Unveiling The Digital Divide
Catalogue no. 56F0004MIE — No. 7
ISSN: 1492-7918
ISBN: 0-662-32226-6
R e s e a r c h P a p e r
Unveiling the Digital Divide
by G. Sciadas
Science, Innovation and Electronic Information Division
7-L, R.H. Coats Building, Ottawa, K1A 0T6
Telephone: 1 613 951-2581 Fax: 1 613 951-9920
This paper represents the views of the author and does not necessarily reflect the opinions of Statistics Canada.
Connectedness Series
The Series publishes analytical studies and research reports in the broad area of
Connectedness. This includes cross-economy activities, such as the penetration and
use of the Internet and electronic commerce, as well as industries in the Information
and Communications Technologies sector, such as telecommunications, broadcasting
and Internet services. It offers a statistical perspective in these emerging phenomena
that are transforming the economic and societal landscape.
All papers are subject to peer and institutional review, as well as review by subject
matter experts, as necessary. They are distributed to Statistics Canada Regional
Offices, Provincial and Territorial Focal Points and are also available on the Internet.
Papers in the Connectedness Series express the views of the authors and do not
necessarily reflect the opinions of Statistics Canada.
The Series is produced by:
Science, Innovation and Electronic Information Division
Director: Fred Gault
For further information:
Science, Innovation and Electronic Information Division
R.H. Coats Building, 7-L
Ottawa, Ontario, K1A OT6
Telephone: (613) 951-2581
Facsimile: (613) 951-9920
Aussi disponible en français
Note of appreciation
Canada owes the success of its statistical system to a long-standing partnership between Statistics Canada,
the citizens of Canada, its businesses, governments and other institutions. Accurate and timely statistical
information could not be produced without their continued cooperation and goodwill.
Unveiling the Digital Divide
by G. Sciadas
Editor: George Sciadas
Assistant Editor: Heidi Ertl
Publishing: Lucienne Sabourin
Technical Support: Robert Trudeau
Review Committee: Ray Ryan, Mike Sheridan, Philip Smith, Tim Davis, Fred Gault,
Maryanne Webber, Louis-Marc Ducharme, Paul Johanis, Philip Cross.
October 2002
Catalogue No. 56F0004MIE, No. 7
ISBN: 0-662-32226-6
ISSN: 1492-7918
Frequency: Irregular
Published by the authority of the Minister responsible for Statistics Canada
© Minister of Industry, 2002
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in
any form or by any means, electronic, mechanical, photocopying, recording or otherwise without prior written
permission from Licence Services, Marketing Division, Statistics Canada, Ottawa, Ontario, Canada K1A 0T6.
Abstract
The digital divide, commonly understood as the gap between information and
communications technology (ICT) ‘haves’ and ‘have-nots’, has emerged as an
important issue of our times largely due to the uneven diffusion of the Internet.
Many variables, including income, education, age and geographical location, exert
significant influences on household penetration of both ICT and non-ICT commodities.
Thus, divides can be defined for any permutation of the above. In the case of ICTs,
divides depend on the specific technology, its timing of introduction, as well as the
variable of interest. This study shows that the digital divide is sizeable; ICT
penetration rates grow with income. Generally, the effect of income is larger on
newer ICTs (Internet, computers, cell phones) than older and established ones
(television, telephone). Then, using the Internet penetration of households by
detailed income level, it finds that in an overall sense the Internet divide is slowly
closing. This, however, is the result of the accelerated adoption of the Internet by
middle-income households – particularly upper middle. The Internet divide is
widening when the lowest income deciles are compared with the highest income
decile.
At the same time, the rates of growth of Internet adoption among lower-income
households exceed those of higher-income households. This is typical of penetration
patterns of ICT and non-ICT commodities. Rates of growth are initially very high
among high-income groups, but at later stages it is the penetration of lower-income
groups that grows faster.
Science, Innovation and Electronic Information Division
CONNECTEDNESS SERIES
Unveiling the Digital Divide
By G. Sciadas
George Sciadas is with the Science,
Innovation and Electronic Information
Division.
1. A DIVIDE THAT UNITES?
The commercial arrival of the
private sector is active through the
Internet, in conjunction with the
Global Digital Divide Initiative of the
convergence of information and
World Economic Forum (2002). The
communications technologies (ICTs),
issue found its highest political
has generated a creative turmoil in all
manifestation through the G-8 whose
walks of life, including a research
Digital Opportunities Task Force (DOT
thirst among the business, policy and
Force) referred explicitly to ‘digital
academic communities. Prominent
opportunities’ and ‘digital dividends’
among the plethora of issues that
and is following through with an
have emerged as worthy of
action plan (2000, 2001a, 2001b).
understanding is the Digital Divide.
With all these efforts underway the
The issue has implications for
issue of the Digital Divide offers many
important public and private sector
research links.
initiatives, including government
1 -- ‘Digital’ is a misnomer.
online and e-commerce.
The term itself1 contains notions of
desired universality and is
Early work examined the gaps
reminiscent of the century-old
between ICT ‘haves’ and ‘have-nots’
policies for telephones. But why?
for a variety of socio-economic
Our world abounds with divides of
groups. It was accompanied by
all kinds. At a time when many
research on aspects of access, use
North American driveways resemble
and impediments, which brought to
small dealerships, many have not
the fore the key role of skills. While
yet driven, and at a time when
such studies initially focused on
many households are already in
internal country divides, the issue
their multiple computer upgrade and
inevitably expanded soon to cross-
increasingly turning into
country divides. Comparisons of
sophisticated local area networks, a
connectivity among developed
good part of the population has yet
countries (e.g. The Conference Board
to touch a computer. While our
of Canada 2000, 2001, 2002) are now
societies still struggle to rid
done in parallel with investigations
themselves of left-over tolerance for
involving developing countries, as the
homelessness, we are capable of
link was made between ICTs and
feeling great empathy for people
development. Today, even casual
without Internet connections - a few
observations make it abundantly clear
short years after its birth. What is
that the issue has made it to the
it, then, about the Internet and
forefront of many an agenda.
other ICTs that touches sensitive
Countless conferences, symposia and
chords, precipitates such reactions
workshops occupy individuals from
and arouses social consciences?
every field imaginable. National and
This study places the Digital Divide
regional governments, international
in perspective (Section 2),
bodies, businesses and non-
quantifies how big it is (Section 3)
governmental organizations are in the
and examines how it is evolving
midst of numerous initiatives. To
(Section 4). Some related matters
name a few, the UN has established
are taken up in Section 5.
the ICT task force (2001), the World
Throughout the study, concepts are
Bank the InfoDev program (2002),
defined and measurement methods
the OECD has published indicators
are suggested. They are then
(2001a, 2001b, 2002), and the
applied to arrive at conclusions.
Statistics Canada - Catalogue No. 56F0004MIE
3
CONNECTEDNESS SERIES
Science, Innovation and Electronic Information Division
2. THE DIGITAL DIVIDE IN
ICTs in one group and expect
PERSPECTIVE
similar patterns in their
penetration either across groups or
The Digital Divide serves as an
over time. The diffusion pattern of
umbrella term for several distinct
each ICT depends on its particular
domains of investigation. Generally,
characteristics, which impact on its
the approaches come in two
actual and perceived uses, as well
variants. One is ICT-centric and
as the relative ease of such use.
focuses mainly on actual
Diffusion patterns are also
connectivity - digital or otherwise. It
influenced by the complex and
reflects the quest to quantify and
shifting relationships among ICTs,
understand the factors that separate
such as the extent to which they
the ‘haves’ and the ‘have-nots’.
are complementary in use or
While findings from this approach
alternatives for the same use. One
point to potential corrective actions,
influential force in this context is
they come with no pretenses to
the process of convergence. Still
study outcomes and economic or
far from complete, its evolution
societal impacts. Examples of this
will continue to determine
approach are the Falling through the
outcomes whose exact nature is
Net series (US 1995, 1998, 1999,
unknown.
2000) and A Nation Online (US
2002), Dickinson and Sciadas (1996,
The television started as a passive
1997, 1999), Sciadas (2000)2 .
receiver of signals – its use was
2 -- This is also discussed in Noll et al.
Another approach ventures beyond
straightforward and unique. Years
(2000). Many other studies worry
about specific issues, such as the race
connectivity and encompasses
later it found an additional role
or the educational dimension of the
divide, and point to research agendas
aspects such as ICT literacy and
with the arrival of the VCR, which
(i.e.Hoffman and Novak 1999a,
skills (e.g. Castells 2001, ETS 2002,
allowed individual users to exercise
1999b, National Science Foundation
2001).
Sciadas 2002), with linkages to
options with regard to both the
knowledge and even social cohesion.
timing and the content of
While the latter approach is broader
watching. With the arrival of
in scope, its effectiveness can be
camcorders, video games and,
enhanced if nested with the former
lately, the digitization of signals
one, that is, if based on rigorous
that makes possible the offering of
analysis of quantitative information.
interactive services, the uses of
This is where the present study is
‘the box’ are many. However, its
situated.
fast and massive penetration in
our lives took place when it was
still a one-use technology – it
needed no more.
2.1 Which Divide?
In reality, many divides exist3 . A
Computers, on the other hand,
3 -- Here we analyze digital divides
proper appreciation of the issues
were initially meant for computing.
associated with people. However,
divides are also found in the business
involved requires an understanding of
Naturally, then, their diffusion was
sector where gaps exist by industry,
firm size and the like.
the role of at least two important
concentrated among those with
dimensions: i) individual ICTs, and; ii)
such a need or desire. With the
variable of interest. There are many
arrival of networks – the Internet,
ICTs and variables, and divides can be
in the case of individuals - the
identified for any permutation of
computer found a new use and its
these. (A key third dimension will be
functionality was radically
added shortly).
transformed. Even within the
Internet era, there are substantial
i) There are old and new ICTs, digital
differences between the early
and analogue. Each is distinct in
years and the recent years – much
its attributes, functionality and
has changed and much is expected
numerous other characteristics,
to change as broadband takes
including pricing. All these matter.
4 -- The computer is still, by and large,
hold. Thus, the diffusion pattern of
the predominant means of accessing
A television, for instance, offers
personal computers among people
the Internet. Should alternative
access become more attractive (e.g.
different services than a cell
has been greatly influenced by
though wireless technologies or Web
phone, which in turn is different
new developments. Everything
TV), diffusion could again be affected.
The same development could be
from coaxial cable. There is no
that affects functionality influences
positive for the diffusion of the
reason, a priori, to lump different
Internet and negative for that of
diffusion4 .
computers.
4
Statistics Canada - Catalogue No. 56F0004MIE
Science, Innovation and Electronic Information Division
CONNECTEDNESS SERIES
Similar arguments can be made
While there is overlap among such
for the cell phone, the Internet on
groupings, as the same individual
cable and other technologies.
or household can be present in
Under the banner of convergence,
many, it is nonetheless important
the reach of their boundaries is
to carry in mind the specific group
continuously redefined. Most
examined and the reasons for
powerful influences are
such examination. There are
technological and emanate beyond
serious policy and business
those related to the more standard
reasons why such itemization of
economic issues of affordability, or
groups may matter - and the two
even social norms. Therefore, in
may well differ. Connecting rural
studying the digital divide we must
areas at ‘reasonable’ cost is non-
be cognizant at all times of the
trivial – especially when
peculiarities of specific ICTs at
broadband is concerned; the use
some point in time.
of the Internet by individuals of a
certain age is not insignificant in
ii) Then, there are many variables of
the deployment of specific services
interest: income, education,
and the associated business
gender, age, geographical location,
investments.
such as metropolitan or rural
areas, and many more. Each of
Clearly, even on the basis of these
these results in the delineation of
two dimensions alone, analyses of
different groupings of people, with
digital divides can be complicated.
different size and other
Specificity is a virtue, with regard to
characteristics.
both ICT and variable used.
NOTE TO READERS
The data utilized in this study originate in several sources at Statistics Canada. A major database contains information
on households, compiled over many years and through different survey instruments going as far back as 1953. Until
recently, the information was collected by the Household Facilities and Equipment survey (HIFE - Statistics Canada
1996). As of 1997, its content has been embodied in the annual Survey of Household Spending (Statistics Canada
2001a). The timing of the surveys differs over the years; while the HIFE was carried out in the spring with
measurements referring to that time, the new SHS is carried out early in the year for the previous reference year.
Other data for household Internet use, including all data that refer to use from any location, come from the Household
Internet Use Survey (HIUS), which has been conducted annually since 1997. In the first three years, the survey was
conducted in the Fall (October/November), but for reference year 2000 it was conducted in January 2001. For details
see Statistics Canada (2001b). Data regarding the use of the Internet by individuals come from the 2000 General
Social Survey (GSS), which was dedicated to the use and impacts of technologies. They refer to individuals 15 years
of age and over (Statistics Canada 2001c, Dryburgh 2001).
The detailed data on the number of households by commodity and by income, both at the level of income deciles and
at $5,000 increments, are available only from 1982 onwards. The reader should be aware that the composition of
households in any income group changes over time. This is particularly true when two decades are involved, as the
household composition of any given income group in the last year of data (2000) would bear little resemblance to its
composition in the first year of data (1982). In the case of income deciles, the number of households in each decile
increases over time as population increases; in the caser of nominal income increments, the number of households
decreases over time for low-income groups and increases for high-income groups.
Data of such long time-series and with this level of detail are subject to various other caveats too. The estimation
methodology of the annual surveys is based on the population counts of the most recent census at the time. When new
census information becomes available, breaks occur in the series which are adjusted through periodic re-weighting. In
the dataset used here, the 1997 data from the first year of the SHS are based on the 1996 census counts, whereas the
data for previous years are based on the 1991 census counts. Therefore, the annual rates of growth between these two
years are not meaningful. However, the effect on the computed penetration rates is minimal, as both the numerators
and the denominators are affected more or less the same. As well, the overall effects on rates of growth over long
periods are also minimal.
Moreover, in earlier periods, data are not available for each year. Annual growth rates for intervening years were
calculated as average annual compounded rates over the period. The same is true for computers, where
measurements were initially made only for 1986, 1988 and 1990. Data for vehicles from 1997 onward refer to
households with at least one vehicle, either owned or leased, and are not directly comparable with data for earlier
years.
Statistics Canada - Catalogue No. 56F0004MIE
5
CONNECTEDNESS SERIES
Science, Innovation and Electronic Information Division
2.2 Divides among households
Table 1 summarizes some of these
Income is always a key variable for
divides for the Internet. Clearly,
analyses of divides. Chart 1 shows the
penetration increases across incomes,
penetration of several ICTs by detailed
but it also increases substantially by
income levels. Clearly, household
the level of education, the presence of
penetration increases with income5.
5 -- Chart 1 reveals a J-curve effect at the
children and urban areas within each
very low-income end for newer
level of income. This is true whether
The effect of income is more
technologies. This has been observed
many times, whether households or
home-use or use from any location is
pronounced on new technologies
individuals are concerned, and to
concerned. The latter is considerably
rather than older and established
some extent it reflects the student
population and those among the self-
higher, indicative of the importance of
ones. However, the income divide is
employed for whom a year is not a
alternative access points (work,
very appropriate unit to report
also present in the case of vehicles - a
income. Depending on the objective
school, library, community resources).
high price-tag example of a non-ICT
at hand, analysis of low income is
perhaps best achieved for levels
Moreover, by 2000, rates of Internet
commodity. This shows that the
immediately after that group.
use from any location were already
effect of income on penetration is not
quite high for several high-income
simply an ICT phenomenon.
sub-groups. For example, the use
Many other variables come into play –
rate among households at the top
more, less or not at all related to
income quintile, headed by someone
income - and can be the foci of
with at least a university degree
analyses. For instance, that education
exceeded 90%, followed closely by
exerts a strong, positive and
households in the same income group
independent influence on Internet
with children less than 18.
use, even when controlled for income,
has been well documented (Dickinson
and Sciadas 1997, Dickinson and
Ellison 2000). The same is true for the
place of residence, with urban
penetration exceeding rural, as well as
for the type of family, where the
presence of children is associated with
higher penetration.
110
Television
100
90
Phone
Vehicle
80
Cable
VCR
70
60
Chart 1.
50
Household penetration,
Computer
40
Internet
by income, 2000
30
penetration rate (%)
Cell phone
20
10
0<10 10-15 15-20 20-25 25-30 30-35 35-40 40-45 45-50 50-55 55-60 60-65 65-70 >70
annual income (thousand $)
6
Statistics Canada - Catalogue No. 56F0004MIE
Science, Innovation and Electronic Information Division
CONNECTEDNESS SERIES
Table 1.
Internet divides, by income, 2000
Regular use from home
Use from any location
income quintiles
income quintiles
bottom 2nd
3rd
4th
top
all
bottom 2nd
3rd
4th
top
all
% %
Education
Education
Less than high school
6.8
10.7 22.4 30.0
40.5 16.1
10.9
15.4
32.5 39.7
50.1 22.5
High school/college
22.5
28.4 42.9 51.0
66.5 42.8
32.2
40.9
56.7 63.9
80.1 55.4
University degree +
42.9
47.4 56.6 63.7
78.9 65.1
56.7
62.3
71.1 81.1
91.0 79.3
Total
16.5
24.4 41.2 50.3 68.2 40.1
23.9 34.5 54.4 63.6 80.9 51.5
Family type
Family type
Single family, children <18
32.6
41.9 50.7 62.4
76.4 57.0
48.0
59.3
67.0 76.6
87.4 71.4
Single family, no children <18
16.2
19.4 35.1 45.8
63.0 37.8
22.0
24.8
43.8 57.3
75.8 46.9
One-person families
8.9
15.9 33.7 25.0
44.6 19.0
13.2
25.4
49.3 58.3
60.2 28.0
Total
16.5
24.3 41.2 50.4 68.2 40.1
23.9 34.5 54.4 63.7 80.9 51.5
Geographical location
Geographical location
Urban (CMA)
18.0
25.9 43.2 51.4
70.1 42.5
25.9
36.4
56.2 64.4
82.7 54.0
Rural (non-CMA)
11.6
19.3 33.9 46.0
57.5 30.9
17.6
28.4
47.7 60.8
70.5 41.8
Total
16.5
24.3 41.2 50.4 68.2 40.1
23.9 34.5 54.4 63.7 80.9 51.5
Notes: The top 18 Census Metropolitan Areas are used as a proxy for urban areas.
2.3 Divides among individuals
The significance of income, whether
100
personal or household, as a divide
90
80
variable can also be seen at the level
Household income
70
of the individual. Chart 2 shows that
Internet use increases steadily with
60
50
income. By 2000, high-income
Chart 2.
40
individuals had achieved use rates
Internet use, by
30
substantially higher than those of
individuals, by income,
Personal income
20
low-income individuals.
penetration rate (%)
2000
10
Age also causes divides. This has
0
0
5-10
15-20
30-40
50-60
80-100
been analyzed by Silver (2001), in
<5
10-15
20-30
40-50
60-80
>100
conjunction with gender issues, and
annual income (thousand $)
is displayed in Chart 3. Internet use
declines dramatically with age, from
over 90% for teenagers to less than
5% for aged individuals. This reflects
a combination of factors, including
100
issues of access opportunities, skills,
90
perceived needs, attitudes and
80
overall lifestyles.
70
60
Whether the analysis focuses on
Men
50
Chart 3.
individuals or households, some
40
Internet use, by age and
Women
variables matter more for some
30
gender, 2000
technologies (e.g. age is important
penetration rate (%) 20
for the Internet but not for cable).
10
0
15-17
20-24
30-34
40-44
50-54
60-64
70-74
18-19
25-29
35-39
45-49
55-59
65-69
75+
age (years)
Statistics Canada - Catalogue No. 56F0004MIE
7
CONNECTEDNESS SERIES
Science, Innovation and Electronic Information Division
2.4 Timing of introduction
decelerating growth7 . On the basis of
A dimension concerning the timing of
the information behind Chart 4, and in
the introduction of individual ICTs
conjunction with the information
must be added for a more complete
contained in Chart 1, the selected
understanding of digital divides. For
ICTs can be classified according to
6 -- From almost nothing in 1952, it
example, a telephone divide today
their overall penetration and growth
reached 10% in 1953 (the first year
for which data exist) and within a
must be seen under the light that the
patterns in the following stages:
decade of its introduction it exceeded
80% (1960). Then, it achieved near-
technology in its basic form, the
complete penetration even before the
twisted copper wire, has been around
i) Saturation stage: Included here
telephone, which had been around
much earlier.
for over a century. This differs from
are older technologies (telephone,
the divide associated with the
television) that have practically
Internet, which has been around for
achieved complete penetration for
7 -- This is known as the S-curve - the
less than a decade in its commercial
some time. For the last two
period of accelerating growth
corresponds to its convex part, up to
incarnation. Thus, analyses of the
decades, the penetration of both
the inflection point, whereas
divide require the explicit recognition
hovers around 98%-99% and
decelerated growth corresponds to the
concave part of the curve. Recorded
of all three dimensions – the specific
their rates of growth have
observations make it difficult to
ICT, the timing of its introduction and
dropped to levels that generally
discern any such shape. For example,
the low part of the S-curve is barely
the variable of interest.
match the rate of growth of
captured during the very early years
households
of the television, while that of the
8.
Historically, the introduction of new
computer (a technology with slower
penetration) is visible from 1986 to
technological commodities in
1996. In vehicles and the telephone
ii) Plateau stage: This refers to
consumption (as well as their diffusion
this part has long given way to a very
technologies whose overall
flat top part. In the case of the
and use by businesses and
Internet the low part is missing,
although early-enough observations
governments) has been gradual.
penetration is quite high for
several years, without having
exist. This means that the period of
Chart 4 presents a collection of
accelerating growth was very short-
reached saturation. Moreover,
lived.
recorded penetration histories, with
many data points over a long period
their penetration is changing
slowly, with growth oscillating
of time. Despite perceptions about the
around the rate of growth of
8 -- This should not be confused with
meteoric rise of the Internet, fast as
market growth, which is higher than
though it may have been, the
households (cable) or somewhat
the rate of growth of households for
higher (VCR)9.
several reasons: replenishment of
penetration of television in people’s
older sets, increase in multiple sets,
lives was faster6 .
and non-household users.
iii) Dynamic stage: Newer ICTs with
The penetration of the VCR was also
lower penetration but very high
9 -- Since 1997, though, the VCR did
very fast, particularly during its first
growth - the Internet, cell phones
reach saturation among the
decade. While the speed of adoption
and computers. Much of the
households in the top income decile.
Cable penetration hovers in the low to
among commodities differs, their
bewilderment and research of late
mid 70% range for over a decade and
penetration is generally characterized
is precipitated by the perceived
is lately slipping. Even among the top
income decile with the highest
by accelerating growth in the initial
opportunities they offer. So is the
penetration, it fell from an all-time
high of 85.5% in 1997 to 82.1% in
periods, which eventually gives way to
issue of the Digital Divide.
2000.
110
100
Television
90
Phone
80
70
Vehicle
60
Chart 4.
Cable
50
Penetration over time
40
Computer
penetration rate (%)
30
VCR
20
Cell phone
10
Internet
0
1953
1960
1966
1980
1985
year
1988
1991
1994
1997
2000
8
Statistics Canada - Catalogue No. 56F0004MIE
Science, Innovation and Electronic Information Division
CONNECTEDNESS SERIES
Stages
ICTs
Saturation
Plateau
Dynamic
Penetration
Practically complete
Very high but
Lower but
and stable
not complete
increasing
Growth
Stagnant
Very low or
Very high
oscillating
The diffusion pattern of ICTs in this
either growing or shrinking depending
stage underscores the previous
on how you look at it” (2001b, p.5).
analysis of the peculiarities of each
To overcome such inconclusive
technology. Moreover, it does not
findings, clarity is needed on two
provide support to the argument that,
fronts: establish well-defined
given a sufficiently long period of
concepts, and; apply appropriate
time, all technologies will reach
measures.
saturation. There are technological,
10 - The uneconomical deployment of
cable in very remote areas, for
economic, behavioural and other
3.1 The relative divide
instance, puts a ceiling on its
penetration. However, as measured
barriers at work that may place a
Quantifying the magnitude and
by households passed by cable, it
ceiling on penetration at levels below
monitoring the evolution of the gap
could have been as high as 95%
(April 2001). Convergence renders
The diffusion pattern of ICTs in this
between ICT ‘haves’ and ‘have-nots’,
the environment more competitive,
stage underscores the previous
as the issue was initially expressed,
with forces pulling or pushing in
different directions. Technological
analysis of the peculiarities of each
can be achieved by the absolute
advances together with the
technology. Moreover, it does not
regulatory regime make possible the
number of users or the overall
provision of more services through
provide support to the argument that,
penetration rates11. As long as from
the coaxial cable (Internet, digital TV
and telephony), which could increase
given a sufficiently long period of
one period to the next more people
its appeal. On the other hand, the
time, all technologies will reach
use, say, the Internet, there are more
competition brought about by
satellite dishes (legally since 1998)
saturation. There are technological,
‘haves’ and fewer ‘have-nots’. When
pulls the other way.
economic, behavioural and other
groups of people are delineated by
barriers at work that may place a
income (or any other variable), each
ceiling on penetration at levels below
one has its own penetration rate. (In
100% (or the 98%-99% that is
effect, the overall penetration rate is a
practically achieved by ubiquitous
weighted average of these). Typically,
11 - While the former is better suited for
market analyses, the latter is a
technologies)10. What is important is
higher income groups have higher
preferred policy measure. Penetration
that all new technologies are subject
rates account for changing
penetration rates but, over time,
demographics. It is conceivable that
to a divide in their early penetration.
penetration rates increase across all
the absolute numbers of both users
and non-users of a technology
Thus, at early stages of diffusion there
income groups. As this happens at
increase from one period to the next.
is strong justification to examine
different rates, though, the digital
If the rate of growth of users was
lower than the rate of growth of
people grouped by characteristic of
divide becomes a relative concept
households or individuals, a falling
interest. Only when ICTs approach
referring to the difference in the
penetration rate would result. This is
unlikely to happen at pre-saturation
saturation do their diffusion patterns
starting penetration rates and the
levels.
start to resemble the population at
growth rates across income groups.
large and the distinction between
Thus, its measurement involves
‘haves’ and ‘have-nots’ fades. Our
comparisons of the ‘haves’ between
analysis of the digital divide will focus
‘have-more’ and ‘have-less’ groups12 .
on income.
It does not involve the ‘have-nots’, as
12 - Alternatively,it oculd involve
is frequently – and inappropriately -
comparisons between the ‘have-nots’
in these groups, but not the ‘haves’.
stated. To understand the relative
3. THE MAGNITUDE OF THE
divide better, it is analytically
DIGITAL DIVIDE
instructive to look at an extreme
situation with no such divide. By
Research to date results in conflicting
definition, then, the penetration rate
messages concerning the size of the
of a certain ICT would be the same
Digital Divide and, especially, its
regardless of the level of income.
evolution. The OECD, for instance,
(Graphically, this would be depicted
identifies sizeable divides of the type
by a line parallel to the horizontal axis
discussed previously and states that
at a value X – Figure 1 in Technical
“…the digital divide could be said to be
Box 1).
Statistics Canada - Catalogue No. 56F0004MIE
9
CONNECTEDNESS SERIES
Science, Innovation and Electronic Information Division
TECHNICAL BOX 1: Trends and Ratios
By definition, the absence of a relative digital divide means that penetration rates are the same across all incomes.
This is depicted in Figure 1 with the straight line having an X intercept and a constant slope (zero). In reality, penetration
increases with income and the lines are sloping upwards. In the simplified case of a straight line, a measure of the relative
digital divide between high and low incomes is provided by its (constant) slope. The steeper the line, the greater its slope,
and therefore the bigger the relative divide.
The evolution of the relative digital divide is then measured by
slope changes. In the hypothetical example, line t shows the
Figure 1. Trend measures of the divide
1
pattern of penetration by income in the initial period and line
t in a subsequent period.
2
N(24)
Clearly, t is steeper than t and therefore b >b , indicating a
t2
2
1
2
1
growing divide. Measured at the same incomes, the slopes are
X
given by: b =(H-L)/(IH-IL) and b =(N-M)/(IH-IL), respectively.
1
2
Y(18)
Since the denominators are the same, the change in slopes can
t3
be approximated only with the numerators, that is, the
H(15)
t1
differences in the penetration rates between high and low
incomes.
penetration (# or %)
At the same time, the ratios of the high- and low-income
penetration rates between t and t are falling (N/M < H/L).
b2
1
2
M(6)
How can this happen at the same time as the slope increases,
b1
and how do all these relate to the rates of high- and low-
L(3)
income penetration growth?
IL
IH
income ($)
The upward movements of both intercepts are the result of
growth, with the rate of growth of the low income exceeding that of the high income - (M-L)/L > (N-H)/H. This is all that is
needed for the ratios to fall, even though t is steeper. The lines would move in parallel (unchanged steepness, slope, and
2
relative divide) only if the proportionate changes in the low and high incomes result in the same absolute changes in
percentage points (line t ), something that would require a much higher rate of growth for the low income group compared
3
to the high income group. How much higher?
With L and H denoting the penetration rates of the low and the high incomes in the initial period, and l and h their
respective rates of growth between the two periods, the following condition would have to hold:
L(1+l) - L = H(1+h) - H ⇒ Ll = Hh ⇒ l/h = H/L. In other words, the rate of growth of the low-income penetration must
be higher than the rate of growth of the high-income penetration by as many times as the ratio of the penetration rate of the
high- to the low-income was in the initial period. Any growth in low-income penetration lower than this would result in a
steeper line. However, any growth in the low income higher than the growth of the high income, however small the margin,
would decrease the ratio of their penetration rates. Both of these can - and do - happen at the same time.
In the numerical example: the penetration of the low income increased by 100% (from 3 to 6) and that of the high income
by 60% (from 15 to 24). Their ratio declined from 5 (15/3) to 4 (24/6), but the relative divide increased from 12 (15-3) to
18 (24-6) points. Although the penetration of the low income grew faster than that of the high income, it would have to be
five times as much for the relative divide to remain unchanged.
Linearity is not the only simplification in the above. Numbers of users or penetration rates can be used in the vertical axis,
while income can be expressed in dollars or percentiles. Practically, the choice matters. Without crowding the Figure, the
following are involved: differences in penetration can serve as a proxy of the relative digital divide if and only if the incomes
used for comparisons between periods are identical. Whether measured in dollar terms or percentiles, incomes generally
increase - especially in nominal terms – and this introduces an upward bias in the measure of the relative divide. This bias is
higher, the higher the income differential between high- and low-income groups – the denominator in the slope equation.
(Even if they both change proportionately the same, high incomes increase absolutely more, and thus the bias is still there).
Alternatively, keeping incomes constant in the analysis from period to period, when penetration rates are used on the
vertical axis, the number of households or individuals (used as the denominator in penetration rates) must stay the same.
This is not the case when incomes expressed in dollars are used, as more households or individuals move up into higher
brackets and fewer remain in the lower incomes. This introduces another bias, also upwards. Even if percentiles are used,
their mean income changes - so we revert to the previous bias. Changes in the composition of households by income
introduce yet another (unpredictable) bias.
That usage of the differences in penetration rates as proxies of the relative divide require the assumptions of constant
incomes and constant number of households or individuals from period to period is crucial. The fact that these assumptions
do not hold introduces biases when trend analysis is applied to real data.
10
Statistics Canada - Catalogue No. 56F0004MIE
Science, Innovation and Electronic Information Division
CONNECTEDNESS SERIES
In such a hypothetical case, we are
that closing divides should not be
immediately confronted by a two-
taken for granted, but they can
fold question: What examples can
regress14.
we think of, and what would be
possible values for X? Certainly not
The figures in Table 2 also indicate
luxury cars, exotic vacation
that the relative digital divide is
packages or stock options. Perhaps
widening across ICT and non-ICT
items that have become staples of
commodities – with the exception of
daily life – foodstuff, mattresses and
the VCR. This, however, cannot be
shoes, ignoring quality
generalized outside the two extreme
considerations. Moreover, the
income deciles, as it is subject to
admissible values for X would be at
several caveats (explained in
the rate of complete penetration -
Section 4).
close to 100%. It is not possible to
imagine anything with an X value at
13 - Income groups are practically
20%, 50% or even 80%
Table 2.
dictated by data availability.
penetration… Consumption patterns
Differences in penetration rates,
Generally, use of income percentiles
is superior to absolute income levels.
simply do not work like that.
top vs. bottom deciles
The distribution of households can be
very uneven when income increments
That penetration increases with
are used (for example, in 2000, 27%
1982 1986 1990 1996 2000
of households had incomes in excess
income is hardly a glamorous story.
of $70,000 – the highest income in
But the uneven Internet penetration
percentage points
Chart 1, whereas only 7% in the
$10,000-$15,000 range). Closely
between high and low incomes did
Telephone
7.4
7.5
4.6
5.2
11.9
related to this, percentiles are more
become a story, precipitating the
suitable for intertemporal
Television
3.9
2.9
2.2
1.5
3.8
comparisons, as the number of
issue of the digital divide, before the
individuals or households are kept
Cable
-
-
-
24.6
23.2
more or less stable, as they are
overall Internet penetration rate was
affected only by the increase in
at 7.4% in Canada, in May of 1996!
VCR
-
47.1
54.3
36.4
33.4
population. On the other hand, as
nominal incomes increase, the
Computer
-
18.8
31.8
48.2
65.2
distribution of households is biased
towards higher incomes.
Internet
-
-
-
18.2
62.5
3.2 Measures of the digital divide
Cell phone
-
-
-
24.8
55.9
The magnitude of the relative divide
can be approximated with the
Vehicle
56.5
56.4
51.3
47.1
58.8
difference in the penetration rates
between high- and low-income
groups. (The explanation is provided
14 - Whether this is the beginning of a
Another measure used for the divide
trend remains to be seen. However,
in Technical Box 1 for the interested
although there has been no
- although with much less
documented evidence so far of
reader). Differences were computed
theoretical justification - are the
substituting cell phone for wireline,
for the top and bottom income
this may just be an example of such
ratios of the penetration rates
a phenomenon, particularly among
deciles13 for selected years over a
the student population present in the
lengthy period and are shown in
among high- and low-income
bottom decile. Evidence provided by
groups. The measure is then
the quarterly telecommunications
Table 2. Again, the findings indicate
survey also points to the same
that the relative divide is very big for
interpreted as the ‘likelihood’ of
direction. In 2001, fixed
being connected. In the case of
telecommunications access declined,
newer technologies and drops for
whereas cellular phone subscribership
perfect equality, the ratio would be
increased by 24% (Statistics Canada
saturated technologies. In 2000, it
1; the greater the number, the
2002).
was bigger in computers and the
Internet, with 65.2 and 62.5
greater the divide. Such ratios were
percentage points separating
computed for 1996 and 2000 and
households in the two extreme
are shown in Chart 5.
deciles, followed by cell phones.
The basic findings are the same as
Considering that the penetration rate
before; the divide is greater in the
of computers was higher than the
newer technologies, especially the
penetration rates of the Internet and
Internet, whereas it barely registers
cell phones, it is evident that their
for saturated technologies.
concentration among higher incomes
(However, more detailed analysis
is heavier. The divide in telephones,
later will reveal regressive
where almost 12 percentage points
behaviours). According to this
in penetration separate households
measure, though, the digital divide
in the top and the bottom income
would appear to be closing,
deciles, is high for such an
contradicting the findings of the
established technology. The fact
previous analysis. For example,
that it widened sharply in the last
households in the top income decile
year of data serves as an example
were 7.6 times more likely than
Statistics Canada - Catalogue No. 56F0004MIE
11
CONNECTEDNESS SERIES
Science, Innovation and Electronic Information Division
2000
1996
8
7
6
5
Chart 5.
Ratios of penetration rates,
4
top vs. bottom income
3
deciles
2
ratio (top over bottom decile)
1
0
Telephone
Television
Cable
VCR
Computer
Internet Cell phone
Vehicle
households in the bottom decile to
time15. But they also felt it
15 - It will become evident shortly that
use the Internet in 1996, but only
important to extend the analysis to
the particular data sets used for 1996
and 1997 paint a different story that
5.3 times more likely in 2000. This
the underlying rates of growth
that of subsequent periods.
apparent contradiction occurs
among the various sub-groups,
because ratios are not measures of
identifying the higher growth of
the relative digital divide.
Internet use among all ‘have-less’
Decreasing ratios will be obtained
groups. Growth differentials,
always as long as the rate of growth
properly placed in perspective, can
of penetration among the low-
provide important clues regarding
income group exceeds that of the
future evolution. Recently, the OECD
high-income group – regardless of
also incorporates this type of
how small the margin may be. (This
analysis (2001b), but proceeds to
is explained in Technical Box 1).
state: “Whether the gap is closing or
They do, however, point to future
widening depends largely on how
evolution.
you choose to examine it. As we
shall see, when examined by
absolute percent access the gap
appears to be widening, and when
4. THE EVOLUTION OF THE
examined by growth rate it appears
DIGITAL DIVIDE
to be closing” (2001b, p.34). As
explained in Section 3, the two
Regardless of the magnitude of the
measures are not equivalent.
digital divide, a more pertinent
Quoting from Dickinson and Sciadas
question is whether it is widening or
(1999, pp. 3.4-3.5) is still relevant:
closing. Early analysis in the US
(1998) looked at the temporal
“…analyzing the evolution of the gap
movement of the differences in
between those who are connected
penetration rates among various
and those who are not involves
groups (high vs. low income, whites
several nuances.
vs. blacks and Hispanics, suburban
areas vs. inner-city cores) and
i) Penetration rates have increased
concluded that the (relative) digital
across income quartiles, levels of
divide was widening. Effectively, the
education, age groups and
methodology relied on trend
geographical locations... That is,
analysis. Dickinson and Sciadas
some of last year’s ‘have-nots’
(1999) employed the same
have turned into this year’s
methodology for Canada and arrived
‘haves’.
at the same conclusions for groups
ii) A more meaningful question to get
based on income, education, age,
a better understanding of the gap
family type and urban/rural areas -
is what population sub-groups did
on the basis of limited data at the
the new ‘haves’ mostly come
12
Statistics Canada - Catalogue No. 56F0004MIE
Science, Innovation and Electronic Information Division
CONNECTEDNESS SERIES
from. Here we must consider both
unmistakably much higher among
absolute and proportionate
the low-penetration population
changes. Penetration is increasing
segments…Should this trend
by more percentage points for the
continue, the percentage point
high- than the low-penetration
difference in penetration rates
population segments….In that
ultimately must begin to fall”.
sense the gap is widening…These
findings show that, in absolute
The nuances involved in the
terms, more of the newly
definition and the associated
connected households come from
measures of the digital divide are
the ‘have’ segments. However,
now apparent. The arguments in i)
such findings mask important
and ii) point to the relative nature of
underlying growth trends.
the divide, whereas iii) identifies
growth patterns as drivers of future
iii) Penetration rates and their
developments. While the precise
differentials are driven by
conditions have now been explained
connectedness rates of growth
(Technical Box 1), more insights on
among population sub-groups.
the evolution of the divide can be
What happens to the gap depends
obtained. The Internet becomes the
crucially on the magnitude of the
centerpiece of such analysis.
initial mass of connected
households, the associated gap in
4.1 Trend analysis
penetration rates by population
Not only does Internet penetration
sub-groups, as well as the
increase with income but, based on
respective rates of growth. Higher
all available statistics to date, its
rates of growth among low-
penetration by income increases
penetration groups may still result
across the entire spectrum - high,
in a widening gap. For instance,
middle and low incomes alike.
starting from a much higher
Trend analysis similar to that of the
number of connected households,
US (1998) and Dickinson and
high-penetration groups can gain
Sciadas (1999) is reproduced here
one percentage point in
on the basis of more detailed data
penetration with only a small
available now (covering the period
percentage increase in user
1996-2000 as opposed to only
households, whereas low-
1996-1997, and broken down by
penetration groups need a much
income deciles as opposed to
higher rate of growth. The
quartiles). Graphical analysis (Chart
underlying growth trends are
6) confirms that penetration rates
80
70
1996
1997
60
1998
1999
50
2000
40
Chart 6.
Trends of Internet
30
penetration, by income
penetration rate (%) 20
10
0
-10
2nd
3rd
4th
5th
6th
bottom
7th
8th
9th
top
Income deciles
Statistics Canada - Catalogue No. 56F0004MIE
13
CONNECTEDNESS SERIES
Science, Innovation and Electronic Information Division
continue to increase every year
The differences in Internet
across all incomes, and that the
penetration were computed for
fitted trend lines are becoming
many pairs of income deciles and for
visibly steeper – although
every year of available data. Then,
progressively less so, particularly
the changes in these differences
between 1999 and 2000. These
were estimated, annually and for
findings would indicate a growing
longer periods. In this specification,
relative divide.
a positive number indicates a
growing relative divide (the bigger
However, the conclusions from this
the number, the bigger the growth)
type of analysis are subject to
and a negative number indicates a
caveats. First, the pattern of
closing relative divide. The results
Internet penetration is not linear
are shown in the top part of Table 3.
(although in recent years the plotted
lines look more like straight lines
The relative divide is smaller, the
compared to earlier ones). Moreover,
smaller the income difference
a line approaching the penetration
between the groups examined.
rate of the top decile from above the
(This can be seen in the first five
trend (concave) would produce the
lines and, more visibly, in the 2000-
same trend as a line approaching it
1996 column). As well, the relative
from below (convex, as is mainly
divide generally increases less over
now). Therefore, the fitted trend
time. The important new finding,
lines mask important movements –
though, is that the pattern of
particularly in the middle incomes.
increase of Chart 6, is no longer
For this, more detailed analysis was
uniform. While there is still a
performed.
growing divide between pairs of very
Table 3.
Evolution of differences in Internet penetration rates
Deciles
1997-1996
1998-1997
1999-1998
2000-1999
2000-1996
2000-1997
top-bottom
22.9
6.1
7.9
7.4
44.3
21.4
9th - 2nd
18.5
6.1
13.6
5.5
43.7
25.2
8th - 3rd
12.1
10.4
4.6
4.8
31.9
19.8
7th - 4th
9.4
4.3
7.3
-4.1
16.9
7.5
6th - 5th
1.3
7.0
-3.3
2.0
7.0
5.7
top - 9th
5.1
1.9
-6.2
0.1
0.9
-4.2
top - 8th
10.5
-4.3
2.1
-2.6
5.7
-4.8
9th - 8th
5.4
-6.2
8.3
-2.7
4.8
-0.6
8th - 7th
1.1
4.7
-4.6
5.3
6.5
5.4
6th - 4th
4.8
6.7
-0.1
0.1
11.5
6.7
5th - 4th
3.5
-0.3
3.2
-1.9
4.5
1.0
4th - 3rd
1.6
1.4
1.9
3.6
8.5
6.9
top 5-bottom 5
12.8
6.8
6.0
2.8
28.5
15.6
income adjusted
Deciles
1997-1996
1998-1997
1999-1998
2000-1999
2000-1996
2000-1997
top-bottom
21.0
4.3
6.9
1.1
32.5
12.3
9th - 2nd
17.1
5.0
11.9
3.7
36.1
20.7
8th - 3rd
11.1
9.6
3.8
3.4
26.9
16.9
7th - 4th
8.9
3.6
7.0
-4.7
14.6
6.0
6th - 5th
1.1
6.6
-3.4
1.8
6.0
5.1
top - 9th
4.4
1.1
-6.2
-1.6
-1.5
-6.7
top - 8th
9.4
-5.2
1.9
-5.0
1.7
-8.3
9th - 8th
4.9
-6.3
7.8
-3.1
3.4
-1.5
8th - 7th
0.8
4.6
-4.7
4.9
5.0
4.7
6th - 4th
4.4
6.1
-0.2
-0.1
9.9
5.7
5th - 4th
3.4
-0.4
3.1
-2.1
3.9
0.6
4th - 3rd
1.5
1.3
1.8
3.1
7.2
6.2
14
Statistics Canada - Catalogue No. 56F0004MIE
Science, Innovation and Electronic Information Division
CONNECTEDNESS SERIES
high and very low incomes (e.g. top
the differences found are so
vs. bottom, 9th vs. 2nd and 8th vs. 3rd
emphatic that existing biases cannot
deciles), a closing divide begins to
possibly change their direction (such
be visible between other pairs (e.g.
as the growing divide between the
4th and 7th deciles from 1999-2000 -
highest and the lowest incomes –
a trend line fitted between these
first three lines in bottom part of
points would be flatter than the year
Table 3), they do not represent
16 - IAlthough graphically each decile is
represented by the same point on the
before). Repeating the exercise for
comprehensive measures of overall
horizontal axis over time, the
pairs of deciles both in the top and
movements.
underlying incomes increase. In
1996, the mean income of the
the bottom halves of the income
bottom decile was $7,989 and of the
spectrum, the mixed pattern that
top $130,656, while they were
$8,686 and $160,041 in 2000,
emerges is more crowded with
4.2 Concentration analysis
respectively. Their difference
increased by a factor of 23%.
closing divides (negative numbers).
While inequalities of the type
The relative divide closes between
encountered in the relative divide
several adjacent deciles, as early as
are difficult to prove conclusively
17 - The differences in penetration rates
1997-1998 (e.g. 8th and top, 8th and
with any single measure, a well-
between any two deciles in period t2
9th). Also, the relative divide
known analytical technique is that
were divided by the ratios of the
income difference between these
between the 4th and 6th deciles
offered by the Lorenz curve
deciles in t over their difference in t .
2
1
disappears in the last two years (two
(Technical Box 2). It was adapted
deciles still separated by more than
and utilized here for the penetration
$18,000 in 2000), as it does
of the Internet across income levels.
between the 4th and 5th deciles in
In the process, the distribution of
1999-2000. Dropping 1996 from
Internet users by income is
the calculations, closing divides
examined and the contribution of
between high-income deciles can be
different income groups to the
detected even over a longer period
expanded number of users is
(2000-1997 column).
identified.
To demonstrate how much such
The distribution of Internet users by
comparisons depend on the exact
income decile, for the 1996-2000
cut-offs chosen, the exercise was
period, is shown in Table 4. The
repeated with only two income
most notable change is observed at
groupings; the top half and the
the top decile, which accounted for
bottom half. In this case, the
18.2% of all Internet users in 2000,
relative digital divide is clearly
down from 28.4% in 1996. The
widening (last line of top part of
decline of its relative importance
Table 3). Therefore, this type of
was steady throughout the period.
analysis is crucially conditional on
the groups compared.
Table 4.
Second, (as explained in Technical
Distribution of Inernet users, by
Box 1) an upward bias is introduced
income
over time in estimating slopes with
data arranged in income deciles,
because of increasing income
Decile
1996
1997 1998
1999 2000
differences between high and low
%
deciles16. These differences matter,
bottom ........... 3.8
2.9
3.6
3.6
3.5
particularly over a period of five
2nd ............ 3.3
2.3
2.4
2.9
3.3
years. An adjustment was made for
3rd .......... 4.4
3.3
3.8
4.2
5.1
this bias and applied to the previous
4th .......... 6.0
4.9
5.5
6.0
7.4
analysis17. The results are shown in
5th ............ 6.5
7.1
7.0
8.1
9.7
the bottom part of Table 3. It
6th .......... 8.3
8.7
10.9
10.0
10.5
7th
becomes immediately evident that
.......... 9.7
11.9
12.2
13.2
12.1
8th ......... 13.4
14.1
15.6
14.4
14.2
measures of the relative divide are
9th .......... 16.3
18.4
16.2
17.3
15.9
getting smaller. More importantly, an
top .......... 28.4
26.5
22.6
20.3
18.2
even more mixed pattern between
all ...... 100.0 100.0 100.0 100.0 100.0
growing and closing divides
emerges, particularly for the last
year.
However, this is not the pattern of
other high-income groups. Only the
Thus more than meets the eye is
share of the 9th decile was somewhat
camouflaged behind trend lines.
smaller in 2000 compared to 1996 -
While the conclusions are valid when
and this after it had increased. The
Statistics Canada - Catalogue No. 56F0004MIE
15
CONNECTEDNESS SERIES
Science, Innovation and Electronic Information Division
relative loss of the importance of the
than the two lowest deciles. For
two highest income deciles
example, the top decile accounted
notwithstanding, the lot of the
for one-quarter of all new users
lowest two deciles did not improve at
between 1996 and 1997, but the 5th
all over the 1996-2000 period –
decile topped the list from 1999 to
indicative of a less than generalized
2000. In the same period, the four
closing of the divide. It is thus the
middle deciles (4th to 7th),
middle incomes that picked up
contributed almost half (48%) of all
share, and accounted for
new users.
proportionately more of the Internet
users in 2000 than in 1996.
Analogous findings hold true when
Table 5.
the income deciles from which
New Internet users, by income
newcomers emanated are explicitly
1996
1997
1998
1999
1996
identified – those who help close the
Decile
-
-
-
-
-
1997
1998
1999
2000
2000
absolute divide. Table 5 shows that
%
while over the entire period there is
bottom .................. 2.2
5.2
3.8
2.9
3.4
a very clear, positive relationship
2nd .................. 1.5
2.6
4.3
4.7
3.3
between income and Internet
3rd ................ 2.5
5.0
5.2
8.4
5.3
newcomers (last column), from
4th ................ 4.1
7.0
7.4
12.2
7.7
year-to-year the relative contribution
5th .................. 7.5
6.9
11.0
15.1
10.3
of the higher-income groups
6th ................ 8.9 16.0
7.6
12.3
11.0
7th
declined (7
............. 13.5
13.1
16.0
8.3
12.6
th to top decile) and that
8th ............... 14.6
19.2
10.9
13.7
14.4
of the others increased. The gains,
9th ............... 20.1
11.2
20.2
11.2
15.8
once again, were more pronounced
top ............... 25.0
13.9
13.7
11.3
16.1
among the middle incomes rather
all ........... 100.0 100.0 100.0 100.0 100.0
TECHNICAL BOX 2: The Lorenz Curve
This is a method commonly used to study the inequality of the distribution of income. Making appropriate modifications
to the standard application, this analytical tool is adapted to fit the context of the digital divide. Rather than plotting
penetration against each income percentile, the cumulative distribution of penetration is plotted against the cumulative
income percentiles, from lowest to highest (Figure 2).
In the case of no relative divide (perfectly egalitarian distribution), the curve would coincide with the diagonal 45o line,
on which X% of incomes account for X% of penetration (0 and 100 are points on the curve, as 0% of incomes account
for 0% of penetration and 100% of incomes account for 100% of penetration). In reality, the curve will always be slanted
below the diagonal. The further away the bend is, the bigger the divide. (The bulge in this case is the measure). The
example shows that the bottom 40% of incomes accounts only for 10% penetration.
Plotting curves for successive periods allows the measurement of
the evolution of the relative divide. When from one period to the
next curves are cleanly inside or outside one another, the
conclusion is unequivocal – the divide is closing or widening,
Figure 2. The Lorenz curve
respectively. In case of crossing lines, however, there are trade-
100
offs involving winners and losers and specific areas must be
90
examined. In such cases, an overall measure is provided by the
80
calculation of Gini coefficients. These are effectively measured by
70
the ratio of area A over A+B. Gini coefficients can assume
60
values from 0 (perfect equality) to 1 (extreme inequality). The
50
larger the area between the 45o line and the Lorenz curve is, the
40
A
further away from perfect equality, and the higher the value of the
e Internet use (%)
Gini coefficient.
30
20
B
Such measures are not free of problems. The Lorenz curve
cumulativ
emphasizes the totality of the situation and does not directly
10
compare percentile pairs (e.g. highest vs. lowest). In addition,
0
10
20
30
40
50
60
70
80
90 100
any measure that tries to encompass the entire Lorenz curve in a
cumulative share of income (%)
single statistic would inevitably contain elements of arbitrariness.
Especially when curves intersect, curves of different shapes (and
therefore different patterns of divides) could generate the same Gini. Clearly, these are aggregate measures best suited
for an overall assessment. They do not replace detailed comparisons of specific groups.
16
Statistics Canada - Catalogue No. 56F0004MIE
Science, Innovation and Electronic Information Division
CONNECTEDNESS SERIES
Computing the cumulative
comparisons involving specific
distribution of Internet use by
income groups, especially those far
income decile (not shown), the
from each other. Therefore, it does
Lorenz curve in Chart 7 was
not invalidate the conclusions of the
constructed. With the exception of
trend analysis concerning the very
1996, curves for each successive
low and the very high incomes.
year are cleanly enveloped by those
Even visual inspection of Charts 7
of the previous year, indicating a
and 8, makes it clear that the inward
closing relative divide. The 1996
bow of the curves over time is more
curve behaves differently, crossing
pronounced in the middle incomes –
other curves, something that
particularly upper-middle. It is
renders the comparison of the
between these incomes and the top
relative between 1996 and other
decile that the relative divide is
years inconclusive (see Technical
closing and this influences
Box 2). This can be rectified by
significantly the overall result (quite
computing Gini coefficients - also
consistent with the findings in Tables
contained in Chart 7. They indicate
4 and 5). At the bottom end of
that the relative divide actually
incomes, it is more difficult to
increased between 1996 and 1997,
decipher movements from this
while it keeps closing from 1997
analysis. The trend analysis of
onwards.
Section 4.1 involving comparisons
This analysis was repeated with a
between the top and the bottom
different data set (HIUS) covering
deciles is very direct, whereas in
Lorenz curves they become two of
the use of the Internet by income
quintile from any location for the
ten deciles and their comparison is
1998-2000 period. The conclusions
subsumed to that of the aggregate
direction.
are the same: the relative digital
divide is closing. Lorenz curves for
These findings jive well with all other
previous years cleanly envelop those
pieces of evidence presented so far.
for subsequent years and the Gini
Collectively, they conclude that the
coefficients decline each year (Chart
relative digital divide is generally
8). Once again, the relative gain of
closing but this is the result of the
the middle incomes (from 40% to
progress middle-income groups have
80% cumulative share) is evident.
made (particularly upper-middle)
when compared to the highest
While revealing, this analysis is not
income group. The lowest income
free of problems either. It offers an
overall assessment of the evolution
groups (the three bottom deciles
here) continue to lose ground vis-à-
of the relative divide across all
vis the very high income groups.
incomes, but it is less suitable for
100
GINI
90
1996 = 0.37
80
1997 = 0.40
1998 = 0.35
70
1999 = 0.32
60
2000 = 0.27
Chart 7.
Evolution of the relative
50
Internet divide, home use.
40
30
cumulative Internet use (%)
20
10
0 0
10
20
30
40
50
60
70
80
90
100
cumulative share of income (%)
Statistics Canada - Catalogue No. 56F0004MIE
17
CONNECTEDNESS SERIES
Science, Innovation and Electronic Information Division
100
90
GINI
1998 = 0.32
80
1999 = 0.27
70
2000 = 0.22
60
Chart 8.
50
Evolution of the relative
40
Internet divide, use from
any location.
30
20
cumulative Internet use (%)
10
0 0
20
40
60
80
100
cumulative share of income (%)
5. RELATED MATTERS
application. ICTs in education,
As the Internet has become the
health, public libraries and the like,
epitome of modern communications,
come with their own specific
there are many pragmatic reasons
research needs. Much remains to be
why the Digital Divide matters. This
understood following these early
study dealt with the Digital Divide
stages. In addition, while in
and arrived at certain conclusions,
developed countries enough
within the scope of its investigation.
attention is paid to connected homes
The divide is generally closing, but
in divide studies, the more
the gap between the highest and the
communal attitudes of less-
lowest incomes persists. However,
developed countries warrant a
this is quite consistent with many
differentiated approach which would
technologies in their early stages of
incorporate telecentres and other
adoption and it remains true that the
community resources more
rate of growth of Internet use at
explicitly.
lower incomes is higher that than of
the higher incomes.
What comes through as the common
thread of any investigation, though,
In addition, this should be placed in
is the issue of skills. For many,
the perspective that the composition
uneven opportunities in acquiring
of income groups changes over time.
ICT-related skills, so crucial in
It is not the same individuals or the
functioning from now on, constitute
same group of families that comprise
the real divide. At the same time,
them. However important income is,
skills are properly perceived as a
there are plenty of closely
continuum whereby their technology
interrelated aspects to this issue.
components are incrementally built
Understanding barriers to access
on cognitive skills and general
and use new technologies can be
literacy. Research in this area is at
improved. While there is ample
an embryonic stage.
evidence that affordability is critical,
it certainly does not explain the still-
In the end, the issue of the Digital
sizeable proportion of non-users at
Divide, like all others, will come
the highest income levels. Numerous
down to outcomes and impacts. “The
other factors are at play, many of
fundamental digital divide is not
which change over time due to the
measured by the number of
evolution of the technologies, falling
connections to the Internet, but by
prices, social norms and much more.
the consequences of both connection
and lack of connection” Castells
Another dimension worthy of
(2001, p. 269). In examining such
examination for the understanding
consequences, though, the degree of
of the Digital Divide is the sector of
connectivity matters.
18
Statistics Canada - Catalogue No. 56F0004MIE
Science, Innovation and Electronic Information Division
CONNECTEDNESS SERIES
REFERENCES
April D. (2001) “Internet by Cable”, Connectedness Series, Statistics Canada, Catalogue No. 56F0004, No. 2.
Castells M. (2001) “The Internet Galaxy”, Oxford University Press.
Dickinson P. and Ellison J. (2000) “Plugging In: The Increase of Household Internet Use Continues Into 1999”,
Connectedness Series, Statistics Canada, Catalogue No. 56F0004, No. 1.
Dickinson P. and Sciadas G. (1996) “Access to the Information Highway”, Canadian Economic Observer, Statistics
Canada, Catalogue No. 11-010, December.
Dickinson and Sciadas (1997) “Access to the Information Highway: the Sequel”, Services Indicators, Statistics
Canada, Catalogue No. 63-016, June.
Dickinson and Sciadas (1999) “Canadians Connected”, Canadian Economic Observer, Statistics Canada,
Catalogue No. 11-010, February.
Dryburgh H. (2001) “Changing Our Ways: Why and how Canadians use the Internet”, Statistics Canada,
Catalogue No. 56F0006XIE.
ETS (2002) “Digital Transformation: A Framework for ICT Literacy”, A Report of the International ICT Literacy
Panel.
G-8 (2000) “Okinawa Charter”, www.dotforce.org
G-8 (2001a) “Digital Opportunities for All: Meeting the Challenge”, Report of the Digital Opportunities Task Force
(DOT Force), including a proposal for a Genoa Action Plan.
G-8 (2001b) “Global Bridges, Digital Opportunities”, www.dotforce.org
Hoffman D. and Novak T. (1999a) “The Evolution of the Digital Divide: Examining the Relationship of Race to
Internet Access and Usage Over Time”, Vanderbilt University, May.
Hoffman D. and Novak T. (1999b) “The Growing Digital Divide: Implications for an Open Research Agenda”,
Vanderbilt University, November.
National Science Foundation (2001) “The Application and Implications of Information Technologies in the Home:
Where are the Data and What Do they Say”?, February.
Noll R., Older-Aguilar D., Rosston G., Ross R. (2000) “The Digital Divide: Definitions, Measurement and Policy
Issues”, California Public Affairs Forum, Stanford University.
OECD (2001a) “Understanding The Digital Divide”, Paris.
OECD (2001b) “The Digital Divide: Diffusion and Use of ICTs”, DSTI/ICCP/IE (2000)9/REV2, October.
OECD (2002) “ICT Diffusion and the Digital Divide”, DSTI/ICCP/IE (2001)9/CHAP6/REV1, prepared as Chapter 6
for the Information Technology Outlook 2002, February.
Sciadas G. (2000) “Digital Households in Canada” in Information, Innovation and Impacts, eds. John de la Moth
and Gilles Paquet, Economics of Science, Technology and Innovation, Kluwer Academic Publishers.
Sciadas G. (2002) “Monitoring the Digital Divide”, an Orbicom-CIDA project, National Research Council of
Canada, March.
Silver C. (2001) ”Internet Use Among Older Canadians”, Connectedness Series, Statistics Canada,
Catalogue No. 56F0004, No. 4.
Statistics Canada (1996) “Household Facilities and Equipment”, Catalogue No. 64-001.
Statistics Canada - Catalogue No. 56F0004MIE
19
CONNECTEDNESS SERIES
Science, Innovation and Electronic Information Division
Statistics Canada (2001a) “Estimates of Household Spending, 2000”, The Daily, December 12, www.statcan.ca
Statistics Canada (2001b) “Household Internet Use Survey”, Catalogue No. 56-M002XCB.
Statistics Canada (2001c) “General Social Survey: Internet Use, 2000”, The Daily, March 26, www.statcan.ca
Statistics Canada (2002) “Quarterly Telecommunications Bulletin”, Catalogue No. 56-002.
The Conference Board of Canada (2000) “Canada Wins Silver in the Race for Connectedness”, January.
The Conference Board of Canada (2001) “Connectedness and the Economy: Measuring Our Performance”,
January.
The Conference Board of Canada (2002) “Connecting Canadians: 3rd Annual Report”, May.
U.N. (2001) “Information and Communication Technologies Task Force”, Report of the Secretary General,
February.
U.S. (1995) “Falling through the Net: A Survey of the ‘Have-Nots’ in Urban and Rural America”, Department of
Commerce, July.
U.S. (1998) “Falling through the Net II: New Data on the Digital Divide”, Department of Commerce, July.
U.S. (1999) Falling through the Net III: Defining the Digital Divide”, Department of Commerce, July.
U.S. (2000) “Falling through the Net IV: Toward Digital Inclusion”, Department of Commerce, October.
U.S. (2002) “A Nation Online: How Americans Are Expanding Their Use of the Internet”, Department of
Commerce, February.
World Bank (2002) “Information for Development Program: Annual Report 2001”, www.infodev.org
World Economic Forum (2002) “Annual report of the Global Digital Divide Initiative”, www.weforum.org
20
Statistics Canada - Catalogue No. 56F0004MIE
Science, Innovation and Electronic Information Division
CONNECTEDNESS SERIES
Connectedness Series
Plugging In: The Increase of Household Internet Use
1
Continues into 1999
P. Dickinson and J. Ellison
2
Internet by Cable
D. April
3
Internet Shopping in Canada
J. Ellison, L. Earl, S. Ogg
4
Internet Use among Older Canadians
C. Silver
5
Electronic Commerce and Technology Use
G. Peterson
6
Embracing e-Business: Does Size Matter?
S. Charles, M. Ivis, A. Leduc
7
Unveiling the Digital Divide
G. Sciadas
Statistics Canada - Catalogue No. 56F0004MIE
21
