Media Aristocracies, Network Resources, and the Global Digital Divide

Seven Bridges Over the Global Digital Divide

Davis Foulger

Visiting Associate Professor
Oswego State University
Oswego, NY

Evolutionary Media

January 12, 2002

Originally Presented at the IAMCR & ICA Symposium on Digital Divide: November, 2001.

This paper can be downloaded page formatted in MicroSoft Word.

Comments to this paper can be posted to The Digital Divide page on my Wiki .


On May 17, 2001, U.N. Secretary-General Kofi Annan called for "a bridge that spans the digital divide". This digital divide separates a few hundred million users of the Internet, most of them located in a small number of "digital-have" countries, from over 5 billion people who are unable to access the wide variety of Internet digital media. There is nothing new about media divides. Innis documents at least four millennia of media divides and priesthoods in his "Empire and Communication". What is new is our willingness to make an issue of the existence of a media divide, digital or otherwise. This paper asserts that there are seven distinct sets of obstacles that must be overcome if we are to bridge the global digital divide. All seven bridges are discussed in detail. Only one, "Choice", is excluded from detailed analysis. The other six bridges, "Social and Legal Constraints", "Economic Priorities", "Basic Infrastructure", "Literacy and Language", "Network Infrastructure", and "Computer Resources", are modeled and and explored using multiple regression. The model, which asserts specific relationships of the six bridges to Internet Use, proves reasonably successful. Indeed, it suggests a general plan of action for bridging the digital divide in digital have-not countries. A simple and easy to understand index of Internet Readiness is derived from that model. This Internet Readiness index, which is built entirely from measures of the six bridges examined in the path model, appears to be able to predict a country's level of Internet Use with a high degree of accuracy (a .91 correlation), and may allow countries that are attempting to bridge the digital divide to predict the growth in Internet Use within their countries based on their success in crossing the intermediate bridges. Indeed, the index suggests that, once a baseline is satisfied, each new bridge that a country crosses should result in an order of magnitude increase in Internet Use.


If we accept the assertion that the first "computer-mediated communication system for dispersed human groups" was designed and implemented in 1970 (Hiltz and Turoff, 1978, p. 43), we are barely thirty years into the age of digital media. By many measures, the reach of such systems has expanded greatly since then, with over 190 million people (or 5% of the world's population) connected to the Internet by 1999 (Annan, 2001). There is little question of the growing importance of computer-media in the conduct of the business around the world. It should not be surprising, however, given the relatively low percentage of the world's population that uses the Internet, that the increasing importance of digital media has been accompanied by an increasing awareness of substantial inequities in the distribution of Internet access, with some 85% of Internet Users, and 90% of Internet hosts, located in a relatively small number of developed countries (Annan, 2001).

There was early awareness of impending inequities in access to networked Computer Resources. Hiltz and Turoff noted in 1978 that "whenever a useful new technology is developed, one policy question that should be vigorously pursued is how to make it available to those who cannot afford to buy it themselves" and propose that "a purposeful public program might be designed to make the new communication medium serve the disadvantaged rather than compound their disadvantages" (p. 167). This call was renewed in 1992 at a National Research Council Workshop on the "Rights and Responsibilities of Participants in Networked Communities". The workshop report (National Research Council, 1994) discusses the need to provide "everyone access to the electronic highway in order to close the gap between the information 'haves' and 'have nots'", but notes that simple access may not make much difference if the information available through the technology must be purchased" (Chapter 1). Another noteworthy outcome of these meetings is a recasting of network access as a basic human right in Sara Kiesler's assertion that "the purest form of censorship is the absence of access. If you can't have access to a network at all, then you are completely censored from that forum" (Chapter 4).

One can debate the overall effectiveness of subsequent efforts to resolve issues of unequal access in the United States and other developed countries, and there is clearly room to do better (Office of Policy Analysis and Development, 1998). It remains the case that, at least in the United States, physical access to the "information superhighway" is no longer the primary issue. Thanks to government initiatives like "e-rate" (Cattagni and Farris, 2001), industry initiatives like "netday" (NetDay, 2001), and foundation initiatives like the "Bill and Melinda Gates Foundation Library Program" (Bill and Melinda Gates Foundation, 2001), the United States is at least approaching universal Internet accessibility through public use computers in community centers such as public schools, libraries, shopping malls, and other publicly accessible locations (Office of Policy Analysis and Development, 1998). Public school access is particularly strong, with over 98% of public schools and 77% of classrooms connected to the Internet (Cattagni and Farris, 2001). Large disparities of access still exist, particularly for the rural poor, but community access centers have made Internet access a possibility for almost anyone in the United States.

One cannot reasonably, however, debate the reality of the far deeper global "digital divide" that U.N. Secretary-General Kofi Annan describes in his May 17, 2001 call for "a bridge that spans the digital divide". Annan's message is extremely effective in describing the extent of this divide:

Today, there are almost as many hosts in France as in all of Latin America and the Caribbean, and there are more hosts in Australia, Japan, and New Zealand combined than in all the other countries in the Asia-Pacific region combined. Perhaps most telling, there are more hosts in New York than in all of Africa.

Less than 10% of the world's population currently uses the Internet, and most of those users (85%) are located in the small number of developed countries "where ninety percent of Internet hosts are located" (Annan, 2001). Bridging this global digital divide will prove much more difficult than providing universal Internet access has proved in the United States and other developed nations. This paper explores those difficulties by discussing seven obstacles that need to be overcome if there is to be a true global bridge across the digital divide.

A History of Media Divides

In 1984 the first of what would become a series of young Russian-immigrant programmers came to work at a major corporation's research lab. After taking a lead role in building a high speed backbone for what we now call the Internet, he subsequently left the major corporation to take a lead technical role at Cisco Systems. Bottom line, he was a highly competent network programmer and architect whose work has literally made the Internet possible and the problem we are considering here, the "Digital Divide", an issue. Like all of the young Russian-immigrant programmers this lab would hire over the next few years, he was trained, in Russia, as a programmer, but he never actually used, or even saw, a computer until he arrived in the United States. He had learned to program from classes and books, like many Russian-trained programmers of that era, and had practiced his art by writing his programs on paper and "bench-checking" them with others looking on. It was an effective system, particularly in a country which had few computers of any kind, and none to spare for classrooms, but it represents a most profound expression of the digital divide, one in which even the digital elite (those who are trained to build, maintain, and be expert users of digital systems), have little or no access to digital systems.

I offer this story because it illustrates several key elements of the digital divide as it exists today. First, the Internet is the domain of the literate. A person who cannot both read and write cannot have any reasonable expectation of making practical use of either a computer or the Internet with any facility. Second, a literate person doesn't need to have full time access to a computer to develop the skills requisite to making use of the Internet. Third, even a person who is literate and trained in digital media may not be able to actually take advantage of that training if computer and network resources are not available. Fourth, a person who makes the effort to learn how to use the Internet can quickly become a productive user of Internet resources once they have access to them. And finally, there are profound disparities of access to Internet resources in this world, in part because of hard economic realities. A bridge across the digital divide can easily be a low priority for a country with limited resources when compared with budget priorities like education and literacy, security, AIDS and other health issues, overpopulation, housing, hunger, and providing jobs for large numbers of undereducated people.

There is, of course, nothing new in this story. Innis (1950) teaches us that restricted access to new media is more the rule than the exception. Innis may be right in presuming that such restricted access reflects a conscious attempt, on the part of a media elite, to preserve power. It might just as easily be the case, based on Innis' own evidence, that media like stone, clay, papyrus, and parchment were simply expensive to use given the technologies of their times. It could easily take more than a person-year of effort to create a single papyrus scroll before it could even be written on, and that at a time when everything people did, including agriculture and food preparation, was far more labor intensive than it is for most of us today. Other Economic Priorities competed with these storage media, much as they compete for the attention today.

The digital divide is not the only information divide we face. There are divides of various sorts for virtually all contemporary media. Newspapers, magazines, and books all require literacy, cost money, and entail substantial publishing and distribution infrastructures. Radio and television only require equipment, but acquiring a high quality signal and wide range of viewing choices will usually require a cable or satellite subscription. Most media entail a media aristocracy which, through some means of selection, has acquired skills and requisite access to a particular mediumís means of production. Even the obvious exceptions to this general rule, including face-to-face interaction, snail mail, and the telephone, entail some degree of unequal access. There are economic, infrastructure, and other prerequisites to telephone use (including the ability to hear) that deprive many people of telephone access. Indeed, it is a sign of the United States success in bridging the digital divide that schools are networked to the Internet at a higher rate (98%) than homes are connected to the telephone network (97%) (Cattagni and Farris, 2001; Office of Policy Analysis and Development, 1998). There are economic, infrastructure, and other prerequisites to use of a postal system (including literacy) that deprive many people of postal service (reliable or otherwise). Even actors in face-to-face interaction face inequities of access based on restricted physical modalities, shyness, varying levels of interpersonal communication competence, and community actions including shunning.

Hence our willingness to make an issue of the digital divide is something new in the history of media. Never in the history of media has the existence of a media aristocracy been a matter of serious debate or change agency. Success in this endeavor will have serious future implications that will affect international law and the meaning of national boundaries. It is difficult to see, for instance, how a truly global Internet can be policed without international laws that reach down all the way to people and organizations, without global enforcement agencies that can operate across international borders, and without an International judiciary that becomes, at some point, the court of highest resort. Those implications, if real, will be resisted by at least some people, including many people in the United States. Indeed, it can be expected that the usual historical practice of trying to preserve existing media aristocracies (Innis, 1950) will continue, in at least some countries and other social collectives (churches, communities, companies, etc.), through mechanisms of social organization and law. While it is already evident that something approaching universal Internet access is becoming reality for large numbers of people, it is doubtful that a bridging of the digital divide will entirely avoid this historical reality.

Seven Bridges over the Global Digital Divide

Annan (2001) identifies "language", "infrastructure", and "cost" as three problems that must be resolved if we are to bridge the digital divide. He is optimistic. There are at least seven fundamental problems to be bridged (Figure 1) if we are to make generalized Internet access a global reality.

Figure 1: Seven Bridges Across the Digital Divide.

The obstacles that must be bridged can be described as follows:

Figure 1 suggests that these difficulties must be resolved sequentially (that one bridge must be crossed before the next matters). This is true to some extent. The combination of digital communication being an Economic Priority, Basic Infrastructure being in place, Literacy attained, Network Infrastructure in place, and computer resources available won't give you the choice to be an equal player on the Internet if legal and social restraints prevent you from using the full range of Internet resources. A fully functional network and computer won't give a literate user the ability to use the Internet if Economic Priorities divert substantially all of a countries power resources to military and industrial use, or if the power grid shuts down because of lack of fuel. It remains, however, that the ordering of the bridges is somewhat arbitrary. The question is not so much of crossing all the bridges in a particular order as of crossing all the bridges. Some level of parallelism is certainly possible.

This paper will explore these bridges using income, literacy, educational expenditure, and media access data from UNESCO, with data on social systems, infrastructure, and other data drawn from additional sources. It will integrate these divides into practical measures of a countries Internet-readiness, and it will propose practical solutions for reducing the global digital divide based on the reality of this data and the availability of social, economic, and technology solutions.


It is possible to regard the bridges outlined above as a set of hypotheses: that Internet Use, in a given country, is a function of Social and Legal Constraints, Economic Priorities, Basic Infrastructure, Literacy and Language, Network Infrastructure and connectivity, and Computer Resources as follows:

  1. Societies that impose fewer Social and Legal Constraints on individuals and organizations will generally experience greater Internet Use.
  2. Societies with more resources, and therefore less pressure to make forced choices between competing priorities, will generally experience greater Internet use.
  3. Societies with more complete and larger capacity Basic Infrastructures for providing such services as electric power will generally experience greater Internet Use.
  4. Societies that are more literate will be more likely to experience greater Internet Use as a function of both the percentage of the population that is literate and the extent of their Literacy (e.g. more education, ability to read and write in multiple languages, etc.)
  5. Societies with more extensive media infrastructures, and especially substantial telephone infrastructures, will be more likely to experience greater Internet use.
  6. Societies with greater levels of computer use are more likely to experience greater Internet Use.

No hypothesis is offered for the relationship of Choice to Internet Use. The effects of Choice are, in a literal sense, a matter of self-fulfilling prophecy. Those who choose not to use the Internet will not do so. Those who can and do choose to use the Internet will.

A variety of secondary hypotheses are also be proposed as a function of the ordering of the bridges. Rather than outline each specifically, we simply note that:

  1. Computer use should be a function of Social and Legal Constraints, Economic Priorities, Basic Infrastructure, Network Infrastructure, and Literacy and Language in the same directions that should be the case for Internet Use. Among these, only Network Infrastructure might be reasonably argued to not be a prerequisite to computer use, as computers can be useful without Internet connectivity. If, however, the purpose, in obtaining a computer, is to access the Internet, than Network Infrastructure is, in fact, a dependency.
  2. Network Infrastructure and connectivity should be a function of Social and Legal constraints, Economic Priorities, Literacy, and Basic Infrastructure in the same directions that should be case for Internet Use. Among these, only Literacy might be reasonably argued to not be a prerequisite to network infrastructure. To the extent, however, that one must be able to read a manual and understand letters and numbers in order to look up a telephone number and use a telephone, then literacy is, in fact, a prerequisite to network infrastructure.
  3. Literacy should be a function of Social and Legal Constraints, economic priorities, and Basic Infrastructure in the same directions that should be the case for Internet Use.
  4. Basic Infrastructure should be a function of Social and Legal Constraint and Economic Priorities in the same direction that should be the case for Internet Use.

No hypothesized relationship is offered for the relationship of social and legal constraint and Economic Priorities. It is tempting to propose that economic priorities will be a function of Social and Legal Constraints, as less constraint provides more opportunities for ideas to compete and distributed political power allows everyone to participate in setting Economic Priorities. One might just as easily counter, however, that reduced competition for economic resources makes it easier to reduce social and economic constraints and encourages the demand for fewer constraints. In truth, both perspectives are correct. While it is tempting to say that the hypotheses should not be connected because economics and governance are different dimensions of a social system, economic growth and minimum constraint governance exist in a positive feedback loop.

These hypothesized relationships are summarized in Figure 2.

Figure 2: Hypothesized relationships among the bridges across the digital divide.



A data set, described below, was assembled that provides representative data for six of the seven bridges outlined above. This data was explored using a combination of multiple regression (path analysis) and descriptive statistics. There were high levels of missing data for some variables (see below). Hence where statistical analysis entailed dealing with missing values, they were dealt with on a pairwise basis. This can be an issue, especially where variables are used to construct new composite variables. As a result, several composite measures, and several variables that might have been used in their construction, ultimately were not used in the analyses. All multiple regressions were stepwise. While conventional standards of significance were used, all accepted significances were at the p<..003 level or better.

Note that the data analysis done here is intended to be exploratory. While hypotheses have been offered and shaped into a formal model, the study does not pretend to have data that is adequate to fully testing the model. Data has been collected, as described in the next section, and will be analyzed as outlined here. The intent, as will be seen in the analysis, is to explore the model as a way to better understand the obstacles to bridging the global digital divide while both suggesting ways in which the model might be more thoroughly tested and exploring any hints the model may provide as to how one might more effectively bootstrap digital have not countries into digital have countries.


UNESCO has supplied several data sets which are available on the UN web site (UNESCO, 2000). I have drawn on two of these data sets: Public current expenditure on education, and Literacy, culture and communication. The former table provides a variety of measures of public spending on education for approximately 190 countries. The latter provides measures of literacy and communication infrastructure for the same countries. These tables were merged, with unrelated and duplicate data removed. Additional data was then added from the Nations of the World section of the on-line edition of the World Almanac and Book of Facts (Anonymous, 2001). Over 190 articles on individual countries provided a wealth of additional data concerning the population, power generating capacity, gross domestic product, and average lifespans of these countries, as well as backup data on communication infrastructure which was used to fill in a number of blanks in the UNESCO data.

These data sources do not line up entirely. Several countries listed in the UNESCO data, including Hong Kong, Netherlands Antilles, the Palestinian Autonomous Territories, the Cook Islands, the British Virgin Islands, and Macau, do not appear in the World Almanac. Because this data is essential to the analysis, these countries have to be excluded from the analysis. Several countries listed in the World Almanac and Book of Facts, including the Marshall Islands, Nauru, Liechtenstein, Taiwan, Palau, and Andorra, do not appear in the UNESCO data. Because the critical UNESCO data is replicated in the World Almanac and Book of Facts, however, it was possible to retain these countries in the analysis. Indeed, a number of other missing data points in the UNESCO data were filled in from this source, and it was probably possible to do the entire analysis based on this one source.

The result is a data set representing 190 countries that contains key data points representing most of the "bridges" outlined above. Variables in the data set include:

Some observations about the variable set:

Correspondence of the Data with the Bridges

The relationship of data set variables to Internet Use and the bridges across the digital divide that are documented in the hypotheses are as follows:

Social and Legal Constraints Democratic government
Economic Priorities

GDP per Capita (Gross Domestic Product)
Average Lifespan

Basic Infrastructure Domestic electric power production per Thousand.
Language and Literacy


Network Infrastructure Main telephone lines per Thousand people.
Computer Resources Personal Computers per Thousand people
Internet Use Internet hosts per hundred thousand people

These correspondences can be used to convert the hypothesized relationships of Figure 2 into a path diagram that maps the bridges across the digital divide to specific variable relationships as shown in Figure 3.

Figure 3: Application of the relevant variables available in the data set to the hypothesized relationships among the bridges across the digital divide.

The First Bridge: Social and Legal Constraint

The genius of Apple's 1984 "Big Brother" superbowl ad is that it captures the obvious irony of the personal computer revolution: the computers and media technology that had once been feared as the enslaving tools of a "big brother" have in fact become liberating tools that enable a level of human expression and freedom beyond anything known before. Anyone with free and open access to Internet digital media has the ability to freely publish and express themselves, without the editorial processes associated with traditional mass media, such that they can reach a mass audience. Anyone with free and open access to Internet digital media has the ability to find and view such content from anyone who has posted such content. Anyone with free and open access to Internet digital media has the ability to interact with any other user of Internet digital media so long as both parties consent to the interaction.

The key to such freedom is in combining "free and open" with "access to digital media". Such is not the case everywhere. Social and Legal Constraints that make access to the Internet less than "free and open" exist in a variety of countries and organizations. Examples include:

This study has collected only one variable that pertains to the issue of social and legal constraint: "Government a Democracy?" This variable should be regarded as an indicator of the extent to which a country is likely to prosecute people for the expression of their ideas, with democracies regarded as being considerably less likely than totalitarian governments (whether expressed as communism, military dictatorship, absolute monarchy, or other form) to formally restrict or prosecute the free expression of ideas. Free expression (i.e. expression without substantial Social and Legal Constraint is a requisite element of democracy. One cannot have a true democracy unless people are free to express a diverse range of competing ideas. It is difficult, moreover, to maintain a totalitarian form of government in the face of such expression. Democracy is not the only possible measure of Social and Legal Constraint, but it is a strong correlate of such constraint. Democracy does not, and arguably should not, assume or assure an equal distribution of resources, but its prevalence in today's world makes it possible to ask, for the first time in human history, if is it possible to create a new dominant media environment without also creating a priesthood of media haves.

This variable is exogenous within the scope of the model depicted in Figure 3. "Government a Democracy?" is, as might be expected given this discussion, positively correlated with all of the other model related variables that are described in this study.

The Second Bridge: Economic Priorities

It is pleasant to imagine that the Internet is free. It is not. Yes, it is possible for individual Internet Users to access a broad array of "free" content and services. E-mail, instant messaging, listservs, newsgroups, portals, personal web sites, electronic greeting cards, streaming audio, streaming video, MUDs, MOOs, wikis, search engines, and downloadable music and software can all be accessed at no charge. Web sites can be built on "free" web servers using freeware and shareware web page editing software. All wonderful, and none it really free, especially when one considers costs in the context of a country's per capita Economic Priorities.

Somebody has to pay the up front costs associated with buying a computer or other digital appliance for accessing the Internet. Somebody has to buy, set up, and operate the servers that Internet services run on. Somebody has to build, operate, and maintain the last mile infrastructure and ISP access infrastructure that allows people in homes, schools, offices, and community access centers to access the Internet. Somebody has to build, maintain, and operate the core network (NSP) infrastructures that tie all of the ISP, corporate, government, education, and other localized Internet infrastructures (e.g. the world) together. Somebody has to pay the ongoing costs, including power, buildings, people, and maintenance, that keep the system running.

Nobody has, at this point, figured out how you can provide a computer for less than $500, a useful digital appliance (device, keyboard, and display) for less than $150, a server for less than $600 ($1800 is probably a better low end figure for practical servers), a commercial grade modem for less than $150, or a secure high speed router for less than $2000. All of these elements and more will be a necessary part of almost any Internet access infrastructure, and all are on top of the ongoing costs (in people, electricity, telephone network charges, etc.) of actually running the network on a day to day basis.

It doesn't matter, when considered from the perspective of a country's per capita Economic Priorities, who pays. Indeed, in normal practice the costs are distributed across a range of economic decision makers, each of which makes prioritization decisions for the portion of per capita resources that they control. These decision makers include individual people, families, businesses (ranging from small to large), schools, community groups, and other institutions, including local, regional, or national government. It only matters that they have to be paid within each country that participates in the Internet, and generally from within the resources that are available within that country.

Figure 4: Distribution of Gross Domestic Product (GDP) per Capita across countries. GDP measures the value of all the goods and services produced within a country within a year. GDP per capita states that value in terms of the value created for each person.


As can be seen in Figure 4, the distribution of per capita gross domestic product across the 190 countries in the dataset is very unequal. The 23 richest countries in terms of GDP per Capita all boast average production of at least $20,000 per capita. The 23 poorest countries all have an average GDP per capita of less than $1000. 45 countries have GNP in excess of $10,000 per capita. 63 have a GNP of less than $2000 per capita. The 75 countries that produce more than $5000 of goods and services per person contrast with 115 countries that produce less.

These disparities are not just differences in income. They are differences in options. The people, businesses, schools, community organizations, and governments among whom the fruits of this production are, even within countries, unequally distributed, must make choices about how they spend their share. Food, shelter, security, health, infrastructure, and education are but a few of the priorities that must be supported to some reasonable level. Management of these competing priorities can be relatively easy for people and organizations in countries with comparatively high GDP's per capita. Choices among competing Economic Priorities becomes more difficult as the average GNP declines.

Figure 5: Distribution of Lifespan across countries.


It isn't clear that a country with limited resources, one in which people are already making forced choices between health, education, food, infrastructure, security and other priorities, can trivially divert those resources to closing the digital divide. A country that has a per capita GDP of less than $1000 and an HIV rate among adults of 25% or higher has far bigger problems to solve than providing access to Internet resources. Indeed, it is difficult to see how this kind of health care crisis can be met given such restricted resources. Unfortunately, this situation is all too real in a number of countries.

There are good reasons why digital media can, should, and will be at least a limited budget priority for some in these countries. Businesses will still need to function, and some of those businesses will need to use digital technology to communicate with customers, suppliers, and business partners in other countries. National governments will want to encourage both tourism and investment from outside the country, and will need to communicate with the international community, including banks, other governments, and various international organizations. Community organizations will still need to raise money and coordinate their efforts with others. Those with international sources and/or affiliations will frequently depend on Internet resources in those efforts. Educational institutions will find the Internet to be one of their most cost effective library resources. Individuals may find Internet access to be essential for business, education, financial, communication, or other reasons.

In countries in which per capita economic resources are limited, however, it remains that digital resources will necessarily be secondary to other spending priorities, including food, shelter, health, safety, and security needs. Such restricted resource environments are a breeding ground for the kind of entrenched priesthoods that Innis' documents in "Empire and Communication". If one needs substantially all of the effort of a population to satisfy one's food, shelter, security, and ceremonial needs, one will, as the Pharaohs did, establish a small cadre of scribes to handle one's writing needs; as the Greeks did, restrict writing to the upper classes; as pre-civil war Americans did, deny education and writing to most slaves and women.

The evaluation of Economic Priorities will necessarily involve two very different kinds of measurement. One set of measurements looks at the resources a country has at is disposal. A second set of measurements looks at the challenges associated with allocating resources. The more challenged a country is in a particular area (e.g. health care, national security, government stability, factional infighting, crime, etc.), the more important it will be to allocate resources in that area rather than somewhere else. The less challenged a country is in all areas, the easier it will be for its economic decision makers to invest in new technologies, including the digital resources.

This study will use one variable of each type. "GDP per Capita", which has already been discussed at length here, will be used to characterize the resources a country has at its disposal. "Lifespan in Years" will be used to look at one of the more general challenges that many countries face. There are, of course, a large number of things that can effect the average lifespan of an individual. As a result, lifespan provides what is in effect a composite measure of nutrition, disease, health care, workplace safety, government stability, freedom from war and terrorism, and challenges. The lower a countries average lifespan, in general, the higher the level of demand for essential services. Economic decision makers in these countries will have a more difficult time finding the resources to invest in digital infrastructure, resources, and services.

These variables are exogenous within the scope of the model depicted in Figure 3. Both "GDP per Capita" and "Lifespan in Years" is, as might be expected given this discussion, positively correlated with all of the other model related variables that are described in this study (e.g. GNP's and longer lifespans are positively correlated with higher levels of Basic Infrastructure, Network Infrastructure, education, computer use, etc.)..

The Third Bridge: Basic Infrastructure

Technology-based media abhor an infrastructure vacuum. This was as true for clay tablet and papyrus-based media millennia ago as it is for paper and printing press or broadcast-based media today. Clay tablet media depended on infrastructures of clay tablet production, firing, and storage. Papyrus-based media depended on infrastructures of papyrus and ink production, storage, transportation, and, for at least some percentage of production, retail sales of both production materials and finished documents. Today's publications depend on far more intricate infrastructures, including buildings, roads, transportation systems, home delivery systems, retailers, postal systems, telephone systems, billing systems and the electrical power generation and distribution infrastructure. Broadcast systems are similarly dependent on various infrastructures, including package delivery, billing, retail distribution (or radios, televisions, etc.), and electrical power generation and distribution.

Internet-based digital media are particularly dependent on electrical power infrastructure. Every device that powers the Internet, from servers to routers to end user workstations, depends on electrical power for operation, and other infrastructures (most notably the Network Infrastructure) are frequently built on top of the local electrical infrastructure. This second use of the electrical power distribution infrastructure as a basis for the telephone Network Infrastructure complicates "alternative energy" solutions to bridging the digital divide. Local power generation capacity in whatever form (solar, geothermal, wind, hydroelectric, batteries, fuel cells, etc.) may very well allow an isolated community to operate one or more personal computers, but it does not provide a ready means for connecting those computers to local, national, or international networks. We are, perhaps, getting ahead of ourselves, but it remains that the electrical distribution infrastructure provides an important building block on which other infrastructure elements can be built.

This study uses "Electricity Production in kMh per capita" as an approximation of electrical power infrastructure. This can be an imperfect measure of infrastructure. At least some countries import some portion of their electricity from other countries. Even where a country produces its own power, it may distribute it unequally. In some cases distribution infrastructure may be constrained to cities. In others electricity production may be focused toward satisfying the needs of business, industry, and other national priorities.

It remains, however, that countries that produce more electricity per capita are in a better position to support the electrical power requirements of an Internet infrastructure than countries that do not; are more likely to have a broad distribution infrastructure than countries that do not. Table 1 shows the results of a regression of Electricity Production in kMh per Capita against the exogenous variables of the first and second bridges. Only one of the three exogenous variables, GDP per Capita, proves to be a significant predictor of electricity production, but its predictive power is substantial, with the resulting model accounting for 56% of the variance in electricity production (F=234.187, p<.000).

Model Unstandardized Coefficients Standardized Coefficients t Sig.
B Std. Error Beta
(Constant) -3.260E-02 .270   -.121 .904
GDP per Capita 4.110E-04 .000 .754 15.303 .000
Table 1: Regression of Electricity Production in mHh per Thousand against GDP per Capita, Lifespan, and Government a Democracy?


The Fourth Bridge: Literacy and Language

At its current stage of development, the Internet is effectively unusable by people who cannot read. While audio and video streaming media have made substantial inroads on the Internet in recent years, Internet media are fundamentally oriented to text. The most widely used media, including e-mail, Instant Messenger-style chats, and web pages, are almost entirely text based, and even Internet radio and video generally depend on text-based user interfaces for users to make listening and viewing choices.

This may change. Voice-synthesis and recognition may well be successfully combined with iconic user interfaces to create Internet User interfaces that minimize or eliminate the requirement that Internet Users be literate. It is certainly the case that experiments in building such interfaces are underway. The experiments are hardly new. Work on computer voice recognition systems goes back to the 1970's, and speculation of imminent practical voice recognition systems goes back to the beginning of the 1980's. Workable voice recognition systems have been in available at reasonable prices since the mid-1990's. For whatever reason, however, they have never really taken off. There are reasons for this:

All of these issues are magnified for Internet digital media:

One could go on. Text and graphical rendering of digital content is generally more forgiving of user inattentiveness than voice rendering is. The rendering of graphics, including icons, is particularly difficult in a voice only environment. The substitution of icons for text does not negate the need for literacy. It simply changes the vocabulary that one needs to be literate in. Where icons are used, moreover, they are considerably more effective in representing nouns (objects, documents, etc.) than they are in representing verbs (actions, commands, etc.). The obstacles to enabling Internet access for the illiterate remain high, and it isn't clear that they can ever be overcome entirely via technological means.

High illiteracy rates in many parts of the world make this a difficult bridge for many countries to cross. Literacy rates for different countries around the world vary from 14% to 100%, but these numbers can be misleading. Many countries claim, as can be seen in Figure 6, to have very high literacy rates. It isn't clear, however, that literacy is defined consistently from one country to the next. Many countries that claim high literacy rates claim those levels of literacy based on very minimal levels of education. In some of these countries literacy is claimed for people who have as little as two years of education. Two years of education is, of course, a considerable improvement over none at all, but a second grade reading level is, at best, a starting point for dealing with Internet content. Only a few students who have achieved a second grade reading level will have language mastery sufficient to reading a newspaper in their own language. Even fewer will be able to read computer manuals or specify URL's to a web browser.

Figure 6: Distribution of Literacy across countries. Red bars show the distribution of Literacy across countries worldwide. Green bars show the distribution of Literacy across countries in Africa.


Indeed, it is very unlikely that many of these students will have learned to read a second language. This matters, as most Internet content is written in one of a small number of languages. The dominant language of the web is English, but large companies will often support a small number of other languages. Common choices include French, German, Spanish, Japanese, Mandarin Chinese, Simplified Chinese, Italian, Brazilian Portuguese, Korean, Russian, and Arabic. The United Nations supports six of these languages. IBM's automated language translation software (International Business Machines, 1991) supports translations between eight languages in a limited set of combinations (English-to-French, French-to-English English-to-Italian, Italian-to-English English-to-German, German-to-English English-to-Spanish, Spanish-to-English English-to-Chinese (simplified) English-to-Chinese (traditional) English-to-Japanese English-to-Korean). Even browsers are limited in their language support. At this writing, the latest version of Internet Explorer supports 24 languages. Netscape supports 6. Opera supports 25.

The highest rates of illiteracy are constrained to a small number of regions around the world, with particularly high rates of illiteracy (as can be seen in Figure 6) associated with countries in Africa. None of the languages typically associated with the Internet is native to the countries with the highest illiteracy rates, making literacy in a second language an effective prerequisite to Internet use in these countries.

It isn't clear how one solves this problem quickly on a worldwide basis. While people can learn to read at any age, the most effective point in time for providing literacy (and especially literacy in a second language) is during childhood. People who don't learn to read during childhood generally never learn to read at all. As a result, improvements in literacy are often generational, with literacy rates increasing both as a function of children learning to read and illiterate adults dying of old age. Teaching the young to read depends on making an investment in childhood education, but this investment can be hard to make in countries with competing Economic Priorities.

Per pupil spending levels on education in the countries with the highest illiteracy rates tends to be fairly high as a percentage of the counties overall GDP, but that spending is frequently extremely unequal, with fairly low per capita GDP investments made in primary school classrooms and fairly high per capita GDP investments made educating a small number of students at the secondary and tertiary education levels. It has also declined somewhat over the last decade. While the U.N. data includes estimates of the level of these these investments for a large number of countries, the data is only partial. Attempts to create a secondary measure of literacy based on this unequal spending have foundered because of the large numbers of missing values.

For this study, we have selected "Literacy as a Percentage of Population" as an indicator of a countries literacy levels. This measure is imperfect. A better measure of literacy would factor in the level of literacy within a given country (e.g. the mean or median grade level at which people are literate). Table 2 shows the results of a regression of Literacy as percent of population against the variables of the first three bridges. Two variables, Lifespan in years and Government a Democracy?, prove to be significant predictors of literacy. The resulting model accounts for 49% of the variance in literacy (F=86.147, df= 2/177, p<.000).

Model Unstandardized Coefficients Standardized Coefficients t Sig.
B Std. Error Beta
(Constant) -15.684 7.279   -2.155 .033
Lifespan in years 1.324 .108 .661 12.237 .000
Government a Democracy? 9.387 3.167 .160 2.963 .003
Table 2: Regression of Literacy as percent of population against GDP per Capita, Lifespan, Government a Democracy?, and Electricity Production in mHh per Thousand.


The Fifth Bridge: Network Infrastructure

Internet digital media do not operate in isolation. The idea of the web, e-mail, instant messaging, computer conferencing and other Internet media is to provide a near real-time global connectivity to what is, in effect, the world's largest library of information and people. A personal computer or other Internet appliance can only provide this level of access if it is connected to the Internet. This connectivity is provided, in general, through a complex layering of data networks.

End users generally connect their computer or Internet appliance to a localized network that is provided through a school, business, government entity, or local Internet Service Provider (ISP). ISPs most often provide dial-up service via modem and the local telephone system, but may leverage other infrastructures, including cable television service, high speed data lines, cellular telephone or data service, and even two-way satellite transmission. Small businesses will, in many cases, leverage a single ISP connection to support Internet connectivity for multiple machines via a local area network and Internet gateway machine. Schools and larger business will provide connectivity for hundreds or thousands of machines through higher capacity Internet connectivity and routers. ISPs interconnect, in turn, through Network Service Providers (NSPs) that provide regional, national, and even worldwide connectivity.

All of this entails not only transmission capacity (copper wire, optical fiber, radio frequency bandwidth), but a large number of routers, gateways, bridges, proxies, modems, and other hardware. Network Infrastructure cost money both in equipment and ongoing maintenance and operation. They are also generally highly dependent on an existing telephone infrastructure, especially for individual users. Many countries lack a telephone infrastructure infrastructure in any but the largest cities, and vandalism of copper wiring frequently makes it difficult to maintain even that Basic Infrastructure. In an increasing number of countries in which cellular telephones are a primary telephone infrastructure, Network Infrastructure problems are particularly severe. Radio frequency bandwidth is not, in general, a practical substitute for the higher bandwidth wireline infrastructure. There simply isn't enough radio frequency bandwidth to handle anything but very small cells which themselves must be interconnected by wireline.

Where there is no wireline telephone infrastructure, the most practical base on which it can be built is the electrical power infrastructure. Where there is no electrical power infrastructure, the practical options for providing Internet connectivity are extremely limited. Radio frequency bandwidth is not, in general, a practical alternative to wireline infrastructure, especially in highly populated areas. Higher bandwidth cellular connectivity provides the best solution for providing scalable radio frequency bandwidth, but that scaling is generally dependent on wireline interconnections between cells and the the wider Internet.

Two-way satellite bandwidth, while more practical for providing connectivity to isolated areas, comes with an even larger set of problems, the most important of which is network latencies that are often measurable in seconds, even for small packets. These latencies don't matter for the large blocks of broadcast data that are most typically associated with satellite communications. They matter a lot, however, in most Internet protocols. A simple web page request, for instance, is actually a series of small transactions. A first transaction (SYN) requests a connection. After that SYN is accepted by the remote host, additional transactions actually request a web page. Most of the requests and responses are very small and can travel almost anywhere in the world via wireline in less than a second. Satellite transmissions from fixed dishes like those used in 2-way satellite must travel much longer distances to and from satellites in geosynchronous orbit. These same transactions take seconds each. Hence even when 2-way satellite offers a higher bandwidth connection, small web transactions will often be much slower than they would be on a much slower dial-up wireline infrastructure.

It should be no surprise, then, that we have selected "Main Telephone Lines per Thousand" as a basic indicator of a countries Network Infrastructure. Table 3 shows the results of a regression of Main Telephone Lines per Thousand against the variables of the first four bridges. Four of the five variables, including GDP per capita, Electricity Production in mHh per Thousand, Literacy as percent of population, and Government a Democracy?, prove to be significant predictors. The resulting model accounts for 77% of the variance in telephone Network Infrastructure (F=147.339, df= 4/173, p<.000).

Model Unstandardized
t Sig.
B Std. Error Beta
(Constant) -128.180 29.176   -4.393 .000
GDP per Capita 1.463E-02 .002 .541 9.451 .000
Electricity Production in mHh per Thousand 13.189 2.760 .266 4.779 .000
Literacy as percent of population 1.393 .391 .155 3.563 .000
Government a Democracy? 70.466 19.778 .134 3.563 .000
Table 3: Regression of Main Telephone Lines per Thousand against GDP per Capita, Lifespan, Government a Democracy?, and Electricity Production in mHh per Thousand, and Literacy.


The Sixth Bridge: Computer Resources

With a Network Infrastructure available, the last physical resource required to access Internet-based media is a personal computer, workstation, or other Internet appliance. There are an increasing variety of Internet-capable appliances available, including pagers, digital cellular telephones, PDAs, Web and e-mail Appliances, thin clients, WebTV's and videogames. Each, at its heart, is a computer, and each provides end users with a different mix of capabilities at a somewhat different price point. The features that differentiate these devices, in general, include processor performance, screen size, on-device storage, and functionality. Any of these devices will provide some level of access to Internet digital media. Table 4 summarizes the capabilities and restrictions associate with each type of Internet capable device:

Device Type Minimum Price Capabilities Restrictions
Digital Pagers ~ $100 Text messages, limited e-mail, limited web. Very small screens and storage capabilities restrict these devices to text. Restricted text entry and navigation capabilities slows messaging.
Digital Cellular Telephones ~ $150 Text messages, limited e-mail, limited web. Very small screens restrict ability to view web pages, which generally require special markup to display at all. Restricted text entry and navigation slows messaging.
Personal Digital Assistants ~ $200 Text messages, e-mail, terminal emulation, limited web, other programmable web services. Small screens restrict ability to view web pages. E-mail and programmable web services are generally restricted by screen size. While limited storage is usually a problem, it is possible to use high capacity storage media in PDAs.
WebTV ~ $100, excluding TV Many Internet services, including many web pages and web-based e-mail. The limited resolution of television displays restricts the size of the browser window, and a minimum function browser can't display the content of web pages that assume plug-ins and other browser extensions. Limited storage, memory and programmability prevent local storage of e-mail and programmed extensions. Generally hard wired for dial-up to a specific Internet service provider.
E-mail Appliances ~ $200 E-mail. Small text only displays allow for the display and composition of small to moderate sized e-mail messages. Limited local storage allows a restricted number of e-mails to be saved locally for future reference.
Videogames ~ $200 As delivered, interactive gaming via optional dial-up modem capabilities. With extensions, any LINUX-based Internet application that can be reasonably viewed on a television display, including streaming audio and video. Videogame consoles are generally fairly powerful computers that display to a television and have limited local storage capabilities. More recent offerings from Nintendo and Sony have proved capable of extension such that they can be converted, at fairly low cost, to run LINUX and a full range of LINUX-based PC applications. It is likely that the same can be done with the $300 MicroSoft X-Box, which already has a built-in hard drive for local storage.
Web Appliances ~ $300 Most Internet services, including almost anything that can be done with a browser, including streaming audio and video. Web appliances are generally low performance personal computers with built-in displays, limited storage capacity, and a PDA class operating system (Windows CE, for example). They usually accept optional peripherals via USB ports and are extensible with device drivers and software upgrades. Generally hard wired for connection to a specific Internet service provider.
Thin Clients ~ $400 Most Internet services, including anything that can be done with a browser. Depending on ones perspective, either a more powerful variation on the web appliance or a personal computer with minimal storage. Thin clients generally load their operating system and services from firmware and a server. Moderate performance processors generally run the same operating systems and applications that a personal computer might run, but with little or no local storage of applications and data. Not terribly practical for individuals, but useful in business environments.
Personal Computers ~ $500 Any Internet service. None. Even at the low end price point, can be used as either a client or server.
Workstations ~ $2000 Any Internet service. Can be used as a high performance server.
Table 4: A comparison of Internet Capable Devices

Table 4 should by no means be regarded as a comprehensive overview of either existing or possible Internet devices. It simply draws a continuum of functionality and price among a range of existing Internet devices. The most successful of these devices, the personal computer, is distinguished primarily by the fact that it is provide access to any currently available Internet service out of the box. Other devices provide more restricted functionality at lower price points. Other devices will have similar tradeoffs of price against functionality. Less expensive Internet access appliances will typically trade off screen size, keyboards, pointing devices, storage, and processing power to achieve a lower price point.

These are, by and large, unfortunate tradeoffs. Limited screen size restricts the amount of material that can be viewed in a window, the number of Internet applications that can be run at the same time, and the percentage of the display that can be dedicated to user interface elements. Inadequate text entry and pointing devices slow down and restrict both selection of content and the creation of new content, including e-mail, responses in Instant Messenger sessions, and web pages. Inadequate local storage makes it difficult for users to customize their environment, add new Internet features and applications, maintain a set of bookmarks, retain copies of e-mail, or maintain a local copy of the content they create. Reduced processing power will inevitably restrict the range of Internet applications that can be used, as new Internet applications frequently originate in the possibilities offered by an increasingly functional microprocessor.

The history of computing is one of increasing functionality at declining prices. Processors, storage, memory, displays, and operating systems grow ever more powerful. In most years prices remain stable as power and functionality increase. In some years prices decline, usually as a response to a temporary oversupply of components. Personal computer prices have declined in what has effectively been a step function, with the minimum price for personal computers remaining stable for a number of years and then declining to a new stable price point. It seems likely that this will happen again, with declines in in the price of flat screen display technology leading the way such that laptop computers become the low price entry point for personal computers. The entry level price point for these machines will probably decline to about $250 at some point in the next 10 years.

Predicting the future is, of course, an exercise in wishful thinking. Ten years is a long time. By the GNP standards of many digital have-not countries, $250 remains a high price for an entry level PC. The technology of full function entry level Internet access devices may be revolutionized over the next ten years. The price of a full function entry level PC's may fall dramatically. Personal computer interfaces have remained fairly stable for over 15 years now. They work, which makes it more likely that they will remain stable. Personal computer prices have fallen dramatically over the last twenty years, but the pattern has been a roughly 50% decline in prices every six years or so, and a continuation of that pattern would suggest that $250 entry level PC's are only a few years away. Displays remain one of the most expensive components of personal computers, but advances in flat screen technology suggest that flat screen display prices will fall dramatically over the next few years, with larger flat screen displays costing considerably less than smaller CRT's. This suggests that low cost entry level PC's will have a laptop or tablet form factor.

If these PC's remain expensive by the standards of low GDP per Capita digital have not countries, and they will, the most workable alternative will remain one of PC sharing via schools, Internet cafe's, libraries, and other community access centers. In this model individuals should still own their own storage, in the form of a rewriteable CD or similar low cost and high capacity storage medium, but PC's could be treated as common community resources, with the price of a machine distributed across many people.

One wishes that there was a true magic bullet here that could solve the problem of personal computer access for everyone who wants such access. One wishes that video game machines could be readily turned into low cost personal computers or that handheld game boy machines, enhanced with Internet access and voice synthesis, could provide full function access to Internet digital media. The truth remains that such devices, while buildable, and probably at reasonable prices, trade Internet functionality for price. While it may well be that some Internet access is better than none at all, it remains that anything less than full access to Internet media represents something less than a bridge across the digital divide.

It is with this reality in mind that we have selected "Personal Computers per Thousand" as a basic indicator of a countries computer resources. Table 5 shows the results of a regression of Personal Computers per Thousand against the variables of the first five bridges. Four of the six variables, including GDP per Capita, Main Telephone Lines per Thousand, Lifespan in years, and Electricity Production in mHh per Thousand, prove to be significant predictors. The resulting model accounts for 84% of the variance in telephone Network Infrastructure (F=145.293, df= 4/107, p<.000).

Model Unstandardized Coefficients Standardized Coefficients t Sig. Collinearity Statistics
B Std. Error Beta Tolerance VIF
(Constant) 77.624 33.674   2.305 .023    
GDP per Capita 1.160E-02 .001 .710 9.468 .000 .258 3.872
Main Telephone Lines per Thousand .283 .046 .469 6.134 .000 .248 4.029
Lifespan in years -1.766 .569 -.162 -3.103 .002 .531 1.883
Electricity Production in mHh per Thousand -4.339 1.851 -.145 -2.344 .021 .380 2.629
Table 5: Regression of Personal Computers per Thousand against GDP per Capita, Lifespan, Government a Democracy?, and Electricity Production in mHh per Thousand, Literacy, and Main Telephone Lines per Thousand.

The standardized betas for two of these variables are counterintuitive, however. The results for GDP per Capita and main telephone lines per Thousand seem sensible, with personal computer use rising as a function of each. The results for lifespan and electricity production make less sense, however, as personal computer use rises as these measures fall. This could be regarded as the basis for an interesting insight. The collinearity statistics of Table five suggest, however, that these results may be an artifact of a high level of collinearity between GDP per Capita and Main Telephone Lines per Thousand.

The results shown in Table 6, which drops GDP as an exogenous variable, are consistent with this interpretation. Indeed, only one variable, Main Telephone Lines per Thousand, proves to be a significant predictor. The resulting model accounts for 71% of the variance in telephone network infrastructure (F=270.925 df= 1/110, p<.000).

Model Unstandardized Coefficients Standardized Coefficients t Sig. Collinearity Statistics
B Std. Error Beta Tolerance VIF
(Constant) -10.407 7.272   -1.431 .155    
Main Telephone Lines per Thousand .509 .031 .843 16.460 .000 1.000 1.000
Table 6: Regression of Personal Computers per Thousand against Lifespan, Government a Democracy?, Electricity Production in mHh per thousand, Literacy, and Main Telephone Lines per Thousand.

The reduction in collinearity associated with this model simplification is a good thing, of course, but while the reduction is accomplished in a theoretically consistent way (e.g. a more proximate bridge is favored over a less proximate bridge), it is accompanied by a large reduction in the variance accounted for. This should not be entirely surprising, as collinearity problems are most common when high levels of variance are accounted for, and the red flag raised by the variance reduction is not nearly so large as the red flag that the collinearity raised, but the red flag remains. It is certainly possible that, in a more refined test of the model, other variables would prove to have predictive value. The result is an interesting one, however, that will clarify the path across the the digital divide (to be seen in Figure 7).

Six Bridges Across the Digital Divide

At the end of every trail is destination. If our purpose, in exploring the bridges across the digital divide, is to better understand how we can minimize or eliminate the digital divide on a global basis, we must allow those bridges to carry us to that destination. This studies uses the variable "Internet Hosts per Hundred Thousand" as its indicator of a countries Internet Use. As with other measures used in this study, there may be better ways to measure Internet Use within various countries. It remains, however, that Internet Hosts per Hundred Thousand people provides a reasonable measure of the extent of Internet use within a given country.

Table 7 shows the results of a regression of Internet Hosts per Hundred Thousand against the variables of the six bridges considered above. Four of the seven variables, including Electricity Production in mHh per Thousand, Personal Computers per Thousand, GDP per Capita, and Main Telephone Lines per Thousand, prove to be significant predictors. The resulting model accounts for 75% of the variance in telephone Network Infrastructure (F=82.831, df= 4/107, p<.000).

Model Unstandardized Coefficients Standardized Coefficients t Sig. Collinearity Statistics
B Std. Error Beta Tolerance VIF
(Constant) 7.233 84.462   .086 .932    
Electricity Production in mHh per Thousand 263.147 26.100 .799 10.082 .000 .364 2.750
Personal Computers per Thousand 11.423 1.277 1.038 8.948 .000 .169 5.900
GDP per Capita -9.714E-02 .022 -.541 -4.515 .000 .159 6.285
Main Telephone Lines per Thousand -2.839 .684 -.427 -4.152 .000 .215 4.642
Table 7: Regression of Personal Computers per Thousand against GDP per Capita, Lifespan, Government a Democracy?, and Electricity Production in mHh per Thousand, Literacy, and Main Telephone Lines per Thousand.

As was the case with Personal Computers per Thousand in Table 5, however, there is strong evidence of collinearity, with several counterintuitive beta coefficients (GDP per Capita and Main Telephone Lines per Thousand. Indeed, one standardized beta exceeds one, an indication of significant computational problems in the regression. The regression of Table 8, which eliminates this collinearity by dropping both GDP per Capita and Electricity Production in mHh per Thousand from the analysis, is consistent with this interpretation. Indeed, only one variable, Personal Computers per Thousand proves to be a significant predictor. The resulting model accounts for 53% of the variance in our measure of Internet Use (F=120.557, df= 1/110, p<.000).

Model Unstandardized Coefficients Standardized Coefficients t Sig. Collinearity Statistics
B Std. Error Beta Tolerance VIF
(Constant) -150.144 89.236   -1.683 .095    
Personal Computers per Thousand 7.956 .725 .723 10.980 .000 1.000 1.000
Table 8: Regression of Internet Hosts per Hundred Thousand against Lifespan, Government a Democracy?, Main Telephone Lines per Thousand and Personal Computers per Thousand.

Once again, the reduction in collinearity is accompanied by a large, and somewhat predictable, decline in the variance the model accounts for. This variance reduction raises a red flag which is only mitigated by the even larger red flag associated with more complex model, with its counterintuitive results and high levels of collinearity. The other variables might have demonstrated predictive value in a more refined test of the model. Nonetheless, the reductions are accomplished in a theoretically consistent way (e.g. a more proximate bridge is favored over a less proximate bridge), and the model simplification, as will be seen in Figure 7, is theoretically satisfying.

The Model in Aggregate

Figure 2 shows twenty general hypotheses concerning the relationship of the six tested bridges over the digital divide. Six of these hypotheses are very specific and make statements concerning the relationship of specific bridges to the digital divide. Fourteen describe relationships between the bridges, and are described in four generalized statement of secondary hypothesis. Figure 2 makes these hypotheses explicit as linear relationships. Figure 3 restates these hypotheses in terms of the eight variables that are explored in the regressions of Tables 1, 2, 3, 5, 6, 7, and 8.

Figure 7 aggregates the results of these these regressions (tables 1, 2, 3, 6, and 8) as a summary path model. Over the course of these regressions, the 25 specific hypotheses of Figure 3 have been reduced to 9 significantly predictive relationships. While only one of the bridges directly predicts Internet Use, all of the bridges play at least an indirect role. Indeed, the predictive relationships from the regressions order themselves in a manner which may be useful in suggesting ways in which countries might usefully sequence their efforts as they attack the problem of bridging their digital divides.

Figure 7: An aggregate model of the contributions of the six bridges over the digital divide. This model combines the results of multiple regressions shown in Tables 1 (Basic Infrastructure), 2 (Literacy), 3 (Network Infrastructure), 6 (PC Use), and 8 (Internet Use).

Specifically, the model suggests that:

  1. Economic Priorities are, on the whole, more important than Social and Legal Constraints, but both matter. Increasing economic resources enables improved Basic Infrastructure. Increasing lifespan promotes increased literacy. It is likely that both of these relationships involve feedback loops such that lifespan promotes literacy promotes lifespan; increasing GDP per Capita promotes improved infrastructure promotes increasing GDP per Capita. The relationship of democracy and literacy is probably also circular, with democracy promoting literacy promoting democracy. Such presumed circularities are not a primary issue here. What matters is that both Economic Priorities and social and legal constraints, while not directly predictive of Internet Use, do have an indirect effect through their influence on Literacy and Basic Infrastructure.
  2. The problems of Literacy and Basic Infrastructure are not interdependent. They can, therefore be pursued in parallel. Each is affected by different factors, with literacy predicted by lifespan and democracy; Basic Infrastructure predicted by GDP. While this parallelism won't speed up the effects of increasing literacy or the cumulative effects of building up infrastructure over time, it at least allows the accumulation of infrastructure and literacy to happen at the same time.
  3. The most fundamental problem associated with bridging the digital divide is that of bridging the Network Infrastructure divide. It is Network Infrastructure that culminates a countries efforts to minimize Social and Legal Constraints, enhance economic decision making, increase literacy, and build up Basic Infrastructure. These elements are the backbone on which universal telephone service can be offered, and the availability of such this basic network service becomes the basis on which Computer Resources can be attached to the Internet.
  4. The best predictor of PC Use is, in fact, the universal telephone service that serially precedes it in the model, and the best predictor of Internet use is PC use. The path from universal telephone service to Internet Use is, in effect, a two step linear path, with telephone service enabling PC Use enabling Internet Use.

These results strongly support the basic model of this paper, as initially outlined in Figure 1, and considerably clarify the hypotheses of Figures 2 and 3. Indeed, they suggest that the bridges can be recombined into a measure of digital readiness that may have considerable predictive value.

Beyond the Model: Assessing Digital Readiness

A general measure of Internet Readiness has been constructed using the predictive variables associated with the regressions. It is possible, given the regression results, to get fancy in building this estimate, with various weightings based on the predictive value of the various measures. This appears to make only a marginal difference to the effectiveness of the measure, however, and given the choice, simpler is better. Hence the measure of Digital Readiness has been constructed as follows:

  1. With the exception of "Government a Democracy?", which is already scaled as zero and 1, all variables were rescaled to have a maximum value of 1. For the current data set this meant:
    1. Dividing GDP per Capita by 32700
    2. Dividing Lifespan by 83
    3. Dividing Electricity in mHh per Capita by 25.77
    4. Dividing Literacy by 100
    5. Dividing Telephone Lines per Thousand by 1014
    6. Dividing PC's per Thousand by 695
  2. The rescaled values were added together to form a composite measure with a theoretical minimum value of zero and a theoretical maximum value of 7. The actually minimum and maximum values were .97 and 5.83.

The correlation of this variable with Internet Hosts per Hundred Thousand is .667 (n=111, p<.001). Examination of the residuals suggested, however, that the relationship between this Internet Readiness variable and Internet Hosts per Hundred Thousand was nonlinear. This turns out to be the case. The correlation of Internet Readiness with the logarithm (base 10) of Internet Hosts per Hundred Thousand is much stronger: .913 (n=111, p< .001). It is possible, with a minor reshaping and weighting of the Internet Readiness variable, to improve this correlation to .918, but the marginally higher correlation hardly seems worth the effort. Indeed, the additional complications turn out to make the variable somewhat harder to interpret.

Table 9 shows the regression of the logarithm of Internet Hosts per Hundred Thousand against Internet Readiness. It should not be surprising, given the base correlation, that this model is highly significant (F=544.586, df=1/109) or that it accounts for a huge percentage (83%) of the variance in Internet Hosts per Hundred Thousand.

Model Unstandardized Coefficients Standardized Coefficients t Sig.
B Std. Error Beta
(Constant) -1.611 .145   -11.077 .000
Internet Readiness .997 .043 .913 23.336 .000
Table 9: Regression of the logarithm of Internet Hosts per Hundred Thousand against Internet Readiness

What is surprising is how easy the regression equation is to interpret. The equation:

-1.6 + .997*the number of bridges crossed

rounds trivially to:

-1.6 + the number of bridges crossed 

In other words:

There are obviously limits to how much improvement can be attained this way. Indeed, there is undoubtedly a ceiling on Internet hosts per hundred thousand that is hit somewhere in the range of 100,000. Nonetheless, the Internet Readiness Index presented here gives a clear sense of how growth in Internet Use can be accomplished within a given country and the manner in which that growth is likely to accelerate.

The Next Study: Ways in which this analysis might be improved

This study was never intended to be the last word on how countries might bridge the digital divide. It was intended only to make a theoretical statement about what needed to be accomplished in order to bridge the digital divide on a worldwide basis and to do a limited test of the adequacy of the theoretical model. The limited test has been surprisingly successful. Indeed, it is a matter of considerable surprise to the author that the available data was sufficient to demonstrate that the model was both workable and useful. Even after performing the analysis, moreover, the author finds the simplicity and ready interpretability of the Internet Readiness index to be almost unbelievable. In other words, this is a study that demands a more thorough test with better data.

Almost every measure in the study can be improved:

The analysis can be improved as well. This study was intended to be exploratory. While a high standard of statistical significance was maintained (every variable relationship that is documented as a significant predictor here is significant at at least the .003 level), each level of the path model was tested in an individual regression. A better test would use structural equation modeling software to test the entire model as a unit. While, moreover, a direction of effect is hypothesized for every bridge in the model, a number of the relationships are obviously circular over time. Again, a better test of the model would collect data for all of the variables over time and use time series modeling (lag 1 feedback) to estimate the extent of that circularity.

The Seventh Bridge: Choice

No model of the digital divide can ever force anyone to actually use Internet resources, and while the model and Internet Readiness index that are explored above should help countries to understand how they can go about bridging the digital divide, there is no escaping the fundamental need to present the Internet to people as something they want to have access to. This is a problem, to some extent, in every country that has crossed the digital divide. Even in the U.S. there is a persistent minority of people who, despite having the ability to access Internet resources, decide not to do so. The issue of Choice is not necessarily a problem. It should be perfectly fine for a person to decide that they are not interested in using computer and Internet resources, just as some people decide not to use cars, telephones, calculators, and televisions. For some people this may be a productive decision, and it is certainly no ones place to judge another's decision to not use a particular technology.

There is a point, however, where such decisions can be a problem, and advocates of a bridging of the digital divide need to be on the look out for patterns of nonuse. A social or religious group may decide, as an article of faith, to eschew computer technology. Indeed, it would be surprising if anti-Internet cults did not arise at some point. To the extent these groups simply practice their beliefs without impinging on the rights of others, there will probably be no problem associated with the decisions of such groups. To the extent, however, that these groups prevent others from using Internet resources or are denied access to Internet resources because of presumed beliefs, there may well be a problem that has to be dealt with. There probably isn't any way to systematically measure Choice as a worldwide influence on Internet use, but it is certainly possible to do local studies of who isn't using the Internet in a particular area and why. These studies will not be terribly interesting when most citizens in a country are unable to choose to make use of Internet resources. They become much more interesting as Internet Use in a country rises to fairly high levels.

Conclusion: Media Aristocracies, Network Resources, and the Global Digital Divide

The idea of unequal access to media resources is hardly a new one. Indeed, it can be argued that there has never been a technology-based medium of communication that did not have an associated priesthood in the form of a privileged few who had access to the technological means of content production within the medium. The exceptions to that argument, including the telephone and C.B. Radio, are only exceptions in the sense that they redefine the concept of priesthood from individual to country. Much as there are media, like television, in which a few produce content for the many, so there are media, like the telephone, for which the medium is effectively ubiquitous in some countries, and fairly exceptional in others. In the telephone-have world, the only people who don't have telephones are the people who choose not to. In the telephone have-not world, which comprises more than half of the world's countries and well over three quarters of its population, most people don't have the choice. It seems odd, as one ruminates on the global digital divide, that we have never made an issue of the global telephone divide, but we have not, and that is, for reasons that have been outlined above, one of the larger problems that will need to be faced if we are to bridge the digital divide.

Indeed, if there is a single most important lesson to be gleaned from the model of Figure 7, it is that a bridging of the global telephone divide is a necessary prerequisite to a bridging of the global digital divide. Indeed, the lessons of this study's data set includes a series of divides that must be bridged if we are to bridge the digital divide.

Our willingness to make an issue of the digital divide IS something new in the history of media. No one has ever declared access to any medium of communication beyond face to face communication as a basic human right before. The challenge we face in actually bridging the digital divide on a worldwide basis is a daunting one. There are are at least seven obstacles that that need to be bridged if we are to reach the goal, and it is likely that it will take generations of effort to achieve. There are dangers that will have to be faced in the wake of such fast paced change, not the least of which will be religious and political reactionary responses to that change. But the challenge can be met and is worth facing. There is no guarantee that a true global village won't be a worldwide distributed set of stratified communities of belief, but it is difficult to see how we can overcome existing belief-based community stratifications without a global mechanism, like the Internet, that enables both personal and interpersonal communication. It is hoped that this paper will make a useful contribution to the effort.


The author would like to make special note of the insights provided by Joan Dyer while this paper was being conceived and written. It would not be the same paper without her helpful suggestions.


Annan, Kofi. (2001). World Telecommunication Day Message. May 17, 2001. (Accessible at and

Anonymous. (2001). Nations of the world. World Almanac and Book of Facts. PRIMEDIA Reference.

Bill and Melinda Gates Foundations. (2001). Libraries - Bill and Melinda Gates Foundation. 2001. (accessible at

Cattagni, Anne and Farris, Elizabeth. (2001). Internet Access in U.S. Public Schools and Classrooms: 1994-2000. Statistics in Brief. U.S. Department of Education: Office of Educational Research and Improvement: May, 2001.

Hiltz, Starr Roxanne, and Murray Turoff. (1978). The Network Nation: Human Communication via Computer. Reading, MA; Addison-Wesley.

Innis, Harold A. (1950). Empire and Communications. Oxford University Press.

International Business Machines. (1991). Enterprise Products: WebSphere Translation Server for Multiplatforms.

National Research Council. (1994). Rights and Responsibilities of Participants in Networked Communities. Washington, D.C.; National Academy Press. (Accessible at

NetDay. (2001). NetDay Home Page. (accessible at

Office of Policy Analysis and Development. (1998). Falling Through the Net: Defining the Digital Divide. National Telecommunications and Information Administration. (Accessible at

UNESCO. (2000). UNESCO's World Education Indicators. Excel data sets associated with UNESCO World education report 2000. (Accessible at