Vodafone survey

Africa:
The Impact
of
Mobile
Phones
The Vodafone Policy Paper Series • Number 2 • March 2005Moving the debate forward
MovingthedebateforwardTheVodafonePolicyPaperSeries•Number2•March2005

Contents
Page
Foreword 00
– Arun Sarin, Chief Executive, Vodafone Group
Introduction 01
– Neil Gough and Charlotte Grezo, Vodafone Group
Overview 03
– Diane Coyle
The impact of telecoms on economic growth in developing countries 10
– Leonard Waverman, Meloria Meschi, Melvyn Fuss
Mobile networks and Foreign Direct Investment in developing countries 24
– Mark Williams
Introduction to the community and business surveys 41
Mobile communications in South Africa, Tanzania and Egypt:
results from community and business surveys 44
– Jonathan Samuel, Niraj Shah and Wenona Hadingham
Linking mobile phone ownership and use to social capital in rural South Africa and Tanzania 53
– James Goodman
Bibliography 66
Africa: The Impact of Mobile Phones Moving the debate forward • The Vodafone Policy Paper Series • Number 2 • March 2005
Foreword
I hope you enjoy our second Vodafone Policy Paper. Our aim in these papers is to provide a platform for leading experts to write on
issues in public policy that are important to us at Vodafone. These are the people that we listen to, even if we do not always agree
with them. These are their views, not ours. We think that they have important things to say that should be of interest to anybody
concerned with good public policy.
Arun Sarin, Chief Executive, Vodafone Group
To keep the environmental impact of this document to a minimum, we have given careful consideration to the production process. The paper used was manufactured in the UK at mills with
ISO14001 accreditation. It is 75% recycled from de-inked post consumer waste. The document was printed in accordance with the ISO14001 environmental management system.
All the steps we have taken demonstrate our commitment to making sustainable choices.
Designed and produced by Barrett Howe Plc
Africa: The Impact of Mobile PhonesMoving the debate forward • The Vodafone Policy Paper Series • Number 2 • March 2005

International Institutions DirectorNeil Gough
Introduction
About 18 months ago we became interested in studies on the
economic and social impacts of mobile telecommunications.
However, a thorough review of the existing literature revealed
surprisingly little systematic evidence. There were many
anecdotes, some interesting sociological research, but few
successful studies looking at the economic impacts on
individuals, businesses and overall economic activity.
This project has its roots in our dissatisfaction with that situation.
It seemed extraordinary that a technology that has clearly taken
the world by storm had attracted so little rigorous research.
It was equally clear that there was widespread interest in the
subject. As we discussed our programme and ideas with people
both inside and outside the industry, the appetite for this work
became obvious.
We wanted the work to be able to survive the scrutiny of a
potentially skeptical audience. Therefore, with advice from the
Vodafone Advisory Panel (a group of academics, officials and
NGO representatives with interests in this field) we developed a
programme of research.
The field was wide open so we could have chosen to focus
anywhere but we started with the impact of mobile in the
developing world. The reason was simple. We were inspired by a
conversation with Alan Knott-Craig, the CEO of our affiliate
company in South Africa, in which he talked about the impact
mobile was having in Africa. The variety of the examples he
mentioned were simply extraordinary.
Vodafone operates around the globe and has a particular interest in
developing markets in Africa, not least because of the success of
our investment in Vodacom, initially operating in South Africa and
now also in Democratic Republic of Congo, Lesotho, Mozambique
and Tanzania. Vodafone also operates in Egypt and Kenya.
At the time we began this work, the fact that Africa was to play
such a leading part in the G8 agenda for 2005 and the work of
the Africa Commission was unknown. We have been fortunate
that the issues we have covered resonate with these important
international initiatives. We hope that these studies will assist in
highlighting the part that mobile telecommunications can play in
the developing world.
We have learned a great deal. Most important is the fact that the
ways in which mobiles are used, valued and owned in the
developing world are very different from the developed countries.
More attention should be paid to the characteristics of how
people actually do use phones in the developing world in policy
debates on increasing access to Information and Communication
Technology (ICT). It is wrong to simply extrapolate our developed
world models of needs and usage patterns to poorer nations.
Understanding the context is vital. In the UK, the ratio of the
01
Neil is currently the International Institutions Director for Vodafone Group, coordinating
international policy and relationships with global institutions throughout Vodafone Group.
Director of Corporate ResponsibilityCharlotte Grezo
Charlotte is Director of Corporate Responsibility for Vodafone and is responsible for
coordinating the Group's approach to managing social and environmental issues.

02
number of outgoing voice calls made to the number of SMS
messages sent is 0.6:1; in South Africa as a whole, the ratio is
3:1 for pre-pay phones; yet in the rural communities we
surveyed, the average ratio was a remarkable 13:1. In Ndebe,
a rural community in South Africa, the ratio was 17:1, but when
one considers this in the context of a community in which
access to education is not universal, the data are more
understandable. The combination of illiteracy and indigenous
languages clearly has dramatic effects on the use of SMS
messaging; the implications of this extend to other types of data
usage (e.g. the internet). Our view is that the policy debates on
ICT policy are not sufficiently informed by this type of evidence.
The value of communications in the developing world is also
different. Imagine you are painter living in a township near
Johannesburg and you are some way from your potential clients.
You are looking for work but the postal service is poor and there
is no fixed-line phone. How does a potential employer contact
you? A mobile provides you with a point of contact; it actually
enables you to participate in the economic system (see
photograph below). Similarly, if you live in a rural community
and you need to go to the nearest town to shop for some
particular goods, a mobile phone call could save you a relatively
expensive return bus fare and the lengthy journey time, if the
goods were out of stock. When other forms of communication
are poor, whether roads or fixed-line telephones, the value of
quality mobile communications is much greater.
We have also learned that people in Africa use mobile phones
very differently. Most striking is the accessibility of mobile.
While penetration rates are by the standards of the developed
countries low, the way in which mobiles are informally shared
between people, the formation of private resellers of mobile
services and the provision of mobile phones for public use, all
increase accessibility, even in rural communities. The impact of
mobile extends well beyond what might be suggested by the
number of subscriptions alone.
The informal arrangements that extend the reach of
telecommunications are very powerful. In the data for the rural
communities in South Africa, we noticed that the ratio of inbound
texts to outbound texts was about 8:1. This imbalance is
attributed to the entrepreneurial activity of some of the more
literate individuals with cell phones who, for a marginal fee,
receive and relay text messages to those without cell phones or
those who cannot read or write. This is apparently a very
common practice in most of the rural areas.
The developed world model of personal ownership of a phone is
not relevant, or indeed appropriate, to the developing world.
With an understanding of this context, one can more easily
appreciate why the usage of the technology is growing so quickly
and in such distinctive ways in Africa. In the UK, there are now
more mobile subscriptions than fixed lines; that cross-over
occurred in 2000 (about 15 years after the first mobile call was
made); in Tanzania, that cross-over point was also reached in
2000 (but just 5 years after the first mobile was sold). The
relative impact of mobile on communications has been much
more dramatic in Africa and the growth is now accelerating at a
tremendous rate. The number of subscribers in Nigeria, the
world’s fastest-growing market according to the International
Telecommunications Union, increased by 143 per cent in the 12
months to June 2003. In Africa, increasingly telecommunications
means mobile telecommunications. Fixed-mobile substitution is
not a relevant concept, because the whole developmental stage
of widespread fixed line service has been leap-frogged by mobile
in many nations.
The mobile telecommunications story in Africa and the
developing world is a remarkable one. There have been large
infrastructure investments, which have enabled millions of people
to communicate better. While there is a lot of focus on low
absolute rates of mobile penetration, this underestimates the real
impact that mobile is having through the innovative and
entrepreneurial ways in which the technology has been extended
beyond the model of individual ownership. Thousands of jobs
have been created and some very successful indigenous
companies have emerged. All of these results were achieved
through enterprise rather than aid. A clear success story in
commercial terms but one that also had a profound impact on
the development of the economy and society.
We have been greatly assisted in this program by the work of
Diane Coyle, who has written the introductory piece and edited
this pamphlet. It would not have been possible without her
efforts and enthusiasm. We would also like to express our thanks
to the various contributors for their papers and the stimulating
discussion that has accompanied the work. We all have a lot to
learn about mobile communications in Africa and the developing
world. This is our initial contribution to that process, which we
hope will stimulate you to explore these issues further.
A mobile enables tradesmen to participate in the economy. Innovative advertising on the
outskirts of Johannesburg.

03
Just 20 years after the launch of the world’s first commercial
mobile services, there were more mobile than fixed-line users
globally, and nearly as many people had a mobile as a television.1
Vodafone’s Socio-Economic Impact of Mobile (SIM) programme
started from the beginning of 2004 to commission research
which would help extend the evidence and develop a better
understanding of the effects of this extraordinary phenomenon.
Mobile communications are experiencing faster growth rates in
low-income countries – more than twice as fast as in the high-
income countries in recent years. Low- and middle-income
countries are therefore accounting for a rising share – now more
than 20 per cent – of the world mobile market. But there is great
variety between countries in mobile phone penetration and use.
Surprisingly, given its extensive poverty, Africa has been the
fastest-growing mobile market in the world during the past five
years. The first cellular call in Africa was made in Zaire in 1987
(the operator was Telecel). Now there are more than 52 million
mobile users in the continent (compared to about 25 million fixed
lines). In 19 African countries, mobiles account for at least three
quarters of all telephones.2
Africa as a whole lags far behind
richer regions of the world. Nevertheless, the rapid spread of
mobile in so many of its countries is a remarkable phenomenon,
especially in the context of their huge economic and social
challenges.
This report describes and summarises the initial research
projects commissioned by Vodafone and carried out in the
second half of 2004. The results described here confirm the
vital social and economic role already played by mobile
telephony in Africa less than a decade after its introduction
there. The research documents its impact both at the macro-
economic level and at the level of particular communities and
businesses. It contributes to the evidence base for the
development of both regulatory policies and business strategies
in Africa. This opening section sets the context with an overview
of the data and of the earlier academic literature on mobile, and
information and communication technologies more generally, in
developing countries.
The African context
At the end of 2003, there were 6.1 mobile telephone subscribers
for every 100 inhabitants in Africa, compared with 3 fixed line
subscribers per 100.3
Mobile penetration is much higher in other
regions of the world – 15 per 100 inhabitants in Asia for
example, 48.8 in the US and 55 in Europe. Even so, there were
51.8 million mobile subscribers in Africa at the end of 2003,
reflecting an increase of more than 1000 per cent in five years.
Access to mobile telephony in Africa is also almost certainly far
more extensive than the subscriber figures suggests, as each
handset and subscription has many users.
Figure 1: Mobiles and fixed lines per 100 people,
1998 and 2003
Enlightenment EconomicsDiane Coyle
Overview
A Vodacom Community Phone Shop bringing new communication possibilities to Dobsonville,
South Africa.

04
Investment in total telecommunications in Africa has been about
5-6 per cent of total fixed investment spending on the continent
in recent years, although with wide variations between countries.
Mobile network coverage is most extensive in the North African
countries and South Africa, where coverage has improved
dramatically. Continuing to improve telecommunications
infrastructure is a priority area of policy for African governments
and organisations such as Nepad (New African Partnership for
Economic Development) and the international community.4
Figure 2: GSM mobile coverage in Africa
As the International Telecommunication Union has pointed out,
the phenomenon of the rapid spread of mobile cuts across many
of the obvious characteristics distinguishing one country from
another, such as GDP per capita, socio-demographic or
geographic criteria.5
Thus Finland and Uganda have a similar
proportion of mobile-only users but are obviously not sensibly
comparable countries. Within Africa, countries as different in
their socio-economic characteristics as Algeria and Lesotho have
similar mobile penetration rates. So there is no simple way to
summarise the penetration patterns across countries. According
to the most recent ITU figures, shown in Table 1, penetration rates
ranged from 0.1 per 100 in Guinea-Bissau and 0.14 in Ethiopia
to 68.18 per cent in Seychelles and 74.74 per cent in Reunion.
In most of the continent’s biggest economies, penetration rates lie
in the 20-40 per cent range, although with exceptions such as
Egypt (8.26 per cent) and Nigeria (2.55 per cent).
However, there can be little doubt that the wildfire spread of
mobile was triggered partly by the liberalisation of the telecoms
markets in many African countries from the mid-1990s,
including the issuing of private mobile licenses, often to
international operators. Those countries which made an early
start down this path – such as Gabon or Mauritius – have mobile
penetration rates which might seem surprisingly high given other
social and economic indicators, and their size; and the converse
is true for countries where there were no early private licences
issued, such as Algeria or Nigeria. Research by the World Bank6
looking at 41 African countries found that the introduction of a
second and subsequent (private sector) competitors accelerates
mobile penetration, whereas the presence of a state-owned
telecoms incumbent in the market inhibits diffusion. Table 2
demonstrates this pattern for a number of countries.
Understanding the differences will be important for the design of
policy by African governments and telecoms regulators, and this
is an important area for further research. Formal competition
policies are in their infancy in Africa, with only Kenya and South
Africa having a clear framework in place at present. Many
countries still have dominant state telecoms operators, with
sufficient political power to ensure the regulatory framework is
designed in their own interest. Given their typical history of
inefficiency and corruption, their dominance is counter-
productive, inhibiting the rapid spread of mobile communication
networks.
Table 1: Mobile penetration rates in Africa
Population, Mobiles,
millions thousands Mobiles/100
Algeria 31.8 1447 4.6
Egypt 70.2 5731 8.2
Libya 5.5 100 1.8
Morocco 30.1 7333 24.3
Tunisia 9.9 1844 18.6
North Africa 147.5 16455 11.2
South Africa 46.4 16860 36.4
Angola 14.4 250 1.7
Benin 7.0 236 3.4
Botswana 1.8 493 28.0
Burkina Faso 12.3 227 1.9
Burundi 7.1 64 0.9
Cameroon 16.3 1077 6.6
Cape Verde 0.4 53 12.1
Cen. African Rep. 4.1 13 0.3
Chad 8.1 65 0.8
Comoros 0.8 2 0.3
Congo 3.5 330 9.4
Cote D’Ivoire 16.6 1236 7.4
DR Congo 52.8 1000 1.9
Djibouti 0.7 23 3.4
Eq Guinea 0.5 42 7.6
Ethiopia 69.4 98 0.1
Gabon 1.3 300 22.4
Gambia 1.4 130 9.5
Ghana 22.4 800 3.6
Guinea 7.8 112 1.4
© Acacia Initiative – IDRC
Population per sq.km
Unpopulated 50 to 100
Less than 1 100 to 200
1 to 10 200 to 400
10 to 25 400 to 500
25 to 50 Greater than 500
GSM Coverage
Mobile (GSM) coverage

05
Population, Mobiles,
millions thousands Mobiles/100
Guinea-Bissau 1.3 1 0.1
Kenya 31.7 1591 5.0
Lesotho 2.2 165 7.6
Liberia 3.4 2 0.1
Madagascar 16.3 280 1.7
Malawi 10.5 135 1.3
Mali 10.9 250 2.3
Mauritania 2.8 300 10.9
Mauritius 1.2 463 37.9
Mayotte 0.2 36 21.6
Mozambique 18.8 429 2.3
Namibia 1.9 190 9.9
Niger 12.3 24 0.2
Nigeria 123.3 3149 2.6
Reunion 0.8 565 74.7
Rwanda 8.4 134 1.6
S Tome & Principe 0.2 5 3.2
Senegal 10.4 783 7.6
Seychelles 0.1 55 68.4
Sierra Leone 5.0 100 2.0
Somalia 10.3 40 0.4
Sudan 33.3 650 2.0
Swaziland 1.0 88 8.4
Tanzania 35.3 891 2.5
Togo 5.0 200 4.0
Uganda 25.6 776 3.0
Zambia 11.2 150 1.3
Zimbabwe 11.8 363 3.1
Sub-Saharan 647.7 18363 2.8
AFRICA 841.5 51678 6.1
Source: ITU African Telecommunication Indicators (2004)
Table 2: Mobile competition in selected African countries
Date of 1st State-
Date of 1st competing owned
mobile private mobile Mobiles/100
Country licence licence operator? population
Algeria 1989 2001 Y 4.6
Benin 1995 2000 N 3.4
Egypt 1987 1998 N 8.2
Mauritius 1989 1996 N 37.9
Morocco 1987 1994 Y7
24.3
Nigeria 1992 2001 Y 2.6
Senegal 1992 1998 Y 7.6
S Africa 1986 1994 N 36.4
Tunisia 1985 2002 Y 18.6
Uganda 1995 1998 Y 3.0
Source: Based on Gebreab (2002), ITU database.
There are of course many other possible explanatory factors
apart from regulatory policy for differences in mobile penetration
rates – factors such as incomes and growth, urbanisation,
education levels, and other aspects of policy including tariffs.
Not surprisingly, as Figure 3 shows, mobile penetration is
strongly positively correlated with income per capita. (The simple
correlation coefficient is 0.75 for the period 1995-2002.8
)
However, it is not strongly correlated with trade, measured as the
ratio of imports plus exports to GDP. The correlation coefficient in
this case is just 0.34.
Figure 3
On the other hand, per capita income is clearly not the only
important explanatory factor, as many African countries have
seen rapid growth in mobile during a period when income
growth has been low. This means there is some trend towards
convergence in access to mobile telephony across countries.9
For example, between 1998 and 2003 the number of mobile
subscribers per 100 rose from 7.92 to 36.36 in South Africa,
which has one of Africa’s highest penetration rates; during the
same period the figure for Rwanda, which has one of the lowest,
increased from 0.12 to 2.52. Mobile seems to be a good
example of a technology that permits leapfrogging of an older
infrastructure.10
What’s more, in contrast to the diversity of
patterns between countries, mobile use within any given country
is characterised by greater uniformity than other ICTs across, for
example, socio-economic groups or gender. The implication of
these two trends – some convergence between countries and
smaller differences within countries – is that the digital divide
could be smaller in the case of mobile compared with other ICTs.
However, there is some evidence that an increase in (fixed line)
telephone density in the past has been correlated with faster
growth in the incomes of the poor but even faster growth in the
incomes of the rich, therefore associated with increasing
inequality.11
It is far from established that mobile is yet affecting
income distribution in either direction. The likelihood is that the
distributional impact will be complicated, depending on the
geographic pattern of rollout between different areas, and
especially as between urban and rural areas.

06
Potential explanations for the rapid spread
of mobile
There are, then, many potential explanations for the universally
rapid spread of mobile in the developing world. Research has
focussed on the following list:
• the shorter payback period on investment compared to fixed
line; lower installation costs and faster build than fixed line eg.
in India up to six times lower than the estimated $1000
variable cost per additional fixed line;12
straightforward
scalability of mobile compared to other infrastructure
investments.13
• complementary with lower levels of skills than needed for
computers or the internet.14
Especially important for providing
technological access to the poorest people, who are much
more likely to be illiterate and speakers of minority
languages.15
• potentially lower social/income entry barriers than the internet,
due to lower up-front expenditure16
, and also compared to
fixed lines because of greater ease of sharing mobile
handsets.
• business model innovations: pre-pay which helps overcome
credit barriers; the Grameen model of micro-entrepreneurship;
mobiles as public telephones (model found in Bangladesh,
Botswana, India, South Africa, Thailand, Uganda)17
; telecenter
models.18
• network effects which generate rapid momentum once critical
mass is reached.19
• greater ability to overcome geographic hurdles eg mountains,
deserts. Bhutan is an extreme example – the mountainous
state was unsuitable for the installation of fixed line telephony
at all.20
Also less vulnerable to natural disasters than fixed
telecoms. Mobility itself is likely to be valuable for some users,
but less so than in developed economies where mobiles are
complementing extensive fixed networks rather than
substituting for them.
• competition with fixed incumbent, stimulating the growth of
the telecommunications market.21
The poorest developing
countries are still substantially less likely to have reformed
their telecoms markets.22
Competition has knock-on effects to
related influences such as operators’ pricing policies.
• rollout requirements in licences.23
Specific requirements for
rollout in rural and low-income areas are to be found in
Ghana, South Africa and Uganda, for example. In a well-
known example in another region, Chile ran a reverse auction
to subsidise bidders for rolling out services to under-served
areas.24
Many of these favourable factors for the spread of mobile have
been present in many African and other developing countries.
At the same time, as noted above, an explanation is needed for
the differences in penetration rates and usage in different
countries. To sum up, the key explanatory factors here are likely
to include:
economic fundamentals such as income per capita, or relative
prices of handsets and calls (there are high price elasticities of
demand, see below). Macroeconomic stability and urbanisation
also appear to have a significant impact on teledensity;25
policy differences such as regulatory structure and the
competition regime; tariff and non-tariff barriers to imports which
raise the price of handsets; the structure of universal access
obligations; government attitude (are mobiles a dangerous
liberty? A frivolous luxury?);26
social and cultural factors such as urbanisation, trends in rural-
urban or overseas migration,27
women’s security, women’s
empowerment, cultural attitude to communication;28
natural differences such as geography, population density.
Although the economics of mobile make this less of a problem
than for fixed lines, thin population density rapidly escalates the
average cost of extending rollout in rural or remote areas.29
Mobiles and economic growth
The spread of telecommunications should improve growth and
consumer well-being in poor countries. Earlier research suggests
that, as might be expected, telecommunications rollout boosts
growth, with a surprisingly strong effect reported in some
studies.30
This kind of evidence contrasts with the difficulty in
demonstrating a positive link between ICTs in general and an
increase in trend growth in most countries.31
Successful once-
developing countries such as Hong Kong, Korea and Singapore
used telecommunications as a key part of their economic
development strategies.32
More recently Malaysia has placed the
same emphasis on telecoms investments.33
Hardy (1980)34
found
that the impact of telecoms investment was greatest in the least
developed economies and lower in advanced economies, which
is entirely intuitive given the much wider availability of fixed-line
telephony and other complementary technologies in the
developed economies. Roeller and Waverman (2001)35
analysed
only OECD countries and found that there was a critical mass
effect – that the impact of increased telecoms penetration was
especially important at near universal service. (Network effects
may also favour larger markets – South Africa over Botswana,
for example.)36
Roeller and Waverman also attempted to analyse
the experience of the developing economies (in an early 1996
draft, looking at the period 1970-1990), but the data limitations
made the results problematic. However, these results suggested
low impacts of telecom advancement for developing countries,
as they are not near universal service. Nor are developed country
approaches to achieving universal service appropriate for
countries where so little of the population yet has access to
telecommunications, despite the rapid spread of mobile.
The impact of mobiles on growth in developing countries,

07
however, is not covered the earlier literature.37
This is the gap
filled by the next section of this report, by Leonard Waverman,
Meloria Meschi and Melvyn Fuss. They confirm that the growth
impact of mobiles is large in both developed and developing
countries, but around twice as important in the latter group,
where there is also a critical mass effect. The policy implication
of their results for developing countries is clear: it will be worth
investing large amounts in telecommunications to get close to
universal service. As wireless technologies are much lower cost
to roll out over large areas than fixed line systems, mobile can
potentially play a vital role in economic development.
World Bank research suggests the internal rate of return
generated by telecoms investments in developing countries of
around 20%. There is also some evidence that telecoms rollout
is linked with higher levels of foreign direct investment.38
This relationship is explored in the section on FDI in this report
by Mark Williams, which assesses the separate impacts of fixed
line infrastructure and mobiles on FDI.
Other economic and social impacts of
mobile
The existing evidence on other impacts of mobile indicates
positive correlations between teledensity and quality of life
indicators – allowing for GNP per capita – such as longer life
expectancy, lower infant mortality and lower illiteracy (although
such correlations must be treated with great caution given the
existence of simultaneity and omitted variables).39
One measure
of the perceived opportunities and benefits provided by mobile or
by telecommunications in general is the amount consumers are
willing to spend on services. The available evidence is that
telecoms services are very highly valued. In all developing
countries, the average spent on telecommunications is 2% of
monthly expenditure. In a sample of Indian villages, the average
was 3% of household income. In Chile poor people spend more
of their incomes on telecommunications than on water, and even
the average household spends more on telecoms than on water
and electricity combined.40
However, estimates of the price
elasticity of demand are typically quite high, which implies that
high call charges could inhibit mobile penetration and usage in
some developing countries. Income elasticities are also high: one
study in India found a 1% rise in household income almost
doubled demand for telecommunications.41
Waverman et al in
this report also confirm that price and income eleasticities of
demand are high.
There is every reason to believe that the economic and social
returns to mobile will be highest of all in rural areas, which are
consistently less well provided with telecommunications services.
Serving rural areas is also closely linked to anti-poverty efforts.
Half the world’s population – 3 billion people – lives in rural
areas, and there is a substantial overlap between poverty and
rural dwelling. Telephone connectivity appears to be highly
correlated with the extent of the non-farm sector, and
consequently average incomes, in rural areas. A study of 27 Thai
villages found that the only non-agricultural activities took place
in the 18 with Public Call Offices (mostly fixed line); the other 9
had no manufacturing businesses. This is consistent with
findings from other countries from Botswana to Ecuador showing
an improvement in non-farm incomes in rural areas.42
To the extent that mobile communications are reaching some
rural areas with little or no fixed line availability, rural people are
better able to stay in contact with family members. Mobiles are
also improving the flow of information available to would-be
migrants from urban centres or from overseas. Survey evidence
from Bangladesh suggests the main reason for calls made via
GrameenPhone mobiles are financial (queries about remittances,
finding jobs in the city) or family-related (staying in touch with
relatives working elsewhere).43
There can be medical or educational benefits from improved
access to expertise, for example in access to medical advice for
a remote villager. Earlier research documented such impacts of
telephony in the remote areas of developed countries, such as
Canada and Australia. Likewise, previous studies on telephony
looked at the importance of social contact for people living in
remote and lightly populated areas – such as the Australian
outback. Researchers suggest this is particularly important
for women.44
There are now several studies documenting the improvement in
prices received by farmers as a result of better access to
telephony in general and mobile in particular, in developing
countries in Asia, Africa and Latin America. One particularly nice
example is the case of fishermen in India using mobile phones to
get information about prices at different ports before deciding
where to land their catch.45
This specific example was confirmed
in a study of fishermen on Mafia Island, off the Tanzanian coast,
where the Vodafone Foundation partners the WWF in a marine
project.46
The improved flow of information evidently reduces monopsony
power in agricultural markets – especially non-commodity
markets such as perishable fruits, where prices were not already
published in newspapers. The impact of an improved information
flow thanks to better telecommunications ought to be apparent
in the dispersion of prices for the same product in different parts
of the same national or regional market. If information flows are
poor, the ‘law of one price’ will not operate: the market will not
work well, and middle-men will be able to discriminate between
different suppliers or customers (although competition amongst
middlemen can limit this). There is evidence from the historical
record that the telegraph and telephone reduced the dispersion
of agricultural prices, and raised farm incomes, in the United
States in the 19th and 20th centuries. Earlier work demonstrated
the impact of the development of a long-distance fixed line

08
network in the creation of a credit market for coffee growers in
Ethiopia.47
The same phenomenon of reduced farm price
dispersion has been documented in China recently.48
Just as in developed countries, mobiles save time and enable
efficiencies in business, especially in terms of coping with
unexpected events (taxis responding to customers in the cities,
dealing with a blocked road or an accident when making a
delivery). There is plentiful anecdotal evidence of this kind, much
of it to be found in newspaper coverage. A few studies report
similar findings – for example, reduced emergency response
times.49
In a few countries – notably China – mobiles are being
used for e-commerce: home shopping or trading in shares.50
Studies also report law enforcement benefits from the ability to
contact police quickly. In Bangladesh for example, law
enforcement agencies give GrameenPhone some credit for
reduced rural crime rates. There are other examples of mobiles
being used to improve security and thus efficiency – for
example, maize farmers in the Democratic Republic of Congo
have provided phones to security guards, increasing their yields
significantly through reducing looting.51
The two final sections of this report contribute to this research
on social and economic impacts of mobiles using the results of
surveys on the use of mobile carried out for Vodafone in rural
communities in South Africa and Tanzania, and of small
businesses in Egypt and South Africa. The community surveys
assess the factors affecting mobile use, and the range of
potential impacts, in relatively poor, rural African communities.
As Jonathan Samuel, Niraj Shah and Wenona Hadingham report
below, the surveys suggest that mobile telephony is frequently
accessed by the poorest people, thanks in part to widespread
sharing. The surveys suggest that gender, age and education do
not present insurmountable barriers to access – nor even the
absence of electricity. Individuals surveyed in rural communities
highlighted savings in travel time and costs and easier
communication with family and friends, in addition to access to
business information and easier job search. A majority of small
businesses reported increased sales and profits, time savings
and greater efficiency. For many black-owned businesses in
Cairo, a mobile phone was the only means of communication
available. The final section of this report, by James Goodman,
looks specifically at the implications of the survey results for
social capital, or the strength of social networks and contacts in
the rural communities. Mobile phone ownership in the
communities surveyed was positively linked to life satisfaction
and a willingness to help others. A clear majority of respondents
said owning a mobile had improved their relationship with family
members living elsewhere.
The studies included here represent the early stages of
Vodafone’s SIM programme, which will continue to contribute to
the growing body of evidence. As more data and more research
become available, it will be important for policy makers and
anybody interested in social and economic development in Africa
to understand the impact of the extraordinary spread of mobile.
All references in this section are to the bibliography at the end of
the report.

09
Notes
1
UNDP HDR database.
2
ITU 2003, ITU 2004, Kirkman and Sachs, World Bank2000, World Economic Forum 2003.
(All references are to Mobile Bibliography).
3
Figures in this paragraph from ITU 2004.
4
DFID (2004); see also www.infodev.org.
5
ITU 2003.
6
Gebreab (2002)
7
The Kingdom of Morocco and Vivendi Universal agreed on November 18, 2004, to the
acquisition by Vivendi Universal of 16% of the capital of Maroc Telecom. The agreement
allows Vivendi Universal to increase its stake from 35% to 51%, thereby perpetuating its
control over the company. Payment for this transaction was made in January 2005.
8
See also Eggleston et al (2002), Forestier et al (2002), World Economic Forum (2003).
9
Grace et al (2001). See also ITU World Telecommunications Indicators 2004, Chapter 4
on the Millennium Development Goals.
10
Grace et al (2001).
11
Forestier et al (2002), Navas-Sabater (2002), Rodriguez and Wilson (2000).
12
Kenny (2002)
13
Dholakia and Kshetri (2002)
14
Mansell (2001), Qiang et al
15
Kenny (2002)
16
Forestier et al, Kenny (2002)
17
Bruns et al (1996), ITU (2002), Navas-Sabater et al (2002).
18
Latcham and Walker (2001), Proenza (2001)
19
Grajek (2003); see also Roeller and Waverman (2001).
20
Dorj (2001)
21
Azam et al – Senega (2002); Bruns et al – Thailand (1996); Forrestier et al (2002);
Gebreab – Africa (2002); ITU (1999) – Bangladesh; Laffont et al – Cote D’Ivoire (2002);
Rossotto et al (2000) – MENA; Rossotto et al (2003); UNDP; Wallsten (1999) – Africa and
Latin America,
22
Beardsley et al. (2002)
23
ITU (2003), Navas-Sabater (2002).
24
Kenny (2002), Wellenius (2001)
25
Forestier et al (2002);
26
Lopez (2000).
27
Bruns et al (1996),
28
Dholakia and Kshetri (2002).
29
Dorj (2001), Kenny (2002).
30
See Röller and Waverman for a careful study using OECD data – this refers back to the
older literature; Madden and Savage (1998) find a stronger result for Central and Eastern
Europe; Nadiri and Nandi (2003) also find a strong link for developing countries.
31
OECD 2003.
32
Saunders et al (2003).
33
Riaz (1997).
34
Hardy (1980).
35
Roeller and Waverman (2001).
36
Qiang et al
37
An exception is Jha, R and S. Majumdar (1999).
38
Mansell (2001), Navas-Sabater (2002).
39
Forestier et al (2002), Kenny, UNDP (2002).
40
Navas-Sabater (2002), Wellenius (2000); Blattman et al (2002); De Melo cited in
Forestier (2002).
41
Grajek (2003), ITU (2003); Blattman et al (2002).
42
Bruns et al (1996); Duncombe and Heeks (1999), Forestier et al (2002).
43
Bruns et a (1996); Bayes et al (1999).
44
Bayes et al (1999), Hammond (2001); Hudson (1995)
45
Bruns et al (1996); Forestier et al (2002); Hudson (1995); ITU (1999); Lopez (2000).
Dholakia and Kshetri (2002).
46
http://www.wwf.org.uk/annualreview/2003-2004/business.asp
47
Hirschman, referenced in Forestier et al (2002).
48
Hudson (1995); Eggleston et al (2002).
49
Bruns et al (1996), Schwartz (2001.)
50
Dholakia and Kshetri (2002), Laperrouza (2002).
51
Bayes et al (1999); Lopez (2000).

Introduction
There is a long tradition of economic research on the impact of
infrastructure investments and social overhead capital on
economic growth. Studies have successfully measured the
growth dividend of investment in telecommunications
infrastructure in developed economies.2
But few have assessed
the impact of telecommunications rollout in developing countries.
Given the importance of telecommunications to participation in
the modern world economy, we seek to fill the void in existing
research. Investment in telecoms generates a growth dividend
because the spread of telecommunications reduces costs of
interaction, expands market boundaries, and enormously
expands information flows. Modern revolutions in management
such as ‘just-in-time’ production rely completely on efficient
ubiquitous communications networks. These networks are recent
developments. The work by Roeller and Waverman (2001)
suggests that in the OECD, the spread of modern fixed-line
telecoms networks alone was responsible for one third of output
growth between 1970 and 1990.
Developing countries experience a low telecoms trap – the lack
of networks and access in many villages increases costs, and
reduces opportunities because information is difficult to gather.
In turn, the resulting low incomes restrict the ability to pay for
infrastructure rollout.
In the OECD economies, modern fixed-line networks took a long
time to develop. Access to homes and firms requires physical lines
to be built – a slow and expensive process. France, which had 8
fixed line telephones per 100 population (the ‘penetration rate’) in
The Impact of Telecoms on Economic
Growth in Developing Countries
Professor & Chair of Economics, London Business SchoolLeonard Waverman
10
Associate Professor of Economics, John Cabot University,
Rome and Affiliate, LECG, LondonMeloria Meschi
Professor of Economics, University of TorontoMelvyn Fuss1

11
1970, doubled this by 1976, and reached 30 main lines per 100
population in 1980. Mobile phones are lower cost and far quicker
to rollout than fixed lines. In 1995, Morocco had 4 fixed lines per
100 inhabitants after many years of slow investment, and zero
mobile phones per 100 inhabitants. In 2003, only eight years
later, the mobile phone penetration rate in Morocco was 24, while
fixed line penetration had stagnated at its 1995 level.
We find that mobile phones in less developed economies are
playing the same crucial role that fixed telephony played in the
richer economies in the 1970s and 1980s. Mobile phones
substitute for fixed lines in poor countries, but complement fixed
lines in rich countries, implying that they have a stronger growth
impact in poor countries. Many countries with under-developed
fixed-line networks have achieved rapid mobile telephony growth
with much less investment than fixed-line networks would have
needed.
We subjected the impact of telecoms rollout on economic
growth in poorer nations to a thorough empirical scrutiny. We
employed two different approaches – the Annual Production
Function (APF) approach following the work of Roeller and
Waverman (2001) and the Endogenous Technical Change (ETC)
approach similar to the work of Robert Barro (1991). The latter
provided us with the most robust and sensible estimates of the
impact of mobile telephony on economic growth. We used data
on 92 countries, high income and low income, from 1980 to
2003, and tested whether the introduction and rollout of mobile
phone networks added to growth.
We find that mobile telephony has a positive and significant
impact on economic growth, and this impact may be twice as
large in developing countries compared to developed countries.
This result concurs with intuition. Developed economies by and
large had fully articulated fixed-line networks in 1996. Even so,
the addition of mobile networks had significant value-added in the
developed world: the value-added of mobility and the inclusion of
disenfranchised consumers through pay-as-you-go plans
unavailable for fixed lines. In developing countries, we find that the
growth dividend is far larger because here mobile phones provide,
by and large, the main communications networks; hence they
supplant the information-gathering role of fixed-line systems.
The growth dividend of increasing mobile phone penetration in
developing countries is therefore substantial. All else equal, the
Philippines (a penetration rate of 27 percent in 2003) might
enjoy annual average per capita income growth of as much as
1 percent higher than Indonesia (a penetration rate of 8.7
percent in 2003) owing solely to the greater diffusion of mobile
telephones, were this gap in mobile penetration to be sustained
for some time. A developing country which had an average of
10 more mobile phones per 100 population between 1996 and
2003 would have enjoyed per capita GDP growth that was 0.59
percent higher than an otherwise identical country.
For high-income countries, mobile telephones also provide a
significant growth dividend during the same time period.
Sweden, for example, had an average mobile penetration rate of
64 per 100 inhabitants during the 1996 to 2003 period, the
highest penetration of mobiles observed. In that same period,
Canada had a 26 per 100 average mobile penetration rate.
All else equal, we estimate that Canada would have enjoyed an
average GDP per capita growth rate nearly 1 percent higher than
it actually was, had the mobile penetration rate in Canada been
more-than-doubled.
Our research also provides new estimates of demand elasticities
in developing countries – we find both the own–price and income
elasticities of mobile phone demand to be significantly above 1.
That is, demand increases much more than in proportion to either
increases in income or reductions in price. We also find that
mobile phones are substitutes for fixed-line phones.
Economists have long examined the importance of social overhead
capital (SOC) to economic growth. SOC is generally considered as
expenditures on education, health services, and public
infrastructure: roads, ports, and the like. Telecommunication
infrastructure, whether publicly or privately funded, is a crucial
element of SOC. We in the west tend to forget what everyday life
would be like, absent modern telecommunications systems. These
networks enable the ubiquitous, speedy spread of information.
Alan Greenspan, the Chairman of the US Federal Reserve Board,
coined the term “New Economy” to represent how the spread of
modern information and communications technology has enabled
high growth with low inflation. This “New Economy” is the direct
result of the networked computer – the ability of higher bandwidth
communications systems to allow computer-to-computer
communications.3
The ”New Economy” enables greater
competition and new means of organising production.
In earlier periods, telecommunications networks helped generate
economic growth by enabling firms and individuals to decrease
transaction costs, and firms to widen their markets; Roeller and
Waverman (2001)4
estimated the impact on GDP of investment
in telecoms infrastructure in the OECD between 1970 and 1990.
They showed it significantly enhanced economy-wide output,
allowing for the fact that the demand for telecoms is itself
positively related to GDP. One must remember that in 1970
telecoms penetration was quite low in a number of OECD
countries. While the US and Canada had near-universal service
in 1970, in the same year France, Portugal and Italy for example,
had only 8, 6, and 12 phones per 100 inhabitants respectively.
It is then not surprising that the spread of modern
telecommunications infrastructure between 1970 and 1990
generated economic growth over and above the investment in
the telecoms networks itself.
Roeller and Waverman also demonstrated that the scale of
impact of the increased penetration of telecoms networks on
growth depended on the initial level of penetration, with the
biggest impact occurring near universal service – a phone in
every household and firm. The standard government policy of

12
universal service was, then, not only a question of equity, but
was also implicit recognition of the growth-enhancing properties
of telephony expansion.
In 1995, just under half of the membership of the International
Telecommunications Union (ITU), an international organisation
comprising 214 countries, had telecoms penetration rates
below 8, the level attained by France in 1970. Much of the world
still lacked a major component – the telephone – of a modern,
efficient economic system in 1995.
In the 1970 to 1990 period analysed by Roeller and Waverman
mobile phones were not important: telecoms networks were fixed-
line systems. Today, when we consider telephone networks, the
importance of mobiles stands out, especially when we examine the
102 members of the ITU that had low phone penetration in 1995.
Table 1 lists these countries (i.e., with less than 8 phones per
100 population in 1995, when virtually all phones were fixed
lines) and the penetration rate in 2003 for both fixed lines and
mobiles. The average fixed-line penetration rate of these 102
countries in 1995 was 2.5 phones per 100 population, and this
level was achieved after decades of investment. With the
subsequent rapid growth of mobile phones in many, but not all,
of these countries, the average penetration rate of mobile
phones alone rose to 8 per cent in 2003. In 22 of the 102
countries, mobile penetration reached double digits in 2003.
And in 7 countries, over one-quarter of the population had
mobile phones in 2003 – Albania, Bosnia, Botswana, the
Dominican Republic, Paraguay, the Philippines and Thailand.
The story is clear. In developing countries, modern telecoms
systems are largely mobile systems and not fixed lines. The
reason is the lower cost and faster roll-out of mobile systems as
compared to fixed lines. It has been estimated that a mobile
network costs 50 percent less per connection than fixed lines
and can be rolled out appreciably faster. The cost advantages of
mobile phones as a development tool consist not only of the
lower costs per subscriber but also the smaller scale economies
and greater modularity of mobile systems.
Table 1: The Emergence of Mobile Telephony in 102 Low and Middle-Income Nations
Main lines per Main lines per Mobile Subscribers Mobile Subscribers
100 population 100 population per 100 population per 100 population
Country in 1995 in 2003 in 1995 in 2003
Afghanistan 0 0 0 1
Albania 1 8 0 36
Algeria 4 7 0 5
Angola 0 1 0 ..
Bangladesh 0 1 0 1
Benin 1 1 0 3
Bhutan 1 3 0 1
Bolivia 3 7 0 15
Bosnia and Herzegovina 6 24 0 27
Botswana 4 7 0 30
Burkina Faso 0 1 0 2
Burundi 0 0 0 1
Cambodia 0 0 0 4
Cameroon 0 .. 0 7
Cape Verde 6 16 0 12
Central African Rep. 0 .. 0 1
Chad 0 .. 0 1
China 3 21 0 21
Comoros 1 2 0 0
Congo 1 0 0 9
Congo (Democratic Republic of the) 0 .. 0 2
Cote d'Ivoire 1 1 0 8
Cuba 3 .. 0 ..
Dem. People's Rep. of Korea 2 4 0 ..
Djibouti 1 2 0 3
Dominican Rep. 7 12 1 27
Ecuador 6 12 0 19
Egypt 5 13 0 8
El Salvador 5 12 0 18
Equatorial Guinea 1 2 0 8
Eritrea 0 1 0 0
Ethiopia 0 1 0 0
Gabon 3 3 0 22
Gambia 2 .. 0 ..
Ghana 0 1 0 4
Guatemala 3 .. 0 ..
Guinea 0 0 0 1
Guinea-Bissau 1 1 0 0

Table 1: The Emergence of Mobile Telephony in 102 Low and Middle-Income Nations – continued
Main lines per Main lines per Mobile Subscribers Mobile Subscribers
100 population 100 population per 100 population per 100 population
Country in 1995 in 2003 in 1995 in 2003
Guyana 5 .. 0 ..
Haiti 1 2 0 4
Honduras 3 .. 0 ..
India 1 5 0 2
Indonesia 2 4 0 9
Iraq 3 .. 0 ..
Jordan 7 11 0 24
Kenya 1 1 0 5
Kiribati 3 .. 0 1
Kyrgyzstan 8 .. 0 ..
Lao P.D.R. 0 1 0 2
Lesotho 1 .. 0 ..
Liberia 0 .. 0 ..
Libya 6 14 0 2
Madagascar 0 0 0 2
Malawi 0 1 0 1
Maldives 6 .. 0 ..
Mali 0 .. 0 2
Marshall Islands 7 8 1 1
Mauritania 0 1 0 13
Mayotte 4 .. 0 22
Micronesia (Fed. States of) 7 10 0 5
Mongolia 4 6 0 13
Morocco 4 4 0 24
Mozambique 0 .. 0 2
Myanmar 0 1 0 0
Namibia 5 7 0 12
Nepal 0 2 0 0
Nicaragua 2 4 0 9
Niger 0 .. 0 0
Nigeria 0 1 0 3
Oman 8 .. 0 ..
Pakistan 2 3 0 2
Palestine 3 9 1 13
Papua New Guinea 1 .. 0 ..
Paraguay 3 5 0 30
Peru 5 7 0 11
Philippines 2 4 1 27
Rwanda 0 .. 0 2
Samoa 5 7 0 6
Sao Tome and Principe 2 5 0 3
Senegal 1 2 0 6
Sierra Leone 0 .. 0 ..
Solomon Islands 2 1 0 0
Somalia 0 .. 0 ..
Sri Lanka 1 5 0 7
Sudan 0 3 0 2
Swaziland 2 4 0 8
Syria 7 .. 0 ..
Tajikistan 4 4 0 1
Tanzania 0 0 0 3
Thailand 6 10 2 39
Togo 1 1 0 4
Tonga 7 .. 0 ..
Tunisia 6 12 0 19
Turkmenistan 7 .. 0 ..
Tuvalu 5 .. 0 0
Uganda 0 0 0 3
Uzbekistan 7 7 0 1
Vanuatu 3 3 0 4
Viet Nam 1 5 0 3
Yemen 1 .. 0 3
Zambia 1 1 0 2
Zimbabwe 1 3 0 3
Average Fixed Penetration in 1995: 2 Average Fixed Penetration in 2003: 5 Average Mobile Penetration in 1995: 0 Average Mobile Penetration in 2003: 8
13

14
The importance of conveying information
Consider what communicating in France must have been like
35 years ago, in 1970, with only 8 phones per 100 people.
The description of Geertz (1978) as applying to developing
countries, “information is poor, scarce, maldistributed,
inefficiently communicated and intensely valued”5
, must have
applied equally to France. Residents of remote villages with no
phone connections would have enormous difficulty in discovering
prices of commodities. Farmers would not have access to
alternative sources of fertilisers or to alternative buyers of their
products. As recent studies on the use of mobile phones in
South Africa show, the substitute for telecommunicated
information would have been physical transport.6
Instead of a
quick phone call, never mind Internet usage, determining selling
or buying prices would require costly, time-consuming physical
contacts and transport. Thus without telecommunications, the
costs of information retrieval and of transacting in general would
be high. Besides greater transaction costs, the range of supply
would be much smaller, or for transactions across large
distances, risks would be higher as prices and conditions of sale
would not be known exactly. Modern telecom networks, then,
are crucial forms of Social Overhead Capital. But how important
are they?
There are two basic ways in which economists determine the
extent of the economic growth impact of some factor such as
increased education or telecoms infrastructure investment –
aggregate production function (APF) estimation and the
endogenous technical change (ETC) approach.
In the first approach – the APF – the level of economy-wide
Gross Domestic Product (GDP) each year is assumed to be
determined by that year’s aggregate capital, aggregate labour,
and other specific factors such as education or the spread of
telecommunications. The growth dividend of telecoms would be
measured by its annual contribution to GDP growth. The second
approach – the ETC – relates the average rate of growth of GDP
over a substantial period (we use the 24-year period 1980 to
2003) to the initial level of GDP, average investment as a share
of GDP during that period, the initial stock of labour represented
in terms of its educational attainment7
, and the initial or average
telephone penetration rate. The contribution of telecoms to
growth is here measured by its boost to the long-term growth
rate. The ETC approach is not an average over time of the APF
approach, as the two models rest on different theoretical
underpinnings.
Empirically, the two methods differ as well: the production
function approach uses annual data, so errors or missing
observations cause significant difficulties. The endogenous
technical change approach uses period averages and initial
period values instead, and it is thus less prone to data errors.
Given the paucity of reliable data in developing countries, the
ETC approach should prove more robust and tractable.
Because demand for telecoms services rises with wealth, it is
crucial in the APF approach to disentangle two effects – the
impact of increased telecoms rollout on economic growth and
the impact of rising GDP itself on the demand for telecoms. This
is called the two-way causality issue, or ‘endogeneity’, as the
demand for telecoms is itself dependent on the level of GDP.
Hence estimating an APF alone would lead to biased and likely
exaggerated measures of the growth dividend of telecoms.
This endogeneity problem is handled in Roeller-Waverman by
developing a four-equation model: the first equation is the output
equation or economy-wide production function; the second
equation determines the demand for telecoms; a third equation
determines the investment in telecoms infrastructure and a final
equation relates investment to increased rollout. In this model,
the explicit causality from GDP to demand is recognised in
equation two, allowing any estimated effect of telecoms on
growth (equation one) to be net of the demand-inducing effects
of rising GDP.
The two-way causality problem cannot be dealt with explicitly in
the endogenous growth model approach but is unlikely to be a
central issue. One cannot, for example, add a demand equation
defined as the average demand over the period. Instead one has
to use data analysis, instrumental variables and statistical tests
to determine whether there is any reverse causality present.8
Existing literature
The notion that telecoms infrastructure is an important part of
SOC is not new. Various researchers beginning with Hardy9
in
1980, Norton10
in 1992 and others11
have all found that there is
an “externality” component in enhanced fixed telecoms
penetration – that is, GDP is higher, and growth faster in
countries with more advanced telecoms networks. Of course, as
noted, one has to worry about reverse causality in richer
countries; there, as income rises, demand for luxuries such as a
universal telephone service rises as well. Although these studies
do not adjust for reverse causality, several facts bear out the
existence of the telecoms externality. First, Hardy examined both
radio and telephone rollouts, since if the telephone simply
provides information, radio broadcasts might be good
alternatives. Hardy found no significant impact of radio rollout on
economic growth, in contrast to telephones. Secondly,
telephones (unlike radios, for example) have strong network
effects – the value of a telephone to an individual increases with
the number of other telephone subscribers.
Hence, as networks grow, their social value rises. This suggests
that the social return – the value to society of an additional
person connected or of an additional dollar invested in the
network – exceeds the private return to the network provider, if
that provider cannot price so as to extract these externality
values. The Roeller-Waverman paper shows strong network

15
effects. In the OECD in from 1970 to 1990, incremental
increases in penetration rates below universal service levels
generated only small growth dividends. Only at near universal
service (30 mainline phones per 100 inhabitants which is near
70 or so mainline phones per 100 households) were there
strong growth externalities from telephone rollout.
Several more recent papers extend this analysis to mobile
phones – among these are Torero, Choudhary and Bedi12
(2002)
and Sridhar and Sridhar13
(2004). Several points need to be
made on this research. First, for economies without many fixed
lines, or where mobiles supplement low fixed-line rollout, there
should be no inherent difference in the growth dividend of a
phone, whether it is mobile or fixed. In developing countries, an
additional phone, whether fixed or mobile, increases the small
network size and adds to the economy’s growth potential.
Secondly, where mobile phones complement fixed lines (in
advanced economies), their externality effects will probably be
different from those found for fixed lines. As individual lifestyles
change and as firms utilise mobiles in productivity-enhancing
ways, we should see new economic growth from mobile
networks as well. For penetration rates of fixed lines are not 100
percent in developed economies. For example, in the USA in
1995, the penetration rate was 60 phones per 100 people.
Mobile phones move the developed economies closer to
universal service because pre-pay contracts allow exact
monitoring of use, something very difficult to manage with fixed-
line phones, making them accessible to other groups of users.
Some of the recent empirical studies specifically examine the
impact of mobile phone expansion on growth in developing
countries, using the Roeller-Waverman (RW) framework.
Three caveats must be mentioned here. First, in many of these
countries, growth has been low due to a host of issues – poor
governance, lack of capital, low skill levels, and the like. It is
difficult to show that mobile telephony increases growth rates
where growth is low. Secondly, advances in telecoms penetration
rates in developing countries are recent, so there is little real
trend as yet. Finally, since mobiles are so new, there has been
extremely rapid growth in mobile penetration starting from zero.
Thus, if one tries to explain economic growth by changes in
capital, labour, education and mobile phones, one could find
either that all economic growth is due to the explosive growth in
mobile phones, or conversely that mobile phones decrease
growth since their use increases so quickly with little underlying
economic growth occurring. Good econometrics requires careful
consideration of underlying facts.
Sridhar and Sridhar (2004) apply the RW Framework to data for
28 developing countries over the twelve-year period 1990 to
2001. The average compounded annual growth rate (CAGR) of
GDP per capita in this period was minus 2.03 per cent, while the
CAGR of mainlines was 6.60 and of mobile phones 78.0
percent. In their regression, they find that mobile phones explain
all growth – a 1 percent increase in mobile phone penetration
increases growth by 6.75 percent. Below, we provide our own
analyses of the RW aggregate production function approach.
We do find more plausible although still exceedingly high
impacts of mobile phones on growth. But the result is not robust
to alternative specifications or to changes in countries included
in the sample, and we do not rely on these estimates to draw
any conclusions. We provide the APF model also to show the
demand equation estimates – these are also most interesting,
and robust.
The Aggregate Production Function
In order to estimate the impact of mobile phones in developing
countries, we gathered information from the World Bank’s World
Development Indicators (WDI) database for basic variables such
as GDP, population, labour force, capital stock and so on for
both low-income and lower-middle-income countries. The
International Telecommunication Union (ITU) produces a World
Telecommunications Indicators database, updated annually, and
we used this for data on our major telecoms-related variables –
such as revenue, investment, and subscriber numbers. We also
relied on the World Bank’s Governance Indicators, so that we
could incorporate some measures of institutional quality, which
most certainly has an impact on growth. We included 38
developing countries for which full data are available for the
period we used, which is 1996 to 2003.14
The framework employed was a three-equation modification of
the Roeller-Waverman approach. Appendix A provides further
details. We summarise briefly the model that we used:
1. The Output equation models the level of output (GDP) as a
function of the total physical capital stock net of telecoms
capital, the total labour force, a variable that captures the
extent of the “rule of law”, and the mobile telecoms
penetration rate. To account for the fact that output generally
increases over time, we included a time trend term. We also
included indicator variables capturing the level of external
indebtedness of the country (there were three levels – High,
Medium and Low). Roeller and Waverman used a dummy
variable for each country (a so-called “fixed effects” or “Least
Square Dummy Variables” approach). This variable controls
for unobservable characteristics or omissions from the
equation that are peculiar to each country; our approach here
is similar in spirit, since it captures the impact of particular
characteristics (such as the indebtedness level) on output.15
2. The Demand equation models the level of mobile telecoms
penetration as a function of income (the level of GDP per
capita), mobile price (revenue per mobile subscriber), and the
fixed-line price (which is revenue per fixed-line subscriber).
The demand equation also allows for a time trend, since
demand for a new product such as mobiles could also
feature a strong trend.

16
3. The Investment equation simplifies the Roeller-Waverman
“supply” and “investment” equations. It assumes that the
growth rate of mobile penetration depends on the price of
telecoms (the relationship should be positive since higher
prices should invite additional supply), the geographic area
(the relationship should be negative), and a time trend term.
We estimated the system of equations described above using the
Generalised Method of Moments (GMM) method.16
This approach
uses all the exogenous variables in the system of equations (i.e.,
those that we can reasonably assume are not determined by the
other variables in the system, such as the amount of labour and
the amount of total capital) as “instruments” for the endogenous
variables (output, the level of mobile and fixed penetration, and
the mobile and fixed prices).17
The results for the output and demand equations from running
this GMM regression are summarised in Tables 2 and 4
respectively (see Appendix A for the full set of results):
Table 2: Output equation (dependent variable is log of
output)
Variable Coefficient T-Statistic
Capital 0.776 13.79
Labour 0.204 3.91
Mobile Penetration18
0.075 3.60
The coefficients obtained above are encouraging at first glance.
The coefficients on capital and labour sum to close to 1, which is
roughly consistent with the standard hypothesis of constant
returns-to-scale for the economy as a whole. The coefficient of
the log of mobile penetration (which is a transformed version of
the original variable) is 0.075. However, the interpretation of this
is not straightforward: the impact of penetration on output
depends on the level of penetration. Table 3 shows the average
levels of mobile penetration and GDP in those countries that the
ITU classifies as “Low Income” and “Lower-Middle-Income” for
1996 and 2002 respectively.19
For the average country, with a
mobile penetration of 7.84 phones per 100 population in 2002,
the coefficient of 0.075 on the transformed mobile penetration
variable implies that a doubling of mobile penetration would lead
to a 10 percent rise in output, holding all else constant.
Table 3: Mobile Penetration and GDP for “average”
developing country, 1996-2002.
Year Mobile Penetration GDP
1996 0.22 $41 billion
2002 7.84 $47 billion
Considering that the average CAGR of GDP in these nations has
been roughly 2 percent, this seems to high an estimate of the
impact of mobile penetration. A growth rate of GDP of 2 percent
over 8 years for the average country would imply total
(compounded) growth of 19 percent. Meanwhile, the average
CAGR of mobiles has been 64 percent in these same countries:
mobile penetration more than doubles every two years in the
average country. Given the estimated impact of mobile
penetration presented in Table 2, if a developing country were
enjoying “typical” growth rates of GDP and mobile telephones,
then increased mobile penetration explains all the growth over
the sample period.
The problem here is the one of weak output growth in many of
the countries, but robust growth in mobile phone penetration.
The model does not adequately control for the other factors
affecting growth in the economy.20
We attempted to extend the
sample – both by adding more countries and increasing the time
period back to 1980,21
and also to modify the specification
somewhat, but the results did not prove robust to either changes
in the sample or changes in the model specification.
On the other hand, the demand equation from the aggregate
production function model always performed well. Table 4 shows
the results of the GMM estimation for the demand equation:
Table 4: Demand equation (dependent variable is mobile
penetration)
Mobile Price -1.50 -6.06
Fixed-line price 0.31 2.79
GDP per Capita 1.95 23.30
Table 4 shows that mobile demand falls when the price of
mobiles increases, but increases when the price of fixed lines
increases, suggesting that there is substitution between fixed
line telephony and mobiles. Mobile demand is also strongly
positively correlated with increases in income. The equation
is in double-log form so the coefficients can be interpreted as
elasticities of demand, at the average penetration rate.
The own-price-elasticity of mobile phones is minus 1.5, which
implies that demand is elastic: a 10 percent price increase
would reduce demand by roughly 11.6 percent for a country in
which mobile penetration is about 8 percent, the average level of
mobile penetration for the developing countries.22
The cross-price
elasticity between mobile and fixed lines is positive, indicating
that in these countries, mobiles and fixed telephones are
substitutes: an increase in the price of fixed-line phones by
10 percent increases the demand for mobiles by 2.4 percent,
assuming mobile penetration at the “average” level of 8 percent.
Moreover, mobiles are ‘luxuries’ (in the technical sense) as the
income elasticity is significantly above one – for the “average”
developing country with 8 percent mobile penetration, a 1
percent increase in per capita GDP is associated with a 1.5
percent increase in the level of mobile penetration. The structure

17
of the demand equation is much simpler than that of the output
equation and since the equation deals with demand for one
particular characteristic – mobile penetration – it is relatively
easier to capture the factors that affect this demand than it is to
capture all the factors which serve to increase or reduce output
over time.
Ultimately, though, in light of the problems with the APF
approach, especially the significant difficulties of obtaining
adequate data across a large group of developing countries, we
turn to the endogenous growth model.
The Endogenous Growth model
We follow the work of Barro,23
who ran growth regressions for a
cross-section of countries for the time period 1960 to 1985.
The basic questions Barro was addressing were two-fold: was
there ‘convergence’ between rates of growth between poorer
and richer countries as economic theory predicts; and how did
differences in skill levels affect growth rates? Barro took average
growth rates of per capita GDP for a cross-section of 98
countries and regressed these growth rates against regressors
which included initial levels of GDP per capita and human capital
stock,24
the average government consumption to GDP ratio for
the period 1970-1985, and measures of stability.25
Barro found that, conditional on the initial human capital stock,
average GDP per capita growth was negatively correlated with
initial GDP per capita.26
Thus, all else equal, poorer countries
should close the income gap with richer countries, albeit only
over long periods of time. The initial level of human capital stock
was positively correlated with GDP per capita growth, so
countries that were initially rich might actually grow faster than
poorer countries if there were sizeable differences in their initial
endowments of human capital. Only by controlling for these
differences could he verify that there is indeed economic
convergence between richer and poorer nations.
Our approach is similar. We took the average growth rate of per
capita GDP from 1980 to 2003 as our dependent variable, and
regressed this average growth rate on variables which included
the initial level of GDP, the average ratio of investment to GDP,
the stock of telecoms in 1980 (measured by the level of fixed-
line penetration in 1980), the proportion of the 15-and-above
population that had completed at least primary schooling in
1980, and the average level of mobile penetration for the period
1996 to 2003 (the period in which mobile penetration increased
rapidly). Our sample consisted of 92 countries – developing and
developed alike. The data came from the same sources – the
World Development Indicators and the ITU – that we used for the
APF estimation.
We are not primarily examining the issue of ‘convergence’ in
income levels but instead in whether the increase in mobile
penetration increases growth rates, and whether it does so
equally in rich and poor countries. As mobile growth starts in
essentially the same recent period for all countries, rich and
poor alike, this is an interesting and important question. Our
hypothesis is that increased mobile rollout should have a greater
effect in developing countries than in rich countries. The reason
is simple: while in developing countries the benefits of mobile
are two-fold – the increase in the network effect of telecoms
plus the advantage of mobility – in developed economies the first
effect is much more muted.
In this model, there are no mobile phones in 1980, as there is
for other stock variables (e.g., we have proxied the stock of
human capital in 1980, and have included the stock of telecom
capital in 1980). We can assume that the 1980 levels of human
and telecom capital are exogenous – that is, they ought not to
be the result of income growth between 1980 and 2003.27
We cannot, however, assume that there is no reverse causality
between income growth in the 1980 to 2003 period and average
mobile penetration over a portion of the same period with quite
the same safety. Thus, mobile penetration is potentially
endogenous, and we must examine whether or not this is so.
We started with an initial specification that did not attempt to
capture differential effects of telecoms between developing and
developed countries. Table 5 (also reported in fuller form in
Appendix B) reports the results of a simple Ordinary Least
Squares (OLS) regression:28
Table 5: Baseline results from the ETC model (dependent
variable is average per capita GDP growth)
GDP80 -0.0026 -4.00
K8003 0.0017 4.73
TPEN80 0.0418 1.63
MPEN9603 0.0003 2.76
APC1580 0.0002 2.43
Constant -0.0289 -3.93
Table 5 shows that the average GDP growth rate between 1980
and 2003 was positively correlated with the average share of
investment in GDP (taken over the entire period), with the 1980
level of primary school completion, and with the average level of
mobile penetration between 1996 and 2003. It was negatively
correlated with the level of initial GDP per capita (GDP80).
The results confirm Barro’s convergence hypothesis: conditional
on other factors such as human capital and physical capital
endowments (captured by school completion rates and telecom
penetration), poorer countries grow faster than richer ones.
Every additional $1,000 of initial per capita GDP reduces
average growth by roughly 0.026 percent. Considering that
average growth is typically in the 1 to 2 percent range, a

18
$10,000 difference in initial per capita GDP would imply growth
that would be 0.26 percent lower, which is a substantial
difference in the light of typical rates of growth.
The initial level of telecoms (i.e., fixed line) penetration was not
significant in this model (TPEN80). However, the average level of
mobiles penetration (MPEN9603) was significant – a unit
increase in mobile penetration increased growth by 0.039
percent, all else being equal. In line with Barro, the coefficient
on primary school completion (APC1580) was positive
and significant.
As mentioned above, we were concerned about a potential
problem of endogeneity of the mobile penetration rate (as a
regressor). We performed a Hausman test,29
which showed that
endogeneity was not likely to be an issue.30
(See Appendix B for
fuller details of the IV estimates and the Hausman test).
Having tested for endogeneity, we then divided the sample into
four income quartiles according to their level of GDP per capita
in 1980. We classified countries as “low income” (or potentially
fast-growth) if they were in quartiles 1, 2 or 3, while quartile 4
countries were classified as “high income.” Our “low income”
sample included a mix of some countries that had (and still have)
much catching-up relative to the highest-income nation, and
some countries (like Hong Kong) that were on the verge of
becoming advanced economies in 1980. We created dummy
variables for high and low income countries and then split the
effects of penetration by generating new variables that were the
product of these dummy variables and initial telecoms
penetration, and the dummy variables and average mobile
penetration from 1996 onwards. Table 6 (reported also in
Appendix B) illustrates the results:
Table 6: Table 5 regression separating out effect of
telecoms variables
GDP80 -0.0025 -3.68
K8003 0.0018 4.67
TPENH80 0.0005 1.92
TPENL80 -0.0002 -0.32
MPENL 0.0006 2.46
MPENH 0.0003 1.99
APC1580 0.0002 2.22
Constant -0.0284 -3.83
Here, we found that the effect of initial telecoms stock in 1980
was not significant for the low-income countries (TPENL80) but
was almost significant (at the 5 percent level) for high-income
countries.31
This is to be expected in view of the fact that fixed
penetration was extremely low for low-income countries in 1980
(an average of 3.3 main telephone lines per 100 inhabitants).
The coefficient on the average mobile penetration from 1996 to
2003 (MPENL for low-income countries and MPENH for high-
income countries) was positive and significant for both cases,
but the impact was twice as large for the low-income countries.
The results suggest a noticeable growth dividend from the
spread of mobile phones in low-income and middle-income
countries.
All else equal, in the “low income” sample32
, a country with an
average of 10 more mobile phones for every 100 people would
have enjoyed a per capita GDP growth higher by 0.59 percent.
Indeed, the results suggest that long-run growth in the
Philippines could be as much as 1 percent higher than in
Indonesia, were the gap in mobile penetration evident in 2003 to
be maintained. The Philippines had 27 mobile phones per 100
inhabitants in 2003, compared to 9 per 100 in Indonesia.
Another estimate of the importance of mobiles to growth can be
seen by comparing Morocco to the “average” developing country.
In 2003, Morocco had 24 mobile phones per 100 inhabitants,
compared to 8 in the typical developing country. Were this gap in
mobile penetration maintained, then Morocco’s long-run per
capita growth rate would be 0.95 percent higher than the
developing country average.33
Thus, current differences in mobile
penetration between developing countries might generate
significant long-run growth benefits for the mobile leaders.
Finally, while Argentina and South Africa both had disappointing
economic performance over the 1980 to 2003 period, both
registering negative average growth in per capita incomes, the
analysis suggests that South Africa’s higher level of mobile
telecoms penetration over the period (17 for South Africa versus
11.4 for Argentina) prevented this difference from being even
larger – South Africa’s negative average per capita growth of 0.5
percent compares with Argentina’s negative average per capita
growth of 0.3 percent, but this difference would have been 0.3
percent wider had it not been for the greater spread of mobiles
in South Africa.
For the high-income countries, mobile telephones still provide a
significant growth dividend. Sweden, for example, had an
average mobile penetration rate of 64 per 100 inhabitants during
the 1996 to 2003 period, whilst Canada had a mere 26 per 100
average penetration rate. All else equal, Canada would have
enjoyed an average GDP per capita growth rate 1 percent higher
than it actually registered, had it been able to achieve Swedish
levels of mobile penetration over the 1996 to 2003 period.
Conclusions
In summary, telecommunications is an important prerequisite for
participation in the modern economic universe. There is a long-
standing literature attempting to gauge the economic impact of
telecommunications, with the findings of Roeller and Waverman
(2001) suggesting a substantial growth dividend in OECD
nations.

19
We have modelled the impact of mobile telecommunications in
poorer countries, since in these countries mobile phones are
fulfilling the same role as fixed lines did previously in the OECD
nations. Initially we attempted to use the Roeller-Waverman
framework, but data problems and econometric problems made
it difficult to get truly sensible estimates of the growth impact of
mobile telecommunications that were also robust to changes in
the sample and small changes in the specification of the model.
We turned to a purely cross-sectional model that looked at long-
term averages of growth, and our results were more robust and
sensible than under the previous framework.34
They suggest the
following:
• Differences in the penetration and diffusion of mobile
telephony certainly appear to explain some of the differences
in growth rates between developing countries. If gaps in
mobile telecoms penetration between countries persist, then
our results suggest that this gap will feed into a significant
difference in their growth rates in future.
• As Romer (1986) and Barro (1991) hypothesised for human
capital stocks, there are also increasing returns to the
endowment of telecoms capital (as measured by the telecoms
penetration rate).
• Given the speed with which mobile telecoms have spread in
developing nations, it is unlikely that large gaps in penetration
will persist for ever. However, differences in the speed of
adoption will affect the speed with which poor countries
converge to rich countries’ level. Relative poverty still poses
serious political problems, such as instability and increased
demand for emigration. Our analysis suggests the need for
regulatory policies that favour competition and encourage the
speediest possible rollout of mobile telephony.
Notes
1
London Business School and LECG; John Cabot University and LECG; University of
Toronto and LECG. Funding for this research was provided by Vodafone and the
Leverhulme Trust. We thank Kalyan Dasgupta for sterling assistance. We are indebted to
Mark Schankerman for suggesting the use of an endogenous growth approach.
2
These studies include Hardy (1980), Norton (1992), and Roeller and Waverman (2001).
Full bibliographical details are given in footnotes 8, 9 and 3 respectively.
3
The “Networked Computer” is the focus of a major research programme at London
Business School funded by the Leverhulme Trust.
4
Roeller, Lars-Hendrik and Waverman, Leonard. “Telecommunications Infrastructure and
Economic Development: A Simultaneous Approach.” American Economic Review, 2001,
91(4), pp.909-23.
5
Geertz, Clifford. “The Bazaar Economy: Information and Search in Peasant Marketing.”
American Economic Review, 1978, 68(2), pp.28-32.
6
See (for example) World Resources Institute. Digital Dividends Case Study: Vodacom
Community Phone Shops in South Africa, www.digitaldividend.org
7
In this, we follow the endogenous growth literature, which postulates increasing returns to
human capital.
8
The data requirements of the full 4 equation APF model are much larger than for the one
equation endogenous growth model.
9
Hardy, Andrew. “The Role of the Telephone in Economic Development.”
Telecommunications Policy, 1980, 4(4), pp. 278-86.
10
Norton, Seth W. “Transaction Costs, Telecommunications, and the Microeconomics of
Macroeconomic Growth.” Economic Development and Cultural Change, 1992, 41(1),
pp. 175-96.
11
Among these others are Leff, Nathaniel H. “Externalities, Information Costs, and Social
Benefit-Cost Analysis for Economic Development: An Example from Telecommunications.”
Economic Development and Cultural Change, 1984, 32(2), pp. 255-76. And Greenstein,
Shane and Spiller, Pablo T. “Estimating the Welfare Effects of Digital Infrastructure.”
National Bureau of Economic Research (Cambridge, MA) Working Paper No. 5770, 1996.
12
Torero, Maximo; Chowdhury, Shyamal and Bedi, Arjun S. “Telecommunications
Infrastructure and Economic Growth: A Cross-Country Analysis.” Mimeo, 2002.
13
Sridhar, Kala S. and Sridhar, Varadharajan. “Telecommunications Infrastructure and
Economic Growth: Evidence from Developing Countries, National Institute of Public
Finance and Policy (New Delhi, India) Working Paper No. 14, 2004
14
Since the production function approach is so data-intensive, the sample used in this
regression consisted of 38 countries and 260 observations. Even from this sample, 95
observations were eliminated in the course of the regression analysis due to missing
data. Of these 38 countries, 19 are low income countries (Bangladesh, Benin, Burkina-
Faso, Central African Republic, Cote d’Ivoire, Gambia, India, Indonesia, Kenya, Lesotho,
Madagascar, Mali, Mozambique, Myanmar, Nepal, Pakistan, Senegal, Tanzania and
Vietnam) and 19 are lower middle income countries (Armenia, Bolivia, Brazil, China,
Colombia, Egypt, Fiji, Iran, Jordan, Morocco, Namibia, Peru, Philippines, South Africa, Sri
Lanka, Swaziland, Thailand, Tunisia, and Turkey).
15
Because we had very few observations for some of the countries in the sample, a model
with full fixed effects collapsed.
16
GMM estimation offers some advantages in terms of efficient estimation and ability to
correct for serial correlation over other methods available for estimating a model
comprised of a system of equations.
17
Instrumenting the endogenous variables essentially involves isolating that component of
the given endogenous variable that is explained by the exogenous variables in the system
(the “instruments”), and then using this component as a regressor.
18
Following Roeller-Waverman, we used a transformed and “unbounded” version of the
penetration variable, namely (PEN/0.35-PEN) in the regression analysis. We do so to
increase the range of the observed penetration rates.
19
It should be noted that this is a larger set of countries than we were able to include in our
actual regression analysis.
20
Appendix A shows the sign on the time-trend term is negative and statistically significant,
implying that there is large-scale technological regression: unlikely and troublesome. This
also suggests that the mobile penetration rate variable is explaining too much growth.
21
Since there were no mobiles in 1980, we ran a model for the effects of total telecoms
penetration with the demand equation adjusted so that both fixed lines and mobile
demand are estimated when mobile penetration is non-zero.
22
Since we use a transformed version of mobile penetration, the impact of an increase in
GDP per capita or increase in the price level varies according to the level of mobile
penetration.
23
Barro, Robert J. “Economic Growth in a Cross Section of Countries.” The Quarterly
Journal of Economics, 1991, 106(2), pp. 407-43.
24
Measured by school enrolment rates in 1960.
25
The average numbers of revolutions per year and assassinations per million population
during the sample period.
26
Standard neoclassical growth theory predicts long-run convergence of income levels
between countries as richer, more capital-intensive countries run into the problem that
the returns to capital diminish beyond a certain level of capital intensity. In the later
growth literature, initiated by Romer (1986), there are increasing returns to particular
factors- such as human capital- that also play a significant role in determining the speed
of convergence. See Romer, Paul M. “Increasing Returns and Long-Run Growth.” Journal
of Political Economy, 1986, 94(5), pp.1002-37.
27
However, it is possible that these variables proxy for subsequent flows of income into
human and telecom capital, a subtlety that Barro (1991) explored for human capital, and
rejected.
28
All results are corrected for heteroscedasticity.
29
Loosely speaking, the Hausman test computes the “distance” between an estimator that
is potentially inconsistent under the alternative hypothesis of endogeneity bias and one
that is always consistent. See Hausman, Jerry. “Specification Tests in Econometrics.”
Econometrica, 1978, 46(2), pp. 1251-71.
30
In this context, the Hausman test compares the OLS estimates with estimates from an
instrumental variables regression (IV). We used average fixed line penetration between
1960 and 1979 as an instrument for average mobiles penetration between 1996 and
2003: the correlation coefficient between the two variables is 0.81.
31
This is also consistent with Roeller and Waverman (2001) who report an inability to derive
consistent results for low-income countries.
32
Because data for more advanced countries is more widely available, and because we only
treated the very advanced nations (top quartile) of 1980 as “high income”, our “low
income” sample probably underweights the most underperforming developing country.
Developing countries and overweights middle-income countries. Clearly, better data
availability – particularly of historical data – would enable us to expand our sample and
thereby gauge how robust our results really are.
33
It should be noted that Morocco is not part of the sample from which our results were
actually derived.
34
However, we need to examine whether our sample can be expanded, and while we have
tested for the endogeneity of the mobile phones penetration variable, we still need to
examine some more subtle issues such as the potential endogeneity of some of the other
regressors. We also need to test for the possibility that some third factor (such as
institutional quality) that we have not captured influences both growth and the level of
mobile penetration, thereby generating a spurious relationship between the two.

20
Variable Variable Description
GDP Real GDP in constant 1995 dollars
PC_GDP Real GDP per capita in constant 1995 dollars
POP Population
TLF Total Labour Force
GA Geographic Area (square kilometres)
TTI Total telcommunications investment in constant 1995 dollars
LAW "Rule of Law"
K1 Physical capital stock (net of telecoms capital) in constant 1995 dollars
MPEN Penetration rate of mobile telecoms (expressed per 100 inhabitants)
MTELP "Price" of mobile telecoms measured as revenues per mobile subscriber (converted to constant 1995 dollars)
FPEN Penetration rate of fixed telecoms (expressed per 100 inhabitants)
FTELP "Price" of fixed telecoms measured as revenues per telephone subscriber
T Time (starting with 1996=1)
Appendix A: The Production Function
Approach
Sources of Data: World Development Indicators (available from the World Bank website), World Bank Governance Indicators (1996-2002) and International Telecommunication Union (ITU),
World Telecommunications Indicators, 2004 CD-ROM.
1. Overview of Data
N Mean St. Dev Min Max
GDP 255 90234.8 192231.2 317.4 1095347.2
PC_GDP 255 1.1 1.0 0.1 5.3
POP 255 100879359 256181098 774000 1312709294
TLF 260 49.7 137.2 0.3 769.3
GA 260 1090.1 1963.4 10.0 9327.4
LAW 260 -0.3 0.5 -1.6 1.2
K1 243 221514.7 460234.2 719.7 3066821.0
TTI 237 1103.0 3523.6 0.1 27629.4
MPEN 260 3.3 5.7 0.0 34.8
MTELP 216 359.9 295.6 20.2 1897.7
FPEN 260 5.6 6.1 0.2 28.5
FTELP 231 518.6 314.6 18.8 1626.5
Table I: Summary Statistics
2. The Production Function Model
The three-equation model that we employ is:
log y=a1.(HIGHDEBT )+a2.(LOWDEBT )+a3.(MEDDEBT )+a4.log(K1)+a5.log(TLF )+
+a6 .log(MPEN)+a7.(LAW )+a8.(t )+U
log(MPEN)=b0+b1.log(GDP_PC)+b2.log(TELP)+b3.log(FTELP)+b4.(t )+U
log(MPENt )–log(MPENt-1)=c0+c1.log(GA)+c2.log(TELP)+c3.(t )+U

21
The results from this regression are reported in Table II (variable names starting with “A” correspond to first equation, “B” to second
equation, and “C” to third equation):
Variable Coefficient Standard Error T-Statistic
AHIGHDEBT 1.544122 0.5984 2.58
AMEDDEBT 1.536928 0.5746 2.67
ALOWDEBT 1.705664 0.5771 2.96
AK1 0.776639 0.0563 13.79
ATLF 0.204081 0.0522 3.91
APEN 0.075426 0.0210 3.60
ALAW 0.060486 0.0656 0.92
AT -0.08871 0.0239 -3.71
B0 1.60262 1.7523 0.91
BGDP 1.951197 0.0837 23.30
BTELP -1.49887 0.2475 -6.06
BFTELP 0.312194 0.1121 2.79
BT 0.492504 0.0765 6.44
C0 -1.50804 0.5285 -2.85
CTELP 0.358958 0.0820 4.38
CGA -0.03535 0.0149 -2.37
CT 0.096033 0.0220 4.37
Table II: Summary of Regression Results
Variable Variable Description
GDP8003 Average growth rate of real GDP per capita (in constant 1995 International Dollars at
Purchasing Power Parity) over the 1980-2003 period.
GDP80 Level of real GDP per capita in 1980 (in 000s of Dollars)
K8003 Average share of investment in GDP for the 1980-2003 period
TPEN80 Level of telecoms (i.e., fixed) penetration in 1980 expressed in terms of telephones per 100 inhabitants
MPEN9603 Level of mobile penetration averaged over the 1996-2003 period expressed in terms
of subscribers per 100 inhabitants
APC1580 Proportion of 15 and over population who had completed at least Primary School in 1980
TPEN80H Variable obtained by multiplying high income dummy with TPEN1980
TPEN80L Variable obtained by multiplying low income dummy with TPEN1980
MPENH Variable obtained by multiplying high income dummy with MPEN9603
MPENL Variable obtained by multiplying low income dummy with MPEN9603
FPEN6079 Average level of fixed telecoms penetration during the 1960-79 period, used to instrument MPEN9603
Appendix B: The Endogenous Growth
Model
Sources of data: GDP and Investment Share from the World Development Indicators; telecoms data from the International Telecommunication Union (ITU). World Telecommunications
Indicators (2004), and data on education from the Barro-Lee dataset (updated to 2000) available from various websites, including www.nber.org.
1. Overview of Data

22
N Mean St. Dev Min Max
GDP8003 92 0.01 0.02 -0.05 0.08
GDP80 92 6.97 6.25 0.56 23.26
K8003 92 20.86 5.48 9.32 45.88
TPEN80 92 10.22 14.11 0.06 58.00
TPEN80L 69 3.32 5.07 0.06 25.37
TPEN80H 22 31.84 11.10 11.42 58.00
MPEN9603 92 19.23 20.95 0.05 67.32
MPENL 69 10.27 14.40 0.05 67.32
MPENH 23 46.10 12.96 16.99 64.99
APC1580 92 45.13 25.97 4.00 97.00
FPEN6079 90 7.52 10.66 0.05 47.67
Table I: Summary statistics for main variables
2. The Endogenous Technical Change Model
The basic specification for our Endogenous Technical Change model is:
GDP8003=a0+a1.(GDP80)+a2.(I/Y8003)+a3.(TPEN80)+a4.(MPEN9603)+a5.(APC1980)+u
Hausman test: H0: OLS is consistent and efficient under the null hypothesis, IV is consistent
H1: OLS is inconsistent, IV is consistent under the alternative.
Result: ( B
^
OLS – B
^
IV ) ‘( V ) –1
( B
^
OLS – B
^
IV ) = 0.34 , P-value=0.9967. Fails to reject H0.
The second specification that we employ is as follows:
GDP8003=a0+a1.(GDP80)+a2.(I/Y8003)+a3.(TPEN80).(LOW)+a4.(TPEN80).(1–LOW)+a5.(MPEN9603).(LOW)+
+a6.(MPEN9603).(1–LOW)+a7(APC1980)+u
GDP80 -0.0026386 0.0006591 -4.00
K8003 0.0017272 0.000365 4.73
TPEN80 0.0418567 0.0256544 1.63
MPEN9603 0.0003851 0.0001397 2.76
APC1580 0.0002249 0.0000927 2.43
Constant -0.0289961 0.0073738 -3.93
Table II: Basic specification, OLS regression
R-Squared=0.545, n=91.
GDP80 -0.0026519 0.0007095 -3.74
K8003 0.0017125 0.0003349 5.11
TPEN80 0.0004352 0.0003466 1.26
MPEN9603 0.0003699 0.0003801 0.97
APC1580 0.000232 0.000103 2.25
Constant -0.0288258 0.0071842 -4.01
Table III: Basic specification, IV regression (Instrument for MPEN9603 is FPEN6079).
R-squared=0.5450, n=89.

23
GDP80 -0.0025463 0.000692 -3.68
K8003 0.0016998 0.0003642 4.67
TPENH80 0.0005329 0.0002769 1.92
TPENL80 -0.0002023 0.000625 -0.32
MPENL 0.0005942 0.0002414 2.46
MPENH 0.0002924 0.0001466 1.99
APC1580 0.0002127 0.0000959 2.22
Constant -0.0284366 0.0074336 -3.83
Table IV: Regression with penetration effects split according to income group
R-squared=0.5501, n=91.
Note: Countries that were ranked in quartiles 1, 2 and 3
according to GDP per capita in 1980 were “low” income,
quartile 4 countries were “high” income.
Countries in the endogenous growth
regression sample
Algeria, Argentina, Australia, Austria, Bahrain, Bangladesh,
Barbados, Belgium, Benin, Bolivia, Botswana, Brazil, Bulgaria,
Cameroon, Canada, Central African Republic, Chile, China,
Colombia, Costa Rica, Cyprus, Denmark, Dominican Republic,
Ecuador, Egypt, El Salvador, Fiji, Finland, France, The Gambia,
Germany, Ghana, Greece, Guatemala, Honduras, Hong Kong,
Hungary, Iceland, India, Indonesia, Iran, Ireland, Israel, Italy,
Jamaica, Japan, Jordan, Kenya, Kuwait, Lesotho, Malawi,
Malaysia, Mali, Mauritius, Mexico, Mozambique, Nepal,
Netherlands, New Zealand, Nicaragua, Niger, Norway, Pakistan,
Panama, Paraguay, Peru, Philippines, Portugal, Rwanda, Senegal,
Sierra Leone, Singapore, South Africa, Spain, Sri Lanka, Sudan,
Swaziland, Sweden, Switzerland, Syria, Thailand, Togo, Trinidad
and Tobago, Tunisia, Turkey, United Kingdom, United States,
Uruguay, Venezuela, Zaire/Congo (DR), Zambia, Zimbabwe.

24
Introduction
Higher investment is central to achieving long-term sustainable
economic growth and poverty reduction in developing countries.
Foreign investors are often seen as an important source of
capital finance and some types of foreign investment may also
bring spill-over benefits to the recipient country in the form of
transfer of skills, tax revenues and formal employment.
Understanding the determinants of the level of foreign
investment therefore has potentially important policy
implications.
In this study, we investigated the relationship between one type
of foreign investment – Foreign Direct Investment (FDI) – and the
characteristics of the recipient countries. We have focused, in
particular, on the relationship between FDI flows into developing
countries and the penetration of mobile telecommunications
networks in the recipient country. We found that both fixed and
mobile communications networks, in addition to other
characteristics including openness of the economy, GDP and
infrastructure, are positively linked with inward FDI; and the
impact of mobile has grown more significant in recent years.
The determinants of FDI
Capital flows from abroad fall into two categories: official finance
and private finance. The private flows in turn can be divided into
three categories: loans from banks or other private sector
lenders; portfolio capital flows for the purchase of securities such
as bonds and equities; and foreign direct investment, overseas
capital invested as equity in businesses in the recipient country.
FDI involves a long-term relationship between the investor and
the entity in which the investment is made and often includes
some management control.1
In practice, FDI includes a range of
different activities and transactions. The privatisation of state-
owned firms in developing countries is often included, as are
programmes of investment in branches or subsidiaries of
transnational corporations (TNCs). Another major type of FDI
particularly important in Africa is related to concessions for
exploring and developing natural resources such as oil, gas or
mineral reserves.
The volume of FDI varies significantly between countries and
regions, as shown in Figure 1, with poorer regions generally
attracting the least inward investment.
Figure 1: Foreign direct investment, net inflows
(% GDP, 2002)
Source: WDI (2004), Frontier Economics2
The country groupings in the figure are based on the following definitions: Least developed
countries, UN definition; Middle Income Countries, World Bank definition – GNI per capita
(2003) between $765 and $9,385;(Non-) OECD high income countries, World Bank definition
– GNI per capita (2003) greater than or equal to $9,386.
Even within these country groups there is substantial variation
in the amount of FDI between countries. Figure 2 shows the
FDI inflows for each of the countries included in this study.
The sample includes 32 of the 48 countries in Sub-Saharan
Africa, and 39 other less developed countries.3
Frontier EconomicsMark Williams
Thanks to Reamonn Lydon and George Houpis who also
worked on this project.
Mobile networks and Foreign Direct
Investment in developing countries

25
Figure 2: Foreign direct investment by country,
net inflows (% GDP 2002)
Source: WDI (2004), Frontier Economics
FDI flows also vary dramatically over time. This is shown in
Figure 3, which demonstrates clearly the relationship between
FDI flows and the global economic upturn of the late 1990s, and
the subsequent decline in FDI.
Figure 3: Foreign direct investment, inflows over time by
country grouping 1977 – 2002
The evidence on the impact of FDI in developing countries is
mixed. Its developmental impact depends on the form of the
investment, the sector of the economy concerned, and the policy
environment in the host country. Even so, it is generally accepted
that FDI can have a number of positive effects on the economies
of developing countries. It can increase formal sector
employment in countries where it is often scarce. Research
indicates that access to employment in the formal sector is the
most important factor in shifting poor people out of poverty4
.
FDI usually involves the transfer of skilled personnel to the
destination country. Companies also employ and train significant
numbers of local staff. FDI is therefore often associated with the
transfer of new technologies and skills to nationals of the
destination country, which helps to raise productivity and
incomes. It also involves medium to long-term commitments by
foreign investors. Their investments are tied up with physical
capital (plant and machinery, fixed assets etc.). It is therefore
harder for the investor to withdraw than in the case of portfolio
investments. This reduces the volatility of foreign exchange
movements and helps to limit exchange rate fluctuations.
In many developing countries, capital is scarce because there is
very little domestic saving and access to international financial
markets is either limited or non-existent. FDI in such cases can
provide a vital source of capital. There is also some evidence to
suggest that FDI stimulates domestic investment in developing
countries5
. Lastly, foreign-owned enterprises in developing
countries are often significant sources of tax revenue in
countries where public finance is often severely constrained.
These potential benefits mean governments in many developing
countries have gone to considerable efforts to attract FDI.
However, some countries have been more successful in this than
others. Understanding the causes of this variation and the factors
that influence the levels of FDI is therefore an important issue for
developing countries. Recent research on this question has been
based on statistical (regression) analysis, using data from a large
number of countries over a number of years, to assess the
empirical importance a range of potential determinants of FDI
flows. Each of the potential determinants is included as an
explanatory variable in the regression analysis.
The majority of the empirical studies focus on average net FDI
flows, specified as FDI/GDP in order to take account the impact
of the scale of the host country. Furthermore, as FDI tends to
vary significantly from year to year, studies using historical data
have generally analysed average FDI over a number of years.
Morisset’s study (2000) also takes account of the natural
resource endowment of the country. 6
Most studies consider explanatory variables including: measures
of economic openness (the importance of trade); the extent and
quality of infrastructure; GDP; GDP growth; indicators of political
stability; and measures of macroeconomic stability.
Several studies have also investigated the relationship between
additional specific variables and the level of FDI flows. For
example, Asiedu (2002) includes a measure of the return on
capital and Morisset looks at the impact of illiteracy and the
degree of urbanisation. Of special interest here, Reynolds et al.
(2004) look at the impact of telecommunications. They note that
telecommunications infrastructure is closely linked to GDP and
therefore look at the impact of unusually high levels of telephone
infrastructure on FDI flows.
Despite similar analytical frameworks, in general the results of
the analysis are mixed and vary significantly between studies
depending on the periods chosen and the specification of the
regression equations. Table 1 summarises the results of
previous studies on the determinants of FDI.
0
2
4
6
8
10
12
14
Guinea
Bangladesh
Haiti
Nepal
Madagascar
BurkinaFaso
Malawi
Zimbabwe
Niger
Kenya
CentralAfricanRepublic
Panama
Guatemala
Guinea-Bissau
Congo,Dem.Rep.
India
SierraLeone
Mauritius
Botswana
SouthAfrica
Thailand
Egypt,ArabRep.
Venezuela,RB
Argentina
Ghana
Cameroon
Morocco
Mauritania
Pakistan
Philippines
ElSalvador
SriLanka
Uruguay
Benin
PapuaNewGuinea
Senegal
Algeria
Coted'Ivoire
Honduras
Mexico
CapeVerde
Gabon
Colombia
Tanzania
Uganda
Chile
Nigeria
Mali
Malaysia
Brazil
Tunisia
Swaziland
China
CostaRica
Peru
Nicaragua
Ecuador
Zambia
Togo
Guyana
Jamaica
Singapore
TrinidadandTobago
Bolivia
Grenada
Congo,Rep.
Mozambique
Gambia,The
FDI/GDP (%) 2002
0
1
2
3
4
5
6
7
8
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
FDI, % GDP
Middle income countries
OECD high income countries
Regression sample
SSA

26
Table 1: Determinants of FDI
Source: Asiedu (2002)
Morrisset finds that GDP growth and trade openness have both
been correlated with FDI, over and above the impact of GDP and
natural resources. Political stability, illiteracy and infrastructure
(as proxied by the number of telephone lines) are not significant
in all specifications. Asiedu also finds that openness to trade is
positively associated with FDI and finds a positive relationship
between FDI and infrastructure in non-Sub Saharan Africa (SSA).
She finds that FDI is generally lower in SSA than in other regions
and also finds that the effect of most of the other explanatory
variables is lower in SSA than in non-SSA regions. Reynolds et al
(2003) focus their analysis on the impact of telephone lines on
FDI flows and find that having more mainlines than would be
expected, given the size of the economy, is linked to a higher
level of FDI.
The variables which emerge as unambiguously positively related
to FDI flows are economic openness and infrastructure (although
infrastructure is statistically insignificant in some studies). In all
the cited studies, the quality and extent of infrastructure is
proxied by the number of main telephone lines per 1000
population. No research that we are aware of has attempted to
disaggregate between the impact of the different types of
infrastructure (e.g. transport, energy, communications). By
studying the impact of mobile networks on FDI into developing
countries, our work is therefore a natural extension of the body
of existing research.
Mobile networks and FDI in developing
countries
Mobile penetration in developing countries has increased
dramatically during the past 10 years, partly as a result of the
liberalisation of telecommunications markets. This is shown in
Table 2 and Figure 4.
Determinant of FDI/GDP Positive Negative Insignificant
Openness Edwards (1990)
Gastanaga et al (1998)
Hausmann and
Fernadez-Arias (2000)
Infrastructure quality Wheeler and Mody (1992) Tsai (1994)
Kumar (1994) Loree and Loree and Guisinger (1995)
Guisinger (1995) Lipsey (1999)
Real GDP per capita Schneider and Frey (1985) Edwards (1990) Lore and Guisinger (1995)
Tsai (1994) Lipsey (1999) Jaspersen, Aylward, Wei (2000) Hausmann and
and Knox (2000) Fernandez-Arias (2000)
Labour cost Wheeler and Mody (1992) Schneider and Tsai (1994)
Frey (1985) Loree and Guisinger (1995)
Lipsey (1999)
Taxes and tariffs Loree and Guisinger (1995) Wheeler and Mody (1992)
Gastanaga et al (1998) Lipsey (1999)
Wei (2000)
Political instability Schneider and Frey (1985) Lore and Guisinger (1995)
Edwards (1990) Jaspersen, Aylward,
and Knox (2000)
Fernandez-Arias (2000)

27
Figure 4: Growth in overall mobile penetration,
indexed (1995 = 100)
Our interest is in the possible links between the penetration
of mobile networks and FDI flows into developing countries.
This relationship is illustrated in a simple way in Figure 5,
which plots average FDI per capita between 1998 and 2002
against mobile penetration rates in a number of developing
countries.
Figure 5: FDI inflows per capita 1998-2002 average
The figure indicates that there is a positive link between mobile
penetration and FDI, but in order to probe further, we tested the
following relationship:
Where:
Net FDI = net inflow of FDI;
GDP = Gross Domestic Product; and
Variables = a range of possible explanatory
variables, including mobile penetration8
.
We included a wide range of possible explanatory variables, in a
number of different combinations, in the regression, using data
on the value of net FDI and the other variables for the period
1993 to 2002.
We also ran the regressions for different time periods within
this span to explore the impact of the period chosen on the
parameter values, as growth in mobile networks accelerated
in most developing countries towards the end of the 1990s.
There were several other methodological issues. FDI values
typically vary significantly from year to year, particularly in
developing countries. The data can be dominated by flows
relating to specific large projects. For this reason, most studies
are based on data averaged over several years, although this
has the disadvantage of reducing the number of data points in
the analysis. We explored the effect on the results of using
different periods for averaging, in addition to using data for 2002
only. A further difficulty is that some of the explanatory variables
are correlated with each other. For example, there is a close
statistical relationship between the penetration rate of fixed lines
and indicators of the quality and extent of the road network.
A similar relationship may also exist between fixed-line
penetration and other general indicators of the quality of a
country’s infrastructure. This can cause difficulties in both
estimating and interpreting the value of regression coefficients,
as it is difficult to separate the effects of closely related variables.
( )Variablesf
GDP
NetFDI
=
0
2000
4000
6000
8000
10000
12000
14000
16000
18000
1995 1996 1997 1998 1999 2000 2001 2002
Percentage growth in
mobile phone
subscribers (mobile
phones per 1000
people), 1995 = 100
Middle Income Countries
Least Developed Countries
OECD high income countries
Regression sample
SSA
0
50
100
150
200
250
300
350
400
450
500
0 50 100 150 200 250 300
Mobilies per 1,000 people
FDI per capita (current
$US)
Grenada
Trinidad &
Tobago
Argentina
Chile
Table 2: Growth in Mobile penetration by country grouping, 1995-2000
Mobile phones per Mobile phones per Average annual
1,000 population 1,000 population growth rate (%)
1995 2002 1995 - 2002
Least Developed Countries 0.13 21.88 109%
SSA 0.74 61.68 90%
Middle Income Countries 5.73 191.29 66%
OECD high income countries 87.33 765.01 37%
Regression sample 5.28 122.83 58%

28
Our analysis is based on a data for developing countries9
, and
we identified separately the countries that are in Sub-Saharan
Africa (SSA). Some researchers10
also control for the level of
natural resources in these countries by including a measure of
natural resources in the regression. We also repeated all of the
regressions in this section for measures of FDI, normalised to
account for the impact of natural resource endowments on FDI
flows11
. In general we found that the main results are robust to
these alternative specifications. The complete list of countries
included in the regression is given in Annexe 1.
Results
As noted above, many of the existing studies find that the
openness of an economy12
is positively related to net FDI inflows.
This result is not surprising. Foreign companies may be investing
in developing countries with the intention of exporting the
products. Countries with open economies are therefore likely
to attract more foreign investment for this type of production.
An alternative explanation is that the openness of an economy
is related to the quality of general economic management, and
well-managed economies attract FDI. We confirmed that there is
a stable, statistically significant and positive relationship between
economic openness and net FDI inflows. This effect is present in
most regression specifications and the value of the coefficient
remains stable. This robustness is a good indication that economic
openness is indeed significantly related to FDI. It is also consistent
with the results of the other studies.13
We next looked at the significance of fixed line penetration as an
explanatory variable for FDI flows, both on its own and together
with measures of mobile penetration. We found that there is a
significant relationship between fixed-line penetration rates and
FDI inflows in many different specifications. In the specifications
which averaged flows across periods (i.e. 1993-2002 and
2000-2002), we found that the statistical relationship between
fixed lines and FDI was significant and positive. In regressions
that included fixed penetration as the only indicator of
telecommunications coverage (with mobile penetration excluded),
we found that a 1 per cent increase in fixed line penetration was
associated with 1-1.3 per cent higher rates of average FDI. This
parameter was statistically significant and relatively consistent
across model specifications14
.
We then looked at the relationship between mobile penetration
and net FDI inflows. When we included mobile penetration rates
but excluded fixed line penetration rates, we found a statistically
significant relationship between mobile networks and FDI flows in
the later period of the sample (i.e. 2000-2002 and 2002
alone)15
. This indicated that a 1 per cent increase in mobile
penetration rates is associated with 0.5-0.6 per cent higher
rates of FDI/GDP. However, we did not find a similar relationship
when we included data from the earlier period (1993-1999).
This is as we would expect since mobile networks did not
develop significantly during this period.
When we included both fixed and mobile penetration rates
separately in the regression, we found that mobiles are not
statistically significant.
In general, the coefficient on fixed penetration rates in our
analysis was higher than for mobile rates. This result may reflect
the fact that the fixed penetration rate variable is capturing some
of the effect of other (non-telecommunications) infrastructure.
For example, Figure 6 shows the clear relationship between fixed
line penetration rates and the quality of the road infrastructure in
the countries under consideration.
Figure 6: Relationship between telephone mainlines and
road quality
Source: WDI (2004), Frontier Economics. The sample of 169 countries includes both
developing and developed countries. The data refers to mainline penetration and road
infrastructure in 1999, the latest year for which data is available in the WDI (2004).
This relationship is even stronger for countries in Sub-Saharan
Africa. It is therefore very likely that the coefficient on fixed line
penetration is reflecting in part the effect of these other forms of
infrastructure. However, it is not possible to separate these
factors in the analysis because of the lack of data. When the
sum of the two penetration rates, fixed and mobile, is included
this does a better job than either fixed penetration alone, or
mobile penetration alone in explaining FDI. However, direct
comparisons of the significance of fixed and mobile networks on
FDI flows into developing countries should be treated with
caution. At present, given the data available, it is not possible to
use regression analysis to separate the effects of all the different
types of infrastructure on FDI.
We extended our basic analysis by exploring four alternatives
estimating the relationship in differences; looking at Sub-
Saharan Africa only; investigating the impact of natural resources
on FDI; and controlling for endogeneity (that is, the possibility
that higher ratio of FDI to GDP itself leads to greater mobile
penetration) in the regression analysis. The analysis of
differences16
(that is, looking at the changes in the variables
rather than their levels) found a significant relationship between
FDI inflows and mobile penetration. This analysis also indicated
that the effect of fixed lines was statistically insignificant. It is
likely that this is because mainline penetration typically did not
change significantly in many developing countries during the
0
1
2
3
4
5
6
7
8
0 20 40 60 80 100
% of the national road network paved
Mainlines per
1000 population
(log)

29
period. However, it should be noted that this analysis is sensitive
to the period over which it is carried out.
The results of carrying out the analysis for the Sub-Saharan
African countries alone are shown in Annexe 3. The sample
consists of only 32 countries so the results need to be
interpreted cautiously. However, compared with the full sample of
around 70 countries, the effect of fixed line penetration on FDI
flows is significantly smaller. Whereas in the full sample we
observe significant coefficients in the range of 1.0-1.3, the
coefficients for the SSA sample, while also significant, are in the
range of 0.6-0.9. Secondly, the size of the positive relationship
between mobile penetration and FDI flows is greater when we
look at the sub-sample of SSA countries in some periods. This
is particularly the case for 2002, where the coefficient doubles
from 0.5 to 1.017
. Overall, the tentative conclusion from the
analysis of the SSA sample is that telecoms infrastructure is
positively correlated with FDI flows. However, relative to other
developing countries, fixed line penetration is less important,
and mobile penetration is more important. This is consistent
with the observed weaknesses of fixed line networks in many
Sub-Saharan African countries.
We estimated the regression analysis using ‘normalised FDI’18
as a means of controlling for the impact of natural resource
endowments such as oil in the recipient country. The results
from this analysis are shown in Annexe 419
. This analysis does
not significantly change the conclusions from our basic analysis.
Telecoms infrastructure remains positively correlated with FDI
flows. In this case, the statistical significance of mobile
penetration is reduced20
. In contrast to the basic regression
results, we find here that natural resources have a significant
and positive impact on FDI flows into SSA. This is a different
result from that found by Asiedu (2002), who finds that the ‘fixed
effect’ for SSA is negative. Our results imply that an important
difference between developing countries in Africa and outside
Africa, in terms of attracting FDI, is in the relationship between
FDI flows and the value of natural resources. However, we find
that, even taking account of this effect, the positive relationship
between mobile telecoms and FDI remains significant.
Finally, we considered the impact of endogeneity, which will arise
if FDI inflows in turn affect any of the variables we are using to
try to explain FDI. If the regression contains some endogenous
variables, then the coefficients on these variables will be biased.
It might be the case that FDI could be affected by mobile
penetration rates and mobile penetration rates simultaneously
affected by GDP, which is in turn a function of FDI. We have
addressed this problem by using a technique known as
Instrumental Variable (IV) estimation, which substitutes the
variable of interest (mobile penetration) with a predicted value,
based on factors known not to be correlated with FDI/GDP.
This means we need to find an ‘instrument’ that affects mobile
phone penetration but which has no impact on FDI flows. If we
use this instrument to ‘predict’ mobile phone penetration, and
subsequently use this predicted value in the FDI/GDP regression,
then the estimated effect of mobile penetration on FDI flows, as
measured by the regression coefficient, is unbiased.
We analysed the effect of growth in mobile penetration on
growth in FDI over the 1998 – 2002 period using fixed line
penetration in 1998 as an instrument for the growth in mobile
penetration between 1998 and 2002. The relationship between
growth in mobile penetration and the number of mainlines per
1,000 population in 1998 is shown in Figure 722
. As we would
expect, in those countries where the 1998 level of fixed line
penetration is low, growth in mobile penetration is significantly
higher, suggesting this is a valid instrument to use in the
FDI regression.
Figure 7: The relationship between growth in mobile
penetration (1998-2002) and mainlines (1998)
Source: WDI 2004, Frontier Economics
The full results are shown in Annexe 5. The first column shows
that the negative relationship between growth in mobile
penetration and the number of mainlines in 1998 is significant
and negative (coefficient –0.299, t-statistic 3.87). The predicted
value for mobile growth from this first stage regression is
included as an explanatory variable in the second regression,
shown in the second column in the table. This shows that mobile
penetration is significantly positively correlated with FDI.
Furthermore, this approach indicates a stronger positive
relationship than the basic results, set out in the third column of
the table. Comparing the results shows that the coefficient
remains statistically significant and has increased from 1.014 to
2.13123
. In other words, the problem of a simultaneous impact of
FDI on mobile penetration means the initial results were biased
downwards in estimating the impact of mobiles on FDI.
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7
Log (mainlines per 1,000 people 1998)
Log mobiles
subscribers per
1,000 (2002) - Log
mobiles per 1,000
(1998)

30
Conclusions
The flow of Foreign Direct Investment into developing countries
depends on a number of variables, including the country’s GDP,
the openness of the economy, and its infrastructure. In the case
of Sub-Saharan Africa, natural resources are an additional
explanatory factor. We have extended earlier findings to show
that mobile telecommunications networks are also positively
correlated with FDI flows.
This relationship appears to be stable across different model
specifications and the impact of mobile on FDI is more
significant in recent years, as mobile penetration in developing
countries has increased dramatically. Taking account of the fact
that mobile penetration may itself be boosted by higher GDP
increases the estimated impact of mobile on FDI. One natural
extension of the analysis would be to explore whether the growth
of mobile networks is related to investment in particular sectors
but sectoral FDI data are unavailable.
References
Asiedu, E. ‘On the Determinants of Foreign Direct Investment to Developing Countries: Is
Africa Different?’, World Development, Vol. 30, No.1 pp 107-119, 2002
Bosworth, B.P. and Collins, S.M. (1999), ‘Capital flows to developing economies: implications
for saving and investment’ Brookings Papers on Economic Activity: 1, Brookings Institution pp
143-169
Jenkins, C. & Thomas, L. ‘Foreign Direct Investment in Southern Africa: Determinants,
Characteristics and Implications for Economic Growth and Poverty Alleviation’, 2002
Morisset, J. ‘Foreign Direct Investment in Africa Policies Also Matter’, Transnational
Corporations Volume 9, Number 2, 2000
Reynolds, R., Kenny C., Liu J., Zhen-Wei Qiang, C. ‘Networking for foreign direct investment:
the telecommunications industry and its effect on investment’, Information Economics and
Policy Vol 16, 159-164, 2004
Notes
1
The precise definitions of FDI vary between countries, usually according to the degree of
share-ownership that is involved.
2
All the data used in this study comes from the World Development Indicators online
database, published annually by the World Bank.
3
In constructing a sample of developing countries for the analysis that follows, we are
constrained by data availability. The full list of countries included in the analysis is shown
in Annexe 1. The average income per capita for the countries included in the sample is
$4,370 in 2002.
4
Jenkins and Thomas (2002)
5
Bosworth and Collins (1999)
6
Morisset does this by calculating a variable referred to as the Foreign Direct Investment
Climate. This is defined as FDI/(GDP*Natural Resource). This is formally equivalent to
assuming that both GDP and the natural resource endowment are determinants of FDI
with an elasticity of one.
7
Since financial markets are either thin or non-existent in most developing countries, it is
difficult to directly measure the returns to capital. Asiedu uses as a proxy for a
measure of the returns to capital. The rationale for this is that GDP/capita is a proxy for
economic output per worker. High GDP/capita is an indication that there are high levels of
capital per worker in the country. This indicates that the returns to capital are relatively
low. In countries with a low GDP/capita, capital is relatively scarce which indicates that
the returns to investment in capital are relatively high. is therefore a proxy for a
measure of the returns to capital.
8
Mobile penetration is measured as number of mobiles per 1000 people. A full description
of each of the variables used in the analysis, along with a detailed list of sources can be
found in Annexe 6.
9
Our full sample of countries includes Low Income, Highly Indebted and Poor and Least
developed countries as defined by the World Bank.
10
See, for example, Morisset (2000).
11
We follow Morisset (2000), in normalising FDI flows by the value of natural resources in
the country in a given year. The value of natural resources is defined as the sum of
output in the primary (agriculture) and secondary (industry) sectors minus output in the
manufacturing sector. Details of the industries included in the primary, secondary and
manufacturing sectors are given in Annexe 6. We have included the normalised measure
of FDI flows as a dependent variable in some of the statistical analysis that follows.
12
Economic openness is defined as (Imports + Exports)/GDP.
13
The regression results presented in Annexe 4, which include a normalised measure of FDI
as the dependent variable in order to control for the impact of natural resources, do not
include either measures of openness or the return on investments (1/GDP) as explanatory
variables. This is because the effects of these variables on FDI flows are already implicitly
included through the normalisation calculation.
14
This includes using residual values. These are residuals from a regression of fixed line
penetration on GDP/capita. This has the effect of removing the effect of collinearity
between GDP and fixed line penetration. It can also be interpreted as being a measure of
countries with ‘unexpectedly’ high rates of fixed-line penetration. This is the approach
taken by Reynolds et al (2004).
15
The coefficient on mobile penetration and the mobile penetration residual were significant
at the 10% level for the 2000 – 02 averages. However, for 2002, only the mobile
penetration residual was significant at the 10% level.
16
This means that we investigated the relationship between the difference in FDI inflows
between 1998 and 2002 and the difference in the values of the explanatory variables
over the same period.
17
Not statistically significant in earlier periods in some specifications.
18
Normalised for the value of natural resources in a country – see footnote 11.
19
Note that because the dependent variable in this case is a measure of FDI flows
normalised by the product of GDP and the value of natural resources, the magnitude of
the coefficients in Annexe 4 should not be directly compared with those in Annexe 2 and
3. Furthermore, because natural resources, which account for a significant proportion of
trade, and GDP are implicitly included on the left-hand side of the regression, we drop
these variables from the right-hand side of the regression.
20
Mobile penetration is not statistically significant when included in the regression analysis
on its own or together with fixed penetration. However, the coefficients on the residual
values of mobile penetration are significant and positive.
21
Page 108, italics in the original
22
The changes shown in the figure are changes in log variables. If the changes in mobile
phone penetration over this period were small, then the log changes could be interpreted
as percentage changes. However, growth in mobile phone penetration was significant
during this period, therefore the numbers on the Y-axis should not be interpreted as
percentage changes.
23
The implication of this is that the endogeneity was contributing to a negative bias in the
estimate of the effect of mobile penetration on FDI.
1
GDP / capita
1
GDP / capita

31
Source: Frontier economics
Notes: The primary data sources for information on investment flows, as well as data on the
characteristics of each country’s economy, is the World Bank’s World Development Indicators
(2004) and the United Nations Conference on Trade and Development (UNCTAD, 2004).
Country Sub-Saharan Sub-Saharan Sub-Saharan
Africa ? Country Africa ? Country Africa ?
Panama ✘ Egypt, Arab Rep. ✘ Mauritius
Uruguay ✘ Venezuela, RB ✘ Central African Republic
Paraguay ✘ Malaysia ✘ Mauritania
Argentina ✘ El Salvador ✘ Cameroon
Costa Rica ✘ Grenada ✘ Congo, Rep.
Sri Lanka ✘ Nicaragua ✘ Cote d'Ivoire
Ecuador ✘ Jamaica ✘ Burkina Faso
Peru ✘ Pakistan ✘ South Africa
Bolivia ✘ China ✘ Swaziland
Nepal ✘ Philippines ✘ Tanzania
Papua New Guinea ✘ Algeria ✘ Mali
Colombia ✘ Haiti ✘ Kenya
Thailand ✘ Morocco ✘ Nigeria
Mexico ✘ Indonesia ✘ Gabon
India ✘ Congo, Dem. Rep. Botswana
Chile ✘ Niger Uganda
Tunisia ✘ Malawi Cape Verde
Brazil ✘ Senegal Zimbabwe
Bangladesh ✘ Guinea Madagascar
Honduras ✘ Mozambique Guinea-Bissau
Guyana ✘ Togo Zambia
Guatemala ✘ Sierra Leone Gambia, The
Trinidad Tobago ✘ Benin Ghana
Annexe 1: A description of the
countries included in the dataset

32
Specification (1) (2) (3) (4) (5) (6) (7) (8)
Open=((imports 0.026 0.018 0.020 0.026 0.019 0.022 0.027 0.020
+exports)/GDP) (3.72)** (2.19)* (2.52)* (3.99)** (2.39)* (2.49)* (3.80)** (2.51)*
Log (fixed + mobile 1.507 1.477 1.395
subscribers) (3.68)** (2.81)** (2.43)*
Log (fixed lines 1.293 1.035 0.906
per 1000 people) (3.36)** (2.29)* (1.61)
Log (fixed lines per 1000 1.233 1.131
People), residual*** (3.49)** (2.91)**
Log (Mobile subscribers)
Log (mobile subscribers,
residual)
Log (1/GDP per capita) 1.622 1.373 1.562 1.340 0.840 0.986 -0.108 -0.223
(2.83)** (2.03)* (2.03)* (2.53)* (1.47) (1.40) (0.38) (0.89)
Dummy variable for SSA -0.512 0.158 0.248 -0.459 0.401 0.433 -0.737 -0.054
(0.88) (0.25) (0.35) (0.78) (0.53) (0.52) (1.32) (0.09)
Constant 6.886 4.551 5.762 6.048 3.397 4.420 0.322 -0.230
(2.37)* (1.59) (1.76)+ (2.14)* (1.23) (1.33) (0.16) (0.13)
Period Average Average Average Average Average Average Average Average
1993-2002 2000-2002 2002 1993-2002 2000-2002 2002 1993-2002 2000-2002
Observations 69 68 68 69 68 68 69 68
R-squared 0.42 0.26 0.19 0.42 0.24 0.16 0.42 0.26
Robust t statistics in parentheses
+ significant at 10%; * significant at 5%; ** significant at 1%
Annexe 2:
Regression results for all countries

33
(9) (10) (11) (12) (13) (14) (15) (16) (17) (18)
0.021 0.026 0.019 0.022 0.031 0.023 0.025 0.031 0.023 0.025
(2.90)** (3.86)** (2.37)* (2.50)* (3.95)** (2.47)* (2.74)** (3.86)** (2.58)* (2.85)**
1.366 1.019 0.890
(3.11)** (2.18)* (1.56)
1.322
(2.81)**
-0.013 0.370 0.356 0.234 0.577 0.512
(0.05) (1.12) (1.07) (0.89) (1.75)+ (1.49)
0.253 0.514 0.570
(1.03) (1.88)+ (1.82)+
0.042 1.426 1.348 1.478 0.285 0.548 0.739 0.028 -0.077 0.203
(0.16) (2.51)* (2.12)* (1.93)+ (0.56) (0.99) (1.22) (0.09) (0.28) (0.68)
0.127 -0.382 0.374 0.398 -1.451 -0.532 -0.397 -1.460 -0.548 -0.431
(0.21) (0.61) (0.49) (0.47) (2.23)* (0.91) (0.63) (2.25)* (0.94) (0.71)
1.223 6.426 5.719 6.666 2.665 3.157 4.243 1.319 0.813 2.368
(0.62) (1.98)+ (1.80)+ (1.79)+ (0.77) (1.03) (1.27) (0.55) (0.40) (1.07)
Average Average Average Average Average Average Average Average Average Average
2002 1993-2002 2000-2002 2002 1993-2002 2000-2002 2002 1993-2002 2000-2002 2002
68 68 67 67 68 67 67 68 67 67
0.22 0.42 0.27 0.19 0.33 0.21 0.16 0.33 0.22 0.18

34
Specification (1) (2) (3) (4) (5) (6) (7) (8)
Open=((imports 0.034 0.027 0.034 0.035 0.030 0.041 0.033 0.029
+exports)/GDP) (3.47)** (2.30)* (2.02)+ (3.36)** (2.26)* (1.93)+ (3.39)** (2.23)*
Log (fixed + mobile 1.024 1.231 1.535
subscribers) (2.21)* (1.87)+ (2.00)+
Log (fixed lines 0.783 0.482 0.161
per 1000 people) (1.87)+ (0.97) (0.22)
Log (fixed lines per 0.883 0.582
1000 people), residual*** (2.96)** (1.82)+
residual)
Log (1/GDP per capita) 1.901 1.808 2.539 1.618 0.939 0.991 0.699 0.388
(3.00)** (1.88)+ (2.17)* (2.72)* (1.14) (0.92) (1.84)+ (0.94)
Constant 8.674 7.487 10.531 7.760 4.850 5.145 4.124 2.814
(2.81)** (1.91)+ (2.08)* (2.55)* (1.22) (0.94) (2.05)+ (1.25)
Country fixed effects No No No No No No No No
Year dummy variables No No No No No No No No
Period Average Average Average Average Average Average
1993-2002 2000-2002 2002 1993-2002 2000-2002 2002 1993-2002 2000-2002
Observations 32 32 32 32 32 32 32 32
R-squared 0.45 0.29 0.26 0.44 0.23 0.19 0.47 0.25
Annexe 3: Regression results
for SSA countries only

35
(9) (10) (11) (12) (13) (14) (15) (16) (17) (18)
0.037 0.036 0.036 0.045 0.041 0.038 0.044 0.039 0.038 0.042
(1.94)+ (3.32)** (2.80)** (2.29)* (3.74)** (2.84)** (2.41)* (3.78)** (2.72)* (2.45)*
0.974 0.465 -0.042
(2.09)* (0.96) (0.06)
0.661
(1.81)+
-0.151 0.509 1.026 0.058 0.694 1.013
(0.47) (1.48) (1.80)+ (0.17) (1.83)+ (1.93)+
0.145 0.578 1.086
(0.43) (2.25)* (2.24)*
0.762 1.732 1.789 2.255 0.941 1.517 2.286 0.873 0.767 1.229
(1.50) (2.74)* (1.98)+ (1.93)+ (1.51) (2.01)+ (2.38)* (1.96)+ (1.69) (2.41)*
4.621 8.167 8.531 10.487 4.765 7.204 10.631 4.499 4.439 7.014
(1.55) (2.38)* (1.96)+ (1.82)+ (1.34) (2.00)+ (2.26)* (1.88)+ (1.86)+ (2.35)*
No No No No No No No
No No No No No No No
Average Average Average Average
2002 1993-2002 2000-2002 2002 1993-2002 2000-2002 2002
32 32 32 32 32 32 32 32 32 32
0.22 0.47 0.35 0.32 0.37 0.33 0.32 0.38 0.33 0.35

36
Specification (1) (2) (3) (4) (5) (6) (7) (8)
Log (fixed + mobile 0.332 0.315 0.307
subscribers) (1.38) (1.64) (1.35)
Log (fixed lines 0.388 0.390 0.344
per 1000 people) (1.63) (2.03)* (1.54)
Log (fixed lines per 2.057 2.038
1000 people), residual*** (4.39)** (4.23)**
residual)
Sub-Saharan Africa 1.928 1.953 1.844 2.074 2.233 2.059 1.566 1.040
(2.49)* (3.12)** (2.71)** (2.66)** (3.40)** (2.77)** (4.83)** (3.65)**
Constant -28.058 -28.290 -28.457 -28.194 -28.419 -28.428 -26.954 -26.962
(24.97)** (27.35)** (23.86)** (27.27)** (32.03)** (27.67)** (77.21)** (76.54)**
Period Average Average Average Average Average Average
1993-2002 2000-2002 2002 1993-2002 2000-2002 2002 1993-2002 2000-2002
Observations 67 68 65 67 68 65 67 68
R-squared 0.11 0.13 0.10 0.12 0.15 0.11 0.29 0.23
Annexe 4: Regression results,
normalising FDI flows to control for
natural resource endowments
The dependent variable in each of these regression is equal to log (FDI/(GDP* Natural Resources)), where each of the components
are in current $US. The value of natural resources is equal to the value of national output in the primary (agriculture) and secondary
sectors (manufacturing and other industry) minus the value of output in manufacturing. This method for normalising FDI inflows is
described in detail in Morisset (2000).

37
(9) (10) (11) (12) (13) (14) (15) (16) (17) (18)
0.646 0.500 0.342
(1.70)+ (1.50) (1.01)
1.911
(3.25)**
-0.239 -0.107 0.006 0.153 0.214 0.222
(0.84) (0.42) (0.03) (0.90) (1.49) (1.37)
1.103 1.340 1.168
(2.19)* (2.74)** (2.05)*
0.997 2.214 2.283 2.007 1.376 1.647 1.534 0.464 0.443 0.491
(3.04)** (2.78)** (3.37)** (2.52)* (2.24)* (2.95)** (2.52)* (2.33)* (2.22)* (2.43)*
-27.172 -28.701 -28.462 -28.446 -26.931 -27.563 -27.803 -26.533 -26.665 -26.834
(71.92)** (25.30)** (32.20)** (27.44)** (46.57)** (38.35)** (36.44)** (67.54)** (70.11)** (66.59)**
Average Average Average Average
2002 1993-2002 2000-2002 2002 1993-2002 2000-2002 2002
65 66 67 64 66 67 64 66 67 64
0.17 0.12 0.15 0.10 0.08 0.12 0.09 0.11 0.14 0.12

38
Instrumental variable estimates of the effect of mobiles penetration (growth) on FDI flows(growth)
Source: Frontier Economics. Absolute value of t statistics in parentheses, + significant at 10%; * significant at 5%; ** significant at 1%
Dependent variable/ Equation to predict IV-regression Non-IV regression –
regression change in mobile phone Change in FDI/GDP Change in FDI/GDP
subscribers 1998 – 2002 (1998 – 2002) (1998 – 2002)
Stage 1 Stage 2
Log of mainlines per -0.299
1000 people 1998 (3.87)**
Change in log of GDP 1.678
per capita (1998 – 2002) (1.49)
Change in openness 0.004 0.039 0.054
(1998 – 2002) (0.39) (1.20) (1.89)+
Log of area of country 0.072
(Km2, 1998) (0.67)
Log of total road -0.200
network (1998) (1.56)
Change in log of mobile 1.014
subscriptions (1998 – 2002) (2.44)*
Change in log of mainlines -0.194 0.326
per 1000 people (1998 -2002) (0.11) (0.21)
Change in log(1/GDP 1.852 -1.202
per capita) (0.42) (0.31)
Predicted change in log 2.131
of mobile subscriptions (2.88)**
Constant 4.470 -5.780 -3.385
(5.14)** (3.14)** (2.95)**
Observations 59 58 64
R-squared 0.32 0.15 0.15
Annexe 5:
Regression results from IV-estimation

39
In this section we present formal definitions of the
variables used in the analysis. The data are all taken from
the World Bank’s World Development Indicators (2004).
Open=((imports+exports)/GDP)
Exports of goods and services (% of GDP): Exports of goods and
services represent the value of all goods and other market
services provided to the rest of the world. They include the value
of merchandise, freight, insurance, transport, travel, royalties,
license fees, and other services, such as communication,
construction, financial, information, business, personal, and
government services. They exclude labor and property income
(formerly called factor services) as well as transfer payments.
Source: World Bank national accounts data, and OECD National Accounts data files.
Imports of goods and services (% of GDP): Imports of goods and
services represent the value of all goods and other market
services received from the rest of the world. They include the
value of merchandise, freight, insurance, transport, travel,
royalties, license fees, and other services, such as
communication, construction, financial, information, business,
personal, and government services. They exclude labor and
property income (formerly called factor services) as well as
transfer payments.
Log(fixed + mobile subscribers)
Fixed line and mobile phone subscribers (per 1,000 people):
Fixed lines are telephone mainlines connecting a customer's
equipment to the public switched telephone network. Mobile
phone subscribers refer to users of portable telephones
subscribing to an automatic public mobile telephone service
using cellular technology that provides access to the public
switched telephone network.
Source: International Telecommunication Union, World Telecommunication
Development Report and database.
Log (fixed lines per 1,000 people)
Telephone mainlines (per 1,000 people): Telephone mainlines are
telephone lines connecting a customer's equipment to the public
switched telephone network. Data are presented per 1,000
people for the entire country.
Log (mobile subscribers)
Mobile phones (per 1,000 people): Mobile phones refer to users
of portable telephones subscribing to an automatic public mobile
telephone service using cellular technology that provides access
to the public switched telephone network, per 1,000 people.
Log (1/GDP per capita)
GDP per capita (constant 1995 US$): GDP per capita is gross
domestic product divided by midyear population. GDP is the sum
of gross value added by all resident producers in the economy
plus any product taxes and minus any subsidies not included
in the value of the products. It is calculated without making
deductions for depreciation of fabricated assets or for depletion
and degradation of natural resources. Data are in constant
U.S. dollars.
Foreign direct investment, net inflows
(BoP, current US$)
Foreign direct investment is net inflows of investment to acquire
a lasting management interest (10 percent or more of voting
stock) in an enterprise operating in an economy other than that
of the investor. It is the sum of equity capital, reinvestment of
earnings, other long-term capital, and short-term capital as
shown in the balance of payments. This series shows net inflows
in the reporting economy. Data are in current U.S. dollars.
Source: International Monetary Fund, International Financial Statistics and Balance of
Payments databases, and World Bank, Global Development Finance.
Agriculture, value added (% of GDP)
Agriculture corresponds to ISIC divisions 1-5 and includes
forestry, hunting, and fishing, as well as cultivation of crops and
livestock production. Value added is the net output of a sector
after adding up all outputs and subtracting intermediate inputs.
It is calculated without making deductions for depreciation of
fabricated assets or depletion and degradation of natural
resources. The origin of value added is determined by the
International Standard Industrial Classification (ISIC), revision 3.
Annexe 6: Variable definitions and
sources (WDI 2004)

40
GDP (current US$)
GDP at purchaser's prices is the sum of gross value added by all
resident producers in the economy plus any product taxes and
minus any subsidies not included in the value of the products.
It is calculated without making deductions for depreciation of
fabricated assets or for depletion and degradation of natural
resources. Data are in current U.S. dollars. Dollar figures for GDP
are converted from domestic currencies using single year official
exchange rates. For a few countries where the official exchange
rate does not reflect the rate effectively applied to actual foreign
exchange transactions, an alternative conversion factor is used.
Industry, value added (% of GDP)
Industry corresponds to ISIC divisions 10-45 and includes
manufacturing (ISIC divisions 15-37). It comprises value added in
mining, manufacturing (also reported as a separate subgroup),
construction, electricity, water, and gas. Value added is the net
output of a sector after adding up all outputs and subtracting
intermediate inputs. It is calculated without making deductions
for depreciation of fabricated assets or depletion and
degradation of natural resources. The origin of value added is
determined by the International Standard Industrial Classification
(ISIC), revision 3.
Manufacturing, value added (% of GDP)
Manufacturing refers to industries belonging to ISIC divisions
15-37. Value added is the net output of a sector after adding up
all outputs and subtracting intermediate inputs. It is calculated
without making deductions for depreciation of fabricated assets
or depletion and degradation of natural resources. The origin of
value added is determined by the International Standard
Industrial Classification (ISIC), revision 3.
Services, etc., value added (% of GDP)
Services correspond to ISIC divisions 50-99 and they include
value added in wholesale and retail trade (including hotels and
restaurants), transport, and government, financial, professional,
and personal services such as education, health care, and real
estate services. Also included are imputed bank service charges,
import duties, and any statistical discrepancies noted by national
compilers as well as discrepancies arising from rescaling.
Value added is the net output of a sector after adding up all
outputs and subtracting intermediate inputs. It is calculated
without making deductions for depreciation of fabricated assets
or depletion and degradation of natural resources. The industrial
origin of value added is determined by the International Standard
Industrial Classification (ISIC), revision 3.

41
The remaining two papers in this report draw on surveys of poor,
rural communities in South Africa and Tanzania, and small
businesses, mainly in Egypt and South Africa. The first paper
looks at patterns and impacts of mobile use, and the second
specifically at the links between mobile use and social capital in
rural communities. The surveys were commissioned by Vodafone
from Environmental Resource Management and Forum for the
Future and were conducted in mid-2004.
In South Africa 10 communities were surveyed, with 252
interviews completed in total. In addition, 140 small businesses
were surveyed. In Tanzania 11 communities were surveyed, with
223 completed in total. Nine small businesses in Tanzania were
also interviewed. In Egypt 150 small businesses were surveyed.
In each case, the mobile services being used were ‘traditional’,
namely voice or SMS text messaging, and no instances of using
more advanced data services were observed during the
research.
The table below presents some summary statistics on the three
countries. The maps show the locations of the communities in
relation to major urban centres. In each community a mobile
phone mast had been erected in the past five years; prior to this
the communities had little or no access to fixed-line telephones.
Typical incomes in the rural communities selected will be below
the national average. In Tanzania, employment in the
communities is mainly agricultural, and often informal. In the
case of the South African communities a higher proportion of
inhabitants will have formal and non-agricultural employment,
but unemployment rates will be higher than the national average
in most cases.
In Tanzania, the typical community surveyed was small, with only
a few hundred inhabitants. In most cases, the roads to the
villages were sealed, although roads within them were not.
Most of the dwellings were self-built shacks, and local services
were very limited. For example, few of the communities had
formal shops, clinics, or official public transport, and in none was
running water or electricity to the house commonplace.
The South African communities were generally more developed,
with facilities such as formal shops being common, and some
had benefited from government housing, electrification and water
and sanitation projects. However, self-built shacks were also
common, with much of the population living in informal
settlements or squatter camps.
Mpumalanga, South Africa, Keuny Maziya on his cell phone.
Introduction to the community and
business surveys
Egypt, South Africa Tanzania – basic information
Fixed Mobile
Per Capita Lines Lines
Percent GDP Per 1000 Per 1000
Country Population Urban (US$, PPP)* People People
Egypt 70.5 42.1 3,810 110 67
South Africa 44.8 56.5 10,070 107 304
Tanzania 36.3 34.4 580 5 22
All Developing Countries 4,936.9 41.4 4,054 96 101
High Income Countries 941.2 77.8 28,741 584 653
World 6225.0 47.8 7,804 175 184
Source: UNDP, Human Development Report 2004. All data are for 2002.
Note: PPP (purchasing power parity) GDP figures are adjusted to reflect the cost of living, so $1000 of PPP income would yield the same standard of living everywhere.

42
The aim of the fieldwork was to interview a broad cross-section
of the community that included a mix between gender,
employment status, age (all respondents were over 16) and
mobile phone ownership, use and non-use.
The surveys identified three distinct sets of people:
• people with their own personal mobile phone (“mobile owners”);
• people who don’t have their own mobile, but do use other
people’s mobiles (“non-owning users”); and
• people who don’t own a mobile phone and never use other
people’s mobiles (“non-users”).
Figure 1. Map showing the location of the communities
surveyed in South Africa
1. Kga Kgapane
2. New Pietersburg
3. Phake
4. Emondlo
5. Oppermans Kraal
6. Msinga
7. Mvenyane
8. Rhodes
9. Butterworth
10. Van Wyksdorp
1. Masasi
2. Nachingwea
3. Tanangozi
4. Mafia
5. Dimon
6. Issuna
7. Ndago
8. Manyara Ranch
9. Ngorogoro
10. Mirerani
11. Mango
Figure 2. Map showing the location of the communities
surveyed in Tanzania
However, due to the focus of the research, the samples for
each community were not randomly selected and exhibit a
bias towards individuals owning a mobile phone. Fieldwork
concentrated on collecting a large proportion of mobile phone
owners and users, and so the profile of respondents in this
research is in no way representative of South Africa and Tanzania
as a whole or even the rural communities where fieldwork was
conducted. Therefore the levels of ownership for both South
Africa and Tanzania are not representative of mobile ownership
in rural communities in these countries. In addition, there is a
higher proportion of females in the sample for each community
in South Africa and Tanzania, as the surveys took place during
the day, when men were more likely to be out working. The
survey samples are also slightly biased towards the younger age
groups.
1
2
3
45
6
7
8
9
10
11
1
2
3
4
5
6
7
8
9
10

43
In South Africa, the majority of the sample interviewed had their
own phone. Mobile penetration in South Africa was 31 per cent
in December 2002 and in rural communities such as those
included in this survey, penetration is much lower than this.
In the surveys, just under a quarter of those interviewed were
“non-users” and 10 per cent had no phone but used other
people’s phones from time to time – usually borrowing the
phone of a friend or relative, for no charge.
In Tanzania, just over 40 per cent of those interviewed in the
fieldwork owned their own phones. Again, because the survey
actively sought out mobile phone users, this is a much higher
proportion than in Tanzania as a whole, where mobile penetration
was two per cent at the end of 2002. Non-users made up 16
per cent of the sample, and non-owners who used other
people’s phones made up another 42 per cent. The majority of
these people were using cheap phone cards, which meant that
they could borrow other people’s handsets at no cost to the
owner.
Profile of mobile ownership and use, South Africa (250 respondents) and Tanzania
(222 respondents)
The surveys of small businesses were undertaken primarily in
Egypt and South Africa, but also in two communities in Tanzania.
A total of 150 people were interviewed in Egypt, 140 in South
Africa and 9 in Tanzania.
Small businesses were defined as having fewer than 50
employees. Only a small portion of those in the sample did not
have a mobile (although this was not a deliberate part of the
survey design), reflecting the high rates of mobile phone
penetration amongst small businesses in South Africa and Egypt.
Businesses surveyed included professional firms, street traders,
tradesmen, a range of service firms, manufacturers, retail traders
and mobile phone related business operators.
The surveys covered both the formal and informal sectors of
business. The people surveyed in Egypt were all located in Cairo,
whereas in South Africa, we interviewed small businesses in
urban areas (most in Cape Town but some in Durban and
Johannesburg) as well as in the same 10 rural communities
described earlier. The surveys involved face-to -face interviews
supplemented by a mailing of questionnaires to the formal sector
businesses in Cape Town, Durban and Johannesburg. Business
surveys were also conducted in two communities in Tanzania:
Ngorongoro and Mafia Island.

44
Many developing country governments and development agencies
are focusing on extending telecommunications services into rural
areas, as they seek to encourage growth, alleviate poverty and
overcome a perceived ‘digital divide’. Mobile technologies are
playing a major role in this effort. However, relatively little is known
about how rural communities and small businesses use mobile
technologies, and what impacts they are having.
Mobile communication services in Africa have expanded rapidly
in recent years. Most of this growth has been in urban areas,
but there are growing rural networks in many countries.
The affluent urban markets have naturally been targeted first,
but in addition there has been a perception that the rural poor
are not able or willing to pay for mobile telecommunications
services. Yet in fact, in many instances, rural demand has greatly
exceeded initial expectations.
Perhaps equally important, the introduction of mobile services
has brought about a change in the business and operating
climate of the African telecommunication sector: competing
mobile operators have helped create an environment that fosters
innovation and competition
This paper presents the results of research into socio-economic
impacts of mobile communications on households, rural
communities and small businesses in Africa. Some of the
questions the research sought to address include:
• Who uses mobile communications services?
• What are the factors that influence ownership, use and
non-use of mobile phones?
Partner, ERMJonathan Samuel
Consultant, ERMNiraj Shah
Consultant, ERMWenona Hadingham
Mobile Communications in South
Africa, Tanzania and Egypt: Results
from Community and Business Surveys

45
• What are mobile phones used for – as a consumer good,
for business or employment purposes, or both?
• What role do mobiles play in the operation of small
businesses in urban and rural areas?
• What social and economic impact are mobile phones having
on communities and small businesses in Africa?
Answers to these questions have important implications for both
governments and the mobile operators as they seek to expand
their networks to cover rural areas. There is very little empirical
information on the impacts of mobile phone use in rural
communities or by small businesses in Africa. The results of this
study are based on data collected in face-to-face interviews,
carried out in South Africa, Tanzania and Egypt.
Profile of mobile users and non-users in
the communities
a) Who uses mobile phones?
In the surveys, 67 percent of the sample of 252 people in
South Africa owned a mobile phone, while a further 8 percent
used mobiles, but did not own one, and 25 percent of
respondents did not own or use a mobile phone at all.
In Tanzania, the figure for ownership was lower, at 43 percent
of the 223 respondents interviewed, with users at 42 percent
and nonusers at 16 percent. The higher proportion of users as
opposed to owners in Tanzania suggests a greater degree of
‘sharing’ of mobiles than in South Africa. This may reflect a
lack of alternative communication facilities for non-owners in
most of the communities in Tanzania and the smaller numbers
who can afford to purchase a phone. Those respondents who
used someone else’s phone usually bought airtime vouchers
to do so.
When we looked at access to mobile phones (regardless of
whether or not the respondent used them), 97 percent in
Tanzania stated that they could access a mobile phone if they
wished to, whereas only 28 percent could access a landline
somewhere in the community. This indicates a very high
awareness of the potential to use mobile phones for
communication, and very high perceived accessibility, even in
these very poor rural communities.
The survey found that the perception of ownership of mobile
phones in Tanzania is different to that in South Africa. When the
respondents stated that they owned a mobile phone, they often
considered it as a household asset rather than a personal or
individual one. This was particularly the case for female
respondents. However cultural norms in rural Tanzania dictate
that ownership of such items lies with male members or heads
of the households. We found that a broadly similar proportion of
males and females were owners and users of mobiles. Whilst
this result is surprising, particularly in the case of Tanzania, it
may be explained by the fact that the sample is skewed towards
females and those with mobile phones. The figures in Table 1 do
show some differences in men’s and women’s use of mobiles,
especially in Tanzania.
We can make no claims for the sample being fully representative
of the rural population in these communities. However in the
sample of owners and users of mobile phones we found a broad
representation of individuals by age, income groups, education
levels and gender. We looked at the breakdown of owner, user,
and non-user status by gender, age, education and income to
see where use or non-use varied with the by these parameters.
Table 1: Mobile status of the interview sample by gender
Gender Male Female
South Owners 39.1 56.8
Africa Users 40.0 60.0
Non-users 51.7 46.6
Census data for 47.2 52.8
communities
Tanzania Owners 50.5 48.4
Users 47.3 52.7
Non-users 20.0 80.0
We found that nearly 57 percent of the respondents who
owned a mobile phone in South Africa were female. Similarly,
60 percent of respondents who were users but not owners were
also female.
Perhaps not surprisingly, almost half of the respondents who
were users in the South African communities came from the
25-45 age groups. The respondents in this age group are
economically active and therefore may be more likely to own a
phone. However, respondents in age groups of 46-55 and over
55 were still well represented in the group of owners and users.
In Tanzania, the patterns of age distribution in the group of
owners and users was similar but was more concentrated in the
age group of 26-45 as shown below in Table 2.
Table 2: Mobile status by age in South Africa and Tanzania
South Africa 25 26-45 46-55 55
Owner 30.2 49.7 10.1 7.7
User 40.0 36.0 20.0 4.0
Non-user 29.3 37.9 6.9 17.2
Tanzania 25 26-45 46-55 55
Owner 14.7 63.2 13.7 7.4
User 25.8 52.7 10.8 10.8
Non-User 28.6 60.0 5.7 2.9

46
We also looked at mobile phone status according to the
respondents’ level of education. In South Africa, we found that
broadly speaking the pattern of mobile phone ownership in the
rural communities surveyed was skewed towards those with
higher levels of education. Table 3 indicates these proportions
according to the census data. However, the survey data also
reveals that there is a large number of owners and users of
mobile phones, who have no or just primary education.
This is also evident in the case of the Tanzanian communities,
although there is a marked difference between ownership
and usage patterns.
Table 3: Mobile status in South Africa and Tanzania by
education level
No Technical
South Africa education Primary Secondary College University
Owner 7.7 14.8 56.2 12.4 4.1
User 16.0 20.0 56.0 8.0 0.0
non-user 15.5 44.8 34.5 1.7 0.0
Population 22 26 45 6 2
(from census)
No Technical
Tanzania education Primary Secondary College University
Owner 5.3 28.4 33.7 20.0 4.2
User 8.6 62.4 18.3 8.6 0.0
non-user 14.3 71.4 11.4 0.0 0.0
Finally we looked at the breakdown of respondents’ mobile
phone status by income brackets and compared it with the
overall breakdown of incomes from the census data. We found
that over 50 percent of the respondents who were mobile phone
users were within the lowest R501-1000 (monthly) income
bracket (approximately $85-170 per month). The data also
confirm that at higher income levels, people are more likely
to own their own phone. The non-users were unsurprisingly
concentrated in the lowest income group. Overall, we conclude
that income is not a significant barrier to access to mobile
telecommunications.
Table 4: Mobile status in South Africa by income level
South Africa 500 501-1000 1001-4000 4001
Owners 51.8 27.7 10.9 9.5
Users 53.4 41.9 2.3 2.3
Non-user 63.1 21.1 10.5 5.3
R401-R1 R1 601- R204 801
R400 600 R6 400 or more
Census data % 22.6 49.6 23.4 4.4
South African census bureau income brackets differ slightly from those
conventionally used for market research purposes and in these surveys.
R=South African Rand. £1=R11.30, US$1=R6.
b) What explains mobile phone use?
The extent to which mobile phones are used, and the ease with
which new users can access them, is crucial in terms of their
economic and social effects. The reason is that there are strong
network effects accruing from phone subscription. A network
effect (or externality) occurs because each existing subscriber
benefits when the total number of subscribers increases.
As the total number of subscribers increases, so does the
value of having a phone, because each individual can contact
more people.
The network effect is well understood in developed markets
where personal ownership of a phone is the common model.
However, the operation of network effects will be different where
mobile phones are not personally owned, but shared among
individuals, or used in a communal facility (such as a Community
Service Telephone centre in South Africa). Ownership facilitates
two-way communication because an individual is uniquely
identified with a number. In a model of shared use two-way
communication is more difficult; a non-owning user can make
calls out but cannot receive spontaneous inbound calls.
Whether this difference is significant depends upon the type of
communication required. Communications which are initiated by
an individual to acquire data or information from a central source
(such as finding out the availability of goods in a shop) are
largely unaffected by the inability for the individual initiating the
call to be reached in turn. According to the field observations,
mobile phones were essential for searching for work, not only for
getting information and making an application, but also as a
means of being contacted by a prospective employer – that is,
inbound communication was important. On the other hand,
family interactions may not be adversely affected by shared
usage if the mobile is shared within a co-located family unit.
The distinction between models of access that facilitate
essentially one-way communication and individual ownership
which permits two-way communications is important from a
policy perspective. Understanding how people want to use
mobiles can inform policies that are seeking to increase access
to ICT services. The survey results shed some light on the nature
of communication individuals in the rural communities require
and the implications of those needs for the models of ICT
access – in particular the difference between one-way and
two-way communications.
The surveys therefore included a number of questions designed
to ascertain the most important influences on the use of mobiles.
One obvious candidate was income, but we did not consider this
relevant in the Tanzanian communities surveyed. Most of them
are dependent on subsistence farming and people earn very little
cash income. In addition, the income they earned was seasonal,
and dependent on the harvests. A more detailed survey would
have been required to get an appropriate estimate of how much
the respondents could earn in a typical month.

47
Our analysis of income for South Africa generally shows a
positive relationship between mobile phone usage and income.
Not surprisingly, the number of calls made and text messages
sent increases with income, as shown in Figure 1. However,
respondents in the lowest income bracket also appeared to be
reasonably frequent users of mobile phones. This suggests that
making calls is also important for those on very low incomes.
Low income users have also found clever ways to minimise their
own call costs. One example is ‘beeping’, where a caller dials
but hangs up before the call is connected, thus avoiding a call
charge. The recipient will then call back at their expense.
More sophisticated versions include giving meanings to specific
numbers of rings. For example, three rings might mean ‘I am
leaving now’ or ‘pick me up now’.
Figure 1: Income levels of survey respondents and
frequency of usage in South Africa
Reviewing this same relationship at the community level in South
Africa gives a similar outcome. In those communities where
average incomes are higher, people also tend to make more calls
on average. This relationship also seems to hold when looking at
the cross-sectional data between the communities.
Figure 2: Mobile and SMS usage ranked by average
community in South Africa
We also contrasted ownership of mobile phones with ownership
of other consumer durables. Mobile phones are one of several
consumer durables that households in the survey typically
owned.
Figure 3: Top 50% and bottom 50% of individuals by income
and the percentage of ownership of assets
Overall, whilst income is obviously an important influence on
ownership and use of mobiles, the survey evidence clearly
suggests that mobile phone ownership is less skewed towards
the better-off sections of the population than other consumer
durables. This is significant, as the survey sample was
deliberately targeted at communities, which could be expected
to be amongst the poorest in their countries. The results
therefore suggest that, on the whole, mobile is very far from a
luxury good affordable by only the rich.
Expenditure on mobile phones as a proportion of total
expenditure can give some broad information on their
importance and impact on household budgets. In South Africa,
134 mobile phone owners were happy to provide information
on their income and their mobile phone expenditure. These
respondents spent on average between 10 and 15 percent of
their income (or 89 to 108 Rand) on mobile phones (estimation
was made using mid-points of income and expenditure
brackets). However, as only one respondent identified mobile
phones in their top three expenditure items, so these figures
should be treated with caution. National data suggest the biggest
items of expenditure for the poorest black South Africans (urban
and rural) are food (about 50 per cent of the household budget),
fuel and energy, and housing (each 7 to 8 per cent). Transport
and communications follow after these categories, however, in
the national statistics1
. The level of spending indicated by the
survey is surprisingly high (as a proportion of income), but it is
interesting to explore the extent to which spending on mobiles
may be substituting for other categories of expenditure, such as
transport. We return to this below.
own radio own mobile own TV own car
Assets
500 501-1000 2001-4000 4000
Income level
0 500 1000 1500 2000 2500
Average income of the sample in each community

48
Impact South Africa (%) Tanzania (%)
Prompted responses*
Improved relationships 78.7 85.3
Call rather than travel to family and friends 77.4 91.1
Un-prompted responses
Easier communication with family and friends 72.2 84.6
Useful in emergencies 25.8 27.1
Assists in job search 15.5 2.7
Avoids problems with public fixed-line phones** 8.8 n.a.
Easier to organise meetings 8.2 9.0
Faster or improved communication 7.7 53.2
Access to business information/business purposes 7.2 34.0
Saves money 5.7 1.6
Easier to contact school/university 4.6 5.9
Contact employer/clients on road 3.6 3.7
Can send cheap messages using SMS 2.6 0.5
Status symbol 2.1 –
Improved access to telecommunications 1.5 1.1
Place an order for groceries or other items 1.0 6.9
Feel safe 1.0 2.7
Make money from lending out phone 1.0 0.0
Health concerns 0.5 0.5
Expensive/costs money – 9.7
* These impacts were identified through specific questions, while the rest of the impacts identified were offered by the respondents without a specific question being asked.
** In Tanzania, in the communities surveyed there was generally no public fixed line phone.
Table 5: Impacts of using a mobile phone
Another potential barrier the surveys explored was lack of access
to electricity, which can inhibit take-up of other technologies in
developing countries. Clearly some form of energy is needed
to recharge mobile batteries and so a lack of electricity could
form a barrier to mobile phone ownership. In Tanzania many
communities had limited or no access to electric power.
Figure 4 shows the relationship between mobile phone
ownership and use and a community’s access to electricity
in Tanzania. The horizontal axis shows the percentage of
households within a community with access to electricity,
from lowest to highest. Whether a respondent was an owner
of a mobile is positively related to whether he or she has access
to electricity.
Figure 4: Access to electricity and mobile phone ownership
Respondents with electricity are more likely to own a mobile
phone. Those without electricity are more likely to borrow
someone else’s.
The communities overcame the constraint of not having an
electricity connection in a variety of ways. For example, at Issuna
Mission in Tanzania, every week someone collected all seven
mobiles in the community and took them to the nearest town
that had electricity, to be charged. In a small community, it is
likely to be easier to charge a small number of phones and share
these rather than each person owning a handset. Communities
without electricity managed to achieve a similar level of mobile
phone usage to those communities with electricity (in terms of
traffic volume), albeit with lower levels of ownership.
In South Africa, communities with and without electricity were
equally likely to own and use mobile phones. This might reflect
greater possibilities for recharging phones using motor vehicles
(motor vehicles were much more prevalent in the communities
visited in South Africa than in Tanzania). In Kwa Phake, South
Africa, a community without access to electricity, a local
hairdresser had a phone charging service using a car battery.
We also looked at whether or not mobile phone ownership and
use might depend on whether or not a community has access to
a post office, as a proxy for access to an alternative means of
communication. However, no relationship of this kind was found
in either South Africa or Tanzania.
c) Impacts of mobiles
Respondents to the surveys identified a large number of impacts
from using mobile phones. Some of these were social in nature,
while others concerned employment or business. The social
impacts were very important in both South Africa and Tanzania.
Greater contact and improved relationships with family and
friends was one of the most significant benefits identified by the
surveys. But reduced travel costs and help in job search were
also highlighted. A limited number of respondents also made
money from renting out their phone.

49
We were surprised by the very small number of negative
comments made. Only 0.5 percent of respondents in each
country mentioned health concerns, and only in Tanzania were
concerns about costs raised.
The main unprompted impact identified by the surveys related
to easier contact with family and friends. In both Tanzania and
South Africa, many people move away from their home to find
work, and mobile phones are now an important means of
keeping in touch with families. In the survey sample, 91 percent
of respondents in Tanzania called friends and relatives rather
than travelling to see them. In South Africa, 77 percent of mobile
users called rather than visited. (These response rates were to
prompted questions.) Indeed, for many families surveyed the
costs of travelling to see relatives would be prohibitive, especially
in the poorest rural communities, and mobile therefore
represented the only option of maintaining contact.
Respondents thought that this generally had a large impact on
travelling time and costs saved. There will also be environmental
and safety benefits associated with avoided travel. Table 6
illustrates the travel time and cost savings identified by
respondents. The impacts were slightly larger for Tanzania,
where roads are worse and public transport less extensive.
The potential importance of mobile as a substitute for travel is
easy to underestimate. Of the communities surveyed in South
Africa, only 4 out of 10 had a regular bus service to the nearest
town and the typical round-trip cost was 15 Rand. In contrast, a
typical pre-paid voice call costs R5 (Average monthly income in
the South African communities was R1271). It is not surprising
that so many respondents identify mobiles as a source of saving
both time and travel costs.
Table 6: Estimates of travel time and cost savings
Saving South Africa Tanzania
Large saving in travel time 52.2% 67.3%
Large saving in travel cost 58.2% 65.4%
Interestingly, the vast majority of those who did travel to see
relatives (85 percent for Tanzania and 79 percent for South
Africa) thought these relationships had improved anyway
because of mobile phones. Only a very small number recorded a
deterioration in the relationship with friends and relatives who
are now phoned rather than visited. A detailed analysis of trips
saved compared with call costs incurred would be an interesting
area of further work.
As an example, one respondent in Mafia Island, Tanzania, said he
was now able to keep in daily contact with his immediate family,
who all lived in Dar es Salaam. Using his mobile phone, he is
able to get information about his children’s progress at school
and what they are doing in their free time, thereby maintaining a
strong relationship with them despite the distance. He felt that
mobile phones had saved him a lot of money as the cost of
going to Dar es Salaam, certainly in relation to calling with a
mobile phone, is high.
A number of respondents also used mobile phones to contact
schools and universities. For example, mobile phones are used
by the students in Kwa Phake, to correspond with various tertiary
institutions such as UNISA (University of South Africa). Instead of
having to travel to these institutions they can easily access
information they need using a mobile phone. Monthly calls for
educational purposes in this particular community were made by
31 per cent of respondents.
There were also examples of parents using phones to contact
children boarding with relatives and attending school in
neighbouring towns. Mobile phones enabled parents living in
towns without fixed-line services to contact their children during
term time. For example, in Rhodes, South Africa, one mother had
a daughter attending school in Barkly East (about 60km away on
difficult roads) who boarded there with a relative. The very poor
public phones in the community and limited public transport
facilities meant that mobile phones were the only way she could
regularly keep in contact with her daughter.
In Tanzania, a strikingly high proportion of respondents
(57 percent) felt that a major impact from mobile phones was
faster and improved communication. The proportion in South
Africa mentioning this as an impact was substantially lower at
8 percent. This probably reflects a greater presence and
reliability of fixed-line phones in South Africa prior to the
introduction of mobile phone services.
Nevertheless, poor public phone services (using fixed-line
phones) were cited by a number of respondents in South Africa
as a key reason for relying on mobile phones. 17 percent of
respondents in South Africa who do not own but use someone
else’s mobile phone noted that problems with the public fixed
line phones mean they now rely on a borrowed mobile if they
need to make a call. Mobile phones also provided peace of
mind to geographically isolated communities with poor fixed
line facilities, with about a quarter of both the South African
and Tanzanian respondents stating that mobiles were useful
in emergencies.
Mobiles can also help to improve services in rural areas.
For example, shared taxi drivers operating in Mango Parish,
Tanzania, used their mobiles to request additional taxis to come
to the taxi stand when there were lots of people waiting for
transport, thus reducing their customers’ waiting time and
increasing their own income.
The responses revealed mobile phones to be important for job
search in South Africa. Altogether, 16 percent of respondents
volunteered this as an impact and 24 percent of owners or users
also said they had made or received a call about an

50
employment, business or training opportunity. Mobile phones
enabled job seekers to ring for information about employment,
and enabled them to be contacted by potential employers.
This was particularly important in South Africa, where fears
about crime would stop many employers visiting potential
employees at their homes.
In Tanzania, 34 percent used mobile phones to access business
information and for business reasons. This could reflect the
importance of agriculture in the economy, with phones being
used to get accurate market price data and order supplies.
For example in Mafia Island one person running a small fuel
supplies operation used his mobile to place an order for more
stocks of fuel when his reserves were running low, and to get a
specific date for shipments from the mainland. He said he was
also now able to source fuel from more suppliers than before.
Business and employment opportunities are an area where
network effects play an important role. Two-way communication
is important in instances where potential employers or clients
would like to contact a prospective employee or supplier.
According to the field observations, mobile phones were
essential for job search, not only for getting information and
making an application, but also as a means of being contacted
by potential employers.
Mobile use by small businesses
In addition to the community surveys, we also explored mobile
phone use by small businesses, surveying businesses in South
Africa (urban and rural businesses) and Egypt (Cairo only).
The breakdown of small business respondents in Egypt and
South Africa by industry is summarised in Figure 5 below.
Figure 5: Respondents by Type of Business
The differences between the samples in the two countries largely
reflect the rural/urban split of respondents. Thus around 42
percent of small business respondents in South Africa were from
the 10 rural communities. These were concentrated in the retail
sector, so there is a greater representation of retail businesses in
the South African sample. There were also very few professional
firms in these communities, and this has resulted in a lower
representation of professional firms in the South African sample
compared with Egypt.
Table 7 presents the impacts identified from the surveys in Egypt
and South Africa.
Table 7: Impacts of Mobile Phones on Small Businesses
Egypt (%) South Africa (%)
Prompted responses
Increased call costs 67.3 47.1
Increased turnover 66.0 56.6
Increased customer numbers 65.3 56.2
Increased profits 58.7 61.8
Unprompted responses
Faster/improved communication 57.3 25.7
Increased efficiency 56.0 21.4
Save time 24.7 10.0
Available to clients all the time 23.3 47.1
Save costs 22.0 15.7
Larger client database 16.7 4.3
Place orders on the job 15.3 21.4
Bad network 11.3 4.3
Assist in breakdowns/emergencies 10.0 20.7
Reduced travelling 8.7 50.0
Contact with the office 8.0 25.0
Less free time/ no privacy 7.3 7.1
Nearly 85 percent of the businesses surveyed in Egypt and 89
percent of businesses in South Africa in the sample used a
mobile. Five years ago just 11 percent of the businesses
surveyed in Egypt, and 34 percent in South Africa, said they
used mobile phones for business purposes.
The number of small business with access to fixed-line
telephones stood at 45 percent in Egypt and 52 percent in South
Africa five years ago. Whilst this number has increased to 80
percent in Egypt and just over 60 percent in South Africa,
mobiles have now overtaken fixed-line phones as the most
important communication tool for businesses in the survey.
Prior to acquiring a mobile, 27 percent of business respondents
in Egypt and 15 percent in South Africa had no telephone access
at all. For comparison, the use of fax machines by the small
businesses surveyed had also increased substantially over the
past 5 years, with 23 percent of small businesses in Egypt and
47 percent in South Africa, now using one (up from 5 per cent
and 31 per cent respectively five years ago).
Nevertheless, the rate of increase in the use of mobile phones
has exceeded the increase in use of other communication tools.
Over the past five years, the number of businesses using
mobile phones increased by over 547 percent in Egypt and
nearly 125 percent in South Africa. This compared with an

51
increase of 325 percent and 53 percent in Egypt and South
Africa respectively for facsimile machines and 71 percent and
15 percent for fixed-line telephones.
However, where small businesses can use fixed-line telephones
and facsimile machines in conjunction with mobile phones,
the survey evidence showed that they often choose to do so.
One of the reasons for this is the higher price for calls from
mobile phones compared with fixed line phones.
In South Africa, mobile phones were the only source of
communication for a large number of small businesses run
by black individuals. Over 85 percent of small businesses
run by black individuals rely solely on a mobile phone for
telecommunications. In addition, for many small businesses in
rural areas mobile phones are the only source of communication
(and most small businesses interviewed in rural areas in South
Africa were run by black people). For these small businesses,
mobile phones are literally essential to their businesses.
In the cities, there were also some examples of mobiles helping
overcome disadvantages. For example, in South Africa, a
manufacturer of children’s dolls based in Cape Town employed
deaf people. Text messaging via mobile enabled the owner and
employees to communicate with each other. The owner felt that
without the technology it would be much more difficult to interact
with his workers and it would not have been practicable to
employ them at all.
In Egypt, the informal sector was more reliant on mobile
phones for running their business than the formal sector
(the informal sector encompasses a wide range of small retail,
small manufacturing, transport and service activities).
Almost 90 percent of businesses in the informal sector used
a mobile phone.
The surveys also revealed that mobile phones played a part
in small business start-ups. In South Africa, 29 percent of
respondents from non-mobile phone related firms were
influenced to some extent by the availability of mobile phones
in starting up their business, while 26 percent in Egypt were
influenced by mobile phones. This was particularly true for small
businesses operating in the service sector. In some cases,
access to mobile phones has increased the range of services
that can be offered. Mobile phones also mean that small
businesses can operate a 24-hour call-out service, which is
important for tradesmen and non-professional service firms.
For some rural communities in South Africa which previously
were without fixed line telephones, mobiles have simply made
running a small business feasible. For instance in South Africa
and Tanzania, operators of spaza shops (informal general stores)
and kiosks are now able to order supplies using a mobile phone
without having to travel to place an order.
Figure 6: A spaza shop operating in Nachingwea, Tanzania
Survey respondents in both Egypt and South Africa said mobiles
had increased their profits: 59 percent in Egypt, with 31 percent
noting a large impact. In South Africa, the figure was 62 percent,
with 27 percent noting a large impact. The reported increase in
profit levels was in spite of respondents generally also saying
that mobile phones had increased their call costs.
Interestingly, none of the businesses in these surveys, including
retailers in the rural communities, suggested that higher
spending on mobile calls by their customers had dented their
profitability. This might suggest that customer spending on
mobiles has at least created additional business opportunities
(such as selling pre-pay vouchers) which compensate for any
lost sales of other products.
Overall, the respondents said increases in profits attributed to
the use of mobile phones were due to a combination of reduced
travelling time and costs, increased customer numbers and
higher turnover.
Reduced travelling was a much more important impact in South
Africa than in Egypt, with 56 percent of businesses in South
Africa identifying this as a beneficial impact, compared with just
10 percent for Egypt. This might reflect the predominance of
rural firms in the South African sample, with 75 percent of small
businesses in rural South Africa indicating that mobile phones
saved them travelling time. Nevertheless, 46 percent of small
businesses based in South African cities also identified reduced
travelling as an important impact.
An increase in efficiency was another widely cited impact.
Some specific examples included being able to run errands
without closing the store, placing orders from the premises
without having to visit supplier (important for shops and those in
the building trade), and keeping in contact with staff and the
office while travelling. Professional firms also noted that mobiles
enabled them to keep in contact with clients while travelling.

52
Conclusions
The results of the surveys suggest that mobiles have brought
considerable benefits to communities and small businesses.
People at all income levels are able to access mobile services,
either through owning or sharing a phone; and gender, age
and education do not seem to constitute barriers to access.
While income certainly explains the level of usage, lack of
income does not prevent mobile use. Even the absence of
electricity does not present an insurmountable barrier, thanks to
the sharing of mobiles and recharging batteries in the nearest
town, or recharging locally by a generator or car battery.
For the residents of the rural communities, mobile phones
have typically had positive economic and social impacts. Mobiles
have reduced travel needs, assisted job hunting and provided
better access to business information. Greater ease of contact
with family and friends has improved relationships. These
benefits were reported even though the communities surveyed
were amongst the poorest in their countries.
Mobile phones have also become an essential tool for small
businesses. A substantial proportion of small businesses have no
alternative method of communication. The proportion is highest
for black-owned businesses in South Africa and informal sector
businesses in Egypt, suggesting that mobiles have become an
important tool for disadvantaged groups. A large majority of
small businesses said mobiles have brought higher profits,
turnover and increased efficiency, although they are also paying
higher call charges.
Notes
1
Income and Expenditure of Households 2000, Statistics South Africa, release P0111, 2002.

53
Introduction: what is social capital?
Human activity uses a variety of resources to achieve different
ends. For example, financial capital is one of those resources,
accumulated over time for drawing on when needed.
Environmentalists use the concept of natural capital to mean the
resources of the natural world such as clean air or trees, which
we can invest in, use, and deplete. The idea of social capital
refers to those social resources that likewise we invest in,
accumulate, draw down and sometimes deplete. Social capital
represents the intangible value of the social group, on whatever
scale, above and beyond the value of its individual members
alone. It can be thought of generally as the social resources
available for human activity.
The term was first used in 1916, when Lyda Judson Hanifan, a
West Virgina school superintendent, published a discussion of
the role of schools as community centres. He noted that high
levels of participation among local people in school affairs not
only led to improved support for the school, but also to general
improvements in the school’s wider community: there was an
unintended social spillover. He coined the term social capital to
describe this, and later defined it as a combination of “goodwill,
fellowship, sympathy, and social intercourse among the
individuals and families who make up a social unit”.
In the past decade social capital has become one of the most
salient concepts in the social sciences. The American sociologist
Robert Putnam has done most to promote the revival of the idea
of social capital. He looked at a wide range of indicators in the
USA such as membership of voluntary associations, participation
in community affairs, trust of strangers and so on, and in his
well-known book, “Bowling Alone: The collapse and revival of
American Community,” used these to argue there had been a
decline in social capital in America. He identified a number of
causes for this decline – chief among them too much television
watching – and proposed means by which the decline might be
arrested. For Putnam, the key to a healthy society is participation
in social groups, and he has advised national governments
around the world (including the UK) on how to promote
community participation.
The concept of social capital appeals to sociologists, economists
and political scientists alike; one of its strengths is to bring these
disciplines together. A weakness, however, is the lack of a
precise definition. Putnam defines it as “features of social life –
networks, norms and trust – that enable participants to act
together more effectively to pursue shared objectives.”1
Francis Fukuyama calls it “an instantiated informal norm that
promotes cooperation between two or more individuals”2
.
Michael Woolcock, a sociologist working for the World Bank,
refers to “the information, trust, and norms of reciprocity inhering
in one’s social networks”3
. Economists are interested in social
capital for its contribution to productivity, and define it as the
spillover from the individual to the group, a sort of social
externality or network effect.4
Most definitions include a structural element supporting a
cognitive element. A parallel might be with the road network and
traffic flowing on it, as the structural element, and the laws and
unwritten rules of the road as the cognitive element. The
structural element of social capital is made up of social networks
and relationships: friendship networks, families, neighbourhoods
or communities, companies, social groups, political groups, and
so on. These are all forms of association, organised in order to
achieve certain ends: to provide support, to distribute products or
disseminate an idea, for example. The cognitive element
comprises a range of social attitudes, relating to a willingness to
trust other people and shared values and norms. Successful
participation in a social network creates trust, which can then be
invested back into the social network to grow the capital,
strengthening and growing the network.
These definitions make it clear that social capital might be a
useful way of understanding the social role of mobile phones.
Mobile phones are used to mediate contact between different
people, and so are likely to have an effect on the size, number
and nature of social networks that people participate in. This in
James Goodman
Linking mobile phone ownership and
use to social capital in rural South
Africa and Tanzania

54
turn may affect levels of trust. Social capital may also provide an
indicator of where take-up of mobile telephony could be higher.
What’s the use of social capital?
The literature generally agrees that high levels of social capital
can result in desirable socio-economic outcomes. According to a
2001 review by the UK Office for National Statistics, “Social
capital has a well-established relationship with the outcomes
policy makers are concerned with e.g. economic growth, social
exclusion, better health and wellbeing.”5
The 2002 Policy and
Innovation Unit (PIU, now the Strategy Unit) report on social
capital for the UK government identified six general benefits,
supported to a varying degree by empirical research:
1. It may facilitate better economic performance, for example
through reducing transaction costs, enabling the mobilisation
of resources and facilitating the rapid movement of
information.
2. It may facilitate the more efficient functioning of job markets,
for example by reducing search costs.
3. It may facilitate educational attainment;
4. It may contribute to lower levels of crime;
5. It may lead to better health;
6. It may improve the effectiveness of institutions of
government.
Research in rural Tanzania has suggested that increased levels
of community participation lead to higher household incomes.6
There appear to be strong correlations between national levels of
social capital, measured in terms of trust, and socio-economic
development. The World Values Survey includes the question,
“Generally speaking, would you say that most people can be
trusted, or that you can’t be too careful in dealing with people?”.
The top-scoring countries on the trust measure also tend to be
countries with high GDP. In 1996, the top scoring country was
Norway, with 65 per cent of people answering that most people
could be trusted. Next came Sweden, Denmark, Netherlands and
Canada, all with scores of over 50 per cent. Ireland scored 47
per cent, Australia 40 per cent, the USA 36 per cent and the UK
31 per cent. In contrast, less developed countries scored much
lower. In Turkey, trust was recorded at six per cent and in Brazil
at just three per cent.7
Consequently, development organisations and governments have
become intensely interested in social capital. The World Bank,
for example, has sponsored a great deal of research on the
relationship between social capital and macro-level outcomes.
It has an extensive area of its website dedicated to social capital,
featuring the results of this research as well as guidance on the
literature, and tools to help with the measurement of social
capital. The Bank aims to promote the use of social capital
analysis to aid social and economic development in developing
countries, and to counteract poverty. The World Bank and other
organisations, such as the OECD, have been followed by a
number of national governments keen to understand how
understanding social capital can aid successful policy-making.
These include the Australian and Irish governments, for
example.8
In the UK, the PIU report identified many areas where an
understanding of social capital could inform public policy.
Suggestions included encouraging mentoring schemes to build
links across communities; promoting schools as community
centres; and reforming the criminal justice system so that
convicted criminals can maintain support networks, and
discouraging the development of “criminal social capital”.
The report also included one proposal involving mobile phones:
“Mobile telephones could have emergency help keys or codes
that would activate the nearest five phones to indicate that the
holder is in danger and needs assistance. Receivers of the
distress signal would be expected to respond, at least to
establish what the problem is or call the police. The scheme
would break down the “diffusion of responsibility” that inhibits
strangers helping each other in times of personal emergencies –
using technology to strengthen social norms of reciprocity and
trust within the wider community.”9
Social capital may be an even more important concept for
developing countries than developed, as in many cases people in
the former have less access to formalised structures of support
such as the legal system or the financial system, and may rely
on informal networks instead.
Social capital and mobile phones
Mobile phones are a communications technology, and as such
they facilitate social networks, so there is likely to be a link with
social capital. However, research on the social role of information
and communication technologies has so far has been heavily
biased towards the internet.
Optimistic assumptions that the development of virtual
communities online would create whole new forms of social
capital have so far been a red herring. However, there is fruitful
research on how internet use affects social behaviour, and how
social tools on the internet – “social software” – can be used to
build social capital.
The US research project, Syntopia, conducted by James Katz
and Ronald Rice from Rutgers University,10
analysed the social
behaviour of users and non-users of the internet between 1995
and 2000. The research identified a clear trend: long-term use
of the internet was associated with more, not less, frequent
socializing; and the same or a higher level of political and civil
society involvement. Internet users were more likely to go out

55
and see friends, although the study found that they also spent
more time away from their local community and generally knew
fewer of their neighbours.10
Syntopia’s findings are supported by
research carried out by the Pew Institute’s Internet and American
Life Project, which concluded that the use of email enhanced
users’ contact with family and friends and that email users
generally had a richer social life.
Keith Hampton of the Massachusetts Institute of Technology
(MIT) led a study specifically into the use of ICT in a community
context. He spent two years living in a suburb of Toronto wired
with wide-band internet access, observing the activity of other
residents to try and find out whether online activity can
supplement offline contact and help revive communities.
The research found that: “The local computer network was used
by residents as a means to exchange introductions, organize
barbecues and parties, search for missing pets [etc]… Rather
than isolating people in their homes, CMC [computer mediated
communication] encourages visiting, surveillance, neighbour
recognition and the maintenance of local social ties.”11
There has been less attention paid to the mobile phone, which
may be explained by the fact that widespread mobile penetration
occurred later and also because mobile is not as prevalent in the
US – where much of the social capital literature originates – as
in many other countries. However, there is a rich literature on
the social impacts of the fixed-line telephone. It describes, for
example, the role of the telephone in empowering middle class
women,12
expanding activities in the local community and
beyond, reducing loneliness and anxiety and strengthening
social ties.13
One European Commission funded study has collected data on
social capital and use of different ICTs in several European
countries. It suggested that access to social capital is becoming
more individualised, with people less dependent on formal
groupings and more involved in loose, spur of the moment
association. “Mobile communication plays into this approach,
since it allows a more flexible form of communication,” writes
the study’s author Richard Ling of Norwegian mobile operator
Telenor. He continues, “It allows one to fit sociation into the
nooks and crannies of everyday life and possibly obviates the
need for social contact in the context of other, more formal
institutions.”14
A later report based on the same data showed a
positive relationship between communication with friends and
quality of life, but a direct link to mobile was not established:
“In no country did acquiring a mobile phone, internet access or
broadband internet have any positive effect on overall quality
of life”.15
It is possible that mobile telephones are having a more
pronounced impact in countries where communications
infrastructure has hitherto been less extensive. Most of western
Europe has had a dense fixed-line network for some decades,
but large numbers of telephones are a very recent phenomenon
in countries such as Tanzania. The introduction of mobile
telephony might therefore be expected to have important
consequences. Many studies have already suggested this, for
example Sadie Plant’s investigation of the mobile phone
undertaken for handset manufacturer Motorola.16
Mobiles and social capital in South Africa
and Tanzania
We aimed to use the concept of social capital as a framework for
understanding the social impacts of mobile phones, theoretically
connecting the localised social impacts with wider socio-
economic changes. The results shed light on the social impact of
mobile, and also suggest the concept of social capital might offer
guidance for companies and governments wishing to understand
the indirect impacts of mobile products and services.
Questions pertaining to different aspects of social capital,
specifically social networks, group participation and social
attitudes including generalised trust, were included in the
community questionnaires, used in surveys in South Africa and
Tanzania.
One of the objectives of the research was to assess the
importance of mobile phones relative to other communication
means. To this end, we asked a number of questions about
general communicaton habits, including the amount of face-to-
face contact respondents thought they had with various different
types of people.
The responses in South Africa and Tanzania were broadly similar:
there was very frequent face-to-face communication with family,
close friends and others living within the community. Face-to-
face contact with others outside the community was less regular,
as was contact with tradesmen and figures of authority.
South Africa Tanzania
(% communicating (% communicating
Face to face “frequently” or “somewhat” or
communication “very frequently” “very often”
with… face to face) face to face)17
Family 81% 87%
Close friends 77% 89%
Others in the community 81% 96%
Others outside of the community 25% 22%
Businessmen or tradesmen 19% 21%
Government services
(inc doctors, teachers) 28% 8%
Police or security 16% 5%
Table 1 Percentage of respondents communicating
frequently or very frequently face-to-face in South Africa
(252) and Tanzania (223).

56
In the Tanzania survey, the amount of face-to-face contact with
“family” and “close friends” was lower than with “others in the
community”. This may reflect the fact that in Tanzania, people
often leave their communities to seek employment or education
in larger cities.
We asked respondents whether they had easy access to a
number of different means of communication. Fixed-line phones
were prevalent in the South African communities surveyed and in
Tanzania three quarters of people said they had easy access to a
post office. But in both surveys, the mobile phone was the
communications tool that most people had easy access to .
Figure 1. percentage of respondents with access to
different communications infrastructure in South Africa
(252) and Tanzania (223)
For those who had access to each means of communications,
frequency of use varied. The figures in Table 2 below show the
average number of times in one week people who had access to
each communications medium used it. Mobile phones were by
far the most frequently used communications means for the
people interviewed, primarily for calls but also, significantly, for
texts. The data suggest that in the communities surveyed,
despite the relatively recent introduction of the technology,
mobile phones are at the very heart of communication.
Mean no. Mean no.
of times Base of times Base
used/week size used/week size
Post Office 1.6 80 1.7 82
Landline phone 3.9 162 2.2 32
Payphone - 1.7 29
Vodacom phone - 2.4 35
Internet 4.9 10 1.8 11
Cell phone to
make calls 9.7 188 6.5 182
Cell phone to send
text messages 7.8 118 5.6 143
Table 2. Mean weekly usage of different communications
tools, for those with access.
Group participation is often used as an indicator of social capital.
It was one of the main areas of investigation for Putnam in his
study of declining social capital in the USA, and features
prominently in most social capital questionnaires. We asked
respondents to tell us which community groups they were
members of, how often they met formally and how often they
communicated outside of formal meetings. Overwhelmingly the
most popular association was with religious groups – 76 per
cent of respondents in South Africa and 95 per cent in Tanzania
said they were members of religious groups.
Top group membership – Top group membership –
Religious group 76% Religious group 95%
Sports group 19% Sports group 7%
Community/charity group 10% Finance/savings group 7%
Finance/savings group 12% Political party 7%
Political party 15%
Funeral society 31%
Table 3. Membership of community groups South Africa
(252) and Tanzania (223)
Membership of associations other than religious groups was very
low in our Tanzania survey, but quite high in the South Africa
survey, indicating a higher degree of formalised socialisation.
On average, over half of our respondents in South Africa were
members of two or more different social or community groups,
double the proportion in Tanzania. However, in the Tanzania
survey, due mostly to the importance of affiliation with religious
groups, there were fewer people who were members of no
group.
Figure 2. Number of groups respondents are a member of,
South Africa (252) and Tanzania (223)
In our South African survey, mobile owners were the most likely
to be members of multiple groups, followed by non-owning
users, with non-users least likely (figure 3).

57
Figure 3. Membership of groups in South Africa, by mobile
ownership and use
Likewise in the Tanzania sample (figure 4), mobile owners were
involved in more community groups.
Figure 4. Membership of groups in Tanzania, by mobile
ownership and use
In South Africa, there were differences in group membership
according to mobile phone ownership and age, income,
education level, gender, the amount of face-to-face contact with
family, and which community respondents lived in. Using
regression analysis, it was possible to identify independent
relationships between group membership and age, highest
educational level achieved, community, and mobile phone
ownership. Older respondents were more likely to be members
of two groups or more. Respondents who had been to high
school or university were more likely to be multiple group
members than those who had stopped their education at primary
level. Mobile phone ownership also had an independent, positive
relationship with multiple group membership: controlling for all
the other factors involved, mobile phone owners were more likely
to be members of two or more community groups.
In the Tanzania survey, there were statistically significant
differences in the number of social or community groups
respondents were members of according to community, age,
income, mobile phone ownership and the length of time
respondents had lived in their communities. Regression analysis
showed that group membership was related independently to the
length of time lived in the community and, as with the South
Africa survey, mobile phone ownership.
Further data showed that mobile phones were used frequently to
communicate with group members outside of formal group
meetings. Although the relationship between mobile ownership
and group membership is a strong one, suggesting that on this
measure mobile owners are more willing to invest in social
capital than non-owners, the direction of the causal relationship
is unclear. Are mobile users more likely to join groups, or are
group members more likely to get mobiles?
How mobiles are used in the communities
We asked a number of questions aimed at understanding how
mobiles were being used for communication with certain groups
and for certain purposes, and how this compared to other
commonly used means of communication. The structural
element of social capital is social networks, made up of people
and the links between them. To use the phraseology of American
sociologist Mark Granovetter, links can either be strong links or
weak links19
. Strong links are those between close friends and
family, people who are regularly in contact and have a lot in
common. Weak links are those between acquaintances or distant
friends in irregular contact. Both types of links are crucial.
The strong links provide support and are particularly important at
the beginning and end of life, while weak links become more
important in adult life, delivering new social and economic
opportunities such as leads about job openings, and creating
competitive advantage.
Ideally there is a balance of strong and weak links: relationships
that offer support as well as relationships that offer
opportunities. If, in a particular community, there is a wealth of
strong links and very few weak links, this can lead to social
exclusion and stagnation. This is characteristic of traditional,
tight-knit communities. In contrast, if a community has very few
strong links but many weak links, opportunities for social or
economic advancement might abound, but there is likely to be
no sense of community cohesion or neighbourhood spirit, as
sometimes found in suburbs. We hoped the surveys in South
Africa and Tanzania would indicate how people were using
mobiles to manage strong and weak links.
The table for the South Africa survey (table 4) shows the
proportion of people using each method of communication
frequently or very frequently with different groups. Face-to-face
communication was the most common method for all groups,
and as expected communication was most frequent with family
members, close friends and others within the community.
Face-to-face communication with others outside of the
community, businesses or tradesmen, teachers, doctors and
police was less frequent.

58
The use of mobile phones to manage these strong links (family,
close friends, others in the community) and weak links (all the
others) was broadly the same as use of fixed-line phones.
Remember that for most communities surveyed in South Africa,
the proportion of people with access to landline phones was only
a little less than those with access to mobile phones. However,
fixed line and mobile phone usage diverged when it came to
contact with family and close friends. In both cases, but
especially with family contacts, mobiles were used significantly
more than landlines. This suggests that mobile phones were
being used more to manage strong links, the links that make up
tight-knit support networks, than for weak links. Although they
were also being used for weak links, the frequency of mobile
use did not differ from the use of fixed-line phones for this
purpose.
This suggests that mobile phones were helping to meet a
demand for more communication with family, friends and
neighbours that is not otherwise satisfied, even if landlines are
present in the community. Although mobiles were being used to
manage weak links too, there is no suggestion from this data
that they were satisfying unmet demand.
Table 4. Percentage communicating frequently or very frequently with these groups, using each communication medium,
South Africa (242-250).
Table 5. Percentage communicating somewhat or very often with these groups, using each communication medium,
Tanzania (223-4)
Key to tables 4 and 5
Others in Others Govt services Police or
South Close the outside of Businessmen (inc doctors, security
Africa Family friends community community or tradesmen Teachers
Face to
face 81% 77% 81% 25% 19% 28% 16%
Using a
landline
phone 16% 18% 7% 11% 4% 6% 9%
Using a cell
phone to call 33% 26% 8% 11% 3% 6% 5%
Using a cell
phone to text 13% 13% 4% 4% 1% 2% 2%
Under 5 per cent using this maximum to
communicate frequently or very frequently
with this group
Between 11 and 20 per cent Between 41 and 70 per cent
Over 71 per centBetween 21 and 40 per centBetween 5 and 10 per cent
Close the outside of Businessmen (inc doctors, security
Tanzania Family friends community community or tradesmen Teachers
Face to
face 87% 89% 96% 22% 39% 66% 38%
Using a
landline
phone 2% 2% 1% 3% 3% 2% –
Using a cell
phone to call 50% 42% 17% 43% 21% 8% 5%
Using a cell
phone to text 33% 32% 13% 28% 13% 5% 2%

59
In the Tanzania survey (table 5) as with South Africa, face-to-face
communication was the most common method of
communicating, and communication was most frequent with
family, close friends and others in the community. There are a
number of notable differences though. In the Tanzanian sample,
face-to-face communication with businessmen, tradesmen,
doctors, teachers and police was relatively higher. It is also
obvious that fixed-line telephones did not play a significant role
in these communities. When it comes to the use of mobile
phones, as with the South Africa survey, mobiles were used a lot
for contact with family and close friends – strong links. But there
were also reasonable levels of usage to contact others outside of
the community and businessmen or tradesmen – weak links.
Responses to the question “In the past year, have you made
a call or sent an SMS about business, education or training
opportunities outside your community?” reinforce the suggestion
that mobiles were used by respondents in our surveys to manage
weak links. Around a fifth of mobile phone users in both surveys
replied in the affirmative. So in the Tanzania survey as with the
South Africa survey, mobiles were used to manage strong links
with close friends and family, but they were also used for weak
links, contacts that may offer social and economic opportunity.
These findings were reinforced by answers to another question
in the surveys, “Do you use cell phones to speak to people
instead of travelling to see them?”. In Tanzania, 91 per cent of
mobile users answered positively. Just over two-thirds of these
calls (68 per cent) were not to family or friends, but fell into the
“other” category. In South Africa in contrast 77 per cent of
people used mobiles to speak to people instead of travelling to
see them, and the majority of these calls were to family, friends
or both (99 per cent in total). Only one per cent of these calls fell
into the “other” category.
If mobiles were being used to manage strong links, it is
legitimate to wonder if calls were replacing face-to-face
communication. In the South African sample, there was a small,
statistically insignificant reduction in the amount of face-to-face
contact mobile owners had with family members. The same was
the case for the Tanzania survey. However, in this case mobile
owners had a lot less face-to-face communication with others
outside the community, a difference that was statistically
significant.20
Therefore there may be a limited substitution effect operating in
these two samples. Investigating this relationship further would
be a fruitful area for further research, as most previous studies
suggest that communication over phones or using the internet,
does not substitute for face-to-face contact, but rather
supplements it. We do know from our two African surveys that
people who use their mobile to talk to people instead of meeting
them said that their relationships with distant people had
improved because of mobiles – 79 per cent in the South African
survey and 85 per cent in the Tanzanian.
The “social halo” effect of mobiles
Results from both surveys showed a high degree of sharing
mobile phones, suggesting that the devices are a social amenity
as well as being a communications tool. This can be an
important contributor to social capital as well. Alex McGillavrey of
the New Economics Foundation, who has written widely on the
role of new technologies in building trust, has talked of the social
facility of his chainsaw in the small French village where he lives.
Many people in the locality need a chainsaw on occasion but not
often enough to warrant owning one, instead borrowing the
chainsaw that McGillavrey bought. The chainsaw therefore
facilitates social contact within the local community but also
initiates a network of reciprocity: McGillavrey is doing people
favours which at some point in the future they are likely to return.
Our survey results show something similar happening with
mobile phones in the communities studied. In South Africa, over
half of mobile owners said that they allowed family members to
use their handset for free, and almost a third did the same for
friends. There was also ample evidence of people making and
receiving calls and texts on behalf of others.
We see a similar pattern in the Tanzania survey, again with over
half of respondents with their own phone letting family members
use it for free, and with a similar proportion doing the same for
friends (higher than in the South Africa survey). In both samples
there was a negligible amount of charging others to use
handsets. However, in Tanzania a large proportion of the non-
owning users were paying to use others’ handsets. This took the
form of paying for phonecards and then using the cards with
other people’s handsets, at no charge to the owner.
Social attitudes
People’s attitudes to others and their feelings about the
community in which they live emerge from the network of
relationships they are part of. In particular, levels of “generalised”
or “extended” trust are an area of research focus: how willing
people are to trust others in general. The level of trust is
considered a key indicator of social capital.
Measuring the cognitive aspects of social capital, as we are
here, also helps us to distinguish between positive and negative
social capital. Social capital should not be understood as a good
thing in itself, but rather as a neutral social resource that allows
people to do things. As such a resource, it can be used for
positive ends or for negative ends, and deciding which is which
is a subjective process. The example of mafia networks is often
given to illustrate how a high level of social capital in a particular
network can have negative results for the wider community.

60
We chose a standard measure of trust for inclusion in our
surveys, and eight other measures that would indicate whether
or not respondents had access to a high level of social capital.
In South Africa, in answer to the question, “Generally speaking,
would you say that most people can be trusted, or that you can’t
be too careful in dealing with people?,” overall 63 per cent of
respondents said they thought that people could be trusted.
This was a surprisingly high level, more than twice the level
expected on the basis of results from the same question asked
in South Africa as part of the World Values Survey. This may be
peculiar to rural communities in South Africa, or the communities
surveyed may be untypical of the country as a whole. There were
no significant differences in answers to this question according
to whether people owned or used mobiles.
Figure 6. Generalised trust by mobile ownership and usage,
South Africa.
Overall the response to this question in the Tanzania survey was
also surprisingly positive. There were small differences in the
amount of reported trust depending on mobile ownership and
use, with mobile owners actually coming out slightly less trusting
than others. The differences were not statistically significant,
however.
Figure 7. Generalised trust by mobile ownership and usage,
Tanzania
Since trust is regarded as a primary indicator of social capital,
the answers to this question in both surveys should be taken as
inconclusive of any relationship between mobile phones and
social capital. However, of the eight other measures we collected
in this research, there were some statistically significant
relationships, all of which indicated a more positive social
outlook from mobile phone owners.
In South Africa, there were significant differences in the answer
to the question, “All things considered, how satisfied are you with
your life as a whole these days?.” Overall, 52 per cent of
respondents told us they were satisfied or very satisfied with
their life these days. Mobile owners were more satisfied and
non-owners who do not use mobiles were a lot less satisfied,
with the difference statistically significant at the 99 per cent level.
Figure 8. Life satisfaction by mobile ownership and usage,
Tanzania (252)
A logistical regression was run on the life satisfaction variable for
the South Africa sample. The analysis suggested significant
relationships between life satisfaction and income21
, age, amount
of face-to-face family contact, membership of social groups and
mobile ownership. The regression showed that mobile phone
ownership had a positive influence on life satisfaction, controlling
for all other factors including income. There was also a
relationship between life satisfaction and the community lived in.
Mobile owners in the South Africa survey responded more
positively to the question, “Do you feel you have control over the
way your life turns out? Do you have no control at all, some
control or a great deal of control?,” a difference that was
statistically significant at the 95 per cent level.
Figure 9. Feelings of control over life, by mobile onwership
and usage, South Africa.

61
In addition to a relationship with mobile ownership, there were
differences in feelings of control according to income, gender,
and the amount of face-to-face contact with family members.
Each of these apart from gender had a positive influence on
feelings of control over the way that life turns out, independent
of each other and other factors.
On the other measures tested in the South Africa survey, mobile
owners tended to answer more positively than non-owning users
and non-users, but the differences were not large enough to be
statistically significant, given the relatively small size of the
overall sample.
Turning to the Tanzanian results, the picture is somewhat
different. We asked seven other social attitude questions as well
as the trust question discussed above. On three of the measures,
mobile owners gave slightly less positive answers than non-
owners and non-owning users. However, the differences in these
measures were not found to be statistically significant.
There were two questions on the Tanzania survey where there
was a statistically robust relationship between positive social
attitudes and mobile phone ownership. The first of these was the
question, “How well do people in your community get along
these days?.”
Figure 10. Perceptions of community harmony by mobile
ownership and usage, Tanzania.
There were differences in perceptions of community cohesion
according to the different communities surveyed, age, income
and level of group membership as well as mobile ownership.
Mobile phone ownership was the only independent factor
influencing views on community cohesion. In other words, mobile
phone owners were more likely to say their community got on
well or very well, independent of other factors such as income
levels. The difference between mobile owners and non users is
significant at the 95% level.
Mobile owners were also much more likely to say that they had
helped somebody in their community in the last six months
(figure 11). The difference was significant at the 99% level.
Figure 11. Willingness to help others, by mobile ownership
and usage, Tanzania.
Apart from differences in community helpfulness between mobile
users and non-users, there were also differences according to
the community surveyed and gender. Further analysis showed
that mobile phone ownership and community were the significant
independent factors explaining whether respondents said they
had helped someone in their community in the past six months.
To review the results of our analysis of the relationship between
mobile phones and social attitudes, in the South Africa survey
mobile phone ownership was positively associated with life
satisfaction, independent of other social and economic factors
tested such as income and age. There was a similar relationship
between mobile ownership and feelings of control over how
respondents’ lives turned out. There were no other statistically
valid differences between mobile owners, non-users and non-
owning users in this survey. To speculate for a moment, these
relationships point towards a role in personal empowerment for
the mobile phone with the people surveyed. Possibly because
fixed-line infrastructure was available in the South African
communities and readily accessible for most people, the value of
the mobile phone in this context might be related to the specific
facilities of the mobile – personal ownership (not necessarily a
household resource) and portability – rather than the simple fact
of connectivity. Whether this is the case, and whether it is
representative of rural South African communities in general,
would have to be tested further.
In the Tanzania survey, there was a significant relationship,
independent of other factors, between mobile phone ownership
and perceptions of community cohesion, as well as whether
respondents had helped somebody in the community in the past
six months. Again, to speculate, these results may suggest that
mobile ownership has a relationship with community
participation. This speculation is supported perhaps by the
existence of a statistically strong relationship between mobile
ownership and membership of community groups in the
Tanzanian sample. Here, it may be the simple connectivity that is
key. In the Tanzanian communities surveyed fixed-line networks
and other types of communications infrastructure were rare.

62
The suggestion is that the specific qualities of mobile phones
(portability, individual ownership) were less important in these
places than the simple fact of remote connectivity. Again, this is
a hypothesis that would need to be tested further.
It is important to note at this point that the relationships explored
here relate to mobile phone owners and not non-owning users.
Non-owning users in the Tanzania and South Africa surveys (the
former a reasonably sized subgroup, the latter rather small) in
general resembled non-users, in their responses to the social
attitudes questions. However, this is likely to be because mobile
owners use their phones a lot more than people who borrow or
pay to use others’, rather than to any separate, intrinsic value of
ownership.
A key question, then, is whether the role of the mobile phone is
the same for mobile owners and people who use mobiles but do
not own them. In both countries, mobile phone owners used
their phones a lot more than non-owning users. In the South
African sample, over three-quarters of mobile owners used their
phone to make or receive calls four times a week or more.
The equivalent figure for non-owning users was just 24 per cent.
The pattern was similar in the Tanzanian sample.
Mobile owner Non-owning
South Africa (169) user (25)
No mobile use 5% 28%
Use mobile
1 – 3 times a week 19% 48%
Use mobile 4 times
a week or more 76% 24%
Table 6. Frequency of mobile phone use by mobile
ownership and usage, South Africa.
Mobile owner Non-owning
Tanzania (95) user (93)
No mobile use – 10%
Use mobile
1 – 3 times a week 24% 74%
Use mobile 4 times
a week or more 76% 16%
Table 7. Frequency of mobile phone use by mobile
ownership and usage, Tanzania.
As expected, there was a strong relationship in both samples
between mobile ownership and how often people said they used
their phone to make and receive calls. The reduced frequency of
usage for non-owners applied to calls made or received in all the
different categories we asked about. For example, table 8 shows
that 78 per cent of mobile owners made or received calls
frequently or very frequently with family members. For non-
owning users this was just 40 per cent. Mobile phone contact
with doctors, teachers and police or security forces was
practically non-existent for non-owning users, whereas an
appreciable proportion of mobile owners were using their phones
for this purpose. Non-owning users also used phones markedly
less to contact others within the community.
Close the outside of Businessmen (e.g. doctors, security
Family friends community community or tradesmen Teachers
Owner (95) Calls 78% 71% 35% 67% 36% 18% 11%
SMS 58% 55% 26% 47% 14% 9% 4%
Non-owning
user (93) Calls 40% 28% 6% 33% 14% 1% 2%
SMS 20% 22% 5% 18% 5% 1% –
Table 8. Use of mobiles frequently/very frequently to contact different groups, by mobile ownership and usage, Tanzania.

63
Mobile owners used their phones for a wider variety of purposes.
Table 9 shows that phone use for religious purposes, arranging
meetings and doing business is quite high for mobile owners,
but much lower for non-owning users, with the exception of
“business reasons”. Although over two-thirds of non-owning
users used their phones on a weekly basis to contact friends and
family, only three per cent did this daily, a much lower frequency
than for mobile owners.
Owners Non-owning users
(95) (93)
Contact with family 96% 69%
and friends (daily – 61%) (daily – 3%)
Religious reasons 12% 3%
Arranging meetings 17% 2%
Business reasons 25% 11%
Safety reasons 2% –
Information on
employment 1% –
Finding out about
community activities
and events 4% –
Educational purposes 4% 1%
Information on
health issues 1% –
Shopping 4% 1%
Table 9. Percentage using cells phones weekly or more,
Tanzania. Differences are statistically significant at the 95%
level or above in the first three cases.
There are also important differences when it comes to managing
weak links. Table 10 below shows that a third of mobile owners
in the Tanzanian sample had made a call or sent a text message
in the past year regarding an employment, educational or
training opportunity, and 28 per cent had received one.
This was much lower for non-owning users.
Owners Non-owning users
(95) (93)
Made a call re
business/training/
education opportunities 33% 8%
Received a call re
business/training/
education opportunities 28% 8%
Table 10. Percentage making and receiving calls re
business, training or educational opportunities, Tanzania.
We can say therefore that, while we saw earlier how mobiles are
used to manage strong links with family and friends, non-owning
users in the Tanzanian communities surveyed did this much less
than mobile owners. Furthermore, where mobiles are used to
manage weak social links, ownership seems to be more
important than simple usage for this purpose.
Conclusions
The research suggests there are some links between social
capital and mobile phone ownership and use in rural
communities in South Africa and Tanzania. Access to mobile
phones was high, as was frequency of usage, even in the South
African communities which had ready access to fixed-line
telephones. This places mobile phones at the heart of
communication in these communities.
In both countries there was a high degree of sharing mobiles
for free with friends and family (and sometimes for money).
This indicates that mobiles may be acting as a social amenity, a
tool to be shared and a focus for social activity, as well as a tool
for communications.
Mobile phones were being used to mediate both strong links
(with family, close friends and others in the community), essential
for maintaining support networks, and weak links (“others
outside the community”, businessmen, tradesmen, government
officials such as teachers and doctors, as well as the police),
providing access to information and possible social and
economic opportunities. Weak links are seen as particularly
important in the relationship between social capital and desirable
macro-level outcomes, and even more so perhaps in a
developing world context, where communities can be very tight-
knit given their paucity of connections to the outside world.
With regard to weak links, mobiles were used for contact with
others outside of the community, businessmen, tradesmen,
doctors, teachers and police. This was particularly prevalent in
the Tanzanian case. Also in Tanzania, over 90 per cent of mobile
users replying to the surveys said they used mobiles to speak to
people rather than travelling to visit them, and two-thirds of
those calls were not to family or friends, suggesting they might
be associated with weak links rather than strong links. Around a
fifth of respondents in both surveys had made and received calls
in the past year relating to business, training or educational
opportunities.
With regard to strong links, mobiles were being used intensively
in both surveys for contact with close friends and family.
Although there was some evidence to suggest that contact by
mobile was replacing some face-to-face contact, a majority of
respondents said that the use of mobiles to contact people far
away rather than travelling to see them had improved their
relationships.

64
We can conclude therefore that, within the parameters of the two
surveys, mobiles were facilitating participation in social networks,
helping to maintain both strong and weak links, including
participation in community group activity. They were thus
enabling people to invest in and draw on social capital.
There was evidence to suggest that mobile phone owners
were more willing to invest in social capital and to draw on it.
Mobile owners were significantly more likely to be members of
community groups such as religious organisations, sports teams
and political parties, in both surveys. In the Tanzania survey,
there was also a statistically robust relationship between mobile
ownership and willingness to help others in the community.
In the Tanzania survey, mobile owners appeared more likely to
think that the community they lived in was functioning well.
In the South Africa survey, mobile owners reported higher life
satisfaction and greater feelings of control over how their lives
turned out. For other social attitude questions, including
measures of generalised trust, no solid relationship with mobile
ownership or use was established. Significant differences were
related to ownership of mobiles, rather than using other people’s
mobiles. This is likely to be due to the fact that in both surveys,
mobile owners used their phones more and for a wider variety of
purposes than non-owning users.
In conclusion, social capital offers a helpful framework for
understanding the social impact of mobile telephones in rural
communities in South Africa and Tanzania. The unrepresentative
nature of the surveys limits the generality of the results.
They concern the individual social capital of mobile owners.
The next research step would be to investigate the aggregate
effect on communities of mobile ownership and use, and
look for examples of a relationship between mobile penetration
and community social capital. This would require survey
work in communities that do not yet have mobile networks,
representative sampling and data collection over an
extended period.
References
Aldridge Halpern (2002), Social capital: a discussion paper,
PIU, UK Cabinet Office http://www.cabinet-
office.gov.uk/innovation/reports/reports.shtml
Anderson, B (2004) Information Society Technologies, Social
Capital and Quality of Life e-Living D11.4b
Buchanan, M., (2002) Small World: Uncovering Nature’s Hidden
Networks, Weidenfield and Nicolson.
Dasgupta, P., (2002) “Social Capital and Economic Performance:
Analytics” in Ostrom, E. and Ahn, T., (ed.s), Social Capital:
A Reader, Elgar
Dasgupta. P., (2004), The Economics of Social Capital,
http::// www.econ.cam.ac.uk/faculty/dasgupta/ecsoccap.pdf
Davies, W., (2003) You don’t know me but…Social Capital and
Social Software, iSociety
De Sola Pool, I. (ed.) (1977). The Social Impact of the Telephone.
MIT press
Fukuyama, F. (1999), The Great Disruption: Human nature and
the reconstitution of social order New York: The Free Press.
Glaesar, E.L. et al., “An Economic Approach to Social Capital” in
The Economic Journal 112
Granovetter, M., (1973), “The strength of weak ties”, in American
Journal of Sociology 78
Hampton, K. (2003) “Grieving for a lost network: collective action
in a wired suburb” in The Information Society” 19 (5), 1-13.
Hanifan, L.J. (1920) The Community Center, Boston: Silver
Burdett.
Harper, R., (2001) Social Capital. A review of the literature,
National Statistics Office
Harriss, J., (2001) Depoliticizing Development. The World Bank
and Social Capital, Anthem
Johnson, S. (2001), Emergence: the connected lives of ants,
brains, cities and software, Allen Lane, London
Katz, James E., Aakus, Mark. (2002) Perpetual Contact: Mobile
Communication, Private Talk, Public Performance. Cambridge
Keser et al, (2002), Trust, the internet and the digital divide, IBM,
New York
Kopomaa, T. (2000) The City in Your Pocket: Birth of the Mobile
Information Society. Helsinki, Gaudemus
Lasen, A. (2002) The Social Shaping of Fixed and Mobile
Networks: A Historical Comparison. University of Surrey
Ling, R., (2004) Social Capital and ISTs, e-Living D11.4a
Narayan and Pritchett, (1997), Cents and Sensibility: Household
Income and Social Capital in Rural Tanzania, World Bank.
National Economic and Social Forum, Forum Report No.28
(2003) The Policy Implications of Social Capital
Perri 6, (1997) Escaping poverty: from safety nets to networks of
opportunity, London: Demos
Plant, S. (2001) On the Mobile: The Effects of Mobile Telephones
on Social and Individual Life. Motorola
Productivity Commission, (2003) Social Capital: Reviewing the
concept and its policy implications, Commonwealth of Australia
Putnam, R. (2000), Bowling Alone: The collapse and revival of
American Community, Simon and Schuster
Szreter, S. (2000) 'Social Capital, the Economy, and Education in
Historical Perspective' in Baron, S., Field, J. and Schuller, T. (ed.)
Social Capital , OUP, Oxford
Uphoff, N (1999) Understanding Social Capital: Learning from
Analysis and Experience of Participation in Sera/Dasgupta
World Bank
Willey, M. and Rice, S. (1933) Communication agencies and
social life, McGraw-Hill, New York
Woolcook, M. (1998), “Social Capital and Economic
Development: Towards a theoretical synthesis and policy
framework” in Theory and Society, 27: 151-208.

65
Notes
1
Putnam, 2000 p.000. Putnam also quotes here L. J. Hanigan
2
Fukuyama, 1999 p.1
3
Woolcock, 1998, p.153
4
Dasgupta 2004, p. 28
5
Harper, 2001.
6
Narayan and Pritchett, 1997
7
Keser et al., 2002
8
A paper was commissioned by the Australian government in 2003 from the Productivity
Commission to investigate the policy implications of social capital. The National Economic
and Social Forum of Ireland produced a similar report for the Irish government in the same
year. It recommended that a government department be chosen to lead on developing
social capital related policies.
9
PIU p.69
10
Katz and Rice, 2002
11
Hampton 2002, p.10
12
De Sola Pool 1977
13
Willey and Rice 1933
14
Ling, 2004, p.2
15
Anderson, 2004, p.24
16
Plant, 2001.
17
The categories offered to respondents were different for this question between South
Africa and Tanzania. Therefore, direct comparison between the two countries is not
possible.
18
Respondents in South Africa were not asked about access to payphones or community
service phones.
19
Granovetter, 1973.
20
12 per cent of mobile owners said they had regular face to face contact with “others
outside of the community” compared to 40 per cent of non-users.
21
Approximately 20 per cent of the sample in South Africa refused to state their income
level. This subgroup was cross-tabulated with all other socio-economic variables to verify
that they had no particular characteristics in common. They were then coded together with
the higher income subgroup so that the answers for this subgroup could be inlcuded in
the regression model.

66
This bibliography contains the specific references cited in the
report, and in addition other references relevant to the role of
mobile telephony on economic and social development.
Accenture, Markle Foundation, and UNDP, Final Report of the
Digital Opportunity Initiative, July 2001.
http://www.opt-init.org/framework/DOI-Final-Report.pdf
Ashurst, Mark, ‘Africa: Now, a 'Quiet Revolution': Mobile Phones
Leapfrog an Obstacle to Development’ , Newsweek International,
August 27, 2001
Asian Development Bank, Towards Universal Access:
Socioeconomic Impact Study of Rural Telecommunications in
Thailand. Volume 1: Main Report, Midas Agronomics Report T.A.
2381, 30 November 1996.
Azam, Jean-Paul, Magueye Dia and Tchétché N’Guessan,
‘Telecommunications Sector Reforms in Sénégal’, World Bank
2002. http://econ.worldbank.org/files/18861_wps2894.pdf
Batchelor, Woolnough and Scott, ‘The Contribution of ICTs to Pro-
Poor Growth’, OECD Discussion Paper
DCD/DAC/POVNET(2004)15, September 2004.
http://www.gamos.org/files/oecd/MDGS_POVENT_15.pdf
Bayes, A, von Braun, J and Akhter, R., Village Pay Phones and
Poverty Reduction: Insights from a Grameen Bank Initiative in
Bangladesh, ZEF, Berlin, 1999.
http://www.zef.de/publications/publ_zef_dp.htm#8
Beardsley, Scott, Ingo Beyer von Morgenstern, Luis Enriquez, and
Carsten Kipping, Telecommunications Sector Reform: A
Prerequisite for Networked Readiness, Chapter 11 of Kirkman
and Sachs (2002).
http://www.cid.harvard.edu/cr/pdf/gitrr2002_ch11.pdf
Blattman, Christopher, Robert Jensen and Paul Roman,
Assessing the Need and Potential of Community Networking for
Developing Countries: A Case Study for India, MIT Working Paper
February 2002.
http://edevelopment.media.mit.edu/SARI/papers/CommunityNetw
orking.pdf
Boston Consulting Group, Options on the Future: The Role of
Business in Closing the Digital Divide, January 2002.
Bruns, Bryan et al, Village Telephones: Socio-economic Impact
and Implications for Rural Futures, paper presented at 6th
International Conference on Thai Studies, October 1996.
http://www.cm.ksc.co.th/~bruns/rurtel.html
Camp, Jean L and Brian L Anderson, Grameen Phone:
Empowering the Poor through Connectivity, The Magazine on
Information Impacts, December 1999.
http://www.telecommons.com/villagephone/Camp_article12_99.
htm
Cohen, Nevin, Grameen Telecom’s Village Phones, World
Resources Institute June 2001.
http://www.digitaldividend.org/pdf/grameen.pdf
Comin, Diego and Bart Hobijn, Cross-Country Technology
Adoption: Making the Theories Face the Facts, Federal Reserve
Bank of New York Staff Report no. 169, June 2003.
http://www.ny.frb.org/research/staff_reports/sr169.html
Department for International Development, Balancing Act: African
ICT Infrastructure investment options, prepared by Balancing Act,
2004.
Department for International Development, The Impact of the
New Economy on Poor People and Developing Countries,
prepared by KPMG Consulting, July 2000.
http://www.globalisation.gov.uk/BackgroundWord/NewEconomyO
nPoorPeopleDevelopingCountriesKPMG.doc
Dholakia, Nikhilesh and Nir Kshetri, The Global Digital Divide and
Mobile Business Models: Identifying Viable Patterns of e-
Development, University of Rhode Island working paper 2002.
http://ritim.cba.uri.edu/working%20papers/Global-Digital-Divide-
e-Development-Models-v7%5B1%5D.pdf
Dorj, Thinley, ‘IP Based Rural Access Pilot Project’, 2001.
http://www.bhutan-notes.com/clif/bt_rural_access_pilot.html
Duncombe, Richard and Richard Heeks, Information, ICTs and
Small Enterprise: Findings for Botswana, Development
Informatics Working Paper no. 7, IDPM, University of Manchester,
1999.
http://idpm.man.ac.uk/publications/wp/di/di_wp07.shtml
Karen Eggleston, Robert Jensen, and Richard Zeckhauser,
Information and Communication Technologies, Markets, and
Economic Development, Chapter 7 in Kirkman and Sachs (2002).
ELDIS ICT Development portal
http://www.eldis.org/ict/ictcountry.htm
Ernberg, Johan, Universal Access for Rural Development, ITU,
December 1998.
http://www.itu.int/ITU-D/univ_access/telecentres/papers/
NTCA_johan.doc
Forestier, Emmanuel, Jeremy Grace and Charles Kenny, ‘Can
Information and Communication Technologies be Pro-Poor?’,
Telecommunications Policy, 26 (2002), pp 623-646.
Gebreab, Amare, ‘Getting Connected: competition and diffusion
in African mobile telecommunications markets’, World Bank
2002. http://econ.worldbank.org/files/15963_wps2863.pdf
Grace, Jeremy, et al, ICTs and Broad Based Development:
A Partial Review of the Evidence, World Bank Working Paper
February 2001.
http://poverty.worldbank.org/files/10214_ict.pdf
Grajek, Michal, ‘Estimating Network Effects and Compatibility in
Mobile Telecommunications’, Wissenschaftszentrum Berlin für
Sozialforschung, December 2003.
http://skylla.wz-berlin.de/pdf/2003/ii03-26.pdf
Bibliography

67
Guislain, Pierre, et al, Connecting Sub-Saharan Africa: A World
Bank Group Strategy for ICT Sector Development, World Bank
January 2005.
http://info.worldbank.org/ict/ICT_ssp.html
Hammond, Allen L., Digitally Empowered Development, Foreign
Affairs, 80, 2, 96-106 (2001).
http://www.digitaldividend.org/pdf/0201ar04.pdf
Hardy, A, 1980 “The role of the telephone in economic
development” Telecommunications Policy, 4(4) pp278-286.
Hudson, Heather, The Economic and Social Benefits of Rural
Telecommunications, unpublished report to World Bank, 1995.
http://www.usfca.edu/fac-staff/hudson/papers/
Benefits%20of%20Rural%20Communication.pdf
ITU, African Telecommunication Indicators 2004, Geneva 2004.
ITU, World Telecommunication Development Report, Geneva
2003. http://www.itu.int
ITU, Mobile Overtakes Fixed: Implications for Policy and
Regulation, Geneva 2003.
http://www.itu.int/osg/spu/ni/mobileovertakes/Resources/Mobileo
vertakes_Paper.pdf
ITU, World Telecommunication Indicators Database, 2003
update.
http://www.itu.int/ITU_D/ict/publications/world/world.html
ITU, World Telecommunication Indicators 2000/01, March 2001.
ITU, World Telecommunication Development Report: Mobile
Cellular, Geneva 1999. http://www.itu.int
ITU, The Mobilization of Bangladesh, 13/10//99.
http://www.itu.int/newsarchive/wtd/1999/iht10/tem-04.html
Jha, R and S. Majumdar, 1999. “A Matter of Connections: OECD
Communications Sector Productivity and the Role of Cellular
Technology Diffusion, Info. Economics Policy 11”, pp243-269.
Kenny, Charles, The Costs and Benefits of ICTs for Direct Poverty
Alleviation, January 2002.
http://inet2002.org/CD-ROM/lu65rw2n/papers/u02-a.pdf
Khalil, Mohsen, Wireless Opportunities for Developing Countries
and the Role of the World Bank Group, (presentation) June 2003.
http://www.w2i.org/pages/wificonf0603/speaker_presentations/
W2i_Khalil_Presentation.pdf
Kirkman, Geoffrey and Jeffrey Sachs, eds, Global Information
Technology Readiness Report 2001-02, Oxford University Press
2002.
Kirkman, Geoffrey, Carlos Osorio and Jeffrey Sachs, The
Networked Readiness Index: Measuring the Preparedness of
Nations for the Networked World, 2002 (Chapter 2 of above).
Laffont, Jean-Jacques and Tchétché N’Guessan,
‘Telecommunications Reform in Côte D’Ivoire’, World Bank 2002.
http://econ.worldbank.org/files/18862_wps2895.pdf
Laperrouza, Marc, China’s Information and Communication
Technology Policy-Making, Evian Group Compendium, April
2002.
Latchem, Colin and David Walker, Telecenters: Case Studies and
Key Issues, Commonwealth of Learning, 2001.
http://www.col.org/telecentres/
Also Development Gateway portal on telecenters at
http://www.developmentgateway.org/node/133831/sdm/docview
?docid=441647
Lopez, Asbel, The South Goes Mobile, UNESCO Courier,
July/August 2000.
http://www.unesco.org/courier/2000_07/uk/connex.htm
Madden, G and SJ Savage, CEE Telecommunications Investment
and Economic Growth, Information Economics and Policy, Vol 10,
pp173-195, 1998.
Mansell, Robin, Digital Opportunities and the Missing Link for
Developing Countries, Oxford Review of Economic Policy, Vol 17,
No. 2, pp282-295, 2001.
http://icg.harvard.edu/~ss98em/Readings/Mansell-
Digitl_misng_links.pdf
Mansell, Robin and W.E. Steinmueller, Mobilizing the Information
Society: Strategies for Growth and Opportunity, Oxford University
Press 2000.
Nadiri, M Ishaq and Banani Nandi, Telecommunications
Infrastructure for Economic Development, NYU working paper,
January 2003.
http://userpage.fu-berlin.de/~jmueller/its/conf/helsinki03/
abstracts/Nandi_Nadiri.pdf
Navas-Sabater, Juan, Andrew Dymond and Niina Juntunen,
Telecommunications and Information Services for the Poor,
World Bank Discussion Paper no. 432, April 2002.
http://rru.worldbank.org/Documents/1210.pdf
Pigato, Miria, ‘Information and communications technologies,
poverty and development in Sub-Saharan Africa and South Asia’,
World Bank working paper 2001.
http://www.worldbank.org/afr/wps/wp20.pdf
Pitt, Alexander and Christine Zhen-Wei Qiang with Seth Ayers,
The Contribution of ICTs to Growth, World Bank, November
2003. Summary at http://www.developmentgateway.org/
download/223526/Contribution_of_ICT_to_Growth_Viewpoint_D
ecember2003.pdf
Pralahad, C.K. and Allen Hammond, Serving the World’s Poor,
Profitably. Harvard Business Review September 2002. Earlier
version
http://www.digitaldividend.org/pdf/serving_profitably.pdf
Proenza, Francisco, Telecenter Sustainability: Myths and
Opportunities, Journal of Development Communication,
December 2001. Available at:
http://www.fao.org/WAICENT/FAOINFO/AGRICULT/ags/Agsp/pdf/P
roenzaTelecenter.pdf
Quadir, Iqbal Z, Connecting Bangladeshi Villages. Unpublished
paper presented at the Telecommunications for Rural
Development Conference, November 1998, University of Guelph,
Ontario, Canada. (Updated Feb 2000).
http://www.telecommons.com/villagephone/quadir.html
Reck, Jennifer and Brad Wood, What Works: Vodacom’s
Community Services Phone Shops (World Resources Institute),
August 2003.
http://www.digitaldividend.org/pdf/vodacom.pdf
Riaz, Ali, 'Telecommunications in Economic Growth of Malaysia'.,
Journal of Contemporary Asia. Vol. 27, No. 4. December 1997.

68
Richardson, Don, Towards Universal Telecom Access for Rural
and Remote Communities, TeleCommons Development Group,
February 2002. http://www.telecommons.com
Richardson, Don, Ricardo Ramirez and Moinul Haq, Grameen
Telecom's Village Phone Programme: A Multi-Media Case Study,
TeleCommons Development Group, March 2000.
http://www.telecommons.com/villagephone/finalreport.pdf
Rodriguez, Francisco and Wilson Ernest.J., Are Poor Countries
Losing the Information Revolution?, World Bank infoDev Working
Paper 26651, May 2000. http://www.infodev.org/library/
WorkingPapers/wilsonrodriguez.doc
Roeller, L.H. and Waverman, L, 2001 “Telecommunications
infrastructure and economic development: a simultaneous
approach” American Economic Review 91 (4), pp 909-923.
Rossotto, Carlo Maria, Khalid Sekkat and Aristomene Varoudakis,
Opening up telecoms to competition and MENA integration in the
world economy, World Bank Discussion Paper 27024, July 2003.
http://lnweb18.worldbank.org/mna/mena.nsf/Attachments/WP+3
3/$File/WP33.pdf
Rossotto, Carlo Maria, Michel Kerf and Jeffrey Rohlfs,
‘Competition in Mobile Telecommunications’, Public Policy for the
Private Sector, Note 184, World Bank August 1999.
http://www1.worldbank.org/viewpoint/HTMLNotes/184/184rosso.
pdf
Saunders, R., Warford, J. and Wellenius, B., Telecommunications
and Economic Development, 2nd Edition. Johns Hoplins
University Press (for World Bank), 1994.
Sciadas, G, ed., Monitoring the Digital Divide – and Beyond,
Orbicom, 2003.
http://www.orbicom.uqam.ca/projects/ddi2002/index.html
Skuse, Andrew, ‘Information and communication technologies,
poverty and empowerment’, Department for International
Development, Social Development Dissemination Note No. 3,
July 2001. http://www.dfid.gov.uk
Sridhar, K.S. and Sridhar, V. “Telecom Infrastructure and
Economic Growth: Evidence from Developing Countries” (2003)
http://www.nipfp.org.in/working%20paper/wp14.pdf
Stiglitz, Joseph, ‘Economic Organisation, Information and
Development’ in J. Behrman T.N. Srinavasan, eds, Handbook of
Economic Development, North Holland, 1989.
Thioune, Ramata Molo, Information and communication
technologies for development in Africa, (3 volumes),
CODESRIA/IDRC 2003.
Torero, Maximo, The Access and Welfare Impacts of
Telecommunications Technology in Peru, Center for Development
Research (ZEF), Universitaet Bonn, Discussion Papers on
Development Policy No. 27, June 2000.
http://www.grade.org.pe/download/pubs/MT-zef_dp27-00.pdf
Torero, M, Chowdhury, S and Bedi, A, (2002)
“Telecommunications infrastructure and economic growth: a
cross-country analysis”, forthcoming in ICTs and Indian
Economic Development, eds A Saith and M Vijayabhaskar, Sage
Publications 2005.
UNCTAD, E-Commerce and Development Report, 2003.
http://r0.unctad.org/ecommerce/ecommerce_en/edr03_en.htm
United Nations Development Program, Human Development
Report 2001: Making New Technologies Work for Human
Development, Oxford University Press 2001.
http://hdr.undp.org/reports/global/2001/en/
Wallsten, Scott, An Empirical Analysis of Competition,
Privatization, and Regulation in Telecommunications Markets in
Africa and Latin America, World Bank Working Paper 2136,
June 1999.
http://www.worldbank.org/html/dec/Publications/Workpapers/wp
s2000series/wps2136/wps2136-abstract.html
Wauschkuhn, Markus, Telecommunications and Economic
Development in China, University of Bremen Institut für
Welwirtschaft und Internationales Management, Berichte aus
dem Arbeitsbereich Chinaforschung 16, May 2001.
http://www.iwim.uni-bremen.de/publikationen/pdf/c016.pdf
Wellenius, Björn, Closing the Gap in Access to Rural
Communication: Chile 1995-2002, World Bank infoDev Working
Paper, November 2001.
http://www.infodev.org/library/WorkingPapers/chile_rural/Chile%
20-%20final%2017%20december%2001%20-%20revised.pdf
Wellenius, Björn, Extending Telecommunications Beyond The
Market: towards a universal service in competitive markets,
World Bank 2000.
http://rru.worldbank.org/viewpoint/HTMLNotes/206/206welle.pdf
World Bank, The Network Revolution and the Developing World,
(report prepared by Analysys), infoDev Group, August 2000.
http://www.infodev.org/library/WorkingPapers/400.doc
World Bank, World Development Report 1998/99: Knowledge for
Development, 1998.
http://www.worldbank.org/wdr/wdr98/contents.htm
World Economic Forum, Global Information Technology Report
2002-2003 – Readiness for the Networked World, 2003.
http://www.weforum.org/site/homepublic.nsf/Content/Global+Co
mpetitiveness+Programme%5CGlobal+Information+Technology+
Report%5CGlobal+Information+Technology+Report+2002-
2003+-+Readiness+for+the+Networked+World
including chapters on ICT in Africa. http://www.weforum.org/pdf/
Global_Competitiveness_Reports/Reports/GITR_2002_2003/ICT
_Africa.pdf and the Arab world
http://www.weforum.org/pdf/Global_Competitiveness_Reports/Re
ports/GITR_2002_2003/Arab_World.pdf
Other Web References
http://www.infodev.org/
ICT data and information for developing countries at a glance:
http://www.worldbank.org/data/countrydata/ictglance.htm
http://info.worldbank.org/ict/
http://www.developmentgateway.org/node/130667/?

Vodafone survey

More Related Content

What's hot (20)

Similar to Vodafone survey (20)

More from CMR WORLD TECH (20)

Recently uploaded (20)

Vodafone survey