and examines the literature, both International and New Zealand originated to determine currently utilised approaches to e-commerce measurement. The relevance of each measure is discussed, along with recommendations of additional measures that are feasible and would assist in understanding the development of the "modern economy". At the conclusion of this section, specific indications are provided to assist in the wider contextual use of these measures within the framework of Connectivity, Capability and Uptake together aiding in the understanding of how changes in the use of information are impacting upon the wider social and economic performance of the "modern economy".

The main focus in this section, in the spirit of the brief for this report, is on measures that reflect utilisation of electronic information processing technologies. However, we have included discussion of measures of non-electronic technologies where these are already available and add insight.

Connectivity

Connectivity provides linkages between individuals and firms. These linkages are achieved via a number of media, on a continuum from purely human on the left to purely electronic on the right, as illustrated in Figure 5. The second line of Figure 5 identifies some of the technologies that enable connectivity, and hence communication, using the media identified in the first line.

Figure 5: Continuum of Information Communications and Processing Technologies

Three key types of connectivity are evident from the continuum:

person-to-person;
person-to-machine; and
machine-to-machine.

While some person-to-person communication is mediated by machine, the origin and destination of the information communicated is human. The other two forms, machine-to-machine and person-to-machine (including its converse, machine-to-person), are determined by the origin and destination of the information contained in the message of which connectivity facilitates the transfer. If, given network externality benefits, more connectivity is better, then higher levels of connectivity via all possible mechanisms should indicate higher potential to exchange, and hence utilise, information. Nonetheless, it is noted (as above) that there is an optimal size beyond which it is inefficient for a network togrow.

Within the types of connectivity are the media via which information is exchanged, principally:

voice;
written; and
electronic (digital).

Voice is exclusively a person-to-person medium, while electronic is limited to machine-to-machine communication. Written media can be used for both person-to-person and person-to-machine/machine-to-person communication interfaces, even though all three may be achieved by means of actual connectivity via an electronic device. Machine-to-machine connectivity may have two purposes: it can be used to automatically manage specific operations independent of human intervention, or it can be used to improve the quality of information received from or subsequently supplied to humans.

Person-to-Person Connectivity

Human communication and connectivity patterns are reflected in patterns of social organisation - families, friendships, geographical communities and other such demographic linkages - as well as associations of communities of interest. Basic demographic measures are contained in national statistics, principally in Census data (for individuals, families and social groupings) and in industry returns (for business groupings). While it is acknowledged that all social and business exchanges are forms of connectivity resulting in exchange of information, statistics collected have been limited to physically observable measures of association, such as family group numbers and composition, population density and firm employee makeup. The extent of communication within each group, or the need to exchange information is not measured, due to the intangible nature of information, and little external justification to monitor such activities.

While it is acknowledged that the dynamics of information communication within and between these groups has been a primary driver of demand for new and better electronic communications equipment, gaps still exist in understanding how information exchange needs of human organisation are driving uptake of human-to-human connectivity tools - is the need to exchange information driving connectivity, or is increasing connectivity driving even higher levels of information exchange by enabling the development of new information products, for which new demand is created, in mutually reinforcing cycles?

Limited understanding of human information processing and how human connectivity influences productivity means that few strong conclusions have been able to be drawn about the relative merits of human connectivity fostered by clusters of like people in cities or businesses, or clusters of like businesses in close geographic proximity, although weak associations have been found (e.g. Porter, Silicon Valley). Little empirical work could be found on the effects of interpersonal connectivity facilitated by technologies providing improved opportunities for communities of like interest but diverse geography, although anecdotes abound about the commercial opportunities offered by the Internet. While evidence that connectivity fostered by tools such as Internet Auctions on E-Bay enables one-off trades to occur that may not otherwise have been achieved (that is, enabling new sales), and sometimes at higher prices (value-enhancing) (e.g. EIU papers), there is less clarity in analyses of widespread one-to-one matching of buyers and seller on the Internet via tools such as Amazon (e.g. Brynjolfsson, Goolsbee). Indeed, Goolsbee suggests that what we are witnessing may be merely substitution of one form of connectivity (flyers, catalogues and mail ordering) by another (the Internet), with possibly limited impact on the number, identity and information content exchange of people actually connected.

Thus, person-to-person connectivity figures need to identify not only the numbers of people and businesses interconnected, but also their identity, connection patterns and information exchange needs. That is, how many are connecting, who is connecting with whom, and why, in addition to how they are connecting. Specific technology usage counts reflect only the "how many" and "how" elements, but ignores the "with whom" and "why". Limiting connectivity measures only to electronic technologies ignores the substantial information exchange already occurring via verbal and written media. We do not measure these well in the existing economic performance frameworks, yet we are expecting to see measurable changes when these are partially or fully substituted by machine-based technologies.

Focused research into these patterns of connectivity should provide greater understanding of observed patterns of diffusion of electronic connectivity, providing foundation and use for the measures of connectivity evidenced in the numbers below.

That said, measures of electronic connectivity do exist, and do provide valuable information with respect to "who" is connected and "how".

Telephones

Telephones provide two key forms of person-to-person connectivity - the ability to exchange information via voice (traditional telephony information communication) and the ability to exchange information via electronic media (e.g. by dial-up access to the Internet). The technology that enables this exchange also enables computer-generated machine-to-machine connectivity. However, the statistics collected on telephone connectivity focus almost exclusively on person-to-person connectivity.

The principal sources of data on telephone connectivity include (for example, but not exclusively⁴⁶):

the annual OECD Telecommunications reports and associated reports (national and international)
telecommunications providers (e.g. Telecom in New Zealand, Telstra in Australia)
telecommunications users associations (e.g. TUANZ in New Zealand)
regulatory authority statistics (e.g. the FCC in the US, OFTEL in the UK)
telecommunications consulting companies (e.g. Ovid, Teligen)
Government statistical agency reports (e.g. Statistics New Zealand, US Census Bureau)
other Government departments (e.g. New Zealand MED's annual technology survey, NOIE's Current State of Play, New Zealand Telecommunications Inquiry (2000)) .

While most of these sources provide publicly available data, some (for example, the telecommunications providers and consultancies) have additional data and analyses available for purchase.

Principal telephone connectivity data comprises:

the number of connections per head of population by country - for both land line and mobile (e.g. OECD reports, telecommunication consultancies)
the number of connections by user type (e.g. business, residential)
the number of second and subsequent lines purchased
the number of connections by specific technology type (e.g. ISDN, XDSL-capable, digital exchanges
(typically available from OECD and telecommunications consultancy reports)
line capacity (sometimes by geographic location within countries - e.g. NOIE Bandwidth Inquiry)
call volumes by number, duration and type (data or voice)
methods of connectivity (wireless, microwave, satellite etc.)
(generally from telecommunications providers and consultancies, or from regulatory agency reports)
prices (usually posted, and very hard to compare across countries due to different attitudes to bundling)
price structures
(OECD pricing comparisons, but note that these are based upon posted prices, and therefore are hard to compare across countries, even when converted to purchasing power parity, given the varying patterns of bundling and usage in different countries)
demographic data

population density
household/business characteristics (e.g. age, income, numbers)
business segment
(typically from national Census statistics)

Key comparisons using connectivity data include comparisons between countries and regions, and within countries and/or regions over time. This requires standardisation, as achieved by the OECD reporting mechanisms. However, inter-country and inter-region comparisons using telephone connectivity figures and prices are fraught with difficulty due to distortions created by differences in product bundles, utilisation practices and policy regimes (see pp 29 The State of E-New Zealand), so care should be exercised when judging relative performance across such boundaries (e.g. Local Access Pricing and e-Commerce shows the effects of pricing policy on usage figures).

Notwithstanding, connectivity statistics provide useful indications of who is connected via telephone, and to some extent, how much the telephone is used for communication by different categories of people. Telephone connectivity for information exchange by voice also indicates potential connectivity for information exchange via machine (e.g. the Internet), but this presupposes that the telephone will remain the predominant method of machine connectivity to the Internet, and this can be by no means assumed over the medium-to-long term, given the rise in use of other methods of electronic exchange. Regional and demographic breakdowns may be used to identify the existence of "potential digital divides" due to limitations in telephone and infrastructure access, but in the absence of analysis indicating information exchange need, connectivity figures alone are insufficient to diagnose divides.

Furthermore, changes in telephone connectivity figures over time reveal technology substitution patterns - for example, domestic usage of second lines being substituted by mobile telephones, and increasing use of mobile communication by individuals rather than fixed communication with someone in a specific location. However, care in interpretation is required here due to the shift in the unit of analysis - for example, from a family or a business at a given location to a specific individual, irrespective of location. Different units of analysis have different information needs, so the substitution may not be complete.

Analysis of substitution of voice traffic by data traffic is feasible, but difficult to assess, due to the absence of publicly available data. While telecommunications providers have access to such data, it is not freely available for regional, national and international comparisons. This reinforces the limitations of connectivity data analysed in isolation from information needs and usage. It is difficult to draw firm conclusions without more detail about the information exchange uses to which these technologies are put, justifying more research into this fundamental underpinning of communication, and hence connectivity, demand.

Television

An analysis of electronic connectivity should also include the use of television as a method of connecting people, and facilitating information exchange, albeit one that until recently has been a one-way exchange from the broadcaster to the recipient. However, growing use of cable and satellite television has raised the prospect of utilising this mechanism for connectivity of individuals via the Internet and Internet-like devices. Television also represents an application of a specific technology to a reasonably tightly defined, but information-rich exchange mechanism - entertainment - to which other connectivity features can be added - e.g. customised downloading of large quantities of content, generated by small quantities of two-way connectivity (interactive requests). In terms of analysis of a model of demand for information exchange, utilisation of pay-per-view satellite and cable television represents both a measure of a different sort of connectivity, and a measure of cultural exposure and acceptance to the buying and selling of information goods, and receipt and payment (or at least, debiting of an account) via electronic exchange.

While national statistics (e.g. Census figures, and New Zealand's annual MED technology survey) record the number of television sets by various demographic breakdowns, analysis of pay TV and other limited-function Internet-capable devices such as Playstation 2 is not typically a feature of economic performance indicators. OECD and NOIE figures, for example, record mobile telephones and computers as primary connectivity devices to the Internet, but not satellite and cable TV connections.

While population-based pay television subscription, utilisation and medium (cable, satellite, UHF) information is typically proprietary to the companies providing the service, survey-based information is available from agencies such as Nielsen and Forrester. While the survey nature of this source poses limitations, in addition to providing a measure of additional potential to connect individuals with each other electronically, the associated usage patterns (i.e. the demographics of purchasers, and the types of information/entertainment purchased) may be useful for the analysis of differences in demand for information between individuals and households, and over time.

Computers

Technological emphasis on computers as the principal tool of electronic information processing has combined with the emphasis on the telephone network as the principal tool of electronic connectivity to create the impression (reflected in the analyses of section 2) that together, computers and the Internet are the key drivers of change in the "modern economy". Computers have been the significant technological tool precipitating changes in the ways in which information is collected, processed, stored, transmitted and utilised. From a connectivity perspective, therefore, it is important to measure the extent to which individuals and businesses are using computers (that is, "connecting" to computers) both as tools of production, and as tools of communication in chains that include other information processing technologies (such as human, telephone, etc.).

Data on the extent to which businesses and individuals use computers are contained principally in:

national statistics (national Census data, surveys of IT capital investment)
OECD surveys (international comparisons)
Government agency reports (e.g. New Zealand's annual MED technology survey, NOIE's (Australia) Current State of Play)
consulting firm surveys (e.g. PWC, BCG, Deloittes)
market research firm surveys and summaries (e.g. Nielsen, Forrester)
sales figures of computer equipment provider firms (e.g. Oracle, Cisco, IBM, Unisys etc.).

The data most generally provided are:

counts of households/businesses with computers
the dollar value of national stocks of business computers
breakdowns of computers and computer and communication equipment produced
sales figures of computer producers may give indications of numbers of specific models/capacities of computer/equipment supplied (but acknowledging that these figures may be unreliable if their principal purpose is as strategic tools to position producers in the market against their rivals).

While these figures provide measures of investment in computer equipment per head of population, with the potential to compare connectivity between countries and regions, and between users of differing demographic profiles (age, gender, industry, income, etc.) counts do not necessarily indicate the specific applications in which the computers are being used. Again, the measures indicate only potential usage.

Usefulness of computer stock figures is further limited by practical difficulties in valuing computer assets for comparative purposes. Computer numbers may be useful in identifying connectivity of individuals to the technology, but not connectivity of individuals to computing capacity. For example, a 486-based processor and a 1GHz Pentium 3 machine are both computers, and both may have had the same purchase cost, but the capacity, and hence potential usefulness for value-generation vastly different. While attempts have been made at a productivity level to record dollars of capital invested in computer stocks, this is also fraught with difficulty, as rapid technological and quality change in computers means dollars invested at different times result in very different information processing capabilities. Although attempts to address this include utilising hedonic price indices (Griliches) and vintage capital models (Greenwood), the more rapid depreciation cycle for computers (three years on average compared to 10 years for other capital items), and the frequency at which new technologies are introduced make this an enormous data collection and analysis exercise, and one which is almost always lagging significantly behind actual developments.

Placing emphasis on computer connectivity also overlooks the fact that complementary investments in software are an equally vital component in effective use. Accounting practises which encourage expensing of software and training in computer use have discouraged collection of this information, although this is now starting to be remedied in the United States, at least (Jorgenson and Stiroh (2000)). However, it is becoming increasingly evident that investment in systems and processes surrounding the use of computers is also a significant component in their effective use (Brynjolfsson and Hitt (2000)). Comprehensive connectivity figures should thus include software and systems spending in addition to hardware spending if the potential of connectivity is to be realised. While this does not address the issue of how effective spending on these items will be, it does at least recognise that computer connectivity requires more than just availability of equipment if benefits are to be yielded.

Another shortcoming of current computer connectivity figures is the emphasis on counting only stand-alone computers used solely for person-to-computer connectivity. None of the measures examined for this report provided counts or separate valuations of computer-controlled capability installed in other capital equipment (for example, microprocessors in other production equipment - such as computer-controlled vehicle assembly machines). Such machine-to-machine connectivity (that is, computers embedded in other production technology) is a significant component underpinning potential productivity gains in sectors other than the Information Processing and Communications sector, and provides another dimension for analysing the extent to which electronic information processing technologies are substituting for other technologies. While it may be reasoned that the macroeconomic effect of this substitution is captured in the outputs of the ICT-producing sectors, and as inputs into other manufacturing sectors, of national accounts, importing and exporting of fully built-up units does not enable components to be separately accounted. Computer connectivity may therefore be under-represented in analyses of electronically-dependent production of non information and communication technology -(ICT) goods, further obscuring the fact that "electronic commerce" is a fundamental part of all production and consumption activities and must be integrated into all economic and social activity monitoring, not maintained as a trading segment set apart from other activities.

A final point regarding computer connectivity concerns the linkage drawn in most analyses between computer connectivity and connectivity to the Internet. While current technologies generally require access to a stand-alone computer to access the Internet, the fact that other technologies (e.g. cable television, WAP telephones, Playstation) are emerging as contenders for this form of connectivity mean that the relative importance of computers as essential requirements for participating in the "internet economy" will diminish. While currently, these statistics assume pivotal importance, a caution is offered - it may be possible, in times of rapidly changing technologies to over-read the importance of one form of connectivity. For just as computers may be superseded by other forms of Internet connectivity, so computers themselves, in conjunction with the Internet, may provide a technology that will supersede the telephone - Voice-Over-IP (Internet Protocol).

Internet

As a network that enables connectivity of individuals via computers and other information processing and communicating technologies, the Internet has been a primary facilitator of increased information exchange potential, between all levels of social and economic organisation. The Internet is the management infrastructure that manages message distribution over the myriad of "pipes" available for moving digital data. Thus, connectivity to the Internet opens up the potential for digital data exchange with any other entity that is likewise connected. As it is a single world-wide network, connectivity to this network is theoretically, by network valuations and network externality effects, the most valuable network connection possible. Connectivity to it, therefore, is the most valuable membership in an economy predicated upon value-creation from information movement and investment.

The position of the Internet as the world's most valuable information exchange network is evidenced in almost all of the taxonomies identified in section 2. However, while in theory it represents the most valuable network, sizing it and placing a value on connectivity to it is problematical, given the variety of ways in which individuals and entities connect (specific technologies, points of entry etc.), the reasons they connect (supplying or seeking information) and the roles in which these activities are undertaken (consumer or producer).

Information about the numbers of entities connected to the Internet is available from a variety of sources, including:

the OECD
national statistical and analysis agencies (e.g. NOIE)
Internet Service Providers (ISPs)
Domain Name Registrars (e.g. Domainz in New Zealand)
Bandwidth and connection providers (e.g. telecommunications providers)
Telecommunications Regulators (e.g. FCC, OFTEL)
Telecommunications consultants (e.g. Ovid)
Telecommunications users (e.g. TUANZ)
Consulting Companies (e.g. KPMG, PWC, Deloittes, Ernst & Young, BCG)
Market research companies (e.g. Nielsens, Forrester, BRC)
Researchers (e.g. FORST study (University of Waikato), BRC)
Government departments (e.g. MED, MAF)
Listing directory providers (e.g. ISCR/Yellow Pages)

Data collected differs, depending upon the principal reasons why individuals and entities wish to be connected to others via the Internet:

domain names per head of population (generally by country - OECD/ICANN)
records unique Internet domains registered, and is thus a measure of entities wishing to identify their willingness for others to find and communicate with them (that is, information traders)
Internet connections per head of population (generally by country - Network Wizards, Telcordia)
records computers connected to the Internet, and therefore provides a proxy for individuals/businesses indicating a desire to use the Internet to access information (noting that this is a representation of computers connected to the Internet, and does not necessarily represent people connected to the Internet, as many people may utilise one computer connection to get onto the Internet)
Secure servers per head of population (generally by country - OECD, Telcordia)
records the stock of secure trading sites physically located in a given geographical location
data volumes transferred
measures the amount of data traffic passing through specific points of the Internet
(although this is generally proprietary to telecommunications providers, ISPs and companies paying for the information, and is not in the public domain)
hours spent online per month (market research companies)
measures of average number of hours spent online by members of a reference group of users, sometimes broken down by demographic characteristics (not generally public - requires purchase) - a "consumption" figure - can measure time connected, but not necessarily the volume of data exchanged, the direction (uploading/downloading) or the purpose to which the information is put
average number of sites visited per session online (market research companies)
another "consumption" figure - similar to hours spent online per month
most popular sites (market research companies)
measured by sample, and sometimes provided/verified by ISPs from data collected by cookies. This is principally an information consumption measure and can identify some information content, but often is only an identifier of the most popular portals for accessing other sites
aggregation of traffic volumes by site (provided by ISPs/web hosting services)
is both an information sourcing and provision measure - it can identify what information is most regularly requested from sites, how many times it is provided, and in many cases the ISP of the end user. However, it does not identify (usually) the identity of the end user, nor the use to which the information is put.

Internet connectivity figures can be used to indicate the capability and intention of both producers and consumers of information to both make it available, and to seek it over the Internet. However, the inability to link information consumption to individual end users, and reliably account for all data traffic at an end user level means the actual extent of Internet connectivity is just an estimate. Furthermore, while geographic location of domain name registrations and secure servers can be reliably determined, the location of actual usage of the domain name (that is, the physical location of the host computer) and the computer from where a user is accessing the Internet may not be easily determined, meaning that comparisons may be biased (although Telcordia has developed a methodology to adjust domain name figures to the country of physical location of the server).

The relevance of Internet connectivity figures in the information communication economy, at the current point in time at least, is drawn into question, however, by statistics revealing that, while the Internet is the fastest growing of all connectivity networks, by traffic volume (in the United States) it is swamped by data communication via other methods, principally EDI (Galbi (2000)). Thus, while the concept of the Internet as a large and ubiquitous network has dominated conceptual thinking and analysis effort, it is still quite insignificant with respect to actual volume of information exchange. Further, the concept of the Internet as the tool via which comprehensive information can be made freely available to all users is also limited by the finding (Bright Planet (2000)⁴⁷) that the data available on the commonly defined "public web" is also swamped by the volume of data on the "deep web" - that is, invisible to current search engines. So while the concept of a huge valuable network is appealing, connectivity figures in isolation from usage and information availability figures can lead to distorting impressions of the value of the network, at least in current applications.

Conceptual problems also limit the usefulness of other Internet-related connectivity statistics (e.g. the NielsenNet statistics). Session numbers and session lengths, for example, are not good proxies for information volumes and values accessed especially if they include idle time awaiting responses or processing data already received. Likewise, valuing information using Internet transactions is problematic. Data volumes transferred cannot be used as a measure of data value, as volume measures are insensitive to content. Yet, data valuations are commonly measured in terms of the price of the medium (e.g. a CD) or the cost of moving volumes of data, rather than the value of the data moved. Furthermore, survey figures measure Internet usage almost exclusively at an individual consumer, rather than business unit level. Hence, the figures reflect recreational use of the Internet rather than commercial. Connectivity figures also ignore the relative sizes of the organisations connected - for example, one connection may represent a single individual, or a multinational firm with several hundred employees, even though usage volumes and values are vastly different.

Connectivity statistics are also insensitive to the end use to which the data is put. Thus, measures of capability to exchange dominate over actual usage. This has the potential, given the imperative that "getting connected increases the value", to lead to possible over-investment in Internet connectivity, tying up resources in Internet capability (with low transaction volumes and values) which may be more efficiently applied elsewhere (Milgrom (2001). Sometimes the most efficient way of transferring data is on a paper form via post or fax, especially for low volume, non-standard transactions.

Thus, Internet connectivity figures must be interpreted with care, and considered in perspective with the rest on the economy in which the Internet forms a part.

Bandwidth Availability and Utilisation

While computer and Internet connectivity figures provide indications of end user ability to collate and process information, electronic transmission of data between end user sites requires sufficient bandwidth to carry these transmissions. Bandwidth is required irrespective of the medium of final processing (both voice and data are carried using this capacity) and the traffic management mechanism (EDI, telephone switching, packet switching etc.). While the quantity of bandwidth required is determined by a complicated interaction of prices (Goolsbee (2000)), pricing structures (Varayia and Varian (1999)), contract mechanisms (Lehr and McKnight (2000)), application availability and information intensity (NOIE Bandwidth Inquiry (1999)), anticipated future demand (Rood (2000)), taxation policy (Goolsbee (2000)) and such factors, available bandwidth quantity, and the locations it links, nonetheless, it has been used as a basic benchmark to assess connectivity and hence capability to take advantage of information connectivity.

Bandwidth figures require analysis from two perspectives: available capacity and utilisation. The principal advantage of using bandwidth utilisation figures for analysing electronic data communication is the ability to measure data transmission independent of the originating and terminating medium (voice, machine), and the controlling mechanism (telephone, Internet, EDI etc.). Available capacity figures in conjunction with utilisation figures and future demand projections provide an indication of potential to grow transmission volumes between specified points, and enable the identification of impediments to current and future transmission in specific areas.

Principal sources of data on bandwidth availability and utilisation are found in:

OECD annual telecommunications reviews
Telecommunications Regulatory Body reports (e.g. FCC, OFTEL)
Government Agency analyses (e.g. NOIE National Bandwidth Inquiry)
Telecommunications Consultancy reports and reviews (e.g. Ovid)
Telecommunications service provider reports
Other bandwidth providers (e.g. Utility companies)
Bandwidth wholesaling company reports

It is noted that figures published by the first four categories above are (anecdotally) based principally upon information volunteered by Telecommunications companies and other bandwidth providers, so may not be a full and comprehensive total due to issues of commercial sensitivity and associated reliability. Further, while detailed capacity and utilisation information is readily available in jurisdictions subject to regulatory disclosure, in New Zealand no such information is available in the public domain due to its proprietary nature and commercial sensitivity. Bandwidth information was procured in Australia in 1999 (Amos Aked Swift (1999)) for the Bandwidth Inquiry, but was subject to specific anonymising and disclosure conditions and is now outdated. Furthermore, Rood (2000) notes that the growing international nature of control of the Internet means that even in jurisdictions where regulatory disclosure is required, accurate utilisation figures are hard to obtain given the fact that traffic can be routed via foreign servers, even when the points of origin and destination are within one country. The increasing range of providers of bandwidth in deregulated and increasingly competitive markets, growing use of privately installed capacity (including LANS) for high volume and mission-critical transfers, and substitution of fibre by wireless options also provides challenges to the ability to collect comprehensive capacity and utilisation data.

While the proprietary nature of much bandwidth information makes it difficult to determine consistent and comparable figures for bandwidth quantities and prices, indicative measures are provided in reports such as Rood (2000), Galbi (2000), Ovid (2001) and the NOIE National Bandwidth Inquiry (1999). These include:

fibre miles (Galbi)
percentage of bandwidth in use (Galbi)
fibre capacity, cost and unit cost by cable (Galbi)
inter-office circuit prices (dollars per month per Mbps) (Galbi)
leased and private line revenue by world region (Galbi, Rood)
leased and private line capacities by world region (Rood)
US packet switching revenue (Rood)
installed and potential capacity by area population size (NOIE)
installed and planned capacity by city (NOIE)
planned capacity by type (fibre, satellite, microwave) (NOIE)
posted prices (OECD).

The NOIE National Bandwidth Inquiry uses figures obtained from their surveys to provide forecasts of future bandwidth demand by location (down to Census block level), highlighting the need to address bandwidth capacity and utilisation at both the backbone (trunk) level and the "last mile". It is noted that the dynamics driving each of these demands are different, and predicated upon different demand patterns for information exchange. The Australian exercise is the only one found in our literature review where analysis has been conducted at both a national and local level, with demand aggregated up in order to inform national policy-making. However, the dependence of the findings upon judgemental forecasts, single level "all you can eat" pricing and the key assumption of no new "killer applications" being developed limits the usefulness of the five year projections created.

Bandwidth availability and utilisation figures thus provide information on the extent to which data exchange is occurring, by medium, and by geographic location. Utilisation to capacity measures, and projections of future use also facilitate identification of present and future capacity constraints by location, depending upon the reliability of future needs-based forecasts. Capacity "divides" can thus be identified and addressed. Analysis by medium (fibre, microwave, satellite etc.) also enables patterns of substitution and complementary investment to be identified, adding to the analysis of specific technology diffusion.

However, application of bandwidth data to future decision-making has limitations. Links between bandwidth pricing and demand are not well understood, particularly with respect to demand for high-speed transmission, making demand forecasting problematical. Flat rate pricing of dial-up access and poor understanding of how to value one's time when connected to the internet are distorting investment decisions by end users, favouring continuation of modem access and depressing demand for high-speed (e.g. cable and XDSL) access, making forward planning difficult (Varayia and Varian (1999)). Regulatory mandating of prices and service quality (Crandall (2001)), limited service quality options (Lehr and McKnight (2000)) and differences in infrastructure taxation policies (Goolsbee (2000)) are also impacting upon demand, making it difficult to assess comparability of capacity and uptake between areas, bandwidth providers and national/regional jurisdictions.

Commercial sensitivity of data and high levels of aggregation make dissection of supply and demand at a local level difficult, meaning (NOIE excepted) most figures are "global" in nature, and hence relatively insensitive to marginal changes in usage at a regional or local level. The nature of cable roll-out also means that new capacity comes available in discrete blocks rather than gradually increasing quantities. Thus, available capacity is often supply-driven rather than demand driven, resulting in cycles of "feast and famine", further distorting pricing patterns and consequently demand (e.g. "fire sales" of surplus capacity). These features can be particularly pronounced in small markets, where the activities of one player can be significant, and where it is difficult to separate out the effects of demand from supply, pricing, availability and dedicated resources. This can result in flow-on investment decisions for other information processing which may not necessarily be optimal (e.g. the dial-up versus cable investment decision - (Varian (1999))).

Other Applications

A complete analysis of electronic connectivity should also consider connectivity to other electronic information processing technologies as identified in section 2. These include (for example, but not exclusively):

EFTPOS utilisation;
microprocessors in domestic technologies (e.g. smart fridges, washing machines, telephones, cars ); and
smartcard use (value cards and access/security mechanisms etc.).

Measures of EFTPOS customers and utilisation in New Zealand and Australia are contained in the annual KPMG Banking surveys. Smartcard utilisation is harder to assess, as figures are maintained only by providers on a case-by-case basis.

Connectivity to these applications illustrates a "seamless" adoption of electronic processing technologies into other production functions. Utilisation of EFTPOS and smart cards, in particular, illustrates how easily consumer populations can adopt and easily adapt to substitution of information (EFTPOS) and information products (represented on the smart card) where the benefits (e.g. time savings, lower risk from not having to carry cash) outweigh the costs (learning to use a new technology). Furthermore, utilisation of these technologies introduces consumers to new skills, such as the use of visual display units, key pads, passwords (Personal Identification Numbers (PINs)) and trusting personal information to a central network, which are important precursors to the ability to use and take advantage of other information processing technologies such as computers and the Internet (The State of E-New Zealand).

The disadvantage in measuring use of many of the other applications, such as computer processors in smart fridges, cars etc. is that microprocessor use within other technologies is not systematically recorded, and the number of products containing such technology is growing very rapidly. Further, as the items in which these microprocessors are contained are not considered "computers", users may be unaware of their construction, and hence unaware of their "second-hand" application of electronic technologies. Just as with EFTPOS, even though individuals may be very comfortable using these items, they do not consider that using them constitutes "electronic commerce" or "electronic production". However, we contend that in a wider economic framework, these applications of computer technology are equally as important as computers per se, or the Internet, as they utilise electronic technologies in order to improve outcomes for both individual users and society.

Capability

As identified in section 2, Capability measures provide insights into the resources available to an economy or society to be applied to productive and welfare-enhancing endeavour. Thus, they are measures of the quantity of resources available (that is, a stock take of skills and assets), along with some analysis of the quality of those resources. Together, quantity and quality measures enable projections of potential yields from applying the resources to be made (that is, estimates of how well we think we could be doing, given the resources and skills available). While capability measures indicate potential, they are nonetheless only indicators of potential, not actual performance. Capability and connectivity measures, combined with uptake, yield actual output performance measures.

Traditionally, capability measures have focused on resource measures:

human capital
physical capital
human capability and
access to infrastructure (a link between connectivity and capability).

The OECD and NOIE methodologies offer a sample of the types of measures typically polled from sources such as:

national accounts
firms' balance sheets
statistical agency data (e.g. census data, time use surveys, company statistical returns)
other government information (e.g. IRD, immigration/emigration data)
educational achievement benchmarks

The NOIE data, summarised in Appendix 2, provides many measures based upon demographic and industry characteristics, such as:

Characteristics of Australians accessing the Internet (age, gender, employment status, income level, family type)
Use of the Internet by children
International benchmarking based on population and household demographics
Business readiness and barriers to use of electronic commerce
International benchmarking of businesses

These are supplemented by other resource information, for example:

population by occupation classification
population distribution by geography
population education levels (by age/gender/geographic location)
workforce numbers by industry/sector classification
hours worked by industry/sector classification
individual remuneration by occupation classification
individual remuneration by industry/sector classification
migration rates by occupation/industry sector classification
migration rates by skill category/education level (international and inter-regional)
industry distribution by location
GDP (output) by industry/sector classification
capital stock by industry sector.

The key use of these statistics has been to assist in tracking performance and to inform the design of policies enabling the capabilities mix to be changed to bring about different (generally improved) uptake and performance outcomes. These measures can also be used to compare New Zealand potential with the levels in other countries (e.g. Australia), as well as acting as identifiers of shortages or surpluses of specific resources, thus generating policy intervention.

The difficulties posed in using these measures are several. Firstly, cause-and-effect relationships between changes in these capability measures and changes in tangible output levels are far from conclusively proven, and often take several years to be revealed (Colecchia (1999)). This makes the link between observations and policy directions tenuous, at the very least. There is still considerable debate about which skills are the relevant ones to measure in determining ability (and hence potential) to operate in an "information-based" economy. The dominance of electronic technologies in statistical measurement has resulted in a premium (higher weighting) being created in policy setting for people with science and engineering skills, but the foundation for this reliance on skills may be dubious. For example, most people successfully drive cars and gain significant benefit from so doing, without having to be fluent in understanding the engineering processes required to construct or repair a car. It is driving skills that are important. Likewise, the current focus on scientific skills for research and development may be at the expense of measuring capability in basic skills, (such as reading, logic and keyboarding), which may be far more influential (in absolute numbers) in enabling individuals to take advantage of the opportunities offered by new electronic technologies. Despite the emphasis given to technical skills in capability statistics, little is understood about the role of generic skills such as keyboarding, let alone intricate technical capabilities. Therefore, over-reliance on currently required technical capabilities and specific skill measures may be misleading when determining long-term capability development.

Secondly, existing capability measures are still analysed within a traditional physical goods-based industry sector breakdown. Classification of staff skills within industry sectors is not necessarily a good indicator of what constitutes a "knowledge worker", as many "knowledge worker" skill investments cross traditional industry borders (Engelbrecht (2000)). This may result in unrealistic expectations being set within a capabilities framework. Furthermore, there is still no real clarity about what a "knowledge worker" actually is. Traditional analyses have focused upon creators of new knowledge as "knowledge workers", confining analysis to those sectors of the economy supporting research and development (i.e. creators of ideas). However, increasingly the role played by the use of information has changed the emphasis onto workers whose primary inputs and outputs are information products. By this definition, knowledge workers permeate all sectors, yet current frameworks are unable to separate out high information intensity workers from low intensity ones within one industry⁴⁸.

Thirdly, the traditional capabilities framework is predicated upon the assumption of "one worker, one job", and that all skills acquired are focussed on meeting needs in that one job. Individual mobile transactors, managing a portfolio of jobs, possibly across industry divisions, may need multiple skills. Limiting analysis to one industry overlooks the need to up-skill specific workers in many tools. Likewise, the growing use of information technology products made available to domestic consumers for own production also underlines the need for measuring the skill base not just in employment situations, but also in non-industrial and unpaid sectors, if full capability is to be recorded. However, it is acknowledged that it is becoming extremely difficult to separate out employment-related activities from other ones, when boundaries become blurred.

In addition, the intangible nature of the items considered as human capital further complicate the ability to measure capability. Proxies such as years of schooling are proving inadequate as technologies are enabling the "uncoupling" and separate storage of some information previously considered human capital from the individual. New frameworks need to be developed if capability measures are to remain relevant and informative.

Uptake

While statistics of connectivity and capability offer measures of potential benefit, the process of converting potential into subsequent benefits requires linking of capabilities and connectivity in uptake of technologies. Like connectivity and capability statistics, uptake statistics track utilisation of specific skills and technologies, but these measure only utilisation, not the consequent benefits (or detriments) that arise from their use.

With respect to the brief for this report, this section surveys uptake measures of applications utilising electronic technologies, as distinct from the infrastructure measures identified in the Connectivity section. However, despite the reliance upon electronic technologies that enable these applications to be used, the key factors underpinning the effects yielded by uptake of these applications are the changes in the prices, speed, medium and method of utilisation of information. Furthermore, uptake of specific electronic applications will have flow-on effects with wider consequences than just within the sectors in which they are applied (for example, improvements in product quality flowing through to subsequent users of those products (Evans (2001))).

Uptake measures alone are thus merely "tracking signals" of what individuals and businesses are using connectivity and capability for. Interpretation of the consequences requires detailed understanding of how these behaviours impact upon wider performance measurements. However, the newness of electronic technologies, the intangible nature of the products and processes which are eventuating, the different characteristics of information ("experience" good, non-rival, non-excludable - see section 1), and the pace at which these new products and processes are being adopted makes drawing direct cause and effect links problematic. Furthermore, the fact that changes to information processing technologies affect all sectors of the economy, as all sectors use information, means that the scale of the changes leading to the "new economy" are far more wide-reaching than other technology changes, such as the introduction of electricity. Consequently, there is a scarcity of research into these effects.

The shortage of research is further compounded by lack of clarity firstly about what indicators should be tracked and measured, and secondly about how to standardise and compare these measures across vastly different economic sectors. As identified by Colecchia (1999), and evidenced above, while measuring infrastructure in connectivity indicators and capability indicators from resource accounts is relatively straightforward, analysis of which applications should be measured is in its infancy. Hence, there are many different approaches to measurement of application uptake.

The commercial impetus to implement electronic commerce in order to reap cost savings and the benefits of new product generation as soon as developments in computer technology make them possible, along with the easily-measured metric of computer investment being captured in most countries annual accounts, has meant that the majority of uptake measurement has occurred at the behest of businesses. This has resulted in collection of uptake statistics that support business decisionmaking surrounding the sale and implementation of specific technologies and applications within the business, rather than supporting analysis of the wider-reaching impacts of their uptake. This is reflected in the fact that most uptake statistics are collected and analysed by market research companies, business consultants and machinery and infrastructure manufacturers. It is only very recently that other agencies, such as the OECD, and government statistics and policy-making bodies (e.g. NOIE, MED, FORST), have begun commissioning and collecting statistics with a wider analytical intention in mind.

While the surveys conducted by commercially-incentivised providers may be coloured by the purpose for which the statistics were intended, they do have some value, as they provide one of the few sources of uptake information available. However, their findings must be interpreted with caution.

Firstly, most focus solely on applications relating to the Internet. While it is acknowledged that the Internet is the fastest growing of the new communication media, it must also be recognised that it still forms only a small component of commercial information movement. Thus, valuable information about non-internet computer applications, such as mature banking products like EFTPOS and ATMs, computer-controlled production and administrative systems such as payroll processing, are ignored. Likewise, information about uptake of non-electronic, but nonetheless important information-based process changes is also overlooked (for example, streamlining information flows through uptake of improved form design), with a double loss of information if these processes were introduced concomitant with the introduction of electronic processing technologies.

In addition, with the exception of the University of Texas Internet Indicators project, they are based upon self-reporting, voluntary participation surveys, raising questions about the representativeness of both the samples and the responses, and hence their ability to be translated to wider populations. Furthermore, many of the questions are subjective, relying upon the interpretations and perceptions of the respondent, rather than specific, independently verifiable aspects. If the respondent within the firm changes between surveys, comparability of surveys is compromised. The extensive use of Likert scales, rankings of lists of characteristics, and estimates of likely future behaviours result in few "hard" numbers, and figures with limited applicability and comparability across countries, firms, industry sectors and time. The firm-based approach also limits the ability to aggregate findings over larger operating entities, such as supply chains, networks and other associations, which are becoming increasingly common units of activity.

The balance of this section canvasses the uptake statistics identified. It is noted that in New Zealand, the principal independent surveys undertaken (BRC/MED in 2000, FORST/University of Waikato in 2001) have utilised frameworks based upon OECD/NOIE indicators in the first instance, and the APEC SME Electronic Commerce Survey undertaken by PricewaterhouseCoopers. Thus, methodologically, they fit and are analysed within in the category of Consultants Surveys.

Market Research and Website Information Collection Surveys

Uptake and usage surveys undertaken by market research companies (such as Nielsen and Forrester) focus principally on Business to Consumer (B2C) applications, and residential consumer uptake. The uptake measures thus reflect end consumption of Internet-based applications, and measure activities such as consumer recreational uptake. While some measures have been placed in the public domain (such as hours spent "surfing" per month and number of sites visited per session), the principal commercial motivation for market research companies to undertake surveys (other than specifically-commissioned research exercises) is to elicit information buttressing the processes of marketing products (e.g. advertising).

Thus it is not surprising that these surveys yield measures focused on the Internet, as it is the dominant electronic method of individual and end-consumer connectivity with sellers (other than the telephone), and on purchasing, as it is the dominant method of consumer exchange which producers wish to evaluate. There is limited incentive for market research companies to collect statistics relating to exchanges for which no remuneration is passed or likely to pass in the future (e.g. information searches for a student's project), as it does not support the information-gathering needs of the purchasers of survey outputs⁴⁹. Hence, such data is not usually collected, despite the fact that the information exchange has value to the consumer.

Market research surveys collect data predominantly from focus groups and groups of selected expert users⁵⁰. It tends to be technology-specific, and requires firstly identification of the technology that warrants uptake investigation, and identification and selection of "representative" users for focus and expert groups. The newness of many applications, and the need to get usage and uptake information rapidly means that these subjects are often "early adopter" candidates, thus the uptake patterns revealed may not be typical of "average" users in the wider population.

Key consumer data collected by market research companies includes:

the amount of time spent "surfing" per month
preferred websites
dollars spent on Internet-based sales
types of information accessed (e.g. games, information buying, etc.)

broken down by demographic characteristics of the survey respondents.

By contrast, web operators and hosting services are able to collect significant amounts of information about visitors to websites as a result of monitoring web traffic. Standard summaries provided to website owners identify the number of visitors, pages visited, data volumes downloaded, referring agencies, the web hosting service of the visitor, and other such information. These statistics are usually anonymised and aggregated into weekly or monthly totals. However, by placing "cookies" on the computers of site visitors, website owners can build a rich amount of data on specific and identifiable individuals. Cookies monitor off-line user activities and report this information back to the web hosting service, enabling sophisticated profiles of individual website visitors to be built. From the perspective of a seller in B2C relationships, this data provides a rich source of customer information, including individual web access habits, detailed search histories leading up to successful (or unsuccessful) sales ("clicks to sale" and lost sales). Sales do not have to be made to obtain this information.

Both survey and website information have the potential to reveal rich data about uptake patterns of Internet applications. In particular, individual habits can be measured, either knowingly by survey, or surreptitiously via cookies. While market research surveys focus on customer behaviour, website statistics enable all activity to be monitored, even if the information exchange process does not lead to a customer relationship being formed. Cookie data also has the advantage of being close to population-based rather than subject to significant sampling error and self-selection bias, although it is recognised that as more people are becoming aware of and voluntarily declining cookies, this data is becoming less representative.

The limitations of survey and website statistics are that the data is almost always proprietary to the firm either paying for market research data or owning the website. Patterns of uptake identified at the firm level become elements of strategic advantage to the owning company, so there is a positive disincentive preventing release of information for analysis. Hence, cookie data remains relevant only at a site level, unable to contribute to informing public understanding of consumer uptake patterns of uptake of services provided on specific websites.

It is noted, however, that Nielsen does release a limited amount of aggregated domestic Internet uptake data on the monthly NielsenNet ratings website. While these data:

relate only to domestic Internet access (it ignores business access),
measure only elapsed time spent on "surfing" sessions thereby ignoring actual downloading time and information volume transferred, and is contingent upon the skills of the user
value sales only, despite the very low proportion of "click to sale" conversion rates reported by cookie data,
are subject to selection bias, and
offer limited ability to detect changes in habits due to emergence of new technologies (surveys are not usually good tools for picking up the use of new technologies - the focus on the Internet and current usage means it will take time for new technologies to be included, by which time substitution patterns will have affected data and ability to interpret)

the information contained in these reports remains the best publically-available measure of consumer Internet access and hence domestic consumer uptake of web-based services.

Consultants' Surveys

While market research surveys and website statistics concentrate upon providing measures of consumer uptake of Internet applications, surveys by consultants such as Deloittes, Boston Consulting Group, PricewaterhouseCoopers, and Ernst and Young provide more detailed information about information technology uptake patterns among firms. These surveys combine both quantitative and qualitative measures, and endeavour to explore both business-to-business and business-to-consumer applications. Some are now into their third year of data collection, enabling patterns of uptake over time to be developed.

While there are minor variations between the firms, the surveys of all the management consultancies cover a similar very broad range of operational and strategic issues relating to uptake of Internet applications. These surveys appear to be designed to "dig below" levels of physical uptake in an endeavour to understand the motivations of firms using specific applications, and the reasons why some firms choose not to use them. Key issues common to all surveys include:

profiles of respondents (can analyse by business sector, firm size (employees, turnover, capitalisation etc.)
perceived benefits/motives for e-commerce (ranked and weighted) e.g.

improve customer service
enhance company image
customer information exchange
improve competitive position
increase customer loyalty
access international markets
increase revenue
reduce costs of information
supplier information exchange
attract new investment
reduce procurement costs

perceived inhibitors e.g.:

low customer and supplier e-commerce use
trust and confidence issues
internal firm capacity issues (knowledge of technology & business models)
infrastructure (high cost of technology, telecoms reliability)

perceptions of key issues for the industry:

competitive strategy
financial management
human resources
alliances
e-commerce strategy

infrastructure spend

by firm capitalisation
by revenue

perceptions of profitability of e-commerce applications

profitable
unprofitable
don't know

spending via e-business applications
using specific applications (by yes/no - could presumably be correlated by industry/firm size etc.)

email
website
EDI
on-line purchasing
on-line sales (to businesses and consumers)
receiving payments
information searching (F)
after-sales service (F)
delivery co-ordination (F)
on-line market research (F)
e-competitor intelligence (F)
on-line staff recruitment (F)
on-line staff training (F)
extranet
(F) represents FORST data identified in addition to consultants' questions

perception of e-business impact on the organisation

none
negligible
unsure
moderate
substantial

Importance of e-business for competitive advantage

none
minor
neutral
important
highly important

perceptions of assistance required:

risk assessment
total solutions
project management
marketing
security
strategy
technical solutions

where respondents would go for assistance

consultants
web page developers
current business advisors
ISPs
Software vendors
Computer services solutions
Hardware vendors

The FORST/University of Waikato study of New Zealand businesses with 10 or more employees, based on the PWC report prepared for APEC, also includes information on:

organisation type (listed, non-listed, sole trader, partnership etc.)
ownership (family, New Zealand, international etc.)
geographic scale (local, regional, national, international)
revenue source (New Zealand, international)
purposes to which websites put
estimates of website activity into the future,

which enables additional analysis of the relationships between aspects of ownership and organisational form and uptake of specific Internet-based applications (although it is noted that, due to the early stage of this project, such analysis has not yet been undertaken⁵¹).

The Boston Consulting Group survey also includes:

counts of B2B marketplaces
percentage of B2B activity by sector
percentage of e-commerce as percentage of all B2B trade
estimates of benefits in percentage cost savings terms

enabling a level of aggregation of firms' individual activities and policies into market indicators.

As with the market research surveys, these consultants surveys, while less than ideal, do represent a start in measuring not just what applications are being implemented, but the rationale supporting their implementation (or lack of implementation). The FORST/University of Waikato survey is particularly important in that it is specifically addressing uptake measures in a New Zealand context, over time. The independence of the research team from any other commercial relationship with the subject firms (unlike the consultants, who have a client relationship with survey respondents), and the ability to include analysis of additional areas of interest that may not be disclosable in the presence of a commercial relationship (e.g. ownership data) add to the potential value of these uptake statistics.

The limitations of the consultants' data remain the centred on the narrow scope of the surveys. The focus on Internet applications to the exclusion of other electronic and non-electronic applications limits the usefulness of the surveys, despite the fact that Internet information is better than no information. Given the relationship between consultants and the responding clients, it is conceivable that obtaining additional information would be feasible. Extending these surveys to include EDI, internal computing applications and non-electronic information processing technologies would be a valuable addition to our knowledge of uptake levels and justifications.

Another limitation is the focus on the firm as the unit of analysis, concealing internal transactions within the firm, and overlooking the dynamics driving uptake of collaborative applications (e.g. across supply chains - an issue when ownership and organisational forms are considered).

As with market research surveys, sampling by voluntary response does induce sample bias, which may reduce the level of trust in some figures, particularly those predicated upon perceptions rather than counts and values. The responses obtained, while relevant and reliable if compared like with like over time, may pose difficulties when extrapolated out to wider populations, as sample bias may render proportions unreliable. Comparing surveys over time may be misleading, as even if the same firms are required to answer the same questions the individuals responding may have changed, rendering comparisons of subjective responses invalid. Furthermore, comparing between surveys of different consultants in the same country is problematic, as the methodologies differ, while comparisons between identical surveys by the same consulting firm in different countries is also compromised by different legal and policy regimes influencing different levels of uptake of the same application.

Nonetheless, the frameworks created by these surveys come closer to surfacing the rationale underpinning strategic responses to the challenges of the "modern economy" than any of the other uptake methodologies examined. It is a framework that can be fruitfully developed to reveal further insights and understanding.

University of Texas Measuring the Internet Economy

The third form of uptake measurements found in literature reviews for this research is provided by the University of Texas Measuring the Internet Economy Project. Again, the focus is limited to measuring uptake of Internet-based applications (it is noted that the project is sponsored by a number of companies with stakes in the Internet and computer industry, such as Cisco, Dell, Intel, Sun Microsystems, Hewlett Packard, KPMG, Ernst and Young, and IBM). The distinguishing feature of this set of uptake indicators is that it measures uptake in quantitative terms, and specifically quantitative measures which replicate those of traditional national accounts frameworks. In this respect, this set of uptake measures provides the set of measures best suited for melding the effects of electronic commerce activities into traditional measurement systems.

Despite the fact that it relies upon survey data extrapolated up into indicative economy-wide estimates, this study differs from the consultants' surveys in that it selects candidates based upon their type of activities, rather than relying upon voluntary participation. Further, the uptake figures used are dollar and quantity-based, rather than being reliant upon perceptions and rankings, and are thus independently verifiable and comparable between surveys. In addition, the methodologies used to collate the uptake statistics reflect many of the aspects of the "modern economy" identified in section 1, such as redesigned aggregation categories, which reflect information usage rather than industry outputs.

The uptake figures collated in this survey, broken down by the four layers (Internet Infrastructure, Applications, Intermediaries and Online Transactions) identified in section 2, include:

employment by layer by quarter, compared to total employment (removing overlaps)
revenue by layer by quarter, compared to non-Internet sector (removing overlaps)
revenue per employee per quarter, per sector
dot.com employment and revenues compared to the rest of the Internet sector

The key advantages of this methodology relate principally to the ability to undertake comparative analysis of the performance of the Internet sector and the rest of the economy, as the measures are translatable. It also enables the application of traditional performance measures (such as productivity - see Performance section) over firms within in each of the four layers over time, enabling easier identification of the layers where productivity gains may be higher or lower. The aggregation of measures across layers rather than industries overcomes the firm-specific limitations of the consultants' surveys, and it enables measurement of the value of information movement to be captured separately (in the infrastructure level) rather than it being subsumed in the final value-added of the products as in traditional productivity statistics.

While these uptake measures offer a significant advance in respect of the ability to link uptake into performance measures, and hence inform Internet-related policy and decision-making, there are at present significant impediments to widespread use of this methodology. The authors acknowledge that it is time-consuming to collect the data required, as it must be separated out from traditionally-collected measures. This requires both expert input from analysts, and co-operation and heavy time and information input from staff of the firms participating in the survey.

The measures created also have limitations. The Internet focus ignores other electronic transactions that would respond to the same style of analysis, and the firm focus omits from consideration all internet-related activities conducted by non-trading individuals. Limiting output measures to revenues perpetuates the emphasis on measuring sales rather profits, value-added or productivity gain from cost reductions, indicating that more work is required before these measures can provide seamless inclusion into standard performance measures. However, the subsequent work of Barua, Whinston and Yin on an "import/export" approach to integrating Internet and non-internet trading figures is promising.

Performance

At the highest possible level of aggregation, performance measures should show whether an economy (or a society) is able to produce more outputs (of all types) for a given level of inputs (of all types) as a result of changing the technologies (human, mechanical, organisational, policy etc.) it applies - that is, productivity gain. While this methodology can be applied at all levels of activity, right down to the level of the productivity of individual transactions, it is recognised that all transactions occur in the context of systems, systems interact with each other, and there will inevitably be flow-on and feedback effects as changes in one system affect the inputs and technology choices of other systems. Furthermore, these effects need to be examined not just within static slices of time, but over time, as the impacts affect different systems in different time cycles (dynamic effects, leads, lags etc.). Micro-analysis of the performance of individual transactions and systems may offer insight, but the net outcomes require a much larger world-view if all of the interactions are to be captured.

The principal decision in analysing performance metrics of systems becomes where to draw the boundaries of the system under examination, and having drawn them, recognising the points of entry of influences that will result in change to the performance of the system (exogenous effects), and the exit points where the system in question feeds on into other systems. Having identified these, then the key indicators of the performance of the system can be identified, and tracked.

Traditionally, the performance of national (system boundaries) economies has been measured and compared using productivity - dollars of value-added created for a given level of inputs. Notwithstanding the difficulties posed by measurement problems (section 2), productivity provides probably the most comprehensive "catch-all" to incorporate the economic consequences of changes in technologies and their interaction over a wide range of systems. Similar methodologies have been used to measure performance in subsystems such as industries and firms, even to the extent of measuring transnational activities (e.g. international firms).

Conceptually, productivity arguably remains the best measure to capture the systemic effects and measure the impact of, changes wrought by any technology. The challenges raised by the changes wrought by changes to the use of information include questions over whether our existing measures of productivity adequately capture what is happening within the traditional economy, in addition to the need to extend these measures to capture the effects of new electronic technologies. Unless these issues are adequately addressed, there is little hope that productivity measures will be able to answer the question of whether economies (or societies, or any other system around which arbitrary boundaries can be drawn) have experienced productivity gains as a result of any change in inputs, outputs or production technologies.

The emergence of information as a separately identifiable commodity has thus caused us to question whether the units by which we have traditionally measured productivity - labour, capital and dollars of value-added to end production traded for cash - are sufficient, given that many of the inputs and outputs created are intangible, and by no means are all traded for cash, even though they may be valued by the recipients. Furthermore, the emergence of the so-called productivity paradox - the enigma that despite the emergence of noticeable changes in individual and firm wellbeing as a result of changes in the technology mix, these do not appear consistent with changes in measured productivity figures - has exacerbated the need to address this issue.

Much of the preceding discussion in this document has focused on the role of information separate from the technologies that process, transmit and utilise it. Changes in the ways in which we use information have always had intangible effects not captured in productivity figures, but so too have changes in technologies. The significant difference brought about by the use of information processing technologies is that they significantly change the ways in which an item that has never been adequately captured in performance measurement is treated, with consequent systemic effects throughout a whole range of systems where its effects have never been overtly analysed. From a fundamental level, performance measurement of changes wrought by changes to information processing technologies can never be rationalised or adequately analysed until this omission of analysis is "corrected".

While we may never be able to derive a measure for a unit of "information" in the same way as we can measure an hour of labour or a dollar of capital, we are now able to start measuring the costs of creating, storing, transmitting, moving and processing bytes of data. We are creating information-based products for which people are willing to pay, meaning that we can start building measures of value-added attributable to the creation of information. When these are used as inputs to production, we can measure returns to the use of information. These methodologies underpin the work being undertaken by projects such as the Brookings Institute's Understanding Intangible Sources of Value⁵².

If, however, we continue to judge the performance of new technologies impacting on intangible and unmeasured inputs and outputs using metrics that overtly exclude them from consideration, then our decision-making is at best flawed and at worst foolhardy. Thus, it is important to recognise the limitations of existing productivity measures.

For example, it has long been recognised that traditional productivity measures do not adequately represent the performance of the services sector (Triplett (2000)), due to the significant component of intangibles involved in the production an valuation of services (e.g. quality issues). Yet as a consequence of changes to technologies and tastes, services are among the largest, and fastest growing sectors of most economies (69.5% of OECD value-added in 1998). The growth of electronic technology use has resulted in an increasing number of commodities becoming differentiated using service complements (e.g. warranties), so even traditional non-service sector outputs are becoming subject to the imprecision which has characterised the service sector, rendering productivity performance even in these sectors subject to inaccuracies. The growth of time-based risk management products has also compounded the problem, as traditional performance measurement methodologies do not cope well with payments made in the present to offset the risk of uncertain costs in the future. The effect of new information-based risk products is thus particularly evident in the finance sector, where time factors and intangibility combine to blur the returns to spending on risk management products such as insurance premiums.

The stock figures from which national returns are calculated are also subject to imprecision. Accounting precedents allow for capital stock to include cash spent on physical items such as computers, but only recently have the conventions been relaxed to allow spending on the computer software required to run computers to be included as a capital item rather than counted as an intermediate item, despite the fact that software can generate value over multiple accounting periods and is effectively a complementary investment to the computer (i.e. the computer is unable to generate wealth without software). This has had obvious distorting effects on productivity figures, as the capital stock figures have not been a true reflection of actual investment. While software may now be more appropriately accounted for as a capital item, spending on organisational restructuring and training necessary to incorporate computers into businesses is equally a complementary asset to the computer (Brynjolfsson and Hitt (2000)), but such spending must continue to be expensed. The emphasis on recording only physical technologies conceals the reality that investment in organisational and human technologies is as much a capital investment as investment in machines. The consequence of focusing solely on equipment means that productivity figures can tell us with much more certainty the returns in the sectors that manufacture chips and computers, than in the sectors that buy and use them (Jorgenson and Stiroh (2000)). And this does not even begin to address the issue of spending on information stocks that can be retained and used (and depreciated?) to lever value from production many years in the future.

Productivity measures also fail to capture exchanges that do not result in cash transactions. Clearly, electronic technologies have profound impacts in the non-traded sector, given the extent to which these technologies are applied domestically (e.g. substitution of leisure time uses). This is further compounded by the fact that evolving work processes (e.g. telecommuting, employment portfolios for individuals replacing one job, "individual mobile transactors") are blurring the distinction between work and non-work activities, making attribution of the costs and benefits of transactions to specific sectors or activities difficult.

This raises the issue of whether productivity remains a valid measure of economic and social system performance. Can proxy measures for intangibles be created that enable their effects to be incorporated into traditional measures? And can these measures adequately deal with a stock such as information that exhibits different characteristics and returns to the rivalous and excludable items (capital and labour) on which the productivity measure has been developed? Are there are any other indicators that separately or together can provide a more comprehensive impression of the net impact of technology and process changes?

Summary

The discussions on Connectivity, Capability and Uptake measures above indicate that, while all provide measures of relative performance within narrowly defined areas, none in isolation can address the systemic effects of interaction, or address the need to integrate the effects of new and changed products, processes and technologies into a wider context, given that the products, processes and technologies they subsequently impact upon may be far from the point at which the change is implemented, and the "currency" in which many of the benefits and detriments are measured may be very different from the currency in which the change is recorded. The "universality" of productivity measurement grants it the potential to aggregate up effects across sub-systems, but its effectiveness is dependent upon the specification of the sub-systems, their points of interaction and the ability to capture all of the flows that occur.

Unless any other measure can replicate all of these effects, then productivity is probably the best we can muster. Further research effort is required to improve our understanding of the systems and sub-systems we are aware of, and better capturing and measuring their inputs and outputs, rather than trying to create a new measure. This means developing a better understanding of the role that information plays as a "currency" that links all systems and sub-systems in ways that we may have previously conceptualised, but not endeavoured to measure. This will both improve our ability to measure performance, and understand the relevance of the connectivity, capability and uptake figures that we currently maintain. It will also reveal new things to track. It is this process of increased understanding which will inform policy and decision-making, as much as the numbers captured and reported. In the meantime, the measures identified above all have some value - the caution is to recognise what they do reveal, along with their limitations, thus avoiding their indiscriminate use in justifying decisions and policy changes that they are incapable of supporting.

⁴⁶Full references for reports of these sources are contained in the References section.

⁴⁷Bright Planet. 2000. The Deep Web: Surfacing Hidden Value. White Paper. BrightPlanet.com.

⁴⁸Although Paul Rappaport (ICFC 2001) has developed regression models for bandwidth demand forecasting by separating out and assigning an information intensity rating to a number of job classifications.

⁴⁹Unless, of course, it includes identifying popular sites which might benefit from selling advertising!

⁵⁰Anecdotal evidence provided by an employee of a market research company collecting Internet data.

⁵¹Research round table, April 2001.

⁵²http://www.brookings.org/

Jump to main page content.

Printed: 26/11/2009