A Simulation model on Industrial Structure, Technological Regime and Organizational Selection: The Case of Korea and Taiwan in frontier industries1 Yoon, Minho2 1. Introduction This paper explores the complex relationship between organizational selection and technological regimes. We develop an evolutionary simulation model will be developed to show organizational selection and evolution of market structure under various technological regimes. Moreover, using this basic model, we lay out a tentative explanation about what determines comparative competitiveness of Korea and Taiwan in frontier industries like FPD (Flat Panel Display), semiconductor, and computer components. Numerous attempts have been made by scholars to explore the evolution of industrial structure (Klepper (1996), Nelson and Winter (1982), Malerba et. al., Malerba and Orsenigo (2001)). In those models, firms are differentiated in terms of entry time and scale economy is the main factor of industrial evolution. However, in many cases, firm organization plays a important role. Specifically, we frequently observe that initial dominance by small specialized firms (hereafter SS firms) shifts to large diversified firms (hereafter LD firms) gradually in R&D intensive industries. FPD and memory semiconductor industry are good examples (Kim and Lee (2003), Mathew (2005)). The memory semiconductor industry was launched by SS firms. However, as DRAM (Dynamic Random Access Memory) market rose to a major sector of this industry, the market share of LD firms increased gradually and SS firms had to find a niche market to survive. FPD followed a similar path. TFT-LCD (Thin Film Transistor Liquid Crystal Display) was the DRAM of the FPD industry. U.S. SS firms are forced to leave the largest TFT-LCD market to Japanese and Korean LD firms, and run into new small 1 Earlier version of this paper has been presented at the monthly seminar of Technology and Evolutionary Economics Research Group, in Seoul. The author would like thank seminar participants and professor Keun Lee for useful comments. 2 Seoul National University, racoon22@freechal.com 1 niche markets. On the contrary, there are a lot of industries dominated by SS firms like computer components and the non-memory semiconductor industry. The simulation model will be built to replicate the shift of market dominance and to show how technological regimes determine the process of industrial evolution. In consideration of this issue, the case of Korean and Taiwanese frontier industries is very interesting. While Korean and Taiwanese firms have been competing in some frontier industries at the same time, market performance has differed. Korean firms have achieved prosperity in the FPD and DRAM industry, but experienced technological stagnation in the non-memory semiconductor and computer components industries (Levy and Quo (1991), Kim (1997), Lee and Lim (2001)). On the contrary, the Taiwanese non-memory semiconductor and computer components industry have flourished whereas market performance of Taiwanese firms in the DRAM and FPD industry has been worse than that of Korean firms, with the exception of some limited fields like the small-size TFT-LCD panel market. Specialization in the R&D intensive industries of two countries cannot be fully explained by existing theories. Traditional Hechser-Ohlin theorem attributes specialization to different endowment across countries, but cannot explain the different performance of Korea and Taiwan. These two countries have similar endowment and there is no evidence that the industries mentioned above need different endowment. Product life cycle theory and endogenous growth theory explain comparative competitiveness as dynamic process in terms of the difference development stages. The case of Korea and Taiwan, however, cannot be fully explained by product life cycle and endogenous growth theory since there was not an important distinction between development stages of two countries. This paper attempts to build a simulation model showing that the difference of firm organization and industrial structure can explain the differing performance of the two countries. The next section discusses earlier studies which considerd explicitly the impact of firm organization on industrial evolution and provides conceptual explanation of our basic model. Section III shows the basic simulation model and Section IV provides the simulation results. In Section V, a tentative argument about the specialization of Korea and Taiwan is provided. Section VI revise the simulation model to replicate the case of 2 Korea and Taiwan and provides simulation results. Finally, Section VII contributes a summary and concluding remarks. 2. Organizational Selection and Technological Regime Some simulation models have been made to show the relationship between organizational selection and technological regime. Swann and Gill (1991) assumed trade-off between economies of scale and organizational inertia. They distinguished small specialized firms with small adjustment costs from large diversified firms with scale economies. In this model, technological trajectory is determined by consumers’ subjective decision and the more unpredictable the technological trajectory, the lower market share LD firms have. While Swann and Gill (1991) provided the first simulation model considering organizational selection, as Kim and Lee (2003) pointed out, their model would produce the same results if their technological trajectory on the X-Y axis is replaced by other unpredictable variables like exchange rate and input cost. Thus, it cannot said that Swann and Gill (1991) explored the relationship between technological regime and organizational selection. Meanwhile, Kim and Lee (2003) introduced heterogeneity of production cost structure. They assumed that LD(SS) firms have low(high) variable cost and high(low) fixed cost. In the history-friendly model of the DRAM industry, technological trajectory is onetracked and scale economies are assumed in relation to production cost as well as R&D efficiency. Thus, initial dominance by SS firms inevitably shifts to LD firms, although the speed of shift depends on the degree of innovation-productivity linkage and cumulativeness. Moreover, the reason for the initial dominance of SS firms is not explained in this model since LD firms are assumed to enter later than SS firms. While cost structure assumption of Kim and Lee (2003) is accepted, our model is differentiated in two ways from Kim and Lee (2003). First, our model brings demand growth in. The change of demand side plays an important role in industrial evolution. This assumption enables initial dominance of SS firms and shift to LD firms. More importantly, however, our model is able to replicate both cases of dominance by SS firms and LD firms introducing product innovation. Whereas Kim and Lee (2003) assumed one kind of innovation, both product and process innovation are considered in our model. Like Klepper (1993), a new product made by product innovation attracts new consumers in the model, and we will call new consumers a new submarket. Process innovation is distinguished from product innovation in that it affects only productivity 3 in manufacturing existing products. While Klepper (1993) assumed that a new product becomes the standard product after one period, a new submarket persists and other firms can run into the new submarket in our model. Meanwhile, it is assumed that experience in production of existing products is hardly useless to manufacture a new product. In the initial period of the model, the market is composed of several submarkets. Firms’ investment decision and structure of production, price and demand in a submarket are similar to Kim and Lee (2003) or Nelson and Winter (1982), except that demand of a submarket grows to a certain limit. As Kim and Lee (2003) assumed, LD firms’ fixed cost is higher than for SS firms and LD firms’ variable cost is lower than for SS firms. Productivity of a firm increases by the firm’s process innovation and imitation efforts. Thus, if there are fixed number of submarket over time, simulation result is similar to Kim and Lee (2003) model; the market share of LD firms grows gradually, then LD firms dominate the market eventually because of scale economies in relation to cost structure as well as R&D efficiency. Although, the speed of market shake-out depends on the degree of easiness of imitation and impact of innovation on productivity. However, product innovation, that is, emergence of a new submarket can be stumbling block for LD firms to dominate the market. SS firms have much more incentive to enter the new submarket because in initial stage the new submarket is too small for LD firms to produce efficiently with higher fixed cost. Of course, as demand size increases, LD firms drive out SS firms eventually because of the advantage of low variable cost and more R&D investment resulting in faster productivity growth. This process explains the shift of market dominance by SS firms to LD firms. Emergence of a new submarket is assumed to decrease existing submarkets. Thus, if emergence of a new submarket occurs very frequently, LD firms can not have enough time to use scale economy. SS firms enter a small emerging market more actively due to relatively small fixed costs. Of course, market growth favors LD firms. However, succeeding emergence of a new submarket decreases demand of existing submarkets. Whereas LD firms, having high fixed costs, enter new submarkets slowly, SS firms enter a new submarket quickly. In the dynamic process, therefore, total share of SS firms can be higher than that of LD firms under the technological regime of frequent product innovation. 4 3. Model As noted by Malerba et. al (1999), showing all the equations of the simulation model certainly obscure readers because of the complex features of the evolutionary simulation model. Therefore, only important equations and an outline of the model will be presented in this section. Interested readers might get the full program through contacting the author. Demand, output and market price In the initial period, there exist several submarkets and some firms in each submarket. Total output at t period in submarket j is defined as nj Q jt = ∑ Qit i =1 where Qit is output of i th firm in period t and n j the number of firms in j th submarket. Demand function of j th submarket in period t is assumed as follows; Pjt = R jt / Q jt Pjt is the price of j th product and R jt represents the size of j th submarket. The initial size of j th submarket, R j 0 , is the probability variable that has uniform distribution. R jt grows at a certain rate to a saturation point that is also a probability variable. That is, there are submarkets that are too small for several firms to survive as well as prosperous submarkets. And the emergence of a new submarket decreases or increases the size of all other submarkets. The extent to which the size of extant submarkets diminishes or expands is determined stochastically, when the case of diminishing are more frequent than the case of expansion. 5 Production and investment The output by i th firm in period t is productivity times capital stock such that Qit = hit ⋅ K it where hit is productivity of i th firm at period t and K it capital stock. When ACK it represents average cost per unit of capital stock, that is, ACK it = Cit K it profit per unit of capital of i th firm at period t is determined as π it = Pjt hit − ACK it − rit with rit representing ratio of the amount of R&D investment to capital stock. The markup of i th firm at period t can be calculated as mit = Pj ACK it / hit We assume that firms aim a markup rate specified by their market share. The higher market share has a firm, the higher markup rate it wants through its monopolistic power. Specifically, target markup of i th firm at period t is determined as following formula in this model; mit* = η /(η − sit ) s it is the market share and η is a parameter of the model. Firms are assumed to invest when current markup rate exceed target markup. Firms’ intended investment is determined by following formula; 6 I it* = (δ + mit − mit* ) K it mit δ is the depreciation rate. In order to assure that amount of investment is nonnegative and bring the assumption of financial constraint in this model, actual investment of a firm determined as I it = max[0, min[ I it* , X it K it ]] where X it means performance index defined by the following formula; X it = βX it −1 + (1 − β )π it where 0 < β < 1 β , the weight given to the lagged performance indicator, is fixed and equal to all firms. Finally, the capital stock of i th firm is determined by the following formula; K it +1 = (1 − δ ) K it + I it Process innovation Productivity of i th firm at period t is determined as follows; hit = hit −1 + ( ρ ⋅ hit −1 + θ ⋅ (h * − hit −1 ))(rit K it + ε ) Although productivity growth depends basically on the amount of R&D expenditure ( rit K it ) , two terms affect the efficacy of R&D expenditure. First, the extent of productivity enhancement is proportionate to the productivity level of the previous period. This means cumulative characteristics of process R&D and results in scale economy in dynamic aspect. ρ can be translated as the degree of cumulativeness. In second term, h * is highest productivity level among all firms in the submarket in which i th firm is operating. The larger the gap between the highest productivity level and that of a firm, the faster productivity level grows up; ρ can be interpreted as easiness of imitation. Finally, ε is a random component. 7 Product innovation In this version of model, product innovation is purely random process. In each period, one new SS firm and one business group succeed in inventing a new concept of product in a certain probability, α . We can think α of easiness of product innovation. This time-invariant parameter is an important characteristic of technological regime in this model since organizational selection depends on this parameter. Each product innovation is assumed to create a new submarket. We assume that productivity in existing submarket is never helpful to new sub-market, that is, all firms entering into a new submarket start at an industry base productivity (a parameter).3 Exit and entry If the output of a firm goes below zero, then it is driven out. Or, a firm is forced to exit the market when the amount of capital stock or performance index of the firm goes below a certain value that is time invariant and equal to all firms. In each period, an SS firm runs into a submarket with a certain probability. LD firms enter a new submarket by diversifying with a certain probability.4 However, a new firm or a new branch firm does not enter when expected profit in next period is below certain level. This assumption is very important because LD firms do not enter a new small submarket due to relatively high fixed cost. In contrast, SS firms run into a new submarket in early stage. Heterogeneous industry structure and product R&D Heterogeneity of firm organizations is presented in cost structure as Kim and Lee(2003) assumed. The total cost of i th firm in period t is defined as C it = VC it + FC i = VCK i ⋅ K it + FC i where VC it , FC i and VCK i represent variable cost, fixed cost and variable cost 3 Whereas this assumption may be too strong and unrealistic, it is appropriate to depict disadvantage form which Korean firms suffered in computer component industry pointed out by Levy and Quo(1993). Moreover, it does not affect the result of simulation runs that we assume that productivity in new submarket is affected by that in existing submarkets if the extent is sufficiently low. 4 In this version of model, diversifying assumption does not have important role. In the Korea and Taiwan model presented in Section V, however, diversifying assumption is very important. 8 per unit of capital stock respectively. LD firms are characterized by large fixed costs and small variable costs; SS firms are conversely characterized, such that VCK L < VCK S FC L > FC S where subscript L and S represent large diversified firm and small specialized firm respectively. SS firms enter into new small market more rapidly than LD firms because of small fixed costs. However, entrance of LD firms makes the market shake out if sufficient time and market size is given. Therefore, industry evolution depends on market dynamics. 4. Simulation Results Since the simulation model has many random components, results of one simulation run would not be consistent. Thus, we use averaged results from 50 simulation runs over a period of 200. Figure 1 provide the average of LD firms’ total market share at each period where α =0 (Figure 1A) and α =0.12 (Figure 1B). The horizontal axis represents the sum of LD firms’ market share and vertical axis time. The case of α =0 is similar to Kim and Lee (2003) except for market size growth. As expected, market dominance by SS firms shifts to LD firms in Figure 1A. In contrast, we find that LD firms’ market share does not exceed 50% in Figure 1B. This means that frequent product innovation disadvantages LD firms, which is also expected. To check robustness of these results, Table 1 shows the average of LD firms’ market share for 50 simulation runs at the final period, varying the degree of easiness of imitation ( θ ) as well as product innovation frequency. According to Table 1, the more frequent product innovation, the lower LD firms’ market share given any value of θ . The impact of the degree of easiness of imitation ( θ ) is not significant or inconsistent given low α . In the case of higher α , however, high θ tend to help LD firms. This is because high θ makes LD firms’ catch-up easier. As for the issue of market centralization, Figure 2A and Figure 2B show the number of firms at each period when α =0 and α =0.12 respectively. The number of firms decreases over time in Figure 2A while it increases in Figure 2B, albeit at diminishing speed. In Table 2, it is also shown that the number increases as α increases given any value of θ . That is, the more frequent product innovation, the less centralized the 9 market. Meanwhile, high θ makes the market decentralized only given high α . Finally, Figure 3A and Figure 3B present that the average revenue of all firms in the market at each period. In both cases, the growth of market size leads to the lager size of firms. In the case of high α , growth of firm size reach the ceiling more quickly compared to the case of no product innovation. In fact, we can find that high α leads to small size of firms’ average revenue given any value of θ in Table3. 5. The Case of Korea and Taiwan The case of Korean and Taiwanese frontier industries provides an interesting and clear example about the relationship between firm organization and performance within different technological regime. We will try to model this case revising the previous model. Levy (1991) suggests that Korean firms needed to be bigger than Taiwanese firms because of a relatively large degree of market failure from the transaction cost theoretical point of view. Meanwhile, Shin and Chang(2003) asserted that Korea adopted substitution strategy while Taiwan adopted complementary strategy at the initial stage of development. The Korean government excluded foreign direct investment and supported big business group, Chaebol, substituting market institutions. Taiwanese firms, on the other hand, entered into international specialization system to offset absence of big firms; Taiwanese firms have become almost small subcontract firms of state-owned or foreign big firms. Moreover, Levy and Quo (1991) observed that firm size determined firm strategy in frontier industries. Korean business groups pursue a high-volume assembly strategy; Taiwanese firms follow a bootstrap strategy, with small size at entry and a high propensity for risk-taking. As such, it is important to consider how the difference in firm strategy (a result of size) has affected the performance of firms in these two countries. Levy and Quo (1991) pointed out that better performance of Taiwanese firms in computer components and assembly industry was due to differentiated strategies. Because of the low degree of standardization of computer components, small size entry and active product innovation was more appropriate than a high-volume assembly strategy for responding rapidly and flexibly to emerging market opportunities. In fact, at the time of interview of Levy and Quo (1991), Korean firms were shocked by the 10 introduction of new range of PC by IBM. Experience of manufacturing previous versions of PC components was hardly useful to produce new PCs. On the other hand, Taiwanese firms with small size at entry and high expenditure on product innovation were successful. Kim (1997) also suggested that an economy composed of small specialization firms with a strong network had the advantage of variety and flexibility in the computer industry, compared to an economy composed of large independent firms. We observe similar pattern in the non-memory semiconductor industry. Whereas Korean DRAM giants tried to foster non-memory semiconductor business, they failed to reach significant market share. Meanwhile, Taiwan has one of the most flourishing nonmemory semiconductor industries. In contrast, Lee and Kim (2003) showed that small specialized firms have been driven out by large diversified firms in DRAM industry by an evolutionary simulation method. In fact, Korean LD firms are the most important player in the DRAM industry. According to Lee and Lim (2001), the DRAM industry has a relatively certain technological trajectory. In the DRAM industry, timely large size investment and subsequent learning-by-doing is important. This kind of industry favored Korean business groups. FPD industry is a similar case. Korean chaebols, like Samsung and LG are the most successful firms in FPD industry. Taiwanese firms also tried to enter this lucrative industry, but were not successful until 2000. After 2000, however, Taiwanese firms have risen to be important players in TFT-LCD industry, which is an interesting case in that SS firms overcame their disadvantage. This will be discussed in Section VII. The distinction of firm organization makes a different pattern of knowledge diffusion in these two countries. According to Lee and Yoon (2005) and Kim (1997), Korean big business groups are more independent in R&D, compared to Taiwanese small firms pursuing a cooperative strategy. In explaining this, the role of government research institutes (hereafter GRI) has been pointed out. Taiwanese GRIs have played more important role since small firms stand at a disadvantage in scale economy of R&D investment. (Lee and Yoon (2005), Shin and Chang (2003)) Lee and Yoon (2005) demonstrated empirically that inter-firm knowledge diffusion was more active in Taiwanese DRAM industry than Korean, using U.S. patent citation data. In the Korea and Taiwan model, mechanism of inter-nation or intra-nation knowledge diffusion is explicitly considered. By a number of studies using patent citation data, such as Jaffe (1997), Jaffe et al. (2002) and Lee and Yoon (2005), some stylized facts 11 about knowledge diffusion of Korean and Taiwanese firms are found. Characteristics unveiled by these studies are modeled explicitly in the model. Usually, inventing new concepts of product needed a variety of production factors and technology. Kim (1997) pointed out that network of knowledge diffusion among related firm is necessary for the development of product in R&D intensive industry. Levy and Quo (1991) reported that Taiwanese firms were more aggressive about product innovation than Korean firms, which may originate from a disadvantage around knowledge diffusion. A new product is assumed to be invented accidentally by a new SS firm (Taiwan) or existing LD firm (Korea). Because an invention of a new concept needs various product elements, the probability of invention is assumed to be affected by other firms’ technological level in other submarket. That is, spillover effect is important. It is assumed that spillover between Taiwanese firms is much easier than that between Korean firms, although spillover in a Korean business group is as easy as that in Taiwan. Taiwanese firms, therefore, can use more various elements than Korean. As such, Taiwanese firms can enjoy a forerunner’s advantage more than Korean firms in a frequent innovation technological regime. 6. Korea and Taiwan: Model and Simulation Results The basic feature of this version of the model is the same as the basic model in Section III except for the product innovation mechanism. It is assumed that Korea has only LD firms and Taiwan SS firms. Therefore, LD firms will be referred to as Korean firms and SS firms Taiwanese firms to avoid confusion. A new submarket is created by new firm (in Taiwan) or a division firm of an existing LD business group (in Korea). At each period, a new firm (in Taiwan) and one of the business groups (in Korea) is assumed to try to develop a new product. The probability of success of product innovation is determined by following formula. n pi = α ∑ β ij T j j =1 T j : The highest productivity level of j th submarket 12 β = 1 (when i and j firm are Taiwanese, or i and j firm are Korean and involved in same business group) β = a (when i and j have different nationality) β = b (When i and j are Korean and involved in other business groups) 0 < a, b < 1 α ranges in arbitrary interval in which a simulation run produce reasonable results. We can think α of easiness of product innovation. This time-invariant parameter is an important characteristic of technological regime in this model since the performance of Taiwanese or Korean firms depend on this. The probability is affected by the productivity level of other submarket, and β determines the extent of this spillover effect. We assume geographic localization of spillover effect found by Jaffe et al.(2002), Lee and Yoon(2003) and Park and Lee(2006) in U.S. patent data, that is, knowledge can be more easily diffused between Taiwanese firms than from Korea to Taiwan. Furthermore, lower spillover effect is assumed between business groups than Taiwanese firms in order to model the fact that Taiwanese firms follow a cooperative strategy. Therefore, b is an important parameter that reflects difference of industrial structure. The lower the value of b , the greater handicap Korean firms have regarding knowledge diffusion. Table 4 shows the result of this version of the simulation model. It presents the market share of Korean firms given various combinations of easiness of product innovation ( α ) and the degree of Korean handicap of knowledge diffusion ( b ). The market share is average of 50 simulations at final period (200). In Table 4, it is shown clearly that the easier product innovation the lower the market share Korean firms have given any value of Korean handicap. This is expected and is the same result as the basic model. As for the effect of the degree of Korean handicap, Table 4 illustrates that when product innovation is not frequent ( α =0.1 and 0.3), there is no impact of changing the degree of Korean handicap on the market share of Korean firms. When product innovation is frequent, however, the lower the degree of the handicap and the higher the market share. Meaning that, handicap of knowledge diffusion actually disadvantage Korean firms. Table 5 shows Korean firms’ market share given various combinations of easiness of product innovation ( α ) and easiness of imitation ( θ ). The result is similar to previous model, although the number is lower in high α cases because b =0.1. 13 7. Summary and Concluding Remarks This paper attempted to explain the relation between organizational selection and technological regime allowing product and process innovation by evolutionary simulation method. The basic model replicated following stylized facts.
When there is no product innovation, initial market dominance by SS firms shifts to LD firms eventually.
The more frequent product innovation, the lower market share LD firms’ have.
High frequency of product innovation leads to more decentralized market structure. We revised the model to explore the factors of specialization between Korea and Taiwan in some frontier industries. This model showed that different industrial structure leads to different performance. Korea was successful in rare product innovation regimes with LD firms and the handicap of inter-firm knowledge diffusion. On the other hand, Taiwan was successful in frequent product innovation regimes with SS firms and no handicap. Which economic system of Korea and Taiwan are more desirable has provoked a great deal of controversy in Korea. There have been the critics about Korean economic system composed of Chaebol compared to Taiwanese economic system. However, it is not appropriate to say that one system is superior to another because merits depend on a technological regime. 5 In the 1980s, when information technology is immature, Taiwanese economic system looked better; thereafter, as information technology get mature, Korean firms have grown faster. Emergence of dominant technology, DRAM in semiconductor industry and LCD in FPD industry, favor Korean economic system. It is another possible question whether Korean firms always have to manufacture products like DRAM and LCD and Taiwanese firms have to give up those markets. We observe the case not following our model. For example, since 2000, the market share of Taiwanese firms has risen rapidly in LCD industry, reaching about 40% in 2005. This was over the Korean firms’ total market share. In relation to this phenomenon, there are three factors to be pointed out. First, technology transfer from Japanese firms, losing the 5 In similar point of a view, Audretsch (1991) discussed that the relationship between industrial structure of U.S. and U.S.S.R and technological regime. 14 market dominance by Korean firms, is critical. Second, although the failure in mid-1990 of Taiwanese LCD industry, ITRI (Industrial Technology Research Institute), a government research institute, have continued to research related technology. (Mathew, 2003) Finally, the alliance among Taiwanese firms enabled timely large investment like Korean LD firms. That is, industrial policy and a new strategy have played important roles for Taiwanese SS firms to overcome size-related handicap in the LCD industry. Some readers may argue that present model can apply to no actual industries including those mentioned in this paper. It is true since all industries have its own technological regime and history. For example, sudden emergence of new generation of production facility is peculiar characteristic of LCD industry. (Mathew, 200) Our model, however, do not consider it. Our model is not history-friendly in that it do not reflect industry – specific details, thus it is not appropriate to explain a specific industry. But, this model tries to capture common feature of evolution process of some R&D intensive industries in Korea and Taiwan. There is broad space between general models and models specific to certain industry like ‘History-friendly’ model. We may say that two different types of models are complementary each other and this version of general model is helpful to lay out history-friendly models of industries mentioned previous like LCD, DRAM and computer component in which Korea and Taiwan are important players. The author is interested in building history friendly model of these industries, either. Reference Audretsch, D. B. (1995), Innovation and Industry Evolution, The MIT Press. Kim, C. W. (1997), ‘Progress and Stagnation of Technological Capacity of Computer Industry’, in K. Lee eds. Technological Capacity and Competitiveness of Korean Industry, Kyungmoonsa: Seoul. Kim, C. W. and K. Lee (2003), ‘Innovation, Technological Regimes and Organizational Selection in Evolution of Industry: A ‘History Friendly Model’ of the DRAM Industry,’ Industrial and Corporate Change, Vol. 12, No. 5, pp. 1195-1221. Klepper, S. (1996), Entry, Exit, Growth and Innovation over the Product Life cycle, American Economic Review, Vol. 86, No. 3, pp. 562-583. Jaffe, A. B., M. Trajtenberg and R. Henderson (2002), Geographic Localization of Knowledge Spillovers as Evidenced by Patent Citations, in A. Jaffe and M. 15 Trajtenberg eds., Patent, Citations and Innovations: A Window on the Knowledge Economy, MIT. Lee, K. and C. Lim (2001), Technological regimes, catching-up and leapfrogging: findings from the Korean industries , Research Policy 30, 459-483. Lee, K. and Yoon, M. (2005), ‘International, Intra-national, and Inter-firm Knowledge Diffusion and Technological Catch-up: Japan, Korea, and Taiwan in the Semiconductor Industry,’ mimeo Malerba, F. and Orsenigo (2001), ‘Innovation and market structure in the dynamics of the pharmaceutical industry and biotechnology: toward a history friendly model,’ paper presented at the DRUID Nelson and Winter Conference, Aalborg, June 15, 2001. Malerba, F., R. Nelson, L. Orsenigo and S. Winter (1999), ‘“History-friendly” models of industry evolution: the computer industry,’ Industrial and Corporate Change, 8, 340. Mathew, J. A. (2005), Strategy and Crystal Cycle, California Management Review, Vol. 47, No. 2. Levy, B. (1991), Transactions costs, the size of firms and industrial policy: Lessons from a comparative case study of the footwear industry in Korea and Taiwan, Journal of Development Economics, 34, 151-178. Levy, B. and W. Quo (1991), The Strategic Orientations of Firms and the Performance of Korea and Taiwan in Frontier Industries: Lessons from Comparative Case Studies of Keyboard and Personal Computer Assembly, World Development, Vol. 19, No. 4, 363-374. Park, K. H. and K. Lee (2006), “Linking the Technological Regime to Technological Catch-up: Analyzing Korea and Taiwan using US Patent Data”, Industrial and Corporate Change, Vol. 19, No.. Nelson, R. and S. Winter (1982), An Evolutionary Theory of Economic Change, Harvard University Press: Cambridge, MA. Shin, Jang-Sup and Ha-Joon Chang, (2003), “Restructuring Korea Inc.,” RoutledgeCurzon. Swann, P. and J. Gill (1993), Corporate Vision and Rapid Technological Change: The Evolution of Market Structure, Routledge: London. 16 Figure1A LD Firms’ Market Share when α =0 1 0.8 0.6 0.4 0.2 25 50 75 100 125 150 175 200 Figure1B. LD Firms’ Market Share when α =0.12 1 0.8 0.6 0.4 0.2 25 50 75 100 125 150 175 200 17 Figure2A. Number of Firms in the Market α =0 25 20 15 10 5 25 50 75 100 125 150 175 200 Figure2B. Number of Firms in the Market α =0.12 50 40 30 20 10 25 50 75 100 125 150 175 200 18 Figure3A Average Revenue of Firms α =0 4000 3000 2000 1000 25 50 75 100 125 150 175 200 175 200 Figure3B. Average Revenue α =0.12 1400 1200 1000 800 600 400 200 25 50 75 100 125 150 19 Table1. LD Firms’ Market Share α 0 0.04 0.08 0.12 0.01 0.88 0.58 0.53 0.47 0.02 0.87 0.59 0.52 0.49 0.03 0.89 0.60 0.52 0.51 0.04 0.89 0.60 0.58 0.52 θ Table2 Number of Firms α 0 0.04 0.08 0.12 0.01 9.3 37.2 42.4 52.1 0.02 9.3 35.0 45.9 49.0 0.03 9.3 35.0 42.3 64.3 0.04 9.4 37.7 58.5 79.4 θ Table3 Average Revenue α 0 0.04 0.08 0.12 0.01 3635 1807 1582 1324 0.02 3500 1861 1502 1354 0.03 3533 1835 1561 1072 0.04 3754 1839 1351 1059 θ 20 Table.4 Easiness of Product Innovation and Korean Handicap of Knowledge Diffusion 0.01 0.03 0.06 0.09 0.1 0.76 0.57 0.37 0.26 0.3 0.72 0.54 0.40 0.34 0.5 0.74 0.56 0.45 0.34 0.7 0.71 0.58 0.47 0.41 1 0.70 0.56 0.47 0.44 α b α =0.02 (50회 실행) Table5. Easiness of Product Innovation and of Imitation α 0 0.03 0.06 0.09 0.001 0.79 0.52 0.41 0.21 0.02 0.89 0.56 0.42 0.25 0.04 0.92 0.60 0.40 0.28 0.07 0.91 0.68 0.41 0.31 θ b =0.1 (50회 실행) 21 Appendix. Initial Value of Parameters Number of submarkets in initial period: 5 Number of firms per submarket: 6 Number of Korean firms per submarket in initial period: 3 ρ =0.001 (parameter of R&D investment and productivity linkage) β =0.75 (lagged performance indicator) λ =0.2 (parameter of R&D investment ratio change) r0 =0.005 (R&D investment ratio in initial period) δ =0.03 (depreciation rate) k 0 =60 (capital in initial period) h0 =1 (productivity in initial period) FC s =0.1 (fixed cost of small specialized firm) VCK s =0.1 (variable cost per capital of small specialized firm) FC L =4 (fixed cost of large diversified firm) VCK L =0.05 (variable cost per capital of large diversified firm) g min =1 (minimum growth speed of a submarket) g max =20 (maximum growth speed of a submarket Rmin =50 (minimum total revenue of a submarket) Rmax =70 (maximum total revenue of a submarket) Rmin =100 (minimum of limitation of submarket growth) Rmax =500 (maximum of limitation of submarket growth) α =0.01; (coefficient affecting probability of emergence of new market) b =0.1; (coefficeint of appro nonlocal firm) θ =0.02 (easiness of imitation) 22 View publication stats