The future direction and rates of technological change are uncertain and therefore need to be explored when developing a range of alternative futures (i.e., scenarios). However, it would be misleading to resort to simplistic parametric variations of scenario assumptions without considering some basic elements of the nature of technological change, briefly reviewed here.
Technological change has often been pictured in linear terms that involve several sequential steps:
However, this model places undue emphasis on the role of basic R&D and scientific knowledge as precursors and determinants of innovation. It also understates the role of interactions among different actors and between the five functions listed above. The emphasis in recent innovation literature is placed more on a "chain-link" model of innovation, exploiting interactions between firms' R&D departments, and various stages of production and marketing (Dosi, 1988; Freeman, 1994). Lane and Maxfield (1995) emphasize the role of "generative" relationships in creativity.
Technological change is linked to the economic and cultural environment beyond the innovating firm in many ways, as described by Landes (1969), Mokyr (1990), Rosenberg (1982, 1994, 1997), Rostow (1990), and Gr�bler (1998a). Innovations are highly context-specific; they emerge from local capabilities and needs, evolve from existing designs, and conform to standards imposed by complementary technologies and infrastructure. Successful innovations may spread geographically and also fulfill much broader functions. The classic example is the steam engine, developed as a means of pumping water out of deep mines in Cornwall, England, but to become the main source of industrial motive power and the key technology in the rail revolution worldwide.
Numerous examples can be used to demonstrate the messiness, or complexity, of innovation processes (e.g., Gr�bler, 1998a; Rosenberg, 1994). But even if the innovation process is messy, at least some general features or "stylized facts" can be identified (Dosi, 1988; Gr�bler, 1998a):
These five features render some individuals, firms, or countries better at innovation than others. Innovators must be willing and able to take risks; have some level of underlying knowledge; have the means and resources to undertake a search process; may need relevant experience; and may need access to an existing body of technology. Many of these features introduce positive feedback into the innovation process, so that countries or firms that take the technological lead in a market or field can often retain that lead for a considerable time.
Technological change may be supply driven, demand driven, or both (Gr�bler, 1998a). Some of the most radical innovations are designed to respond to the most pressing perceived needs. Many technologies have been developed during wartime to address resource constraints or military objectives. Alternatively, some innovation (e.g., television) is generated largely through curiosity or the desire of the innovator to meet a technical and intellectual challenge. Market forces (including those anticipated in the future) can act as a strong stimulus for innovation by firms and entrepreneurs aiming either to reduce costs or to gain market share. For example, Michaelis (1997a) shows the strong relationship between fuel prices and the rate of energy efficiency improvement in the aviation industry; Michaelis (1997b) also discusses the effects of the introduction of competition on the organizational efficiency of the British nuclear industry.
All innovations require some social or behavioral change (OECD, 1998a). At a minimum, changes in production processes require some change in working practices. Product innovations, if they are noticeable by the user, demand a change in consumer behavior and sometimes in consumer preferences. Some product innovations - such as those that result in faster computers or more powerful cars - provide consumers with more of what they already want. Nevertheless, successful marketing may depend on consumer acceptance of the new technology. Other innovations - such as alternative fuel vehicles or compact fluorescent lights - depend on consumers accepting different performance characteristics or even redefining their preferences. An important perspective on technical change is that of the end-user or consumer of products and services. Technology can be seen as a means of satisfying human needs. Several conceptual models have been developed to describe needs and motivation, although their empirical foundations are weak (Douglas et al., 1998; Maslow, 1954; Allardt, 1993). In many cases, a given technology helps to satisfy several different types of need, particularly evident in two of the most significant areas of energy use: cars and houses. This tendency of successful technologies to serve multiple needs contributes to lock-in by making it harder for competing innovations to replace them fully. Hence, many attempts to introduce new energy efficient or alternative fuel technologies, especially in the case of the car, have failed because of a failure to meet all the needs satisfied by the incumbent technology. Different individuals may interpret the same fundamental needs in different ways, in terms of the technology attributes they desire (OECD, 1996). Deep-seated cultural values or "metarules" for behavior can be considered to be filtered through a variety of influences at the societal, community, household, and individual level (Douglas et al., 1998; Strang, 1997). Commercial marketing of products usually aims to adjust the filters, and encourages people to associate their deep-seated values with specific product attributes (Wilhite, 1997). These associations are likely to be more flexible than the values themselves, and provide a potential source of future changes in technology choice.
Technology diffusion is an integral part of technical change. Uptake of a technology that is locally "new" can be viewed as an innovation. Often, when technology is adopted it is also adapted in some way, or used in an original way. Just as technology development is much more complicated than the simple exploitation of scientific knowledge, Landes (1969), Wallace (1995), Rosenberg (1997), and others emphasize that technology diffusion is highly complex. Wallace emphasizes the importance of an active and creative absorption process in the country that takes up the new technology. The implication of this complexity is that no general rules define "what works." The process of technology adoption is as context dependent as that of the original innovation. Rosenberg (1997) also emphasizes the role of movements of skilled people in the diffusion of technology. Transnational firms often play a strong role in such movements. Other factors that influence the technology transfer process include differences in economic developmental, social and cultural processes, and national policies, such as protectionist measures.
Grossman and Helpman (1991), Dosi et al. (1990), and others have attempted to capture some of the complexities in "new growth" and "evolutionary" economic models. They have been able to demonstrate the flaws in some of the simpler solutions to technology diffusion often advocated - for example, they show how free trade might sometimes exacerbate existing gaps in institutions, skills, and technology.
The complex interactions that underpin technology diffusion may give rise to regularities at an aggregate level. The geographical and spatial distribution of successive technologies displays patterns similar to those found in the succession of biological species in ecosystems, and also in the succession of social institutions, cultures, myths, and languages. These processes have been analyzed, for example, in Campbell (1959), Marchetti (1980), Gr�bler and Nakicenovic (1991), and Gr�bler (1998a). An extensive review of the process of international technology diffusion is available in the IPCC Special Report on Methodological and Technological Issues in Technology Transfer (IPCC, 2000). That report provides a synthesis of the available knowledge and experience of the economic, social and institutional processes involved.
Many attempts to endogenize technical change in economic models rely on a linear approach in which technical change is linked to the level of investment in R&D (e.g., Grossman and Helpman, 1991, 1993). More importantly, this linear model has been the basis of many governments' strategies for technological innovation. As mentioned above, important additional features of technological change include uncertainty, the reliance on sources of knowledge other than R&D, "learning by doing" and other phenomena of "increasing returns" that often lead to technological "lock in" and hence great difficulties in introducing new alternatives.
These features can be captured to some degree in models and a great deal of experimentation has taken place with different model specifications. However, the first feature, uncertainty, means that models cannot be used to predict the process of technical change. This uncertainty stems partly from lack of knowledge - the outcomes of cutting-edge empirical research simply cannot be predicted. It also stems from the complexity of the influences on technological change, and in particular the social and cultural influences that are extremely difficult to describe in formal models. Recent attempts to endogenize technical change in energy and economic models are reviewed by Azar (1996). Optimization models usually treat technology development as exogenous, but technology deployment as endogenous and driven by relative technology life-cycle costs. A few GHG emission projection models (e.g., Messner, 1997) were developed to incorporate "learning by doing"- the reduction in technology costs and improvement in performance that can result from experience (Arrow, 1962). Models have also been developed that explicitly include technological uncertainty to analyze robust technology policy options (e.g., Gr�bler and Messner, 1996; Messner et al., 1996). Other models developed more recently incorporate the effects of investment in knowledge and R&D (Goulder and Mathai, 1998). Economists and others who study technological change have developed models that take a variety of dynamics into account (Silverberg, 1988). Some models focus on technologies themselves, for example examining the various sources of "increasing returns to scale" and "lock-in" (Arthur, 1989, 1994). Other models focus on firms and other decision-makers, and their processes of information assimilation, imitation, and learning (Nelson and Winter, 1982; Silverberg, 1988; Andersen, 1994). Few of these dynamics, apart from "increasing returns to scale," have been applied to the projection of GHG emissions from the energy sector.
Other reports in this collection