation is similar to the one of the DOI. Table 1 Skills DS1 Adult literacy rates (Source:
summarizes the characteristics of the described indexes. UNESCO)
In the following subsections, a more detailed descrip- DS2 Gross enrolment rates
tion of recently proposed indexes (ICT-OI and IDI) is (Source: UNESCO)
provided. Info-use Uptake IU1 Internet users per 100
inhabitants
IU2 Proportion of households
The ICT-OI with a TV
IU3 Computers per 100
The ICT-OI was designed to monitor the global digital inhabitants
divide and to track a country’s progress over time in Intensity II1 Total broadband internet
comparison to other countries of similar income levels. subscribers per 100
Based on the Orbicom Infostate conceptual framework, inhabitants
the ICT-OI includes two categories—info-density (with II2 International outgoing
subcategories networks and skills) and info-use (with sub- telephone traffic
categories intensity and uptake). This combined frame- (minutes) per capita
work brings together a total of ten indicators. Table 2
20 G. BRUNO ET AL.

ten. In this way, the ICT-OI can be calculated for a larger groupings into categories and the weights assigned to each
number of countries (183) compared to its predecessor, indicator and category.
the Infostate Index (139 countries). The IDI for a given country i(IDIi ) is computed as
The ICT-OI for a given country i(ICT-OIi ) is computed follows:
as follows:
 I DIi = 0.4∗ (I CT Access)i + 0.4∗ (I CT U se)i
ICT-OIi = I nf oDensity i ∗ I nf oU sei
+ 0.2∗ (I CT Skills)i
in which:
 in which:
InfoDensityi = Networksi ∗ Skillsi
 (I CT Access)i = 0.2∗ (I CT A1i + I CT A2i
InfoUsei = I ntensityi ∗ Uptakei
 + I CT A3i + I CT A4i + I CT A5i )
Networksi = 3 DN1i ∗ DN2i ∗ DN3
 (I CT U se)i = 0.33∗ (I CT U 1i + I CT U 2i
Intensityi = I I 1i ∗ I I 2i + I CT U 3i )

Skillsi = DS1i ∗ DS2i
Downloaded By: [University of Huddersfield] At: 09:44 30 January 2011

(I CT Skills)i = 0.33∗ (I CT S1i + I CT S2i



Uptakei = 3 I U 1i ∗ I U 2i ∗ I U 3i + I CT S3i )

Each category score is calculated as the arithmetic av-


In practice, each category score is calculated as the erage of the respective indicators. Then the three category
geometric average of the respective indicators. scores are aggregated utilizing a weighted sum method.

The IDI Shortcomings of the Extant Indexes


For several years, the ITU has been publishing two in- Using synthetic composite indexes to represent a complex
dexes to track digital divide (the ICT-OI and DOI) among phenomenon provides, in general, advantages and limi-
countries. In 2007 it started considering the possibility of tations. Saisana and Tarantola (2002) propose a list of
merging the two indexes and creating a single index. The pros and cons summarized in table 4, also cited in OECD
process ended in 2009 with the publication of the IDI. (2008).
The IDI includes three categories (Access, Use, Skills) As previously mentioned, shortcomings of composite
and eleven indicators according to the schema illustrated indexes have been highlighted in the literature. One of the
in table 3. The table also shows the indicators and their most severe critiques is by Van Dijk (2006), who argues

TABLE 3
Indicators used for the calculation of Information and Communication Technologies Development Index (IDI)

Category Category weight Indicator symbol Indicator Indicator weight

ICT access 0.40 ICTA1 Fixed telephone lines per 100 inhabitants 0.20
ICTA2 Mobile cellular telephone subscriptions per 100 inhabitants 0.20
ICTA3 International Internet bandwidth (bit/s) per Internet user 0.20
ICTA4 Proportion of households with a computer 0.20
ICTA5 Proportion of households with Internet access at home 0.20
ICT use 0.40 ICTU1 Internet users per 100 inhabitants 0.33
ICTU2 Fixed broadband Internet subscribers per 100 inhabitants 0.33
ICTU3 Mobile broadband subscribers per 100 inhabitants 0.33
ICT skills 0.20 ICTS1 Adult literacy rate 0.33
ICTS2 Secondary gross enrolment ratio 0.33
ICTS3 Tertiary gross enrolment ratio 0.33
Note. ICT = information and communication techonology
DIGITAL DIVIDE INDEXES 21

TABLE 4 developing countries. It results in missing data, which ul-


Pros and cons in using composite indexes (OECD, 2008) timately has an adverse effect on the rankings produced
by composite indexes. Indexes including fewer indica-
Pro Cons tors will help mitigate such problems. Menou and Taylor
(2006) point out that bias is introduced by data standard-
Can summarize complex, May send misleading policy ization and normalization operations. They also note that
multidimensional messages if poorly national average measures are not very meaningful un-
phenomena with a view to constructed or less they can be disaggregated at a more detailed level.
supporting decision makers. misinterpreted. Barzilai-Nahon (2006) criticizes indexes for only measur-
Are easier to interpret than a May invite simplistic policy ing the digital divide at the national and international level
battery of many separate conclusions. while ignoring community level inequalities.
indicators May be misused, for example, Barzilai-Nahon (2006) and James (2007) point out that
Can assess progress of to support a desired policy, the choice of the aggregation methodology of individual
countries over time. if the construction process indicators is responsible for significant biases. This aspect
Reduce the visible size of a set is not transparent or lacks is crucial in the definition of weights when linear and/or
of indicators without sound statistical or geometric aggregations are performed (Munda and Nardo
dropping the underlying conceptual principles. 2005). In particular, since indicators are often weighted
information base. The selection of indicators
Downloaded By: [University of Huddersfield] At: 09:44 30 January 2011

according to equal values, statistical or empirical evidence


Thus make it possible to and weights could be the explaining that all indicators are “worth” the same should
include more information subject of political dispute. be provided (OECD 2008). Thus, Vehovar et al. (2006)
within the existing size limit. May disguise serious failings suggest the use of statistical tools (such as multivariate
Place issues of country in some dimensions and analysis) to understand the interactions among indicators
performance and progress at increase the difficulty of to avoid data redundancies.
the centre of the policy identifying proper remedial
arena. action, if the construction
Facilitate communication with process is not transparent. METHODOLOGY
general public (i.e., citizens, May lead to inappropriate A composite index is intended to describe how an observed
media) and promote policies if dimensions of phenomenon (e.g., the digital divide) is simultaneously de-
accountability. performance that are termined by several factors (indicators). In the definition of
Help to construct or underpin difficult to measure are such an index, indicators should be selected while consid-
narratives for lay and literate ignored. ering the interrelationships (correlations) among them, to
audiences avoid the presence of some overweighted factors (Saisana
Enable users to compare et al. 2005). This way it could be possible to obtain almost
complex dimensions the same results using a reduced number of indicators that
effectively. better represent the influence on the phenomenon. Also,
the OECD (2008) suggests that minimizing the number
of indicators in an index may be desirable for promoting
that attempts to keep track of and measure the digital di- transparency and parsimony.
vide suffer from a lack of adequate theoretical frameworks. With regard to the most recently released digital di-
Indeed, most of the proposed indexes emphasize income, vide composite indexes, we propose a methodology for
education, age, sex, and ethnicity, while not fully address- selecting the most significant indicators and minimizing
ing the deeper social, cultural, and psychological causes redundancies based on appropriate statistical tools. Let
behind access inequalities. Additionally, there tends to be Idx be an index calculated through the aggregation of a
a lack of conceptual elaboration and definition of the indi- set I of n indicators (Idx = f (I1 , . . . , In )). We calculate
cators used in composite indexes (e.g., computer literacy, and analyze the correlation among each pair of indicators
Internet use). Statements supported by theory and valid on a reliable data set of adequate size. In the presence
operational definitions for empirical research could con- of significant correlations, a principal component analysis
siderably help in building reliable indexes. Also, Fuchs is performed. The steps entailed in the methodology are
(2009) criticizes the choice of the indicators in the current described next.
indexes as they reduce the role of socioeconomic, political,
cultural, and social factors.
The Principal Component Analysis
From a methodological perspective, Braithwaite’s
(2007) critique is that existence and use of too many Within a data set, the principal component analysis method
indicators makes data collection difficult, especially for (Kim and Mueller 1978a, 1978b; Stevens 1986) detects a
22 G. BRUNO ET AL.

set of variables (principal components) able to signifi- TABLE 5A


cantly represent the phenomenon. In practice considering Composition of the Information and Communication
the set I of n indicators and a data set, it is possible to Technologies–Opportunity Index (ICT-OI) sample based
define n principal components Ck (k = 1, . . . , n) as: on the Human Development Index (HDI)

Ck = bk1 I1 + · · · + bkj Ij + · · · + bkn In High Medium Low


Variable HDI HDI HDI Total
for which the generic bkj is a weight associated with the Countries in the data set 61 69 19 149
indicator j of the component k and is “optimally” calcu- Percentage (data set) 40.94 46.31 12.75 100.00
lated through appropriate algorithms. The principal com- Percentage (all countries) 41.42 45.56 13.02 100.00
ponents are then ranked in descending order according to
their significance in the explanation of the phenomenon
(in terms of variance within the data set). Methodologies observations from 149 countries for the ICT-OI indicators
for the selection of principal components are described in and from 154 countries for the IDI indicators.
Joliffe (2002) and OECD (2008). In this context we se- To show the general characteristics of the data set,
lect the first p < n components such that the cumulative tables 5A and 5B indicate the distribution of the avail-
percentage of the total variance is at least equal to a given able countries on the basis of the Human Development In-
Downloaded By: [University of Huddersfield] At: 09:44 30 January 2011

threshold. dex provided by the United Nation Development Program


(UNDP 2003). Following Efron and Tibshirani (1993),
countries are classified as high, medium, and low with
Indicators Selection respect to the Human Development Index. The tables re-
The performed principal components analysis is used to veal that the distribution of the data set is similar to the
reduce the number of indicators. However, principal com- distribution of the total number of countries. Thus, we can
ponents do not have physical meaning, as they represent affirm that the sample is not affected by biases due to a
some sort of artificial variables. For this reason, we cor- nonhomogeneous selection of the countries.
relate each indicator and each of the p selected principal In tables 6A and 6B the correlation matrix for the ICT-
components. Then we individuate the indicators with the OI and IDI is reported. These matrices reveal that there is
highest values of correlation for each principal compo- significant correlation among the indicators. In particular,
nent. This way we select a subset I = ik : k = 1, . . . , p of for the ICT-OI, the indicator “Main telephone lines per
p indicators that are more strongly correlated with each of 100 inhabitants” has a correlation coefficient greater than
the principal components. By doing this, a good approx- 0.70 with five other indicators. Indicators “mobile cellu-
imation of the principal components is obtained through lar subscribers per 100 inhabitants,” “computers per 100
the use of some of the original indicators. inhabitants,” and “total broadband subscribers per 100 in-
habitants” show the same high correlation levels with the
other components. The degree of correlation among the
Verification Process skills indicators (“literacy rate,” “enrollment rate”) and
On the basis of the obtained results, we can define a “re- all the others indicators is quite low (0.11 < rij < 0.29).
duced index” Idxreduced = f (i1 , . . . ,ip ) in which the set I This result reveals that the relationship between educa-
of n original indicators has been replaced by the reduced tion levels and technological opportunities is weak. Fur-
set I of p < n indicators. To verify whether the reduced thermore, the two skills indicators are strongly correlated
index is a good synthesis of the original index we calculate (rij = 0.79) with each other; this could suggest that the
the correlation index between Idxreduced and the original
Idx. In the presence of a significant value of correlation, TABLE 5B
indeed, we can affirm that the reduction of indicators does Composition of the Information and Communication
not produce a significant loss in the representation of the Technologies Development Index (IDI) sample based on
phenomenon. the Human Development Index (HDI)

High Medium Low


EMPIRICAL ANALYSIS Variable HDI HDI HDI Total
The methodology described in the previous section was Countries in the data set 62 71 21 154
employed using 2005 data for the ICT-OI and 2007 data Percentage (data set) 40.26 46.10 13.64 100.00
for the IDI, which are the most recent available data sets Percentage (all countries) 41.42 45.56% 13.02 100.00
for the two indexes (ITU 2009). The data sets consist of
DIGITAL DIVIDE INDEXES 23

TABLE 6A
Correlation matrix for indicators of Information and Communication Technologies–Opportunity Index (ICT-OI)

Indicator DN1 DN2 DN3 IU1 IU2 IU3 II1 II2 DS1 DS2

Main telephone lines per 100 inhabitants DN1 1.00


Mobile cellular subscribers per 100 inhabitants DN2 0.83 1.00
International internet bandwidth (kbps/inhabitant) DN3 0.60 0.53 1.00
Adult literacy rates (Source: UNESCO) IU1 0.89 0.82 0.66 1.00
Gross enrolment rates (Source: UNESCO) IU2 0.72 0.75 0.30 0.65 1.00
Internet users per 100 inhabitants IU3 0.87 0.74 0.73 0.88 0.53 1.00
Proportion of households with a TV II1 0.84 0.70 0.75 0.85 0.46 0.90 1.00
Computers per 100 inhabitants II2 0.54 0.38 0.27 0.44 0.31 0.56 0.48 1.00
Total broadband internet subscribers per 100 inhab. DS1 0.20 0.21 0.12 0.20 0.20 0.17 0.21 0.12 1.00
Intern.outgoing telephone traffic (minutes) per capita DS2 0.25 0.26 0.23 0.28 0.24 0.22 0.29 0.11 0.79 1.00

skills category can be described using only one of the two subscribers”) and some ICT access indicators (“House-
Downloaded By: [University of Huddersfield] At: 09:44 30 January 2011

indicators. hold with a computer,” “household with Internet access”),


With regard to the IDI, the indicator “fixed telephone as can be expected.
lines” correlates significantly (rij ≥ 0.70) with all the indi- The significance level for these calculated correlation
cators except the “international Internet bandwidth”,” the coefficients is shown in tables 7A and 7B. According to
“mobile broadband subscribers,” and the “tertiary gross a t-Student test, p values less than or equal to .005 allow
enrollment ratio.” There is a very high correlation be- us to reject the null hypothesis (that the two variables are
tween “household with a computer” and “household with not correlated). For the ICT-OI, all the correlation coeffi-
Internet access,” which means that users working on a per- cients between technological indicators (networks, uptake,
sonal computer (PC) also use the Internet. In addition, a and intensity) are associated with very low p values. The
high correlation can be observed among some ICT use in- correlation between the two skills indicators is also signif-
dicators (i.e. “Internet users”,” “fixed broadband Internet icant (p value of 5.00E-32). Regarding the IDI, Table 7B

TABLE 6B
Correlation matrix for indicators of Information and Communication Technologies Development Index (IDI)

Indicator ICTA1 ICTA2 ICTA3 ICTA4 ICTA5 ICTU1 ICTU2 ICTU3 ICTS1 ICTS2 ICTS3

Fixed telephone lines ICTA1 1.00


Cellular subscriptions ICTA2 0.72 1.00
Internet bandwidth ICTA3 0.16 0.12 1.00
Households with a ICTA4 0.89 0.72 0.17 1,00
computer
Households with ICTA5 0.87 0.68 0.18 0.97 1.00
Internet at home
Internet users per 100 ICTU1 0.88 0.71 0.18 0.94 0.94 1.00
inhabitants
Fixed broadband ICTU2 0.87 0.64 0.18 0.92 0.94 0.92 1.00
Internet subscribers
Mobile broadband ICTU3 0.64 0.52 0.27 0.71 0.72 0.67 0.68 1.00
subscribers
Adult literacy rate ICTS1 0.74 0.71 0.07 0.71 0.66 0.71 0.62 0.43 1.00
Secondary gross ICTS2 0.72 0.62 −0.06 0.69 0.67 0.71 0.67 0.42 0.77 1.00
enrolment ratio
Tertiary gross ICTS3 0.55 0.53 0.07 0.49 0.43 0.48 0.40 0.26 0.70 0.55 1.00
enrolment ratio
24 G. BRUNO ET AL.

TABLE 7A
Significance levels for correlation coefficients of Information and Communication Technologies–Opportunity Index
(ICT-OI) indicators
DN1 DN2 DN3 IU1 IU2 IU3 II1 II2 DS1 DS2

DN1 0.0000
DN2 0.0000 0.0000
DN3 0.0000 0.0000 0.0000
IU1 0.0000 0.0000 0.0000 0.0000
IU2 0.0000 0.0000 0.0002 0.0000 0.0000
IU3 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
II1 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
II2 0.0000 0.0000 0.0005 0.0000 0.0001 0.0000 0.0000 0.0000
DS1 0.0100 0.0060 0.0820 0.0090 0.0095 0.0239 0.0080 0.0740 0.0000
DS2 0.0010 0.0009 0.0030 0.0004 0.0022 0.0038 0.0003 0.0950 0.0000 0.0000
Downloaded By: [University of Huddersfield] At: 09:44 30 January 2011

confirms that “International Internet bandwidth per Inter- TABLE 8A


net user” indicator is not at all related to all the other ones, Principal component extraction for Information and
as p values are extremely high. Communication Technologies–Opportunity Index
(ICT-OI) indicators
Principal Component Analysis and Components Principal % explained % cumulative explained
Selection component variance variance
The presence of significant correlations among pairs of 1 58.01 58.01
indicators of ICT-OI and IDI necessitates the use of tech- 2 16.29 74.30
niques such as principal component analysis to reduce 3 8.67 82.97
the redundancy. Table 8A shows the extracted principal 4 7.72 90.69
components for the ICT-OI sorted in decreasing order of 5 2.74 93.43
the percentage of explained variance for ICT-OI. The table 6 1.98 95.42
suggests that 58.01 percent of the variance in the ICT-OI is 7 1.81 97.23
explained by component 1, while the first two components 8 1.16 98.38
together explain 74.30 percent of the variance. Similarly, 9 0.87 99.25
table 8B suggests that 66.45 percent of the variance in 10 0.75 100.00
the IDI data set is explained by component 1, while the

TABLE 7B
Significance levels for correlation coefficients of Information and Communication Technologies Development Index
(IDI)
ICTA1 ICTA2 ICTA3 ICTA4 ICTA5 ICTU1 ICTU2 ICTU3 ICTS1 ICTS2 ICTS3

ICTA1
ICTA2 0.0000
ICTA3 0.0261 0.0620
ICTA4 0.0000 0.0000 0.0187
ICTA5 0.0000 0.0000 0.0127 0.0000
ICTU1 0.0000 0.0000 0.0132 0.0000 0.0000
ICTU2 0.0000 0.0000 0.0117 0.0000 0.0000 0.0000
ICTU3 0.0000 0.0000 0.0005 0.0000 0.0000 0.0000 0.0000
ICTS1 0.0000 0.0000 0.2007 0.0000 0.0000 0.0000 0.0000 0.0000
ICTS2 0.0000 0.0000 0.2364 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
ICTS3 0.0000 0.0000 0.1958 0.0000 0.0000 0.0000 0.0000 0.0007 0.0000 0.0000
DIGITAL DIVIDE INDEXES 25

TABLE 8B Indicators Selection


Principal component extraction for Information and
Communication Technologies Development Index Tables 9A and 9B show the correlation matrix between
(IDI) indicators each ICT-OI and IDI indicator and the four selected com-
ponents. We observe that extracted components are often
Principal % explained % cumulative explained highly correlated with indicators belonging to different
component variance variance categories. This result suggests that the grouping into cat-
egories would require further considerations and attention.
1 66.45 66.45 Based on the findings detailed in table 9A, we selected
2 11.20 77.66 ICT-OI indicators with the highest correlation levels to
3 8.02 85.67 each of the selected principal components:
4
5
3.79
3.36
89.47
92.83
r International Internet bandwidth per Internet user
6 2.85 95.68 r (DN3).
7 1.69 97.37 r Proportion of households with a televisions (IU2).
8
9
1.23
0.62
98.60
99.22
r International outgoing telephone traffic per capita
Adult literacy rates (DS1).

10 0.62 99.83 (II2).


Downloaded By: [University of Huddersfield] At: 09:44 30 January 2011

11 0.17 100.00 Similarly, with regard to the IDI, to explain the four
selected principal components, we considered:

first two components together explain 77.66 percent of the


r Fixed broadband Internet subscribers per 100 in-
variance. r habitants (ICTU2).
The number of components that could help explain a
significant portion of the variance in the ICT-OI was se-
r Mobile broadband subscribers per 100 inhabitants
Tertiary gross enrolment ratio (ICTS3).

lected by fixing an optimal threshold of 90 percent for the


cumulative variance to be explained and individuating the
r (ICTU3).
International Internet bandwidth per Internet user
number of components that allows obtaining an amount (ICTA3).
of cumulative variance as close as possible to 90 percent.
Consequently, for either the ICT-OI or for the IDI, four Verification Process
components turn out to be relevant. For the ICT-OI they
explain 90.69 percent of the total variance, while for the For the ICT-OI for each country i we calculated a reduced
IDI they explain 89.47 percent. This suggests the possi- version of the index as the geometric average of the
bility of using a restricted number of indicators to obtain selected indicators.
“similar” results to the one provided by the original ICT- 
OI and IDI. (I CT −OIreduced )i = 4 DN3i ∗ I U 2i ∗ I I 2i ∗ DS1i

TABLE 9A
Correlation analysis among Information and Communication Technologies–Opportunity Index (ICT-OI) indicators
and selected principal components

Category Subcategory Indicator Component 1 Component 2 Component 3 Component 4

Info-density Networks DN1 0.601 0.672 0.094 0.319


DN2 0.480 0.781 0.116 0.119
DN3 0.923 0.099 0.084 −0.010
Skills DS1 0.032 0.093 0.943 0.071
DS2 0.159 0.107 0.930 −0.001
Info-use Uptake IU1 0.696 0.600 0.115 0.199
IU2 0.134 0.931 0.118 0.093
IU3 0.782 0.423 0.069 0.352
Intensity II1 0.833 0.344 0.134 0.27
II2 0.225 0.176 0.048 0.945
26 G. BRUNO ET AL.

TABLE 9B
Correlation analysis among Information and Communication Technologies Development Index (IDI) indicators and
selected principal components

Category Indicator Component 1 Component 2 Component 3 Component 4

Access ICTA1 0.788 0.435 0.259 0.070


ICTA2 0.433 0.606 0.466 0.012
ICTA3 0.084 0.018 0.089 0.987
ICTA4 0.849 0.333 0.335 0.071
ICTA5 0.883 0.254 0.328 0.084
Usage ICTU1 0.863 0.336 0.266 0.086
ICTU2 0.905 0.213 0.255 0.092
ICTU3 0.504 0.091 0.787 0.150
Skills ICTS1 0.496 0.774 0.113 −0.004
ICTS2 0.669 0,575 −0.034 −0.164
ICTS3 0.163 0.901 0.070 0.067
Downloaded By: [University of Huddersfield] At: 09:44 30 January 2011

TABLE 11A
The same process can be applied to IDI, as stated next:
Linear regression analysis among
Information and Communication
(I D Ireduced )i = (I CT U 2i + I CT S3i Technologies–Opportunity Index
+ I CT U 3i + I CT A3i )/4 (ICT-OI) and gross domestic
product per capita scores
Thus, we performed a linear regression between ICT- Variable Factor
OIreduced and ICT-OI and between IDIreduced and IDI,
whose results are highlighted in tables 10A and 10B, re- rICT−OI,GDP 0.942
spectively. The presence of a high and significant corre- 2
RICT−OI,GDP 0.887
lation, in both the cases, between the original versions of p(1-tailed) 0.0000
the indexes and the reduced ones suggests that the current Observations 139

TABLE 10A versions of the indexes harbor redundancies in terms of


Linear regression analysis among number of used indicators.
Information and Communication
Technologies–Opportunity Index
(ICT-OI) and ICT-OIreduced scores FURTHER RESULTS
rICT−OIreduced ,ICT−OI 0.946 The application of the methodology based on correlation
2
RICT−OI 0.896 evaluation and principal component analysis reveals sig-
reduced ,ICT−OI
p (1-tailed) 0.0000 nificant redundancies in the structure of the two digital di-
Observations 149 vide indexes under investigation. This evidence confirms
the shortcomings that can affect digital divide measures,
as already highlighted by Vehovar et al. (2006) and, more
generally, by OECD (2008) best practices for constructing
TABLE 10B synthetic indexes. To further verify the efficiency (ability
Linear regression analysis among to represent the digital divide without using redundant
Information and Communication variables) of ICT-OI and IDI, further analysis was per-
Technologies Development Index formed that evaluated the degree of correlation among
(IDI) and IDIreduced scores the two indexes scores and gross domestic product (GDP)
rIDIreduced ,IDI 0.916
values for each country.
2
RID 0.839
Empirical results show that both the indexes are
Ireduced ,IDI
p(1-tailed) 0.0000
strongly and significantly correlated with the gross do-
Observations 153
mestic product, as reported in tables 11A and 11B. In par-
ticular, ICT-OI and IDI present a correlation coefficient
DIGITAL DIVIDE INDEXES 27

TABLE 11B vestigation that employs principal component analysis to


Linear regression analysis among Information highlight that both indexes present redundant indicators.
and Communication Technologies This means that it is possible to increase their efficiency
Development Index (IDI) and gross domestic by reducing the number of indicators and using the same
product per capita scores technique of aggregation (average in the case of ICT-OI
and geometric mean in the case of IDI). The reduction in
Variable Factor the number of indicators could also enlarge the sample of
countries involved in the analysis.
rIDI,GDP 0.921 Our analysis also showed that both indexes are strongly
2
RICT−OI,GDP 0.845 correlated with gross domestic product. This finding high-
p(1-tailed) 0.0000 lights that although digital divide is a complex concept,
Observations 145 main indexes describe a phenomenon that appears highly
correlated with the difference in income among countries.
Thus, the results of this analysis provide an analytic val-
equal to .946 and .916, respectively; p values are, in both idation of critiques by Van Dijk (2005, 2006) and Fuchs
cases, less than 10−5 . Thus, even if digital divide is a still (2009), who argue that current digital divide research is
evolving and complex concept, the main indexes seem to affected by a “reductionistic” approach to measurement
describe a phenomenon that appears highly and signifi-
Downloaded By: [University of Huddersfield] At: 09:44 30 January 2011

that does not emphasize the role of factors other than


cantly correlated to the difference in income among coun- technological access and use.
tries. Therefore, besides presenting significant redundan- This study could represent a starting point for the defini-
cies in their formulation, both the indexes seem to provide tion of digital divide composite indexes that shun indicator
a “reductionistic” approach to digital divide measurement redundancies and include ecological, socioeconomic, po-
(see for instance Van Dijk 2006; Fuchs 2009) that does not litical, and cultural factors that influence digital divide.
emphasize the role of socioeconomic, political,cultural, Further analysis on the dynamic evolution of the indexes
and social factors. to evaluate their ability to capture differences among coun-
tries across time could be also performed.
CONCLUSIONS AND DIRECTIONS FOR FUTURE
RESEARCH REFERENCES
As noted earlier, digital divide is no longer seen as the Abuqayyas, A., and C. Audin. 2008. Measuring the information society.
simple divide in the access to technological resources, In Proceedings of ITU Regional Workshop on Universal Service
but increasingly it is being perceived as a multidimen- Funding (Damascus, Syria). Geneva: ITU.
sional phenomenon that includes a set of complex divides Barzilai-Nahon, K. 2006. Gaps and bits: Conceptualizing mea-
(global, social, and democratic) caused by a variety of surements for digital divides. The Information Society 22(5):
factors (digital resources, gender, income, etc.). In line 269–78.
with this evolution, the literature has proposed numerous Bélanger, F., and L. Carter. 2009. The impact of the digital divide on
e-government use. Communications of the ACM 52(4): 132–35.
definitions, which range from simply the “gap between in-
Bertot, J. C. 2003. The multiple dimensions of the digital divide: More
dividuals who use computers and Internet and individuals than the technology “haves” and “have nots.” Government Informa-
who do not” (Mehra et al. 2004) to the more compre- tion Quarterly 20(2): 185–91.
hensive definition that defines digital divide as “unequal Braithwaite, S. 2007. Measuring the information society: Proposals for
patterns of material access to, usage capabilities of, and measuring the information society in Guyana and the wider world
benefits from computer-based information and communi- (Technical report). Georgetown, Guyana: University of Guyana.
cation technologies that are caused by certain stratification Campbell, D. 2001. Can the digital divide be contained? International
processes that produce classes of winner and losers of the Labour Review 140(2): 119–41.
information society, and of participation in institutions Cette, G., J. Mairesse, and Y. Kocoglu. 2000. The diffusion of infor-
governing ICTs and society” (Fuchs 2008). mation and communication technologies in France. Measurement
In this context, several composite indexes have been and contribution to economic growth and productivity. Economie et
Statistique 339–40:73–91.
proposed to measure the digital divide among countries.
Chinn, M. D., and R. Fairlie. 2006. The determinants of the global
In this article we focused on two indexes: ICT-OI and digital divide: A cross-country analysis of computer and Internet
IDI. The ICT-OI is the result of the merging of two in- penetration. Oxford Economic Papers 59(1): 16.
dexes, DAI proposed by ITU and Infostate developed by Colecchia, A., and P. Schreyer. 2002. ICT investment and economic
Orbicom; IDI is the most recent index and represents the growth in the 1990s: Is the United States a unique case? A compar-
evolution and the synthesis of the previous indexes re- ative study of nine OECD countries. Review of Economic Dynamics
leased by ITU. In particular, we present an empirical in- 5:408–42.
28 G. BRUNO ET AL.

Corrocher, N., and A. Ordanini. 2002. Measuring the digital divide: A Kim, J. O., and C. W. Mueller. 1978a. Introduction to factor analysis:
framework for the analysis of cross-countries differences. Journal What it is and how to do it. Beverly Hills, CA: Sage.
of Information Technology 17:9–19. ———. 1978b. Factor analysis: Statistical methods and practical is-
Cuervo, M. R. V., and A. J. L. Menendez. 2006. A multivariate frame- sues. Beverly Hills, CA: Sage.
work for the analysis of the digital divide evidence for the European Mansell, R. 1999. Information and communication technologies for
Union—15. Information and Management 43:56−76. development: Assessing the potential and the risks. Telecommunica-
Dewan, S., and F. J. Riggins. 2005. The digital divide: Current and tions Policy 23(1): 35−50.
future research directions. Journal of the Association for Information Mansell, R., and U. When, eds. 1998. Knowledge societies: Infor-
Systems 6(12): 298–337. mation technology for sustainable development. New York: Oxford
Dewan, S., D. Ganley, and K. L. Kraemer. 2005. Across the digital University Press for the United Nations Commission on Science and
divide: A cross-country multi-technology analysis of the determi- Technology for Development.
nants of IT penetration. Journal of the Association for Information Mehra, B., C. Merkel, and A. P. Bishop. 2004. Internet for empower-
Systems 6(12): 409–32. ment of minority and marginalized communities. New Media and
Efron, B., and R. Tibshirani. 1993. An introduction to the bootstrap. Society 6(5): 781–802.
Boca Raton, FL: Chapman and Hall. Menou, M., and R. Taylor. 2006. A grand challenge: Measuring infor-
Fuchs, C. 2008. The implications of new information and communica- mation societies. Information Society 22(5): 261–67.
tion technologies for sustainability. Environment, Development and Mossberger, K., C. J. Tolbert, and M. Stansbury. 2003. Virtual in-
Sustainability 10(3): 291–309. equality: Beyond the digital divide. Washington, DC: Georgetown
———. 2009. The role of income inequality in a multivariate cross- University Press.
Downloaded By: [University of Huddersfield] At: 09:44 30 January 2011

national analysis of the digital divide. Social Science Computer Re- Munda, G., and M. Nardo. 2005. Constructing consistent composite
view 27(1): 41–58. indicators: The issue of weights (EUR 21834 EN). Luxembourg:
Guillen, M. F., and S. L. Suárez. 2005. Explaining the global digital Office for the Official Publications of the European Communities.
divide: Economic, political and sociological drivers of cross-national Norris, P. 2001. Digital divide: Civic engagement, information poverty
Internet use. Social Forces 84(2): 681–708. and the internet worldwide. New York: Cambridge University Press.
Gunkel, D. 2003. Second thoughts: toward a critique of the digital Organization for Economic Cooperation and Development, European
divide. New Media and Society 5(4): 499–522. Commission, Joint Research Centre. 2008. Handbook on construct-
Ida, T., and Y. Horiguchi. 2008. Consumer benefits of public services ing composite indicators: Methodology and user guide (OECD pub-
over FTTH in Japan: Comparative analysis of provincial and urban lication code: 302008251E1). Paris: OECD.
areas by using discrete choice experiment. The Information Society Orbicom. 2003. Monitoring the digital divide and beyond. Ottawa:
24(1): 1–17. National Research Council of Canada.
International Telecommunication Union. 2003. Measuring the in- Rao, P. M. 2001. The ICT revolution, internationalization of
formation society. Annual report of International Telecom- technological activity, and the emerging economies: Implica-
munication Union, Geneva, Switzerland. http://www.itu.int/ITU tions for global marketing. International Business Review 10:
-D/ict/publications/ 571–96.
———. 2005. Measuring the information society. Annual report Saisana, M., and S. Tarantola. 2002. State-of-the-art report on current
of International Telecommunication Union, Geneva, Switzerland. methodologies and practices for composite indicator development
http://www.itu.int/ITU-D/ict/publications/ (EUR Report 20408 EN). Ispra, Italy: European Commission.
———. 2006. Measuring the information society. Annual report Saisana M., S. Tarantola, and A. Saltelli. 2005. Uncertainty and sen-
of International Telecommunication Union, Geneva, Switzerland. sitivity techniques as tools for the analysis and validation of com-
http://www.itu.int/ITU-D/ict/publications/ posite indicators. Journal of the Royal Statistical Society 168(2):
———. 2007. Measuring the information society. Annual report 1–17.
of International Telecommunication Union, Geneva, Switzerland. Sciadas, G. 2002. Monitoring the digital divide. Montreal: Orbicom.
http://www.itu.int/ITU-D/ict/publications/ Sciadas, G. 2005. From the digital divide to digital opportunities:
———. 2009. Measuring the information society. Annual report Measuring infostates for development. Geneva: ITU.
of International Telecommunication Union, Geneva, Switzerland. Stevens, J. 1986. Applied multivariate statistics for the social sciences.
http://www.itu.int/ITU-D/ict/publications/ Hillsdale, NJ: Erlbaum.
James, J. 2007. Cumulative bias in the new Digital Opportunity Index: United Nations Development Programme. 2003. Millennium develop-
Sources and consequences. Current Science 92(1): 46–50. ment goals: A compact among nations to end human poverty. New
———. 2008. Digital preparedness versus the digital divide: A confu- York: Author.
sion of means and ends. Journal of the American Society for Infor- Van Dijk, J. A. G. M. 2005. Deepening digital divide: Inequality in the
mation Science and Technology 59(5): 785–91. information society. Thousand Oaks, CA: Sage.
Joliffe, I. T. 2002. Principal component analysis. Berlin: Springer- ————. 2006. Digital divide research, achievements and shortcom-
Verlag. ings. Poetics 34:221–35.
Jollands, N., J. Lermit, and M. Patterson. 2004. Aggregate eco- Vehovar, V., P. Sicherl, T. Hüsing, and V. Dolnicar. 2006. Methodolog-
efficiency indices for New Zealand: A principal components analy- ical challenges of digital divide measurements. Information Society
sis. Journal of Environmental Management 73: 293–305. 22(5):279–90.
Jorgenson, D. W. (2001). Information technology and the U.S. econ- Warschauer, M. 2003. Dissecting the digital divide: A case study in
omy. American Economic Review 91:1–31. Egypt. Information Society 19(4):297–304.

Вам также может понравиться