تکامل ساختار بازار در نیمه رساناها: نقش استانداردهای محصول

This paper looks at the role of product standards in determining the evolution of market structure in semiconductors. Theory suggests that product standards, along with learning-by-doing effects, lead to market concentration and persistence of leadership in product innovation. Empirical evidence supports this with some qualifications. The market concentration hypothesis is confirmed for all submarkets of the semiconductor industry. Persistence of leadership is obtained for markets based on proprietary product standards. In the market for microprocessors, we have persistence of leadership in product innovation that is accompanied with overall market dominance. For microcontrollers, we observe persistence of leadership in product innovation, but not concomitant overall market dominance. Whereas learning-by-doing effects seem to be strong in the small market for microprocessors, they are weaker in the much larger market for microcontrollers.

Explaining the evolution of market structure is a major research programme in the industrial economics literature (Sutton, 1995). The objective is to construct a general class of models that can explain the evolution of market structure across a broad range of industries. Easily identifiable and observable variables should help to partition the space of possible outcomes for market structures and explain their evolution. The empirical investigations adopting this approach have shown that several propositions are testable for both a broad range of industries (Sutton, 1991) as well as for single industries. Examples for the latter are the role of R&D in the pharmaceutical industry (Matraves, 1996) and learning-by-doing in the semiconductor industry (Gruber, 1994). A related issue is the explanation of persistence of market leadership over a long period of time, especially if the industry is concerned with repeated innovation. The extreme patterns found in the theoretical literature concern leapfrogging or persistence of leadership. What type of pattern emerges ultimately depends on features such as the type of price competition or the beliefs of the agents (Vickers, 1986; Beath et al., 1987; Budd et al., 1993). The problem is that most of these features are not directly observable and this makes it difficult to empirically test theoretical propositions. Nevertheless, detailed analysis of industry-specific characteristics, such as learning effects, allows making of some progress. Gruber (1994)has shown the conditions under which the learning curve generates persistence of leadership in product innovation in semiconductor memory chip markets. The interplay of learning-by-doing and market size is crucial for generating this pattern: persistence of leadership is obtained if market size does not exceed a certain threshold level. This paper explores a further feature with a strong bearing on market structure in the semiconductor industry: proprietary product standards. Starting off with the seminal papers by Katz and Shapiro, 1985 and Katz and Shapiro, 1986and Farrell and Saloner, 1985 and Farrell and Saloner, 1986, the theoretical literature on product standards has become vast. We have learned that the number and types of standards produced by market forces depend very much on the basic assumptions on firm behaviour and consumer beliefs: a firm that is able to establish early an industry standard for a product should have a large market share in this. This advantage can be consolidated through repeated product innovation with backward compatibility. The standard can be protected in several ways, such as technological competence and by invoking intellectual property. This paper elaborates on the evolution of market structure in other types of semiconductors such as microprocessors and microcontrollers. These chips are somewhat different from memory chips as they perform logic operations instead of simply storing information. They are what is referred to as a `computer on a chip'. The manufacturing process of logic chips is similar to memory chips. The main difference is in the design. Memory chips are very much a commodity with little scope for differentiating products of the same generation. The above logic chips are based on proprietary standards which deliver scope for product differentiation and `lock-in' of customers. Standards are an additional element bound to strengthen the persistence of leadership patters. The learning curve is pervasive in the production of semiconductors and, given certain conditions of appropriability of learning-by-doing and market size, naturally produces persistence of leadership over a sequence of product innovations (Gruber, 1992). The firm, Intel, has maintained its leadership for all of the products analysed here, except for dynamic random access memories (DRAMs). The main hypothesis of this paper is as follows: given that with logic chips, we observe the persistence of leadership, this may not only be due to the learning curve, but also to product standards. In other words, the learning-by-doing effect is reinforced by the positive feedback on sales due to the leading standards. This hypothesis is supported by the observation that the market for microcontrollers is larger than that for DRAMs. But differently from DRAMs, the market for microcontroller displays persistence of leadership in product innovation. This paper is arranged as follows. Section 2presents the analytical framework of the model of the evolution of market structure and shows the relevance of learning-by-doing and standards. Section 3is a brief description of the main market characteristics of the semiconductor industry. Section 4further explores the memory and logic chip markets and then discusses the main propositions concerning the evolution of market structure. Section 5draws the conclusions.

This study has matched the evolution of market structure in some product markets of the semiconductor industry with recent theoretical models. Even though most types of semiconductors are based on the same technology and production method, large differences in the evolution of market structure can be observed in submarkets of the industry. Technological characteristics of the industry such as learning-by-doing and proprietary product standards unfold in different ways. Both characteristics should, in principle, lead to concentration, and this is actually observed. However, there are different ways of achieving this. With microprocessors, we observe persistence of leadership which translates itself in high overall market shares for the leader. For microcontrollers, we observe persistence of leadership too, but not a concomitant overall market dominance. Whereas learning effects seem to be important for microprocessors, they are smaller for microcontrollers. The results of the study complement the findings of an earlier study on products not based on standards, such as memory chips (Gruber, 1996). There, the strong learning effects explain the persistence of leadership pattern in EPROMs. That pattern breaks down in the DRAM market where learning effects are dominated by economy of scale effects. This study shows that with logic chips, there is a second mechanism at work, reinforcing the persistence of leadership in product innovation. This is related to lock-in effects with downstream products such as computers with respective operating system standards. Advantages due to established standard and reputation with a generation can be carried over to a new generation. For microprocessors, the learning effects and standards mutually reinforce each other, leading to persistence of leadership with very high levels of concentration. For microcontrollers, however, the learning effect is weaker and the market is even bigger than for DRAMs. In the DRAM market, which is large and where learning effects are secondary, persistence of leadership is not observed. Persistence of leadership observed in the market for microcontrollers should thus be attributed to the role of product standards.