• Nenhum resultado encontrado

Applicability of and exceptions to the theory

2.5 Limitations

2.5.1 Applicability of and exceptions to the theory

35

Biopharmaceuticals (Rothaermel 2000), on the other hand, are labelled high-tech, which serves as the explanation for innovations tending to take place in partnerships. The absence of economies of scale is mentioned as a special feature of the encryption software industry (Giarratana 2004).

As specificity for a non-manufacturing industry Fein (1998) concludes that no dominant design emerged in pharmaceutical wholesaling. Instead, he conceptualised a ―dominant business model‖

that standardised channel functions within such a service industry. However, the business model was not selected from among alternative business models but evolved gradually through the one-by- one adoption of new procedures, such as automation in warehouses and generation of data about anticipated customer ordering patterns. Furthermore, Fein (1998) found that increasing returns to size operate similarly to manufacturing industries, but firms tend to exit through mergers and acquisitions instead of closing down the firm or the department completely. This is due to the geographic nature of the wholesaling industry. Growing national wholesalers were willing to buy regional players to extend their coverage to new areas.

In summary, industry life-cycle theory is based on research on mainly manufacturing industries.

The empirical backing is vast, as a very large number of industries have been studied. Studies on non-manufacturing industries are scarce and some treat non-manufacturing industries as similar to manufacturing industries whereas others argue that there are fundamental differences in their life- cycle dynamics. Many high-tech industries have been studied, but they are generally treated just as any other manufacturing industries. On the other hand, the acceleration of the aging patterns predicted by the industry life-cycle theory is proposed by some due to the transition to high tech (e.g. Day et al. 2003).

36

the industry life-cycle model is not applicable to industries where development is very rapid. For example, chemical products and other continuous flow processes tend to have advanced, elaborate and large-scale processes virtually from the initial product introduction. These do not leave much room for radical and incremental product innovations. Furthermore, the applicability of the model is limited in industries where the available process technology defines the mode of operation and may have made the product feasible in the first place. Examples of such products include simple metal and plastic products. In summary, Abernathy (1978) rules out industries where the products or processes are too simple, but no industry is deemed too complex.

Teece (1986), on the other hand, limits the industry life-cycle model to apply to mass markets where consumer tastes are relatively homogeneous. Thus the model is not applicable in small niche markets where the absence of scale economies and cumulative learning does not penalise firms for having multiple designs. Similarly, Windrum (2005) states that the model is applicable only to markets where there is a homogeneous set of consumers. Likewise for Nelson (1995) the industry life-cycle model is applicable only to industries where customers have similar demands. As examples of industries with a high degree of variety in demand he mentions chemical products, pharmaceuticals and computers. In addition, he argues that the model best applies to industries where the product is a ‗system‘ which rules out very simple products. Malerba and Orsenigo (1996) list several limitations. First, the industry life-cycle model best fits consumer durables, such as automobiles and televisions. Second, the model applies to industries defined by a single product.

Broader definition of industries or systemic products limit the applicability. Third, the model does not apply to capital-intensive industries like plastics or petrochemicals, where innovations tend to be only of the process kind. The model also does not apply to customized investment products like machine tools, where innovation is mainly of the product kind. Fourth, the model does not apply to industries where major discontinuities come regularly at a fast pace, as in the semiconductor industry. Finally, Malerba and Orsenigo (1996) point out that the pre-existing conditions vary greatly among industries. For example, the automatic tellers first introduced by major banks have very different background compared to the cars first introduced by entrepreneurs.

Cases that deviate from the general industry life-cycle theory have attracted some research interest and several explanations for the inexistence of dominant design or shakeout in an industry history have been developed. For example, Anderson and Tushman (1990) suggest that when technological discontinuities take place every few years the selection mechanism does not have enough time to work out a dominant design. Phillips et al. (1994) found that there was no shakeout in the manufacturing of business jets. This was explained by the variability of preferences of the buyers.

Furthermore, shakeout may not occur if new uses for the product are discovered continuously and improved products open up new markets which offer opportunities for entrants (Klepper and Thompson 2006). Jin et al. (2004) also argue that effective spillovers can keep shakeouts from happening. In general, industry growth helps new firms to survive (Audretsch 1991).

More systematic approaches to the deviations from the industry life-cycle model have been proposed by Klepper (1997) and Bonaccorsi and Giuri (2000). Klepper (1997) defines three kinds of cases in which the shakeout does not take place. In the first one, specialist firms emerge to serve

37

the manufacturers of the final products. These specialists develop new process technology, new production equipment and supply key inputs for the manufacturing process. An example of this is the evolution of the petrochemicals industry. In the second type of deviation, the division of labour takes a different kind of turn. The product innovators concentrate on developing the product and license manufacturing to others. Examples of this include medical diagnostic imaging products. The third group of deviant cases is characterised by specialisation in customer segments. When demand is heterogeneous small competitors survive as no leader covering all segments emerges.

Bonaccorsi and Giuri (2000) build on this classification by merging the first two cases and elaborating on the third. They classify the violations of the industry life-cycle model into violations of appropriability and violations of increasing returns. In the former class either product or process technology becomes non-appropriable and the incentive to invest in R&D weakens. This happens as product and process R&D and manufacturing are performed by different firms and no one is able to achieve monopolistic appropriation. In the latter class no increasing returns are found in the firm‘s activities, such as manufacturing, marketing or R&D. This threatens the basis of cumulative advantage of incumbents. For example, the lack of increasing returns in advertising may appear due to persistent heterogeneity in customers‘ preferences. In their view, mere demand segmentation is not sufficient to stop a shakeout from happening but the explanations should take comprehensive account of the cost structures in R&D, manufacturing and marketing. Bonaccorsi and Giuri (2000) classify their findings on the life-cycle of the turboprop engine industry into the second class. This industry shows a high level of concentration and a continuous presence of a dominant leader but there is no shakeout visible in the industry history. This is explained by the continuous existence of generalist and specialist firms which is made profitable, on the one hand, by the lack of economies of scope in marketing as the market is fragmented and the submarkets are independent and, on the other hand, the lack of economies of scale in R&D and manufacturing due to the nature of the turboprop technology. Thus the generalist strategy does not have an advantage over the specialist strategy.

Windrum and Birchenhall (1998) argue that the emergence of the dominant design may well be a special case and just one of a number of possible market outcomes. In many industries the patterns of innovation have been far richer and several distinct market niches have appeared. They claim that instead of focusing on the artefact dimension of the technology one needs to investigate the learning and knowledge dimensions which ultimately determine whether a dominant design or some other market outcome emerges. Srinivasan et al. (2006) also argue that a dominant design may never emerge in some industries. This is due to certain characteristics of the products. First, weak appropriability is associated with a greater probability and earlier emergence of a dominant design.

Thus in industries where there is strong appropriability a dominant design is less likely to emerge.

Second, the more firms there are in a value net the sooner the dominant design emerges. Third, in radical product categories dominant designs are less likely and slower to appear. The first two characteristics run counter to the two deviant classes summarised by Bonaccorsi and Giuri (2000).

In the study on the aircraft, helicopter, motorcycle and microcomputer industries by Frenken et al. (1999) the emergence of a dominant design is found to be conditioned by the range of

38

services a technology offers. Helicopters and microcomputers offer a narrow range of services and product variety tends to decrease. This causes the industry to concentrate around one dominant design. Aircraft and motorcycles, on the other hand, offer a wide range of services which allows the creation of many niches. This allows many dominant designs to emerge as each is dominant in respective niche. The emergence of specialist niches has also been identified in beer brewing. Even though it is a model example of an industry with scale economies in manufacturing and advertising, there are over 1400 microbreweries, brewpubs, and larger specialty brewers in the USA (Tremblay et al. 2005). This is due to consumer demand for variety, which makes the production of local traditional products as well as craft-style products worthwhile. For this reason the beer brewing sector has remained a dynamic, competitive market.

Other deviations from the industry life-cycle studies have been reported by Filson (2001), Windrum (2005), Krafft (2004) and Murmann and Homburg (2001). According to Filson‘s (2001) study, the pattern of quality innovation early and cost innovation later on in the life-cycle does not hold for high-tech industries. On the contrary, the microelectronics industry showed cost innovation first and quality innovation later. Filson (2001) argues that opportunities for innovation evolve systematically in different ways in modern high-tech industries. Windrum‘s (2005) study on the amateur camera industry also calls into question the order of product and process innovations. This industry is a particularly messy example as it has experienced several rounds of radical product and process innovations. Krafft‘s (2004) study on the other hand reports a local non-shakeout pattern in the info-communications industry which was caused by specific local knowledge dynamics. The interaction and complementarity that developed among companies, research institutes and dedicated policies contributed to attracting new entries and to limit exits over time. Similarly, Murmann and Homburg (2001) found that in the synthetic dye industry France was the only country to experience a shakeout and this was because of their patent laws and the decision the grant an important patent to one particular producer. Other countries did not experience shakeouts in the research period of 1857-1914. Thus, the patterns detected at the global level may differ from those at the local level.

Many industry life-cycle studies conclude with a call for studies on counterexamples. Anderson and Tushman (1990) suggest the conditions under which dominant designs do not emerge as a topic for further study. Suárez and Utterback (1995) state that contemporary industries and counterexamples should be studied to isolate factors or characteristics of an industry with the help of which the model could be developed to a higher degree of validity. Filson (2001) states that richer environments with heterogeneous choices, network effects and brand effects deserve further study.

Also Filson‘s (2001) proposition of the opportunities for innovation evolving systematically differently for modern high-tech industries compared to previous ones should be elaborated. For Klepper (2005) the ultimate question is why certain industries are susceptible to being dominated for long periods by the same firms, for Murmann and Frenken (2006) whether the emergence of a dominant design is a cause or a consequence of industry evolution.

39