|
Article Excerpt
1. Introduction
Critical to understanding contemporary differences in market share and profitability among firms within an industry is systematic knowledge of how those differences arose in the first place. Understanding the structural evolution of industries--the rate of change in output and prices, the rates of entry and exit (turnover), and the growth and decline of individual firms (mobility) and industry participation--is widely recognized as fundamental to identifying the origins of profitable market leaders who can sustain performance over time. Industry evolution provides important contingencies that affect the viability of various firm strategies. Without a keen grasp of the underlying mechanisms driving industry evolution and the resulting changes that occur at the industry level over time, we are less able to identify why certain firms in an industry are the winners and others losers (Agarwal and Gort 2002). In this paper, we advance an alternative explanation for observed patterns of industry evolution based on interdependencies in productive activities (Milgrom and Roberts 1990, Levinthal 1997, Rivkin 2000) and identify novel predictions based on this underlying mechanism.
Over the years, researchers have endeavored to develop clear characterizations of industry evolution (see Gort and Klepper 1982, Agarwal and Gort 1996). Central to these investigations has been the observance of shakeout, i.e., a rise and then fall in the number of competitors over time. Following the inception of an industry, new entrants rush in, often driving up the rate of innovation and leading to a diverse set of ways to deliver value. Competition intensifies and industry exit increases. Over time the rate of entry decreases, eventually stabilizing at a low level. As a result, the number of firms within the industry grows exponentially at first, then peaks, and then declines, typically settling in with a few dominant firms. Although shakeout has dominated most discussion about industry dynamics, it is far from a universal pattern. Gort and Klepper (1982) report wide variance in the prominence of shakeout across industries. In some industries, barely any shakeout at all occurs; in others, upwards of 80% of the firms in the industry exit.
A number of models have been advanced to explain the classic rise and fall associated with shakeout and the frequency and nature of observed deviations from this pattern. The most prominent formal models of industry evolution rely on convex adjustment costs and scale advantages in innovation and learning to generate the classic life-cycle with shakeout pattern (Ericson and Pakes 1995, Klepper 1996). For example, a firm with a small initial efficiency advantage or a chance innovative success may expand faster than its rivals. If adjustment costs are convex and there are scale advantages in exploiting further innovation, this firm's previous innovation success and expansion will justify larger subsequent investments in innovation and amplify its lead over time.
In contrast, we propose that interdependence in productive activities provides a robust, alternative explanation for the stylized facts about industry evolution and shakeout. By "interdependency in productive activities," we refer to the potential for the value of one activity to depend on whether a firm engages in another activity (Levinthal 1997). In the presence of interdependent and potentially complementary activities, firms face barriers to search and often struggle to discover valuable configurations of productive activities (Rivkin 2000). We advance that industries can vary in their potential for interdependencies and that these interdependencies are fundamental to productive activity. Interdependencies provide a powerful mechanism by which to introduce bounded rationality and search in the spirit of evolutionary economics (Nelson and Winter 1982b) and behavioral theories of the firm (Cyert and March 1963) to models of industry evolution.
In this paper, we construct and analyze a model of industry evolution in the presence of interdependency among firm activities. We first show that the model provides a set of sufficient mechanisms to explain varying levels of shakeout and differing patterns of entry and exit among industries. Specifically, we show that when the potential for interdependencies within an industry is low, entry slows down and incumbent survival is all but assured; in industries where the potential for interdependencies is high, however, shakeouts are severe and the rates of entry and exit remain high over longer time periods, with decreasing survival rates for incumbents.
Differences in interdependency not only provide an explanation for why incumbents are more or less successful in different industries, but they also can explain why incumbents contribute more to technological progress in some industries while in other industries recent entrants make larger contributions. In this way, our model provides a potential resolution to the debate between the "old" and "new" Schumpeter about the growth contributions of incumbent firms versus new entrants. In industries where the potential for interdependency is low, incumbent firms are more likely to innovate and drive growth. In industries where the potential for interdependency is high, new entrants (entrepreneurs) are more likely to drive innovation and growth.
We argue that interdependency is at least as compelling as scale-based explanations (those relying on convex adjustment costs and scale advantages in appropriating returns to innovation) as a way of explaining industry dynamics. Interdependency complements these scale-based explanations by providing a separate mechanism to explain industry dynamics in general--and shakeout in particular--in markets where adjustment costs are low or linear or where innovation returns are independent of the previous scale. Interdependency, however, also conflicts with scale-based explanations, as they give rise to empirical predictions that in several cases are directly opposed to those of past models. Ultimately the two explanations differ in a fairly simple but fundamental way. In the presence of interdependency, past innovation conditions future innovation as it does in previous models, but not because they affect incentives to innovate, as in Klepper (1996) and Ericson and Pakes (1995). It is rather because they affect the content of what will be learned from innovative effort.
The key to this explanation is not differences in the amount of past innovative success, but differences in the content of past innovative successes, which lead to different technological trajectories (Dosi 1982). Our model focuses on the importance of what firms learn rather than solely on how much firms learn and how the nature of learning varies among industries.
2. The Evolution of Industry Structure
Shakeout has long been observed as industries evolve and has come to dominate managerial and scholarly discussion of industry dynamics (Klepper 2005). Shakeout takes a central place in Gort and Klepper's (1982) early and rich documentation and description of industry dynamics. Gort and Klepper studied the life cycles of 46 industries originating between 1887 and 1960 and representing a diverse mix of consumer, industrial, and military products. Their analysis for each industry began with the introduction of a substantial new innovation. They first looked at patterns of industry participation and observed that industries for new products pass through a brief period with few firms, followed by a rapid increase in the number of firms, which then falls rapidly to a relatively stable level (p. 639). During the evolution of the industry, Gort and Klepper also observed that output growth is initially high but declines steadily (p. 645); prices fall rapidly but at a decreasing rate (p. 647); and the rate of both major innovations and minor innovations rise, peak, and then remain stable over time, with major innovations peaking earlier (p. 648).
[FIGURE 1 OMITTED]
Gort and Klepper divided an industry's life-cycle into five stages defined by different rates of net entry (see Figure 1). The first stage is one with little net entry, where a few initial firms are alone in the industry, followed by a second stage defined by rapid entry, causing the number of firms to rise rapidly. Slowing entry and rising exit lead to a third stage, where industry participation remains fairly constant. Later, entry rates fall and exit rates increase, defining a fourth stage, where industry participation falls rapidly. Eventually exit rates slow and the industry enters a fifth stage defined by roughly stable industry participation. Gort and Klepper found that while industry participation rises, falls, then stabilizes, output increases quickly and prices fall rapidly with changes in both prices and output occurring at a decreasing rate over time.
While this full life-cycle pattern of industries has become widely recognized as a useful way of thinking about industries, Gort and Klepper's data show that the pattern is far from universal or uniform. As discussed in Figure 1, the length of each stage varies considerably from industry to industry. In some cases, the stages are so short as to be excludable (e.g., the first stage lasting less than a year) or so long that the question arises whether the following stages are part of the same industry dynamic (e.g., the third stage, with stable participation, lasting nearly four decades).
The magnitude of the rise or fall in industry participation also varies quite a bit across industries. In 19 of the 46 industries that Gort and Klepper (1982) studied and 7 of the 16 industries studied by Agarwal and Gort (1996), there is no dramatic fall in the number of competing firms during the fourth stage defined by net exit (see Klepper 1997, pp. 165, 166). In addition to the presence or absence of a clear shakeout in industry participation in this fourth stage, we observe differences in finer patterns, such as how much industry progress is driven by incumbents or recent entrants and whether early entry conveys considerable or little survival advantage (Hannan 1998).
A number of theories have been advanced to explain these patterns. Jovanovic (1982) provided a model to explain why small firms grow more rapidly and are more likely to fail than larger firms. The model is based on a "theory of noisy selection," where firms are initially and randomly endowed with different levels of efficiency but only learn and update their beliefs about their type over time with noise in the signals. In this model, efficient firms recognize their efficiency over time and grow, and inefficient firms become aware of their inefficiency, leading them to contract and exit. To produce analytic results, the model assumes an infinite number of firms, so firms are all price takers that limit their output due to convex production costs rather than strategic concerns about the effect of their own output prices and the output of other firms.
Utterback and Abernathy (1975) provided an alternative explanation for the shakeout of firms, where firms take a more active role in determining their efficiency levels. Utterback and Abernathy based their explanation on a period of technological uncertainty that ends with the emergence of a dominant product design, after which firms that are unable to produce this design exit the market. Klepper (1996) pointed out that the appearance of a dominant design is far less predictable than the other regularities observed in industry life-cycles and proposed and explored a more complete analytical model to explain these patterns, along with an additional regularity involving a shift in emphasis from product to process innovation over time. Klepper's model is based on (a) advantages of scale in exploiting product and process innovations that lead larger firms to innovate more and lead to a shift from...
|
