Inter-Firm Technology Spillover
and the “Virtuous Cycle” of
Photovoltaic Development in Japan
Despite a locational disadvantage as a mid-latitude country, Japan has taken a leading role in world photovoltaic (PV) development. Two main factors have been critical to the successful development of PV technology in Japan. First, like semiconductors, PV technology is central to a complex web of related technologies and can therefore benefit from both learning effects and economies of scale. Second, because of the interdisciplinary nature of PV development, technology spillover benefits are high, in turn further stimulating learning effects (Watanabe 1997).
The above-mentioned interdependent success factors in PV development highlight the importance of both an endogenous technological innovation perspective and the inducement mechanisms of technological change. In the case of PVs, Japan's Ministry of International Trade and Industry (MITI) initiated development under its Sunshine Program, a research and development (R&D) program on new energy (MITI 1970–1990). In particular, the Sunshine Program aimed to encourage broad cross-sector industry participation; to stimulate cross-sector technology development; and, finally, to induce substantial industry investments in PV R&D, leading to a rapid increase in the industry's knowledge stock on PV technology (Watanabe 1995c; Watanabe and Clark 1991).
Supported by a PV R&D acceleration strategy adopted in 1979 (Industrial Technology Council 1979) and the creation of a research association for PV development (established in 1990), the technology knowledge stock arising from proprietary PV R&D and via spillover effects increased dramatically. The increase in this technology knowledge stock contributed to a significant decrease in the cost of solar cell production, which induced a further increase in demand for solar cells (and hence production). In turn, this increase in demand/production induced further PV R&D, thus creating a “virtuous cycle” (a positive feedback loop) between R&D, market growth, and price reduction (Watanabe 1995b, 1997). The achievement of a virtuous cycle (see Figure 6.1; see also Watanabe 1995a) has significant techno-economic implications. However, to date only limited research has been undertaken to elucidate Japan's policy framework in this area. Grübler analyzed this virtuous cycle using the author's data and noticed the learning effect triggered by MITI's inducement strategy (Grübler 1998). However, this analysis was limited to the aggregate behavior of the entire PV industry and did not consider the impacts of inter-firm technology spillover effects. Although some pioneering studies have attempted to link learning and technology spillover effects (e.g., Griliches 1957; Mansfield 1961; and Jovanovic and Lach 1989), further research is necessary for analyzing the complicated mechanisms behind the creation of a “virtuous cycle” in the advancement of a new technology and the policy framework conducive to such developments.
This chapter presents an empirical analysis of Japan's firm-level PV technology development over the past two decades. The analysis focuses on the impacts of inter-firm technology spillovers to provide an empirical demonstration of the industrial dynamism of this virtuous cycle created by a directed policy initiative. This chapter also illustrates the role of technology spillovers in stimulating these systemic positive feedbacks, resulting in an endogenous technological innovation–development–diffusion process. A framework for relating technology spillover effects and learning/experience curves in the PV innovation process is developed in Section 6.2. The current state of cross-sector industry involvement in PV technology development in Japan is reviewed in Section 6.3. In Section 6.4, the state of inter-firm technology spillovers in Japan's leading PV firms is analyzed. The virtuous cycle between R&D, market growth, and price reduction is assessed in Section 6.5. Finally, Section 6.6, discusses the implications of the present analysis for an induced technological change perspective and presents final observations and conclusions.
6.2.1 R&D, technology spillovers, and learning effects: Maximizing system efficiency on the market
R&D has become increasingly expensive. With such increased costs come opportunity costs, particularly in terms of choosing an optimal technology trajectory for sustainable development. Experience curves and learning effects often optimize a system that is linked to an external technological infrastructure that may not necessarily be the best choice from an environmental perspective. For these reasons, it is difficult for policy makers to choose which environmental technologies to support in using R&D incentives as an effective policy tool. Our work demonstrates the critical importance of “learning through the market,” that is, making decisions based on market signals rather than trying to change the marketplace to achieve sustainable development.
In this context, technology spillovers not only alleviate some of the burden of huge R&D expenditures, but can also enhance the learning effects involved in assimilating environmentally friendly technologies and processes. The objective of the analytical framework presented here is to highlight the construction of a virtuous cycle in Japanese PV development. A learning/experience curve mechanism plays an important role in the policy-formation process. However, policy is only the tool that guides the technology trajectory by inducing a virtuous cycle. Its role is simply to stimulate learning and technology spillovers through—not against—the marketplace.
To illustrate these principles, an analytical framework is developed to elucidate empirically the development of PV technology in Japan. The framework incorporates not only technology spillovers within the PV technology sector, but also technology spillovers that have resulted in energy-efficiency improvements across industrial sectors, thereby further enhancing the positive effects of PV development. Figure 6.1 shows the links between these components. Japan's policy approach has created a virtuous development cycle in which energy-efficiency improvements from technology spillovers enhance learning effects across sectors, leading to effective innovation, particularly in PV technology. This effective innovation has contributed to economic growth, which in turn has stimulated technology spillovers. Thus, the positive feedback cycle continues. There is a loop not only between economic growth and technology spillovers, but also between technology spillovers and learning effects.
Technology Improvement and Learning Effects
In the following equation, a newly generated technology [T] is a function [ω] of current technological level T and R&D investment R:
The learning function φ(T) generally has the following characteristics (Arrow 1962):
With the generated [T], the technology level will improve to T2, as seen in Equation (6.4):
Since φ′(T) > 0 and φ(T2) > φ(T), the learning effects increase as the technological level increases.
Based on Jaffe (1986), provided that technology Td spills over from the donor (D) to the host (H), assimilating technology in the host [Td]h can be expressed using an assimilation capacity function θ as follows:
where [Td]h is assimilating technology; θ(Th,α) is assimilation capacity; Td is spillover technology flow; Th is technological level of host; and α is learning and assimilation capacity.
Equation (6.5) can be further developed as follows:
where Th is learning capacity, which is dependent on the host's technological level; and α is assimilation capacity. Equation (6.6) demonstrates that assimilating technology depends on learning capacity.
Technology newly generated by the host through assimilation of spillover technologies can be expressed as follows:
Since the technology level of the host at this stage is Th2 = Th +[Th2] > Th, φ(Th2) > φ(Th), learning capacity increases as technology spillovers increase.
On this basis, we note a potential virtuous cycle between technology improvement, technology spillover, and learning effects, as illustrated in Figure 6.2. This concept is the basis of our analysis.
6.3.1 Intensive R&D by leading cross-sector firms
In line with its PV development strategy of encouraging the involvement of leading cross-sector firms, in 1990 MITI established an R&D consortium for PV development called the Photovoltaic Power Generation Technology Research Association (PVTEC). Table 6.1 lists the firms participating in PV development under the Sunshine Program and identifies the subset of companies that are also members of PVTEC. The table shows the broad spectrum of firms participating in PV development in Japan—from manufacturers of textiles, chemicals, petroleum and coal products, and ceramics to public institutes and the iron and steel, nonferrous metals, and electrical machinery industries. In addition to PVTEC, MITI has entrusted PV development to broad industrial sectors, including the electric power, housing, and construction industries. A total of 65 firms participated in MITI's PV development program in 1999.
Figure 6.3 illustrates the trends in Japan's R&D expenditures for PVs. Both public and private funding is shown; that is, the figure includes both expenditures via MITI's Sunshine Program as well as those by Japan's PV industry over the 1974–1998 period. As the figure indicates, R&D expenditures increased dramatically after the early 1980s. This increase was in line with the recommendation by the Industrial Technology Council (one of MITI's advisory bodies) to accelerate PV R&D as a priority under the Sunshine Program (Industrial Technology Council 1979).1 Although R&D expenditures began to decline in 1987—during Japan's “bubble economy” period—they began to rise again as of 1993 and have since increased steadily, primarily from industry sources. This further increase can be attributed to MITI's new policy under the New Sunshine Program, an R&D program on energy and environmental technologies. The New Sunshine Program, which began in 1993, accelerated PV R&D activities further, particularly in light of the global environmental consequences of carbon dioxide emissions resulting from energy use (Watanabe 1995d, 1996).
|Chemicals||Kaneka Corp||Mitsubishi Chemical|
|Mitsui Toatsu Chemicals|
|Matsushita Battery Industrial Co., Ltd.|
|Petroleum and coal products||Showa Shell Sekiyu K.K.
|Japan Energy Co.|
|Nippon Sheet Glass|
|Iron and steel||Kawasaki Steel||Japan Steel Works|
|Nonferrousmetals and products||Mitsubishi Materials
|Electrical machinery||Sanyo Electric Co., Ltd
Sharp Corp.Fuji Electric Co., Ltd.
Mitsubishi Electric Corp.
Sumitomo Electric Industries, Ltd.
Matsushita Electric Industries Co., Ltd.
Oki Electric Industry
Japan Measurement and Inspection Institute
|Japan Quality Assurance Organisation|
Central Research Institute Institute of Electric Power Industry
Shikoku Electric Power Research Institute
Japan Electric Safety & Environment Technology Laboratory
|Japan Weather Forecast Association|
|Electric Power||Okinawa Electric power|
|Kashima North Joint Electric Power|
|Housing and construction||Misawa Homes
National House Industrys
YKK Architectural Products
Note: Firms indicated in bold are discussed in this analysis; those listed in middle column are also members of PVTEC.
Table 6.2 summarizes trends in PV R&D expenditures by eight leading Japanese PV firms and the government financial support provided under MITI's Sunshine Program. Approximately 40 percent of MITI's PV R&D budget was allocated to eight leading PV firms.2 Table 6.3 summarizes the results of a correlation analysis of the PV R&D expenditures of these eight firms and MITI's financial support for PV R&D. The data indicate that, in all firms examined, MITI's financial support significantly induced private PV R&D expenditures with a time lag of one year. As demonstrated in Watanabe (1999) and (2000), MITI's energy R&D policy induced private R&D expenditures by firms in the following ways:
• It encouraged broad cross-sector industry involvement in national R&D programs such as the Sunshine Program (energy supply technologies) as well as the Moonlight Program (energy efficiency and conservation) (Watanabe and Honda 1992).
• It fostered cross-sector technology spillovers and inter-technology stimulation.
• It induced substantial industry activity in the broad area of energy R&D.
• Enhanced R&D efforts led to an increase in industry's technology knowledge stock, which stimulated further research activities across technologies and sectors (Watanabe et al. 1991).
• This inducement mechanism played a catalytic role for industry's “technology substitution” for energy; that is, the substitution of efficiency improvements and conservation through the application of new technologies for energy as a factor input.
Note: Column headings are as follows: (A) Sanyo Electric Co., Ltd.; (B) Kyocera Corp.; (C) Sharp Corp.; (D) Kaneka Corp.; (E) Fuji Electric Co., Ltd.; (F) Hitachi, Ltd.; (G) Mitsubishi Electric Corp.; (H) Sumitomo Electric Industries, Ltd. Figures in parentheses indicate amount of financial support from government (MITI). Figures in square brackets include support to non-industry sectors (universities and national research institutes).
*Figures for 1996–1998 are estimated values based on interviews and statistics in the Report on the Survey of Research and Development (special issue on energy R&D), Agency of General Coordination and Management, Japanese Government, Tokyo, Japan.
Regression model: PVR = A × SSPV αt−1
Note: SSPV = PV R&D budget from Sunshine Program; PVR = industry's PV R&D expenditure; α = learning and assimilation capacity; DW = Durbin–Watson statistic. See Table 6.2 for firm names.
The industry R&D expenditure induced by the Sunshine Program has furthered the PV-related technology knowledge stock. In line with previous research (Watanabe 1992), the technology knowledge stock of PVs is measured using the following equation:
where Tt is technology knowledge stock of PV R&D at time t; Rt−m is PV R&D expenditure at time t − m; m is the lead time of (time lag between) PV R&D expenditures and PV commercialization; and ρ is the rate of obsolescence of PV technology.
To identify m and ρ, 19 leading PV firms were surveyed in 1993 with the support of MITI's Agency of Industrial Science Technology (AIST). Survey responses were received from 15 firms (79 percent of firms surveyed), including the 8 firms examined above. From these responses, 57 valid samples for estimating the time lag (m) and 28 for estimating the technology lifetime parameter (ρ)were obtained (see Appendix 6.A). As both samples are well balanced with respect to firms and technology life-cycle stages, the time lag and technology lifetime in leading Japanese PV firms over the past two decades were estimated by taking the average of the respective samples.
The average time lag between PV R&D and PV commercialization was estimated to be 2.8 years, and the average lifetime of PV technology was estimated to be 4.9 years. Assuming that technology depreciates and becomes obsolete over time, the annual rate of PV technology obsolescence was estimated to be 20.3 percent per annum (inverse of the lifetime of PV technology).3 The estimated time lag and rate of technology obsolescence were evaluated by means of comparative statistical analysis using patent data; both proved to be statistically significant (see Appendix 6.A). Figure 6.4 illustrates the trends in the technology knowledge stock as measured by the lagged R&D expenditures as explained above and relates it to PV production volumes and declines in PV prices. Table 6.4 summarizes the same trends in eight leading PV firms. Figure 6.4 and Table 6.4 demonstrate that the technology knowledge stock as measured by PV R&D increased dramatically as of 1983/1984 in Japanese industry while PV R&D expenditures rose sharply from 1980/1981. This corresponds to a 2.8-year time lag between PV R&D and its translation into an increased technology knowledge stock enabling reduced costs and commercialization of PV technology in niche markets.
Table 6.5 summarizes the development of solar cell production in Japan's PV industry. Japan's solar cell production in 1999 amounted to 80.0 MW, or 40 percent of a world production total of 200 MW (see Figures 6.5 and 6.6).4 Figures 6.7and 6.8 show the development of solar cell production by eight leading Japanese PV firms and their distribution. The shares of the top three firms accounted for nearly 90 percent of solar cell production in Japan in the 1995–1999 period.
A comparison of Tables 6.2 and 6.5 indicates that the cross-section of firms involved in PV R&D is broader than that of firms involved in solar cell production alone. Building on this observation, a comparison of Tables 6.4 and 6.5 indicates the following:
• Although the production levels of firms A and B are similar, the estimated technology knowledge stock of firm A is much greater than that of firm B.
• There are no substantial differences in the estimated technology knowledge stocks of firms B, C, D, E, F, G, and H; their production levels can be classified into three groups: firm B (production level much higher than those of the other firms considered); firms C and D (production levels similar to each other's, both firms among the top four); and firms E, F, G, and H (production levels much lower than those of other firms considered).
Note: See Table 6.2 for firm names.
These observations suggest that broad inter-firm or cross-sector technology spillovers have prevailed in Japan's PV industry. The following subsection attempts to corroborate this hypothesis.
Figure 6.9 illustrates the relationship between PV production (SCP; average over the 1991–1995 period; vertical axis) and the estimated PV technology knowledge (TPV; horizontal axis). This map of the “techno-production structure” of leading Japanese PV firms can be divided into the following four clusters:
Note: See Table 6.2 for firm names.
• Cluster 1: Consists of firms A and B, which have the highest production levels in Japan. The technology knowledge stock level of firm A is much higher than that of firm B.
• Cluster 2: Includes the top four firms—firms A and B plus firms C and D—which together account for nearly 90 percent of solar cell production in Japan. Among these four firms, the technology knowledge stock levels of firms B, C, and D are almost the same, while that of firm A is much higher.
• Cluster 3: Consists of firms B, C, D, E, F, and G, which share similar technology knowledge stock levels. This cluster can be classified into two groups, firms B, C, and D, and firms E, F, and G. The former also belong to cluster 2 and rank higher in terms of production levels; the production levels of the latter are much lower.
• Cluster 4: Consists of firms E, F, G, and H. It includes firms of the lowest production levels. The technology knowledge stock levels of firms E, F, and G are reasonably high; that of firm H is extremely low.
An analysis of these clusters suggests the following possible technology spillovers among leading Japanese PV firms:
• Cluster 1 suggests that firm A plays the role of a technology donor while firm B enjoys the role of a technology receiver (host).
• Similarly, cluster 3 suggests that firms B, C, and D play the role of a technology donor, while firms E, F, and G enjoy the role of a technology receiver.
• Clusters3 and 4 suggest that firm H acts as a technology receiver.
Even in the absence of detailed data on the mechanics and flows of PV technology knowledge between firms, the above mapping nonetheless suggests a high level of technology spillover between the firms analyzed.
6.4.1 Impact of inter-firm technology spillovers on
Intensive PV R&D expenditures and the resulting improved technology knowledge stock are the sources of PV innovations. Consequently, R&D expenditures and the resulting increased technology stock will generate a number of patent applications in the field of PVs. Table 6.6 and Figure 6.10 summarize trends in the number of patent applications in Japan's PV industry, including applications submitted by the eight leading PV firms. In line with previous research (e.g., Griliches 1984), the number of PV patent applications (PVPA) can be estimated by the following regression equation:
Note: See Table 6.2 for firm names.
where PVRr is PV R&D expenditures at constant (1985) prices and TPV is the estimated PV technology knowledge stock.
Given that the essential requirement of a patent application is the novelty of an idea and that this novelty generally decreases as time passes (Freeman 1982), a third factor t representing a time trend should be incorporated into the regression equation for PV patent applications as follows:
Furthermore, considering the interdisciplinary nature of PV R&D, MITI's PV R&D policy for stimulating cross-sector technology spillovers, and the resulting broad inter-firm technology spillovers [see Section 6.3.4 above and Watanabe (1999)], the PV technology knowledge stock TPV should be decomposed into proprietary knowledge (PV R&D performed by a given firm) and assimilated technology knowledge (spillovers of technology knowledge generated by the R&D of other firms). Provided that a firm makes every effort to maximize the contribution of assimilated technology knowledge, TPV can be decomposed as follows:
where TPVi is the technology knowledge stock of PV R&D in firm i; Σj TPVj is the total industry technology knowledge stock of PV R&D; and Z (0 > Z > 1) is a measure of the assimilation capacity (Cohen and Levinthal 1989; Watanabe and Griffy-Brown 1999; Watanabe et al. 2000).
On the basis of the above, Equation (6.9) can be estimated using the following simple Cobb-Douglas-type production function for Japan's PV industry patent applications over the 1976–1995 period:
where A is a scale factor and λ, α, and β are elasticities.
In the case of leading PV firms that generate a reasonable portion of their own technology knowledge stock through proprietary R&D, and also considering that, generally, Z 1, the ratio of assimilated spillover technology knowledge to proprietary technology knowledge stock is less than 1 [Z (Σj TPVj − TPVi)/TPVi− 1], Equation (6.12) can be approximated by
where γ = Zβ.
Note: See Table 6.2 for firm names; λ, α, and β are elasticities; γ = Zβ, with Z being the assimilation capacity [see Equation (6.13)]; DW = Durbin–Watson statistic; t-values are given in parentheses.
In Table 6.7, Equations (6.12) and (6.13) are used to identify factors governing patent applications by leading PV firms over the 1976–1995 period. The table suggests that PV R&D expenditures make the most statistically significant contribution to PV patent applications (except for firm C, where the correlation is insignificant), followed by our measure of the technology knowledge stock. In many firms, the technology knowledge stock of proprietary R&D and assimilated spillover technology knowledge are statistically significant contributors to PV patent applications.5 The elasticity of the time trend (λ) is negative and statistically significant for all firms examined. These findings suggest the following interpretation of the factors governing patent applications in Japan's PV industry over the past two decades:
• R&D expenditures—representing R&D activities at the forefront—make the most significant contribution.6
• R&D expenditure flows and the estimated technology knowledge stock of proprietary R&D make an additional contribution. In many firms, the assimilated technology knowledge stock also makes a significant contribution toward explaining PV patent applications.
• Considering the general downward trend of novel ideas worthy of patent applications, a general decrease in the number of patents over time was observed in all firms examined.
Trends in PV production depend inter alia on the improved technology knowledge stock (arising from R&D). As suggested in Section 6.3.4, this knowledge stock consists of the technology knowledge stock from proprietary PV R&D plus the assimilated technology stock acquired via spillovers from the technology knowledge generated by other firms. Meyer-Krahmer (1992) suggests that the extent of internal (R&D) and external knowledge acquisition (assimilative capacity) also depends on price signals.
On the basis of these observations, an equation describing the governing factors of solar cell production in Japanese industry over the past two decades is estimated using the following simple Cobb-Douglas-type production function:
where SCP is solar cell production and Pey represents relative energy prices.
where γ ≡ Zβ
In Table 6.8, Equations (6.14) and (6.15) are used to identify factors Governing solar cell production of leading PV firms over the past two decades. The table indicates that the technology knowledge stock—both from proprietary PV R&Das well as from assimilated technology knowledge spillovers—contributes significantly to increases in solar cell production. In addition, an increase in relative energy prices contributes significantly to a production increase (except for firm D, where the influence of this variable is statistically insignificant).A comparisonof Tables 6.7and 6.8 reveals that inter-firm technology spillovers have a more Table 6.9. significant impact on solar cell production than on patent applications. This suggests that technology spillovers contribute to solar cell production directly rather than by stimulating PV innovation. Therefore, technology spillovers play a critical role in determining the trajectory of technological change of PVs, linking it to experience curves and the existing web of technological infrastructures.
Source: Table 6.9.
As is commonly pointed out, any system of equations should be carefully estimated by multiple regression analysis using ordinary least squares, because this sometimes demonstrates statistical coincidental correlations. Although the interpretation presented here is based on a cross-evaluation of empirical observations, further statistical tests are necessary before drawing definitive conclusions.
Increased solar cell production resulting from an increase in the PV technology knowledge stock and induced by high energy prices can be expected to lead to falling costs; that is, learning curve effects for both solar cell producers and customers. In addition, economies of scale effects can be expected in solar cell production operations. Higher production levels have led to a decline in solar cell prices, as shown in Table 6.9 and Figure 6.11. The solar cell production price in 1974, the year the Sunshine Program was started, was 20,000 yen per watt (W); by 1999 it had decreased by a factor of 40 to 490 yen/W (in current prices). In constant 1985 yen, prices decreased from 26,120 yen/W to 590 yen/W between 1974 and 1999; that is, by a factor of more than 44. This process can be explained by two factors: (1) an improved technology knowledge stock coupled with inducement mechanisms through changing relative energy prices, and (2) effects due to learning and economies of scale.
1985 constant prices
Solar cell production prices in Japanese industry over the past two decades are therefore estimated using the following Cobb-Douglas-type production function:
(1) Inducement by Technology Knowledge Stock and Energy Prices
where PSC is solar cell production price (in constant prices).
Similarly, Equation (6.16) can be approximated by
Note: See Table 6.2 for firm names; λ, α, and β are elasticities; γ = Zβ, with Z being the assimilation capacity [see Equation (6.13)]; DW = Durbin–Watson statistic; t-values are given in parentheses.
where γ = Zβ.
(2) Effects due to Learning and Economies of Scale
In Table 6.10, Equations (6.16), (6.17), and (6.18) are used to identify factors governing the decline in solar cell prices by leading firms. Table 6.10a shows that technology knowledge stock—represented both by proprietary PV R&D and assimilated technology knowledge through spillovers—contributed significantly to a dramatic decrease in solar cell production prices (except for firm A, where the statistical influence is insignificant). In addition, increases in relative energy prices contributed significantly to solar cell price decreases (except for firm F). Table 6.10b suggests that over the past two decades the effects of economies of scale have contributed significantly to a decline in solar cell prices in all the leading Japanese PV firms examined, whereas the impacts of learning effects seem to have been rather limited.7
Model: PSC = A CMSCPη
Note: PSC = solar cell production price (fixed price); A = constant; CMSCP = cumulative solar cell production; DW = Durbin–Watson statistic. See Table 6.2 for firm names.
Because the lack of data precluded an estimate of all variables of the production function for all firms, and because the estimated influence of learning and economies of scale effects is extremely variable among the limited sample of firms analyzed (see discussion below), definitive conclusions await further empirical and statistical corroboration.
Learning effects can be clearly observed at the aggregate industry level in line with the increase in PV R&D technology knowledge stock and its embodiment in production facilities, as illustrated in the learning curve shown in Figure 6.12. In Table 6.11, the learning coefficient among leading PV firms in Japan over the 1979–1999 period is compared with that over the 1980–1990 period. A comparison of the interdependency of technology spillover and solar cell price decreases (Table 6.10a) and the difference in learning coefficients (rates) in Table 6.11 indicates a clear correlation between the two. Figure 6.13 illustrates this correlation in leading PV firms. The figure suggests that firms with a higher dependency on technology spillovers, as indicated by the higher “potentiality of technology spillover assimilation” indicator in Figure 6.13, also seem to demonstrate better performance with respect to learning curve effects. This seems to corroborate the theoretical proposition set forth in Section 6.2 that there is a mutually stimulating interaction between technology spillovers and enhancement of learning curve effects.
6.5.1 Feedback loop to a further production increase
As demonstrated in Table 6.9 and Figure 6.11, a dramatic decrease in solar cell prices induces further production increases. This can be attributed to both a demand- and a supply-side response. Demand for PV cells increases with falling prices. Suppliers aim to maintain sales volumes; that is, to compensate for price declines with increased production volumes. In addition, changing relative prices (in the case of increasing energy prices) also induces a production increase. Equation (6.19) depicts this behavior. In this simple Cobb-Douglas-type production function, both solar cells and aggregate energy prices in the previous year are used as explanatory variables:
Model: SCP = A × PSC α t−1Peyβt−1
Note: SCP = solar cell production; A = constant; PSC = solar cell production price (fixed price); Pey = relative energy prices; α and β are elasticities; DW = Durbin–Watson statistic; t-values are given in parentheses. See Table 6.2 for firm names.
In Table 6.12, Equation (6.19) is used to demonstrate this feedback loop and the factors that have induced it in leading PV firms over the past two decades. As the table indicates, for any given year both a decrease in solar cell prices and an increase in energy prices in the previous year provide significant inducement for a production increase in all firms examined (except for the energy price inducement effect for firm B, which turns out to be statistically insignificant).
Model: PVR = A × SCPα.
Note: PVR = industry PV R&D expenditure; A = constant; SCP = solar cell production; α = learning and assimilation capacity. See Table 6.2 for firm names.
Similar to the feedback loop between falling PV prices and rising PV production, stepped-up PV production induces further R&D. Equation (6.20) illustrates this inducement mechanism:
In Table 6.13, Equation (6.20) is used to demonstrate this feedback loop in leading PV firms over the past two decades. The table illustrates that solar cell production simultaneously induces PV R&D in all firms examined.9
The analysis presented above demonstrates the creation of a virtuous cycle in PV development in Japan. The simultaneous involvement of cross-sector industry collaboration, MITI's inducement policies for R&D (and niche market incentives), the creation of a continuously rising PV technology knowledge stock, technology spillovers, and the interaction between these factors leading to the formation of a PV technological trajectory characterized by dramatically falling costs were analyzed. Figure 6.14 illustrates the virtuous cycle between R&D triggered by the Sunshine Program, steady market growth, and the resulting dramatic price reduction in Japan's PV industry over the 1976–1995 period. A noteworthy element of this cycle is the “double boost” to solar cell production from the increased technology knowledge stock resulting from PV R&D and from falling solar cell prices. A similar “double boost” effect can be observed in PV R&D—the source of the increasing technology knowledge stock—arising from both increased solar cell production volumes and MITI's PV R&D budget's stimulation of further private PV R&D. This virtuous cycle of PV development in Japan suggests that a variety of policy mechanisms exist for inducing endogenous technological change.
On the basis of the empirical analysis presented in this chapter, the following policy implications emerge:
• Institutional and technological demonstrations of PV development and utilization should be carried out.
• Technology improvements and organizational learning should be accelerated.
• R&D on renewable energy technologies (RETs) and the development of market incentive structures that promote the industrial dynamic mechanism of a virtuous cycle involving market growth and price reductions for technically proven advanced RETs should be intensified (Williams et al. 1996).
• Coherent systemic policies aimed at unleashing the industrial dynamic mechanism of a virtuous cycle involving market growth and price decreases for RETs should be formulated.
• Market opportunities for RETs developed with private-sector resources should be pursued and viable RETs industries should be established.
As these suggestions have been extracted from an empirical analysis of the relationships between PV R&D and production in Japan, they may not be directly applicable to other technologies or countries. However, the following general observations could be helpful in considering the relationships between inter-firm technology spillovers, R&D, market growth, and price reductions for the development and diffusion of innovative environmental technologies.
Creating a virtuous cycle in the development of a new technology trajectory depends on a number of factors within the endogenous technological innovation process. The analysis presented here has demonstrated the complex and important role policy can play in inducing technological change if critical success factors are acting in concert. In the case of PV development in Japan, it was critical that the targeted technology was embedded in a complex web of related technologies. Like semiconductors, PVs can maximize the benefits of an improved technology knowledge stock resulting from both public and private R&D, as well as from proprietary knowledge generation combined with knowledge assimilation (spillover effects), learning effects (via experience curves), and economies of scale. This induced technological change appears particularly successful because of the interdisciplinary nature of its development, which maximized the benefits of technology spillovers.
The creation of a virtuous cycle has promising policy applications in terms of induced technological change and the environment. The case of a virtuous cycle in PV technology development demonstrates how critical factors can alter technology trajectories in ways that benefit the environment and how policy, market forces, and R&D can work together to this end. Interestingly, our analysis also indicates the existence and importance of network externalities arising from technological interrelatedness. This highlights the critical importance of considering and coordinating entire “technology networks” as targets for technology policy. Of particular importance in this context is interpreting and responding to market signals rather than creating “false” signals from a policy perspective. In view of a multitude of positive effects on the economy and the environment, further examination of these issues is certainly worth continued study and policy consideration.
Measurement of Time Lag between PV R&D and Commercialization and Rate of Obsolescence of PV
Measuring the technology knowledge stock via PV R&D expenditures requires reliable up-to-date estimates of the time lag between R&D expenditures and increases in the technology knowledge stock, and of the rate of obsolescence of that knowledge. However, no reliable survey exists estimating these factors for PV R&D. Therefore, in 1993, with the support of AIST of MITI, the authors prepared a questionnaire for 19 leading PV firms from the 26 member firms of the Photovoltaic Power Generation Technology Research Association (PVTEC). The survey included questions related to the time lag between R&D and commercialization and the lifetime of PV technology. Responses were received from the following 15 firms (including the 8 firms examined here): Sanyo Electric Co., Ltd.; Kyocera Corp.; Sharp Corp.; Kaneka Corp.; Fuji Electric Co., Ltd.; Hitachi, Ltd.; Mitsubishi Electric Corp.; Sumitomo Electric Industries, Ltd.; Daido-hoxan Co.; Matsushita Battery Industrial Co., Ltd.; Showa Shell Sekiyu K.K.; Tonen Co.; Japan Energy Co.; Osaka Titanium Co., Ltd.; and Matsushita Electric Industries Co., Ltd.
Time lag (years)
Number of Observations
|Total||57 Average: 2.8 years|
Number of Samples
|Total||28 Average: 4.9 years (20.3 percent p.a.)|
To evaluate the estimated time lag (m = 2.8 years) and the rate of obsolescence of technology (ρ = 20.3 percent), a comparative evaluation was made using the following equation:
where PVPA is the number of PV patent applications, PVRr is PV R&D expenditure (in constant prices), and TPV is technology stock of PV R&D.
TPV is a function of PVRr , m, and ρ, with m(2.8 ± ε) and ρ(20.3 ± ε). Results of a sensitivity analysis varying m and ρ are reported in Table 6.A3. The case with m =2.8 and ρ = 20.3 is the most statistically significant.
Note: m = lead time of PV R&D and its commercialization; ρ = rate of obsolescence of PV technology; λ, α, and β = elasticities; DW = Durbin–Watson statistic; AIC = Akaike Information Criterion; t-values are given in parentheses.
This chapter is based on work originally presented at the International Workshop on Induced Technological Change and the Environment (IIASA, Laxenburg, Austria, 1999).
1 Facing a second energy crisis in 1979, the minister of MITI consulted with the Industrial Technology Council about a priority policy menu. In response, the Council prepared a recommendation entitled “Strategy for Acceleration of the Sunshine Program,” which identified certain R&D priority areas, including PV R&D. In response to this recommendation, MITI introduced new policies in 1980, including a Law for the Promotion of Development and Introduction of Oil Alternative Energy; created a new funding system by means of special accounts for energy security; and established NEDO (the New Energy Development Organization, MITI's affiliate responsible for energy R&D). Consequently, R&D activities, particularly in priority areas such as PVs, were accelerated.
2 The remaining 60 percent was appropriated to national research institutes, to universities for basic research, and to other industries such as the electric power, housing, and construction industries for application-oriented research.
3 The rate of technology obsolescence for energy R&D in Japan's manufacturing industries was estimated to vary between 14.5 percent (1970) and 22.2 percent (1994), while total R&D intensity was estimated to be between 8.2 percent (1970) and 12.1 percent (1994) (Watanabe 1999). Table 6.A3 in Appendix 6.A gives more details on these estimates.
4 The world's total solar cell production in 1999 was 201.3 MW, including Japan with 80.0 MW (39.7 percent); United States, 60.8 MW (30.2 percent); Europe, 40.0 MW (19.9 percent); and other countries, 20.5 MW (10.2 percent).
5 A significance of 2.5 percent for two firms, 15 percent for two firms, and 20–25 percent for three firms. The contribution is insignificant for firm A's technology stock of proprietary R&D, which is much greater than that of the other firms.
6 An exception is the case of firm C, where the influence of factors on PV patent applications is statistically insignificant. Looking at Tables 6.3 and 6.6, we note that, contrary to its recent increase in solar cell production, firm C's PV R&D expenditure share has decreased. This suggests that firm C depends on its proprietary technology knowledge stock based on its previous R&D and the assimilation of technology knowledge via spillovers rather than by proprietary R&D for PV patent applications. Table 6.8 supports this view.
7 Contrary to the significant contribution of learning effects to a decrease in solar cell production prices in firms A, E, and F, contributions in firms B, C, and D are not statistically significant. This is due to a rapid production increase in later years in firms B, C, and D that provided significant opportunities for benefiting from economies of scale with limited opportunities for learning effects.
8 The learning coefficient α can be defined by the following equation: PSC = A × CMSCP α, where PSC is the solar cell price (in constant money), and CMSCP is cumulative solar cell production. The “potentiality of technology spillover assimilation” indicator in firm i, TSPi , can be defined as follows: where TPVi is the technology stock of PV R&D in firm i. Firms shown are as follows: (A) Sanyo Electric Co., Ltd.; (B) Kyocera Corp.; (C) Sharp Corp.; (D) Kaneka Corp.; (E) Fuji Electric Co., Ltd.; (F) Hitachi, Ltd.; (G) Mitsubishi Electric Corp.; (H) Sumitomo Electric Industries, Ltd.
9 This can be imputed by the following simple identity: ΔR =ΔR/Y +ΔY , where R is R&D investment, R/Y is R&D intensity, and Y is production.
10 PVR is industry PV R&D expenditure; SSPV is PV R&D budget from the Sunshine Program; TPV is technology knowledge stock of PV R&D; m is time lag of PV R&D to commercialization; ρ is the rate of obsolescence of PV technology; SCP is solar cell production; Pey is relative energy prices; and PSC is the solar cell price (in constant 1985 prices).
Arrow, K.J., 1962, The economic implication of learning by doing, Review of Economic Studies, 29:155–173.
Cohen, W.M., and Levinthal, D.A., 1989, Innovation and learning: Two faces of R&D, Economic Journal, 99:569–596.
Freeman, C., 1982, The Economics of Industrial Innovation, The MIT Press, Cambridge, MA, USA.
Griliches, Z., 1957, Hybrid corn: An exploration in the economics of technical change, Econometrica, 25:501–522.
Griliches, Z., ed., 1984, R&D, Patents, and Productivity, The University of Chicago Press, Chicago, IL, USA.
Grübler, A., 1998, Technology and Global Change, Cambridge University Press, Cambridge, UK.
Industrial Technology Council of MITI, 1979, Strategy to Accelerate the Sunshine Program, Tokyo, Japan.
Jaffe, A.B., 1986, Technological opportunity and spillovers of R&D: Evidence from firm's patents, profits, and market value, The American Economic Review, 76(5):984–1001.
Jovanovic, B., and Lach, S., 1989, Entry, exit, and diffusion with learning by doing, American Economic Review, 79:690–699.
Mansfield, E., 1961, Technical change and the rate of imitation, Econometrica, 29:741–766.
Meyer-Krahmer, F., 1992, The German R&D system in transition, Research Policy, 21(5):423–436.
MITI, 1970–1990, Annual Report on MITI's Policy, Ministry of International Trade and Industry, Tokyo, Japan.
Thomson, R., 1993, Learning and Technological Change, St. Martin's Press, New York, NY, USA.
Watanabe, C., 1992, Trends in the substitution of production factors to technology, Research Policy, 21(6):481–505.
Watanabe, C., 1995a, The feedback loop between technology and economic development: An examination of Japanese industry, Technological Forecasting and Social Change, 49(2):127–145.
Watanabe, C., 1995b, Identification of the role of renewable energy: A view from Japan's challenge, Renewable Energy, 6(3):237–274.
Watanabe, C., 1995c, The Interaction between Technology and Economy: National Strategies for Constrained Economic Environments: The Case of Japan 1955-1992,
Watanabe, C. 1995d, Mitigating global warming by substituting technology for energy: MITI's efforts and new approach, Energy Policy, 23(4/5):447–461.
Watanabe, C., 1996, Choosing energy technologies: The Japanese approach, in Comparing Energy Technologies, OECD/IES, Paris, France.
Watanabe, C., 1997, A Technometric Approach to the Dynamic Mechanism of Technological Innovation, Abstract of the Annual Conference of the Japan Society for Science Policy and Research Management, Tsukuba, Japan.
Watanabe, C., 1999, Systems option for sustainable development: Effect and limit of MITI's efforts to substitute technology for energy, Research Policy, 28(7):719–749.
Watanabe, C., 2000, MITI's Policy as a System to Substitute Technology for Energy: Lessons, Limits and Perspective, Paper presented to a Joint Meeting of the Energy Modeling Forum, International Energy Agency and International Energy Workshop, Stanford, CA, USA.
Watanabe, C., and Clark, T., 1991, Inducing technological innovation in Japan, Journal of Scientific & Industrial Research, 50(10):771–785.
Watanabe, C., and Griffy-Brown, C., 1999, Inter-firm Technology Spillover and the Creation of a “Virtuous Cycle” between R&D, Market Growth and Price Reduction: The Case of Photovoltaic Power Generation (PV) Development in Japan, Paper presented at the International Workshop on Induced Technological Change and the Environment, International Institute for Applied Systems Analysis, Laxenburg, Austria.
Watanabe, C., and Honda. Y., 1992, Inducing power of Japanese technological innovation, mechanism of Japan's industrial science and technology policy, Japan and the World Economy, 3(4):357–390.
Watanabe, C., Santoso, I., and Widayanti, T., 1991, The Inducing Power of Japanese Technological Innovation, Pinter Publishers, London, UK.
Watanabe, C., Wakabayashi, K., and Miyazawa, T., 2000, Industrial dynamism and the creation of a “virtuous cycle” between R&D, market growth and price reduction, Technovation, 20(6):299–312.
Williams, R., Karekezi, S., Parikh, J., and Watanabe, C., 1996, The Outlook for Renewable Energy Technologies, Public Policy Issues, and Roles for Global Environment Facility, STAP/GEF, Nairobi, Kenya.