Some evolutionary scholars stress the heterogeneous nature of innovation and 1 enduring inter-industry differences between structures, origins and processes of innovation (e.g. Pavitt and Rothwell, 1976). Pavitt (1990), for example,argues that distinct modes of innovation can be observed across four groups of sectors: (i) science-based; (ii) scale intensive, (iii) information intensive;and (iv) specialized supplier dominated. Hayes and Wheelwright (1984) distinguish five major groups of production innovation; project, job-shop, batch line flow, assembly line and continuous flow. Nelson and Rosenberg (1993) point to differences between complex systems (CSs), fine chemicals and bulk commodities such as steel.
By contrast, another influential body of work stresses similarities in the innovation process (Utterback and Abernathy, 1975; Abernathy and Clark, 1985; Clark, 1985; Utterback and Suarez, 1993). As Part I shows, this model argues that product and process technologies tend to follow life cycle patterns from birth to maturity. Firms compete by engaging in a technology race while consumers in the marketplace decide which innovations will be successful. For the purpose of this paper, this technology race/market contest approach is referred to as the 'conventional model'. The latter has influenced evolutionary theories of technical change as well as policy views on how the West should respond to the East Asian challenge in areas such as automobiles, semiconductors and consumer electronics.
The paper argues that while the conventional model may apply to mass market commodity products it is highly unlikely to apply to another important group of products and industries, classified here as CSs. As Part I argues, CSs account for a significant proportion of industrial output. In contrast with commodity goods, complex product systems are large item, customized, engineering intensive goods which are seldom, if ever, mass produced.3 Examples include flight simulators, telecommunications exchanges, electrical power equipment, military systems, airplanes, helicopters, flexible manufacturing systems, chemical process plant, intelligent buildings and nuclear power equipment. In contrast with mass production industries,the West retains a general lead over Japan and other East Asian countries,although this cannot be taken for granted.
As Part I argues, CSs can be defined partly by the number of customized, interconnected components and the amount of feedback between them. Typically, they exhibit emerging properties through time as they respond to the economic environment and the innovation demands of large users. CSs involve a high degree of precision and customization in design and production. They are invariably intermediate goods industries which supply large user firms rather than mass market consumers.In contrast with large volume industries, CS industrial structures tend to be characterized by persistent bilateral oligopoly. Products are oriented to the needs of large sophisticated business users which depend on CSs for their survival, growth and profitability. Consequently, users involve themselves intimately in the innovation process.
While attention has been paid to the nature of innovation in individual CS industries, they are rarely treated as a distinct analytical category or compared with mass production industries. Therefore, the purpose of this paper is to examine, in detail, the logic of innovation in one complex industry — flight simulators (FSs)—and to contrast this with the conventional model. Further research is being conducted to compare the differences and similarities among a variety of CS industries. To highlight its distinctive characteristics, the paper contrasts origins, processes and structures of FS innovation with those typically found in the conventional Schumpeterian model. The study shows how the key actors in the FS innovation process (manufacturers, users, regulators and professional groups) collaborate to resolve their competing innovating needs and negotiate with each other to arrive at specific innovation outcomes. The study, carried out by a research team over a two year period, involved more than 120 interviews with industry representatives in North America and Europe, continuous feedback through industrial working groups and two questionnaire surveys.
Part I contrasts the main elements of the conventional evolutionary model with those likely to be found in C5 industries, offering a working definition of CS products and some preliminary ideas on how a taxonomy might be developed (Hobday, 1994). Part II summarizes the results of a detailed study on the international FS industry (Miller et al., 1993). The structure, mechanisms and determinants of innovation are explored and related to the key transformation points in the industry's history. Contrary to the conventional model, despite radical technological discontinuities a persistent pattern of stability occurred among major FS suppliers, while significant upheavals and adjustments occurred in the supply chain. The paper explains this pattern and shows how an institutional superstructure was created by the main actors to coordinate and prosecute innovation, producing a selforganizing industrial system designed to cope with uncertainty and risk. Accepting the limits of a single case study, the conclusion summarizes the main findings and suggests that other CS industries are likely to devise analogous institutional mechanisms for facilitating innovation.
2. Part I: Two Contrasting Models of Innovation
The Conventional Market Contest Model of Innovation
The conventional model of industrial innovation is intimately linked to the production paradigm of mass market commodity goods. Firms and markets tend to clearly defined, recognizable entities. Large and small firms create markets and redefine industries by skilfully exploiting technical opportunities (Schumpeter, 1947). The creation and diffusion of new technologies are usually sequential activities: first, the R&D laboratory develops; then the market selects (Utterback and Abernathy, 1975)...