Counting the Net: Internet Access Indicators

Michael MINGES <minges@itu.int>
International Telecommunication Union
Switzerland

Abstract

The Internet has had a major impact on society and business during the last decade of the 20th century. Yet, despite the popular interest generated by the Internet, there is a lack of comparable data on its spread across the world. Though the quality and quantity of information has recently improved, there are still wide variations in definitions, comparability and scope. Market analysts, particularly keen on the size of electronic commerce, have generated many recent estimates. This type of information -- though it varies widely depending on the source -- is increasingly compiled for developed countries. However, there is a shortage of publicly available data on Internet accessibility, particularly for developing countries. Standard indicators and definitions are needed to measure Internet access across countries. This document outlines statistics that are presently being used, as well as their limitations, and proposes a set of harmonized Internet access indicators.

1. Infrastructure

Infrastructure statistics measure the stock of communication equipment needed for accessing the Internet [1]. This includes Internet host computers and supporting hardware such as telephone lines and personal computers.

1.1 The host dilemma

The most commonly (ab)used indicator to compare Internet development between countries is the number of host computers. The best known survey of Internet hosts is carried out twice a year by Network Wizards for the Internet Software Consortium [2]. Some national Internet network administrators compile data on the number of hosts in their countries [3].

Networks Wizards uses the following definition of hosts: "A host is a domain name that has an IP address (A) record associated with it. This would be any computer system connected to the Internet (via full or part-time, direct or dialup connections). ie. nw.com, www.nw.com." [4] While hosts might be a useful infrastructure indicator of the number of computers in a nation that are connected to the Internet, it is a poor indicator of accessibility since it does not measure the number of users [5]. Furthermore, the definition of host is vague. What exactly are they? It is easy to conceptualize a Web server that provides access to the Internet or disseminates information or is used as a conduit to sell products and services. Yet hosts also include name servers, mail servers, file servers and other automated devices. They enable the Internet to operate, but are less meaningful in terms of analysis [6]. The measurement of host data is also subject to criticism. For example, a single computer may host several domain names and a single domain name might be hosted by a group of computers. Also, the physical connection to a host may not be operational when the survey is carried out. As a result, the data have been prone to major revisions and there are often anomalies between surveys.

A major drawback with hosts is that they are assumed to be located in the country shown by their two-letter ISO country code Top Level Domain (ccTLD) (e.g., .nl for Netherlands). However, "There is not necessarily any correlation between a host's domain name and where it is actually located. A host with a .NL domain name could easily be located in the U.S. or any other country. In addition, hosts under domains EDU/ORG/NET/COM/INT could be located anywhere. There is no way to determine where a host is without asking its administrator." [7] This is a major shortcoming and results in misleading interpretation of the data. For example, the top country in terms of Internet host penetration at July 1999 was the tiny Pacific island of Niue with more than three hosts per inhabitant (see Table 1). Clearly few of those host computers are located in Niue where the number of Internet users was estimated at 200-300 in November 1999 [8]. Another Pacific nation, the Solomon Islands, had no hosts according to the July 1999 Network Wizards survey. Yet that country has had hosts connected to the Internet at least since 1996 [9]. The United States, which is typically assumed to be one of the most Internet-connected countries in the world, only ranks 44th in Internet penetration based on the .us ccTLD. Clearly, the overwhelming number of hosts in the United States are using something other than .us.

As long as Internet host data are based upon the ccTLD name rather than the actual location of the host, they should be interpreted carefully. Although there have been attempts to adjust host counts according to where they are located, these efforts are carried out infrequently or are not widely available [10]. In any case, a more understandable indicator than hosts would be the number of organizations with an Internet site in the country. For example, a study of the largest 1,000 firms in Singapore found that 46 percent had their own website [11]. This analysis could be extended to the number of governmental, educational, and other establishments with websites [12].

1.2 Telephones and personal computers

Telephone lines and personal computers are key components for Internet access. Both have significant impact on the take-up of Internet in a country. Dial-up Internet access requires a telephone line and a personal computer (with a modem). These hardware components thus constitute an upper limit for Internet access. For example, if 25 percent of households have personal computers with modems, then Internet access from households cannot exceed 25 percent. Pro-Internet policies will not be successful if they do not address these fundamental access requirements.

The International Telecommunication Union (ITU) compiles both telephone line and personal computer statistics [13]. Telephone lines for a country include telephone subscribers plus the number of pay phones. These data are regularly supplied by telecommunication authorities or operators. Personal computer data require more effort to compile. Though estimates are generally available for developed countries, they are usually compiled by market researchers and available only for a relatively high price. Most of the ITU data are compiled by estimating the stock of personal computers from sales or import data. This is problematic for many developing countries where shipment data are scarce and a significant portion of imported personal computers can evade statistical reporting (e.g., smuggling, gray market, local assembly). It is useful to disaggregate telephone line and personal computer data by households and business to obtain a more insightful analysis.

2. Access

This group of indicators measures access to the Internet. It is critical to distinguish between different aspects of access. The indicators in this category are often used interchangeably, making comparisons difficult. However, there are key distinctions that should be observed. One way is to start with the total potential Internet universe and gradually burrow into deeper layers. The outer shell is the number of inhabitants in the country. This figure is the total potential Internet universe and is necessary for deriving penetration figures. Most market research limits the data to the adult population, which can affect comparability and often ignore a significant portion of users. The next layer is the number of people that are aware of the Internet. People who do not know about the Internet are not going to use it. The next critical statistic is those covered by the Internet -- that is, the number of people within easy access of the Internet regardless of whether they are using it. After that is the number of users. The frequency and sophistication of use are important qualifiers. Finally, at the core are the number of subscribers, those paying for access to the Internet. This is the most verifiable statistic, but not necessarily a good measure of usage because most users do not themselves pay directly for access [14].

2.1 Users

The number of users is a basic and seemingly comprehensible measure of Internet access. However, comparisons of user data are misleading because there is no standard definition of frequency (e.g., daily, weekly, monthly) or services used (e.g., e-mail, World Wide Web). Although there are several well-known sources of Internet users for different countries, they are often collected from various national surveys that are not comparable [15] and typically ignore developing countries completely [16]. The commercialization of the Internet and increasing interest in electronic commerce has resulted in a growing number of market researchers collecting national user statistics [17]. These are typically based on sample surveys projected for the population as a whole. Table 2 shows the results of user surveys now available for about half the countries in Western Europe. Plans by experienced market research firms to conduct Internet surveys in a number of countries could enhance availability and comparability [18]. One danger is that since these surveys are carried out by private firms, much of the data will remain restricted to clients or be prohibitively expensive for widespread analysis. Also, such "snapshot" surveys are often not repeated on a systematic basis, restricting the analysis of trends over time. Growing involvement of national statistical agencies in compiling Internet user data will help to make such data more widely available to the public. The Australian Bureau of Statistics compiles estimates on the number of Internet users for that country based on household surveys [19]. The U.S. Census Bureau has also started to compile Internet usage statistics derived from its Current Population Surveys [20].

2.2 Subscribers

The number of Internet subscribers -- those paying for access to the Internet -- is a more precise indicator of access than users. Subscription suggests a certain intensity of use since it is unlikely one would be paying for Internet access unless it is being utilized regularly. It should be noted that the number of subscribers measures those who are paying for a subscription and not the number of users. Many users obtain access for free either as the member of a household, or from work or school. On the other hand, the number of subscribers is useful for estimating the number of users when user data are unavailable. The number of subscribers thus sets a minimum threshold for the number of users in a country. Data about the nature of the subscription such as paid or free, business or consumer, and the access method (dial-up, leased line, cable modem, or broadband telephone line (ISDN, xDSL)) also are useful.

Many Internet service providers (ISPs) report their subscriber counts. In countries where ISP service is not a monopoly, individual ISP data needs to be aggregated to obtain a national total. If not all ISPs report, country-level data may be hard to obtain. Also, some ISPs originated as on-line providers to their own proprietary content. Some subscribers may still be accessing only the provider's content and thus not truly be Internet users. Some ISPs report their market share which allows a country-level statistic to be derived. Some national ISP organizations also report the total number of subscribers for the country [21]. A number of government agencies, typically communication regulators but also increasingly national statistical agencies, are compiling country-level subscriber data. The latter typically focus on household subscribership. Examples of Internet subscriber data from different sources are shown in Table 3.

2.3 Coverage

An ideal indicator would be a coverage statistic -- that is, the portion of the population of a country within easy access of the Internet, whether they use it or not. This indicator would express the potential Internet user market and is the fundamental measure of universal access to the Internet. High coverage would denote a significant policy achievement in that technically, people can access the Internet even though they may choose not to do so for other reasons (e.g., high cost, lack of interest). One problem is that few organizations measure those who have access to the Internet. When this is done, it is often confusing, with the term "access" often used synonymously with "user." Surveys also generally ask if access to the Internet is from home, work or school and ignore public access points such as cybercafes. It would be useful to have statistics on the number of community locations with Internet access. Another challenge is the definition of "easy" access. Practically every country today has a cybercafe where theoretically any inhabitant could walk in and access. Obviously the utility of this is influenced by the distance of the potential user from the place of Internet access. Thus a very useful indicator would be distance from Internet access measured in time or length (e.g., minutes to reach or kilometers from an access point).

Data from the few surveys that do distinguish between having access to the Internet and using the Internet are shown in Figure 1. Another measure of potential Internet accessibility would be the number of households with both a telephone line and a personal computer, the basic requirements for home Internet access [22]. This indicator is analogous to what is used to define universal service for telephones and could have the same application for the Internet.

Figure 1

With all the publicity surrounding the Internet, it is hard to believe that there are people that do not know about it. Yet awareness of the Internet is far from universal. For example, in the United Kingdom, some 1.6 million adults had not heard of the Internet according to a September 1999 survey [23]. This type of indicator is important since those who do not know about the Internet are not going to use it. It is also useful to provide a breakdown of awareness as shown in Figure 2.

Figure 2

3. Policy

Some indicators are related to access and thus analytically important from a policy perspective. For example, a competitive ISP market should theoretically put pressure on pricing and thus enhance accessibility. Tariffs have a strong impact on accessibility while the amount of use reflects the intensity of access.

3.1 ISPs

Some argue that competitive ISP markets tend to reduce usage prices, enhance availability of different services and improve quality, all of which impact access. The number of ISPs in a country is often taken to be an index of market liberalization. One problem with counting the number of ISPs is that in many countries, there is no legal requirement for them to register. That, combined with the fact that barriers to entry are relatively low, makes an accurate count of ISPs difficult in large, liberalized markets. In countries where ISPs must register, a distinction should be made between "licensed" and ISPs actually in operation. ISP associations in several countries report the number of members, but the figures may not include all providers. Some government authorities keep track of the number of ISPs. For example, in Hong Kong, the telecommunication regulator has a time series on the number of licensed ISPs (see Figure 3).

Figure 3

3.2 Tariffs

Tariffs are an important indicator of accessibility since if people cannot afford the Internet they will not use it. They may also reveal why some countries have a high level of potential Internet access (e.g., high levels of telephone lines and personal computers) but low user levels. Internet tariff comparisons are complex for a number of reasons. First, there are different prices depending on the access infrastructure (e.g., dial-up telephone line, leased line, cable television, mobile phone). Second, there can be a variety of different prices both across and within service providers (e.g., depending on time of day, hours of use). Third, in competitive markets, prices can change rapidly. From a policy viewpoint the most significant Internet tariffs are those associated with dial-up telephone access since this is the most heavily used method of Internet access by individuals at the present time.

Dial-up Internet tariffs consist of two components: (1) telephone usage charges (monthly subscription (line rental) and call charge paid to the telephone company), and (2) Internet access charges (paid to the ISP). There are a variety of issues to consider. One has to do with local call charges. Some telephone operators do not charge directly on a usage basis for local calls. This is usually based on historical reasons and includes most operators in North America and New Zealand. Other operators include a certain number of local calls in the subscription charge; after these "free" calls have been used up, users pay for each call. Some operators charge a one-time amount regardless of the length of the call. Some operators have different local call charges depending on the time of day or day of the week or whether the call is for Internet access or provide discounted calls to certain user-specified numbers. In some countries, a nationwide dialing prefix has been assigned to ISPs so all calls are treated as local. ISP charges also vary from no charge to different charges depending on the volume of use, day of week or time of day.

It is useful to compare Internet access prices within and between countries. Figure 4 compares the price of 20 hours off-peak Internet access for several countries. The Organization for Economic Cooperation and Development (OECD) does this for its member countries [24]. There are also a number of national studies on Internet pricing [25].

Figure 4

3.3 Traffic

Internet traffic measures usage, an important aspect of Internet access in a country. From a policy perspective, it is interesting to know how long users are spending on the Internet and comparing this over time and between countries. Low usage time might reflect affordability issues or lack of sophistication of use. Two problems with traffic indicators are data and concept. Most operators do not report Internet traffic and few government agencies compile this data. In terms of concept, what should be measured? Is the volume of data transferred across the Internet important or the user session time? From an access perspective, the latter is more relevant. For user connection time, however, there is the additional complexity of whether it is over the fixed telephone network, leased circuit or other means such as cable television or mobile telephone networks.

When available, Internet traffic data are compiled by telecom operators, ISPs, and some government agencies. Deutsche Telekom, the incumbent telecom operator in Germany, provides data on the volume of connection time in minutes over the telephone network to its T-Online service. America Online, the largest ISP in the world measured by the number of subscribers, provides data on the average daily usage in minutes of its subscribers. The Office of Telecommunications in Hong Kong provides statistics on the total number of minutes spent on dial-up access to the Internet each month. Data from these sources reflect wide variations in average usage times (Figure 5).

Figure 5

4. The way ahead

This document has outlined a modest but standard list of Internet access indicators to enhance country comparability and assist analysis and policy-making (see Table 4). Examples are provided to illustrate these statistics are presently being compiled in some countries. Most Internet data are amassed by various organizations, difficult to locate, often expensive and frequently available only in national languages. While many governments in the world take an active interest in promoting the Internet, they need to be more involved in collecting, compiling and disseminating Internet-related information in order to evaluate the situation in their country. This would improve availability, enhance standardization, and make information more widely available to the public. Furthermore, while this paper has concentrated on country-level indicators, governments also need to have disaggregated data to identify bottlenecks and minimize "digital divides." [26] As countries are at different stages of development, governments also need to have statistics that are relevant to their situation. This is often not the case with data compiled by market researchers.

The logical choices for collecting the data are the national statistical office and the communications authority. The national statistical agency focus would be on the user side (e.g., statistics on household ownership and access to Information and Communication Technology) while the communication authority would be on the supplier side (e.g., statistics on number of ISPs, number of subscribers). The national statistical office already has the statistical skills and generally carries out regular household surveys. Internet-related information could easily be added to the survey questionnaires. As industry regulator, the communication authority should gather infrastructure and subscriber statistics, tariffs and other related information.

Until the necessary systems can be put in place, an interim solution is for a government agency to be responsible for locating and disseminating the information on a national Internet monitoring website. When necessary, an English version could also be made available to ease access to the statistics by the international Internet research community as well as potential investors in the market. This "one-stop" shop [27] would enrich analysis [28] and policymaking by making it easier to retrieve data. Quality and transparency would improve by having the data exposed to a larger group of users. One example is the Research on Internet in Slovenia project, coordinated by the University of Ljubljana, which provides a wide variety of useful Internet-related statistics for that country on its website [29]. Another example was a European Union project that compiled Information Technology statistics for its member countries using a standard format [30]. The theme of ISOC '99 was "Internet is for Everyone." Standard and timely "Internet Statistics for Everyone" would help measure progress toward that goal.

Acknowledgments and disclaimer

Tim Kelly and Larry Press provided valuable comments. The views expressed are those of the author and may not represent those of the ITU or its members.

References

[1] For another perspective on infrastructure indicators see Organization for Economic Cooperation and Development (OECD). Internet Infrastructure Indicators. October 1998. Paris. Available at <http://www.oecd.org/dsti/sti/it/cm/prod/tisp98-7e.htm>.

[2] The surveys are available on the Internet Software Consortium website at <http://www.isc.org/ds/. Réseaux IP Européens (RIPE) conducts monthly surveys of European host computers (as well as some countries in Asia and North Africa). The surveys are available on the RIPE website at <http://www.ripe.net/>.

[3] For example, the Korea Network Information Centre (KNRIC) compiles a variety of Internet statistics including hosts for the Republic of Korea. See the KRNIC website at: <http://www.krnic.net/english/net/2_93_00.html>.

[4] See http://www.isc.org/ds/defs.html which also provides definitions of other terms.

[5] The number of users has often been estimated based on the number of hosts. For example, "Such guesses are based on counting the number of computers connected to the Internet, guessing how many people use each computer, multiplying the two..." See Peter H. Lewis. "The Internet and Gender." New York Times. 29 May 1995. Available at <http://www.mids.org/nytgen.html>. However, this practice is unreliable as the number of users per host varies widely among countries. Nonetheless there is typically a relationship between the number of users and hosts.

[6] For example, Netcraft collects data on the number of Internet hosts providing Web (http) services. It found there were 6.6 million Web servers in July 1999 compared to a global host count reported by the Internet Software Consortium of 56.2 million. It makes one wonder what the other some 50 million hosts are doing. See the Netcraft Web Server Survey at <http://www.netcraft.com/survey/>.

[7] See <http://www.isc.org/ds/faq.html>.

[8] All of whom are provided with free Internet access. See Richard Saint Clair. Full Internet in Niue (NU). E-mail message. 16 November 1999. Available at < http://www.nsrc.org/db/lookup/operation=lookup-report/ID=942799095332:488918744/fromPage=NU >. One reason for the popularity of the .nu domain name is that it is marketed as an alternative to the popular .com. See .NU Domain Ltd. Surge in .NU Domain Registrations Funds A Tiny Nation's Free Internet Services. Press Release. 3 February 1999. Available at <http://www.nunames.nu/Press/surge.cfm>.

[9] See New IP Hosts SB - Solomon Islands. Available at < http://www.nsrc.org/db/lookup/operation=lookup-report/ID=890202505692:497423945/fromPage=SB >

[10] For example, the OECD mapped host registrations to countries under the generic .com name. In the case of Canada, it found there were over 70 percent more hosts than those represented by the Canadian .ca country domain name. See OECD. Internet Traffic Exchange: Developments and Policy. 1998. Available at: <http://www.oecd.org/dsti/sti/it/cm/prod/traffic.htm>.

[11] See  James Wong &   Eric Lam.  Measuring E-Commerce in Singapore - Methodological Issues and Survey Findings. Singapore Department of Statistics. December 1999. <http://www.singstat.gov.sg/EC/papers/D5.pdf>.

[12] The Netcraft Secure Server Survey gives an indication of the number of websites conducting electronic commerce. The location is derived from the address in the certificate rather than the domain name. See <http://www.netcraft.com/ssl>.

[13] These data are compiled on an annual basis. See http://www.itu.int/ti/.

[14] Numerous user surveys confirm that most users outside of the United States do not pay for access themselves. For example see GVU's WWW User Surveys at <http://www.gvu.gatech.edu/user_surveys/survey-1998-10/graphs/general/q23.htm >.

[15] One popular online source of the number of Internet users is provided by the Irish company NUA. Although it defines user as "adults and children who have accessed the Internet at least once during the 3 months prior to being surveyed" the national surveys it cites do not always use that methodology. If only subscriber data is available, NUA multiplies this by three to estimate the number of Internet users. See their website at <www.nua.ie>.

[16] Data on Internet users in Africa and the Middle East are being compiled. Mike Jensen has tried to overcome the lack of Internet user statistics for Africa by regularly estimating data for that region. See his website at <http://www3.sn.apc.org/africa/afrmain.htm#user-pops>. User data for the Arab world, derived from subscriber statistics, are available at <http://www.ditnet.co.ae/itnews/me_internet/users.html>. A similar compilation of Internet data for other regions such as Asia or Latin America is sorely lacking.

[17] For example in Finland, Taloustutkimus Oy (http://www.toy.fi/tuotteet/internet/inet5e.htm) has been providing a breakdown of the number of Internet users in that country since 1995. The intensity of use is measured (e.g., every three months, weekly, daily) as well as the location (home, work, school).

[18] One of the world's leading market researchers, ACNielsen, plans to extend its Internet surveys from North America to more than 30 countries representing about 90 percent of the world's Internet audience by the end of 2001. See ACNielsen. "ACNielsen, NetRatings Launch First Global Service for Measuring the Internet. " News Release. 15 November 1999. Available at <http://acnielsen.com/news/asiapacific/sg/19991115.htm>.

[19] An estimated 5.6 million Australian adults (41percent of the adult population) accessed the Internet in the year ending August 1999. See Australian Bureau of Statistics. Use of the Internet by Householders, Australia. ABS Catalogue No. 8147.0. August 1999. http://www.abs.gov.au/websitedbs/D3110122.NSF/66b4effdf36063e24a25648300177cd5/4bb5280f4d9b96694a2567a00005636c?OpenDocument

[20] For example, 22.6 percent of the population between ages of 3-17 and 22.1 percent of the population older than 18 used the Internet in 1997. Although the data are richly disaggregated by location of use, gender, age, region, income and other variables, they are irregularly issued and not up to date. See U.S. Department of Commerce. "Computer Use Up Sharply; One in Five Americans uses Internet, Census Bureau Says." Press Release CB99-194. Washington D.C. 14 October 1999. <http://www.census.gov/population/www/socdemo/computer.html>.

[21] For example see Belgian Internet Service Providers Association. ISPA market survey gives the update of the Belgian Internet market. 17 September 1999. <http://www.ispa.be/en/c030213.html >.

[22] "The obligations set out in the Voice Telephony Directive comprise the provision of voice telephony service via a fixed connection which will also allow a fax and a modem to operate...By including network access within the scope of universal service, users are given the possibility of accessing not only the defined voice telephony service but all services that can be provided over today's telecommunications networks (i.e., every citizen will be able to access interactive and on-line information services including the Internet, provided they have a computer and a subscription with an Internet service provider." http://www.ispo.cec.be/infosoc/legreg/9673.html#RTFToC4

[23] See CommerceNet/Nielsen. "Ecommerce survey shows 27 percent of British adults now use the Internet on a regular basis." Press Release. London, October 27, 1999. <http://www.q4.com/cnet.objects/Templates/pr/pr271099.html >.

[24] See OECD Internet Access Price Comparison at <http://www.oecd.org/dsti/sti/it/cm/stats/isp-price99.htm >.

[25] Not surprisingly, national studies often refute international comparisons. For example, France and Portugal have been shown to have relatively high Internet access prices in international comparisons. Yet studies carried out by those countries suggest they have  lower Internet access prices than comparable countries. This reflects the difficulty of making meaningful Internet tariff comparisons. See Institut de l'audiovisuel et des télécommunications en Europe - Idate. Le coût de la connection à Internet: comparaison entre quelques grands pays. October 1997.  <http://www.telecom.gouv.fr/francais/activ/techno/technweb1ia.htm> and Instituto das Comunicações de Portugal. Custos de Acesso à Internet em Portugal 1998.  <http://www.icp.pt/publicacoes/estudos/internet/index.html>.

[26] A U.S. government report disaggregates Internet access data by age, sex, income, race and region. See U.S. Department of Commerce. Falling Through the Net: Defining the Digital Divide. July 1999.   <http://www.ntia.doc.gov/ntiahome/digitaldivide >.

[27] An example of this "one-stop" shop for economic statistics is the International Monetary Fund's  (IMF) Special Data Dissemination Standard (SDDS). The IMF's SDDS website provides standardized and up-to-date economic and financial data for participating countries in order to enhance transparency and comparability. See <http://dsbb.imf.org>.

[28] This document has focused on the  indicators useful for analyzing Internet development. For an example of the type of analysis that uses these indicators see  "Tracking the Global Diffusion of the Internet" website at http://som.csudh.edu/fac/lpress/gdiff/index.htm.

[29] See "Overview of Basic ICT Indicators in Slovenia" at <http://www.ris.org/ict.html>.

[30] See "Basic Indicators" at <http://www.ispo.cec.be/esis/Basic/HomeBasic.htm>.