Surveying the Global Diffusion of the Internet

Larry Press, lpress@isi.edu
California State University, Dominguez Hills
USA

William A. Foster, wfoster@bpa.arizona.edu
University of Arizona
USA

Seymour Goodman, sgoodman@leland.stanford.edu
Stanford University
USA

Table of Contents

Efforts to Track Internet Diffusion
A Pilot Survey
Toward an On-Going Global Survey

Abstract

Policy makers and infrastructure planners need an understanding of the global diffusion of the Internet. This paper summarizes work that has been done, presents the results of a recent questionnaire study, and discusses options for the future.

A number of people have been tracking the global diffusion of the Internet. Some focus on single indicators such as the existence of IT connectivity in a nation (Landweber and Crepin-Leblond) or the number of hosts in a nation (Lottor and Quarterman). We (Press and Rodriguez) reported on a survey of several Internet-related indicators at INET 96. Others study one aspect of the Internet in depth (Demchack, government Web sites, CAIDA, global backbone networks), focus on one aspect of the Internet in a nation (Boardwatch Magazine, United States ISPs and backbone networks, Mike Jensen, African ISPs) or do comprehensive, in-depth studies of the state of the Internet in a nation (PAN Asia, Mosaic Group).

In a recent study we used a questionnaire based on the Mosaic Group multidimensional framework for assessing the state of the Internet in a nation. We invited attendees of the 1998 ISOC Developing Nations Workshop to participate, and the results of that survey are presented. The paper also discusses the feasibility of continuing such surveys on a regular basis, asking what information is practical to gather on a global basis, and speculating on the establishment of an online database maintained by a decentralized group of respondents from each nation.


Over the centuries, we covered the globe with cities then linked them with railroads, highways, telephone lines, power grids, canals, and so forth. We are now deploying the Internet. Like other infrastructure before it, the Internet is of interest to policy makers. As Ithiel de Sola Pool emphasized, telecommunication infrastructure planning is implicit social planning. Policy makers may see the Internet as an economic and cultural resource, a threat to political stability and cultural values, or both, but none can ignore it infrastructure and society are inextricably interdependent. While this is the case for all nations, a relatively small networking investment may have a significant impact in developing nations [1].

Policy makers and investors require information about the state of the Internet. Before the Internet became commercialized, the United States National Science Foundation could track its growth and diffusion because they operated the backbone for most of the world, and the Internet was relatively small. This is no longer the case, and today, several organizations and projects are tracking the global diffusion of the Internet [3]. In doing so, they must choose a balance between breadth and depth.

Efforts to Track Internet Diffusion

One of the first to chronicle Internet diffusion was Larry Landweber, who simply noted whether or not a nation had an international IP link. He produced well known maps between 1991 and 1997, graphically showing the Net's growth (ftp://ftp.cs.wisc.edu/connectivity_table/). Olivier MJ Crepin-Leblond has maintained a similar list of connected nations (http://www.nsrc.org/oclb) since December, 1992. Keeping track of only one variable allowed Landweber and Crepin-Leblond to maintain a global perspective at a reasonable cost. Network Wizards, http://www.nw.com, also produces a concise representation of the Internet, automatically counting the number of hosts in each top-level domain every 6 months. Matrix Information and Directory Services (MIDS) begins with Network Wizards' host and domain counts, and analyzes them further, determining the geographic location of hosts. They present the information in a variety of graphic and tabular formats (http://www.mids.org).

Others compile in-depth information on a limited geographic area. For example, Boardwatch Magazine, http://www.boardwatch.com, concentrates on the United States, using interviews, questionnaires, and automated techniques to compile data on every ISP and each IP backbone network. The result is a 560-page directory which requires a professional staff. Mike Jensen, http://www3.sn.apc.org/africa/, surveys ISPs in Africa, gathering less information than Boardwatch, but covering the entire continent. The Costa Rican National Research Network (CRNET) automatically counts Latin American and Caribbean hosts and domains, http://ns.cr/latstat/. Reseaux IP Europeens (RIPE) automatically counts European hosts and domains, http://www.ripe.net/statistics/. Press and Rodgriguez conducted a questionnaire survey of Latin American and Caribbean research and academic networks [2]. The MOSAIC Group (http://www.agsd.com/gdiff) has done in-depth studies of the state of the Internet in several nations by sending teams to interview members of the academic and commercial networking community, telecommunication vendors and regulators, interested political figures, etc. and extensive electronic and print literature review [4]. The Pan Asia Networking (PAN) Program of the International Development Research Centre (http://www.panasia.org.sg/) publishes a yearbook with information on 24 Asia-Pacific nations. The Pan Asia Networking Yearbook has a chapter on each nation with a political, geographic and demographic overview and a description of the regulatory environment, Internet connectivity, local content initiatives, PAN activities, and a Web site and contact list. These chapters less detailed then the MOSAIC studies, but more nations are covered.

Chris Demchack and her colleagues at the University of Arizona maintain a global perspective, but focus on one aspect of the Internet -- government Web sites. They have compiled data on the Web sites of national agencies in nearly every nation of the world and rate them on openness and transparency. (For details on the coding scheme and project, see http://w3.arizona.edu/~cyprg/). CAIDA, The Cooperative Association for Internet Data Analysis, (http://www.nlanr.net/Caida) tracks backbone networks globally. They have created a backbone-link database (link starting and ending points, speed, and operator) and software for graphically viewing and updating it.

We have been discussing organizations that track the global diffuison of the Internet. Other organizations track related information on telecommunication infrastructure (e.g., the International Telecommunication Union, http://www.itu.ch/, and Telegeography, http://www.telegeography.com/), social and economic factors (e. g., the World Bank, http://www.worldbank.org, and the United Nations Development Programme, http://www.undp.org/), and network traffic and performance measurement (e. g. MIDS, CAIDA and Boardwatch).

A variety of techniques are used in these studies. Landweber relied on personal knowledge and reports from the field when a nation connected to the Internet. Network Wizards, MIDS, CRNET, and RIPE use automated techniques to identify hosts and domains. Boardwatch Magazine staff use questionnaires, interviews, and automated tools in their work. Jensen uses personal knowledge, surveys and interviews. Press and Rodriguez used a questionnaire distributed at a meeting and over the Internet. The MOSAIC group does expensive, in-depth research using travel and extensive analyst time. Chris Demchack works with a team, identifying and characterizing each government Web site. CAIDA relies on backbone providers to keep their database current.

A Pilot Survey

One would like to combine the MOSAIC Group's comprehensive understanding of the state of the Internet in a nation with the global coverage of some of the others; however, global coverage using their approach would be very expensive, and it would be difficult to update the results over time. In an attempt to achieve multidimensional coverage while keeping cost reasonable, we devised a questionnaire (http://som.csudh.edu/cis/lpress/gdiff) based on the MOSAIC Group analysis framework.

The framework characterizes the state of the Internet in a nation along six dimensions: pervasiveness, geographic dispersion, sectoral absorption, connectivity infrastructure, organizational infrastructure, and sophistication of use (Table 1). Each dimension has five ordinal values ranging from zero (non-existent) to four (highly developed).

In addition to these dimensions, the framework includes an open-ended list of determinants -- factors which influence the development of the Internet, i. e., the values of the dimensions in a nation. One view of these determinants is shown in Table 2, which organizes them into government policies and non-governmental determinants of Internet success.

A call for participation in this survey was sent to a list of attendees of the 1998 Internet Society Developing Nation Workshop. They were referred to a Web site with information on the study, and, for the convenience of those without IP connections or for whom such connectivity was expensive, we offered to email the questionnaire. The questionnaire was completed by 22 people representing 20 nations. Table 3 shows the values they assigned to the dimensions for their nations. Their full replies with comments and assessments of key determinants are at http://som.csudh.edu/cis/lpress/gdiff/.

Note that in the cases of Pakistan and Uganda we received two questionnaires, and that the respondents for Pakistan differed in their assessments of sophistication of use. This was possible because the questions are involve some judgement and estimation in mapping some aspect of Internet diffusion onto a 1-5 scale. For example, in assessing pervasiveness we do not try to pin down the exact number of users per capita, which would be impractical, but are satisfied with a rough, order-of-magnitude estimate. Our goal is categories which accommodate a wide range of nations, and yield a high degree of consensus among Internet experts from a nation. Two qualified respondents disagreeing leads to further discussion regarding both the nature of the question -- a better, less ambiguous definition of the five levels -- and the respondent's assessment of actual state of affairs in their nation. This discussion was conducted using email and the respondents agreed on a value of 1 for sophistication of use.

Toward an On-Going Global Survey

This experience has led us to consider what might be done to refine the process and scale it up to a global level. Doing so would involve revision to both the questionnaire and the process.

A number of respondents had difficulty resolving ambiguity in the initial wording of the question regarding the pervasiveness dimension. This was resolved by a rewording of the definitions of the five levels. It is important that this sort of adjustment and common understanding take place; however, significant readjustment would call into question the validity of comparisons of the results from different periods. The rapid change in Internet diffusion and technology introduces similar concerns. Will new technologies require redefinition of some dimensions and their levels in the future? Is it reasonable to use the same questionnaire for all nations? We attempted to devise scales which would accommodate both developing nations which were new to the Internet and those of developed nations, but perhaps there should be separate versions for each. This version of the questionnaire also kept the discussion of determinants open ended, and the results are subject to content analysis; however, in the future, we may wish to at least partially codify them as we have done for the dimensions.

While some may see the discrepancy in the initial input of our two Pakistani respondents as a bug, we see it as a feature. Multiple inputs from qualified experts in a nation would lead to greater richness and reliability of results. We asked INET workshop attendees to participate in this case, but there are other groups of qualified experts, for example, representatives of ISOC chapters or organizations administering top-level domains, journalists, or government telecommunication officials. Having several (say 3) respondents from each nation would allow a Delphi approach in which they complete questionnaires, and then have an opportunity to refine their responses after discussion.

Language is also an issue if we wish to reach global coverage. Several people on the initial mailing asked if a French language version were available, and did not continue after learning that it was not. Colleagues at the Centre International Pour Le Developpement de L'Inforoute en Francais are now preparing a French version, but other languages would be desirable.

Regardless of language, it is not likely that a voluntary survey can scale to cover all nations and be repeated on a regular basis. Completion of the questionnaire requires approximately one hour, and our population, Internet experts in a nation, does not have a lot of spare time. As such, it would be helpful if a small honorarium could be offered to a relatively stable group of participants. Rather than passive survey respondents, they should think of themselves as participants in the project with responsibility for project management and refinement of the survey instrument as well as providing data. This relatively stable group could maintain a database on the Internet showing dimension values, with brief explanations for those values, explanation of predictions for changes in value over the coming year, and description of key determinants for their nation. The database would contain tables for respondents, nations, dimension predictions, dimension values with explanations, and dimension explanations in terms of key determinants. Examples of the latter two tables, with Cuban data, are shown in Table 4 and Table 5. (Note the increased dimension values between March, 1998 when those tables were compiled and the recent survey).

We would like to see an ongoing survey of the state of the global Internet. Such an effort would have to reach a balance between the information we would like to have and what is feasible to collect and maintain. A decentralized group of experts from each nation, working across the Internet might be able to accomplish this at a reasonable cost.

 


References

  1. Press, Larry., "The Role of Computer Networks in Development," Communications of the ACM, Vol. 39, No. 2, pp. 23-30, February, 1996, http://som.csudh.edu/fac/lpress/devwins.htm.
  2. Press, Larry. and Rodriguez, Luis., "Toward an Internet Census for Developing Nations," Proceedings of INET '96, Montreal, June, 1996, http://www.isoc.org/inet96/proceedings/f2/f2_3.htm.
  3. Press, Larry., "Tracking the Global Diffusion of the Internet," Communications of the ACM, Vol. 40, No. 11, November, 1997, pp 11-17, http://som.csudh.edu/fac/lpress/articles/worldacm.htm, surveys several of these.
  4. Seymour E. Goodman, Grey E. Burkhart, William A. Foster, Laurence I. Press, Zixiang (Alex) Tan, Jonathan Woodard, _The Global Diffusion of the Internet Project: An Initial Inductive Study_ (Fairfax, VA: The MOSAIC Group, March 1998); Seymour E. Goodman, Grey E. Burkhart, William A. Foster, Arun Mittal, Laurence I. Press, Zixiang (Alex) Tan, _The Global Diffusion of the Internet Project: Asian Giants On-Line_ (Fairfax, VA: The Global Information Technology Assessment Group, December 1998).
  5. United Nations Development Programme, Human Development Report, 1997, Oxford University Press, New York, 1997.

 

 

Table 1: Internet diffusion dimensions

Dimension

Description

Pervasiveness

This is based on the number of hosts and users per capita.

Geographic Dispersion

Over 200 nations now have IP connectivity, but in many of these, access is restricted to one or two large cities. This dimension measures the concentration of the Internet within a nation, from none or a single city to nationwide availability with points-of-presence or toll free access in all first-tier political subdivisions and pervasive rural access.

Sectoral Absorption

While widespread access is desirable, the payoff is in who uses the Internet in a nation. This dimension assesses the degree of Internet utilization in the education, commercial, health care, and public sectors. These sectors are seen as key to development, and were suggested by the measures used by the United Nations Development Programme Human Development Index [5].

Connectivity Infrastructure

This measure is based on international and domestic backbone bandwidth, exchange points, and last-mile access methods. A highly rated nation will have high speed domestic and international backbone connectivity, public and bilateral exchange points, and a high proportion of homes with last-mile access using CATV, DSL, or some other technology that is faster than analog modems.

Organizational Infrastructure

This dimension is based on state of the ISP industry and market conditions. A highly rated nation would have many ISPs and a high degree of openness and competition in both the ISP and telecommunication industries. It would also have collaborative organizations and arrangements like public exchanges, ISP industry associations, and emergency response teams.

Sophistication of Use

This variable ranks usage from conventional to highly sophisticated and driving innovation. A relatively conventional nation would be using the Internet as a straight forward substitute for other communication media like telephone and FAX, whereas in a more advanced nation, applications may result in significant changes in existing processes and practices and may even drive the invention of new technology.

 

 

Table 2: Factors and policies influencing Internet success within a nation.

Internet Success Determinants

Government Policies

 

 

Table 3: Internet-diffusion dimension ratings

 

Nation

P

GD

SA

CI

OI

SU

Evaluator

Armenia

2

2

2

1

2

2

Tigran Nazarian

Benin

1

1

1

1

2

1

Ken Lohento

Burkina Faso

1

2

1

2

2

1

Kiswendsida Kisito Kabore

Cameroon

1

1

1

1

2

2

Derek Ajesam Asoh

Chile

3

2

2

2

3

2

Claudio Araya

Cuba

1

2

1

1

2

2

Marta Ruiz

Dem. Rep. of Congo

1

1

1

1

1

1

Eric Nzita

Dominican Republic

2

3

1

1

3

2

Daniel Pimenta

Guinea

1

2

1

1

3

1

Hadja Binta Keita

Guyana

2

1

1

1

2

2

Andrew Mancey

Mauritius

1

1

2

1

1

2

Yann Kwok

Pakistan

1

3

1

1

2

2

Fasih Ahmed Masood Sindhu

Pakistan

1

3

1

1

2

1

Naseem A. Bhatti

Peru

2

4

3

1

2

3

Jose Soriano

Solomon Islands

1

1

1

1

1

2

Samuelu Taufao

Sri Lanka

1

2

2

1

3

3

Priyantha Pushpa Kumara

Tunisia

1

2

2

1

3

3

Lamia Chaffai Sghaier

Uganda

1

1

1

1

-

2

Michael Sserunjoji

Uganda

1

1

1

1

-

2

Dorothy Okello

Uruguay

3

3

3

2

2

3

Ida Holz

Venezuela

2

3

1

1

3

2

Luis German Rodriguez

Zimbabwe

2

3

3

1

3

2

Joyce Chidora

 

Table 4: Explanation of Cuban dimensions, March, 1998.

Dimension

Value

Explanation for value selection

Pervasiveness

1

IP connectivity is minimal, with perhaps as few as 100 users. Even if we were to include UUCP email accounts, there are less than 1/1,000 population. It is noteworthy that email use extends well beyond the network technician community.

Geographic Dispersion

1

The only IP point of presence offering network connectivity in Cuba is at CENIAI in Havana. If; however, we were to consider email connectivity, we would find access in every province and nearly every municipality. Cuba is clearly interested in geographic dispersion.

Sectoral Absorption

1

IP connectivity is rare (under 10%) in the health and government sectors, and nonexistent in education and commerce. If we were to consider UUCP, email is used in the health sector throughout the nation, more than 10% of the government ministries, and Youth Computer Centers have educational accounts in all municipalities.

Connectivity Infrastructure

1

While Cuba has an international IP link, they have no domestic backbone and barely any leased line access. They are severely hampered here by poor telephone infrastructure and their historical concentration on X.25.

Organizational Infrastructure

2

CENIAI and Teledatos are both in the business of providing connectivity to organizations with networks, and there is some degree of competition between them (either by design or historical development). There is also a degree of coordination provided by the Inter-ministerial Commission for Networking.

Sophistication of Use

1

As there is little IP connectivity, Cuba must be ranked at level 1; however, email and information retrieval from email-driven servers have reached level 2 in the health care and biotechnology communities.

 

 

Table 5: Key determinants and the dimensions they affect, using Cuba as an example.

Determining Factor

Dimensions most directly affected

Poor telephone infrastructure

Pervasiveness, geographic dispersion, connectivity infrastructure, and sophistication of use are all inhibited due to the difficulty in connecting end users and networks.

Difficulty attracting capital

Connectivity infrastructure cannot be improved without capital.

Cultural values stressing health, education and equality

Health and educational sectoral absorption is emphasized as is geographic dispersion outside the capital.

Centralized planning

Pervasiveness is reduced by planning delays, and the Inter-ministerial Commission was formulated) is an element of increased organizational infrastructure.

Concern for national security given US hostility

Pervasiveness is reduced by access restriction.

Protection of embargoed business activity

Pervasiveness is reduced by content restriction.

Propaganda to and from US

Pervasiveness is reduced by content and access restriction.

Fear of use by subversive organizations

Pervasiveness is reduced by access restriction.

Non-commercial economy

Commercial sectoral absorption is inhibited as resources are shifted elsewhere.

Populist history

In seeking to server rural areas and small towns, geographic dispersion is increased and connectivity infrastructure extended outside the capital.

Emphasis on human capital

Education sector absorption is increased.