An Analysis Of Cooperative R&d In The United States

Technovation, 16(3) (1996) 123-131 Copyright© 1996 ElsevierScienceLtd Printed in Great Britain. All rights reserved 0166-4972/96 $15.00 + 0.00

ELSEVIER ADVANCED TECHNOLOGY

An analysis of cooperative R&D in the United States Yukio Miyata Department of Commerce and Economics, Osaka University of Commerce, 4-1-10, Mikuriya Sakae-chou, Higashi Osaka-shi, Osaka-fu 577, Japan

Abstract This paper investigates the willingness of US firms to participate in cooperative R&D by cross-industry data and a case study of SEMATECH. The cross-industry analysis results indicate that there is no strong relationship between cooperative R&D and R&D diversification; however, there are strong positive relationships between cooperative R&D and industry's R&D resource costs, the necessity of basic research and standardization, both of which have a spillover problem, and the existence of foreign competitors. SEMATECH, which was organized by major US semiconductor manufacturers, was the first federally subsidized cooperative R&D project. Sharing R&D costs, competing with Japanese firms, and mitigation of spillover problems are important reasons for firms to participate in SEMATECH, agreeing with the finding from the cross-industry study. SEMATECH helped the United States to develop semiconductor processing technology in parity with the Japanese. However, as the threat of Japanese competitors declined, US firms decided to scale down SEMATECH rather than leapfrog the Japanese technology. Copyright © 1996 Elsevier Science Ltd

1. INTRODUCTION Cooperative research and development (R&D) is an agreement through which two or more firms jointly acquire technical knowledge. In the United States, the National Cooperative Research Act (hereafter NCRA) of 1984 relaxed antitrust enforcement for cooperative R&D. In 1987, the US government started subsidizing the private sector's cooperative R&D - - this was a major departure from traditional laissez-faire policy, moving toward Japanese-style industrial policy. The purpose of this paper is to analyze the willingness of firms to participate in cooperative R&D in the United States by a cross-industry study and a case study. Although the performance of cooperative R&D is an interesting topic, it is difficult to measure this performance. This study therefore focuses on participation in cooperative R&D. The cross-industry empirical study uses data of cooperative R&D notified under the NCRA of 1984; therefore, the study pro-

01664972(95)00033-X

vides us with the initial impact of the deregulation by the NCRA of 1984. In the next section, a brief review of the background of the National Cooperative Research Act is given. Then the hypotheses regarding the willingness of firms to participate in cooperative R&D are discussed in Section 3. After explaining the data of the cross-industry empirical study in Section 4, I give the results of a regression analysis in Section 5. In Section 6 I discuss whether or not these same hypotheses can explain the motivations of firms to participate in the industry-wide R&D consortium SEMATECH which was established in 1987 as the first federally subsidized consortium. Finally, I give some conclusions.

2. THENATIONALC00PERA1]VERESEARCHACTOF 1984 Cooperative R&D is expected to promote innovations by sharing costs and technological knowledge

TedmvalknVoL16 No.3 123

Y. Miyata

among firms, and thus to increase social economic efficiency. However, cooperative R&D faces management difficulties because it has to manage several firms which differ in corporate culture and strategies. Moreover, each firm wants to be a free-rider, contributing less but trying to obtain the fruits of R&D done by other participants. If the management difficulties are large, cooperative R&D cannot innovate new technologies as fast as expected. Because of the following three negative aspects, the net effect of cooperative R&D on social economic efficiency may become negative and thus conflict with antitrust laws. First, the decrease in competition pressure may slow down innovation. In addition, if an innovation such as pollution control is socially desirable but privately unprofitable, participants of cooperative R&D may intentionally delay the innovation. Second, cooperative R&D may result in collusion in the output market. Even if cooperative R&D is only organized for R&D, cooperation may provide participants with the opportunity to exchange information about production price and quantity of new or even existing products. Third, cooperative R&D's ancillary restraints may be used as restraints of commerce or business. Because participants do not want their R&D knowledge or the results of R&D to diffuse to non-participants, cooperative R&D may have to restrict business of participants with non-participants. Without this kind of restriction, participants would hesitate to share knowledge with each other. If, however, such a restriction is severe, it hinders business between participants and non-participants. Prior to 1983, there were only three government cases and one private case against cooperative R&D. 1 Although the number of antitrust cases against cooperative R&D was small, if lawsuits occurred, the courts would use these cases as precedents against cooperative R&D. In a private treble damage lawsuit, a plaintiff can claim three times as large as the actual damage caused by a defendant; firms may therefore have been deterred from organizing a cooperative R&D project, particularly an innovative project, because such a project would cause greater disadvantages to non-participants and thus the firm may have had to pay a greater amount of damage if it lost an antitrust treble damage lawsuit. Even if defendants (cooperative R&D participants) won their cases, participants could not recover their litigation costs, in particular costly attorney's fees. Furthermore, participants may have had to reveal their research activities to show that the cooperative R&D would not be Ewing [16] and the author's search from CCH Trade Cases [17]. Three government cases were: US v. Automobile Manufacturing Association et al. (1969), US v. Wisconsin Alumni Research Foundation (1969), and US v. Manufacturing Aircraft Association (1972). The only private case was Berkey Photo v. Eastman Kodak (1979).

1~4 THhnovationVol.16 No.3

anti-competitive. Non-participants could, therefore, disturb the progress of cooperative R&D by suing the participants. In 1980, the last year of the Carter administration, the Department of Justice (Do J) issued its Antitrust Guide Concerning Research Joint Ventures [1] (hereafter the Guide) claiming that the DoJ was not hostile to cooperative R&D. However, in order to preserve the DoJ's flexibility for enforcement, the Guide has to be ambiguous and cannot completely eliminate the uncertainty of antitrust enforcement. In its preface, the Guide admits the enforcement policy may change over time. Even if the current Attorney General is generous toward cooperative R&D, the next Attorney General may strictly enforce the antitrust laws against it. With this uncertainty, firms may still have to hesitate before starting a cooperative R&D project, particularly a long-term project. Therefore, a new law, which clearly requires antitrust enforcement to treat cooperative R&D differently from other horizontal coordinations, was necessary. The National Cooperative Research Act of 1984 deregulated antitrust law enforcement for cooperative R&D in the following ways: •

•

•

The court should use the rule of reason principle, in which the net effect of conduct on economic welfare should be considered, rather than the per se illegal principle, in which the conduct itself is illegal. Private lawsuits can seek single (actual) damages rather than treble damages as long as the cooperative R&D notifies the DoJ and the Federal Trade Commission (FTC) within 90 days of its establishment. Successful defendants as well as successful plaintiffs can receive attorney's fees from the losing side when the court judges that suits are frivolous, unreasonable, or in bad faith.

Treble damages and awards of attorney's fees to only successful plaintiffs had originally been designed so that even small firms could sue large firms in antitrust lawsuits. Because the government cannot find every antitrust violation, private lawsuits were expected to complement government enforcement of antitrust laws. These schemes were, however, increasingly criticized as a reason for rent-seeking frivolous lawsuits. It was a significant change in the US antitrust history when the National Cooperative Research Act mitigated these private lawsuit incentives, although the NCRA was applied only to cooperative R&D cases. Because the punishment was reduced from treble damages to single damages by the NCRA, the antitrust violations had to be easily detected to keep a balance. This is why the NCRA requires the notification of cooperative R&D to obtain exemption from treble damages.

An analysisof cooperaUve R&D in Ute United States

3.

Ht OTHESES

Even if the legal uncertainty has significantly declined, firms organize cooperative R&D only if the expected profit from cooperative R&D is greater than that from in-house or non-cooperative R&D. In cooperative R&D, firms can share the costs of R&D but they also have to share the results. As a result, a critical point is whether or not cooperative R&D can be of higher quality than in-house R&D, which means that the same amount of spending innovates new technologies more frequently in cooperative R&D than in in-house R&D. From this viewpoint, I can derive the following hypotheses regarding the factors promoting firms' participation in cooperative R&D.

(H1) R&D diversification promotes cooperative R&D. Diversification (entering a new industry or starting a new product line) requires fresh knowledge and a new approach to R&D. However, because it may be costly for a finn to acquire such a new knowledge by itself, it may want to cooperate with firms of different industries. Through inter-industry cooperative R&D, the complementarity of expertise is fully exploited, so cooperative R&D will have a higher probability of success than in-house R&D. As a result, finns are willing to organize cooperative R&D when they need to diversify.

(H2) Costly R&D resources promote cooperative R&D. If R&D resource cost - - the cost of experimental equipment or research personnel salary - - is high, individual firms cannot easily finance such resources. In cooperative R&D, however, participant firms share costs, so that they can use expensive equipment or hire high quality research personnel. As a result, cooperative R&D can have a higher probability of successful innovations than in-house R&D, and finns are willing to organize cooperative R&D when resources are costly. Of course, an individual firm can purchase costly R&D resources by using the capital market. However, this market is imperfect, with asymmetric information. A firm knows the probability of success of a R&D project better than investors. Investors, therefore, do not trust the firm even if the firm says that the project is promising. As a result, it is difficult for firms to obtain the necessary capital to finance R&D resources from the capital market. Even if the participants of cooperative R&D do not purchase a single expensive equipment, cooperative R&D can arange for each participant to use different research equipment. As a result, cooperative R&D can maintain the diversity of research paths or strategies, increasing the probability of success for all the participants. On the other hand, under competitive R&D, individual firms tend to choose a single research path

which appears to be the most promising and may not be able to try other methods because of the cost of the R&D resources. As firms may choose the same research path, the diversity of the research path cannot be maintained and the probability of success will decrease [2]. If, therefore, R&D resources are costly, firms are willing to organize cooperative R&D in order to coordinate their in-house R&D and to keep the diversity of the research path.

(H3) The necessity of basic research promotes cooperative R&D. Because patents cannot effectively protect the results of basic research and imitation is easy, basic research has a large spillover effect. Due to this low appropriability, individual firms may hesitate to invest in basic research. Cooperative R&D, however, internalizes potential imitators as members who pay the R&D cost from the beginning. Therefore, when firms have to conduct basic research, they are willing to organize cooperative R&D. However, as the number of participants increases, the management difficulties increase. Therefore, it is impossible for cooperative R&D to include all of the potential imitators as its members, so there are always firms which do not join cooperative R&D but can imitate its results. Hence, cooperative R&D cannot be the perfect solution for the spillover effect of basic research. In addition, managing basic research itself encounters management difficulties, because it is difficult for all participants to agree with the research strategy. In spite of these problems, as a hypothesis, I assume that when firms need to conduct basic research they will organize cooperative R&D.

(H4) The necessity of standardization R&D promotes cooperative R&D. Like basic research, standardization R&D has a spillover problem, because the benefit of standardization effort can be enjoyed by firms which do not pay the R&D cost. Therefore, if firms need to conduct standardization R&D, they are willing to organize cooperative R&D to internalize potential beneficiaries as members. However, like basic research, because of management difficulties which increase with the number of participants, it is impossible to include all of the potential beneficiaries. Cooperative R&D cannot be the perfect solution to the spillover problem of standardization R&D. Compared with basic research, however, standardization R&D tends to have a clear research goal and strategy, so managing cooperative R&D for standardization may be easier than cooperative basic research.

(H5) The threat of foreign competitors promotes cooperative R&D. Where there exist strong foreign competitors, domestic firms are willing to organize cooperative

TedmovatwVd. 16 No.3

Y. Miyata

R&D to compete with these competitors. Whether or not cooperative R&D can successfully generate a new technology against foreign competitors is critical to the participants' profits. Under such conditions, participants are expected to be less opportunistic and thus cooperative R&D can have a high probability of successful innovations, so firms are more willing to organize it.

4.

DATAOF THE CROSS-INDUSTRYEMPImCAL ANALYSIS 4.1 Depwdentvari~le

The dependent variable, COPARTk, is the number of cooperative R&D participants per 1000 firms in industry k. I have categorized the participants of cooperative R&D listed in the Federal Register under the National Cooperative Research Act of 1984 between 1985 and 1987 into 25 manufacturing sectors, which is the National Science Foundation's (NSF) categorization. Data covers 66 cooperative R&D projects and 385 US manufacturers. Then I have divided the number of participants by the number of firms in industry k which conduct R&D; this number is obtained from the National Science Foundation [3]. I have pooled cooperative R&D projects over a threeyear period because firms may have delayed starting projects for one or two years to observe whether or not the US antitrust enforcement actually tolerated cooperative R&D after enactment of the NCRA of 1984. A possible problem of using cooperative R&D projects listed in the Federal Register is that they notified the DoJ and FTC (and thus were listed in the Federal Register) only if they wanted to obtain treble damages exemption. As a result, the cooperative R&D projects listed are the ones which, so the participants thought, have a large potential for antitrust lawsuits. The Federal Register does not disclose much about the actual R&D plan in cooperative R&D but non-participants at least can find the initiation of cooperative R&D and its members and purpose. Firms wanting their initiation of cooperative R&D to remain secret would not have to notify the DoJ and FTC as long as the firms accepted the possibility of treble damage lawsuits. Therefore, some cooperative R&D projects, those consisting of a small number of small firms belonging to different industries, were probably overlooked in my data, because they had a small possibility of treble damage lawsuits and participants might not notify the DoJ and FTC. On the other hand, an advantage of using data from the Federal Register is to mitigate the reverse causality problem whereby the formation of cooperative R&D affects independent variables such as basic research or export. Because cooperative R&D projects listed in the Federal Register were difficult, if not

126 Teclmeva~Yd. 15No.3

impossible, to form without the National Cooperative Research Act of 1984, the number of these cooperative R&D projects would be small before 1985. It would be unlikely that similar cooperative R&D projects existed before 1985 and influenced the independent variables in that period. Therefore, if I use independent variables from the pre-1985 period, the reverse causality problem can be mitigated.

4.2 Independentvadables I use the values of independent variables between 1980 and 1984 (with the single exception of the extent of diversification of cooperative R&D members) to explain the dependent variable, the number of cooperative R&D projects, between 1985 and 1987. I assume that the economic conditions of the early 1980s influenced firms' decisions to participate in cooperative R&D once deregulation occurred at the beginning of 1985. I measure the extent of R&D diversification by the diversification of cooperative R&D participants. I define the variable SHARE in the following way. Suppose a cooperative R&D project consists of four firms: two from the aerospace industry, one from the computer industry and one from the automobile industry; I would give 50 points to each of the two aerospace firms, 25 points to the computer finn, and 25 points to the automobile firm. I have calculated SHAREk as the average of the SHARE points of firms belonging to industry k. A small SHAREk indicates that firms of industry k tend to cooperate with firms from other industries. Because of the availability of data for other variables, I have to use rather coarse (25 industries) NSF industry categorization; however, if firms were categorized as two different industries in this coarse categorization, they would really differ in products. Cooperative R&D between firms from different industries was probably organized to exploit the complementarity of their R&D expertise. Hypothesis H1 suggests that the coefficient on SHAREk should be negative, because firms wanting to cooperate with firms from different industries should participate in more cooperative R&D. Hypothesis H2 states that as R&D resources become costly, firms are more willing to participate in cooperative R&D because sharing expensive R&D resources is expected to have a higher probability of successful innovation. R&D resource costs (RDCSTD are measured by the industry k's R&D spending per R&D personnel (the average value for 1980-84), which is obtained from the NSF [4]. RDCSTk approximates the cost of equipment or personnel salary, both of which are a measurement of R&D resource costs. The NSF data do not include federal government contract research funds. The exclusion is appropriate for my study because federal contract research makes RDCST appear to be larger than the actual industrial R&D resource costs. Because of the exclusion of federal R&D funds, however, communi-

An analysisof cooperat e R&D in Ule UnitedStates

cation equipment, electric devices and aerospace industries, which receive a great amount of federal (military) research funds, have small values of RDCST even though they are high-tech industries. Because the NSF [4] separately lists federal research funds to these three industries, I make a separate regression analysis in which federal funds are included in the~RDCST of these three industries and a dummy variable is given to them. The coefficient on RDCSTk should be positive because, as R&D resources become more costly, firms are more willing to join in cooperative R&D. Hypothesis H3 states that a firm joins cooperative R&D to internalize potential imitators as members sharing the cost of basic research. The industry's interest in basic research, BASICk, is obtained from a survey conducted by Yale University during the early 1980s. The survey covered 630 R&D managers in the United States, and the respondents scaled the importance of basic research from one (lowest) to seven (highest). As industry k requires more basic research, which has a large spillover, firms of industry k are more willing to organize cooperative R&D to internalize potential imitators. Therefore, the coefficient on BASICk should be positive, although, as I mentioned before, cooperative R&D may not be ideal for basic research because of the large management difficulties associated with basic research. Hypothesis H4 states a positive relationship between the necessity of R&D for standardization and cooperative R&D. R&D for standardization was conducted by industry associations even before 1984. But non-participants whose products would be different from standardized products could have sued these standardization efforts in treble damage lawsuits. Cooperative R&D for standardization is, therefore, expected to be a reason for cooperative R&D under the National Cooperative Research Act of 1984. The importance of R&D for standardization in industry k, STAND~, is obtained from the Yale survey. The coefficient should be positive. To test Hypothesis H5, the existence of foreign competitors is measured by import/output and export/output ratios (the average values for 1980-84), denoted IMPORTk and EXPORTk respectively. IMPORTk measures the existence of foreign competitors in the US market, and EXPORTk measures the opportunity of US firms facing foreign competitors in the world market. The ratios are calculated from the Department of Commerce's US Industrial Outlook [5]. Although I categorized only US firms into 25 industries, about a quarter of cooperative R&D projects I covered included at least one foreign firm or foreign subsidiary. As US firms face more foreign competitors in the US market or in the world market, they may cooperate with other US firms against foreign firms, or US firms may cooperate with foreign firms. In either case, the predicted signs for the coefficients on IMPORTk and EXPORTk are positive. The overall data are presented in Table 1.

5.

RESULTS

The results are summarized in Table 2. In order to solve a heteroscedasticity problem, the weighted least square is used. The coefficients on SHARE~ have unexpected positive signs. Diversification is not, therefore, a strong motive for cooperative R&D in the United States. This might be because, as I mentioned, the Federal Register might overlook cooperative R&D projects consisting of small firms from different industries which have a small possibility of treble damage lawsuits. On the other hand, US firms may not really be interested in using cooperative R&D as a strategy for diversification. When these firms need the technological expertise of other industries for diversification, they would probably choose merger and acquisition of the firm which has the expertise, or directly hire research personnel with the expertise through the labor market. 2 In order to be successful in a new business field, a firm needs to obtain expertise through hiring new research personnel or by merger and acquisition rather than just borrowing it from other firms through cooperative R&D. The coefficients o n RDCST k are significantly positive. When federal funds are included in three hightech industries with dummy variables, the t-score reduces but it is still significant at the one-tailed 10% level, suggesting that federal funds overstate the burden of R&D resource costs on firms) The coefficients on BASICk and STANDk have expected positive signs and are significant. When firms need to mitigate the spillover problem associated with basic research or standardization R&D, they are willing to organize cooperative R&D. Because my cooperative R&D data tend to consist of a large number of major firms in the same industry, cooperative R&D might be organized to internalize potential imitators or beneficiaries. The coefficients on IMPORTk and EXPORTk have the expected positive sign and are significant. When US firms face foreign competitors in either the US market or the word market, they cooperate with other US firms as well as foreign firms in R&D.

2 In 1988, the Japanese Ministry of International Trade and Industry (MITI) conducted a survey regarding corporate strategies for obtaining a new technology, receiving responses from 56 US firms and 202 Japanese firms. According to this survey, in both countries, finns listed in-house R&D as the most important strategy. Japanese finns ranked cooperative R&D higher than mergers and acquisitions of finns which had a desired technology, but the US finns ranked these two strategies the other way round [18]. 3 When I used the growth rate instead of the average value of R&D spending per research personnel between 1980 and 1984, the coefficient on the growth rate of R&D resource costs became insignificant without affecting the estimates of other variables. I also used R&D/sales ratio or R&D/sales/research personnel instead of RDCSTk, but the coefficient became insignificant without affecting the estimates of other variables.

TechnovationVd. 16 No. 3 ]27

Y. Miyata

TABLE 1.

US cross-industry data

Industry

SIC

food/tobacco textiles/apparel lumber/furuiture paper industrial chemicals drugs other chemicals petroleum refining rubber clay/stone/ceramics steel/iron non-ferrous metals fabricated metals computers general machinery radio/TV communication equipment electric devices other electric equipment automobile and parts aerospace other transportation equipment measurement instruments medical instruments other manufacturing* mean standard deviation

20XX, 21XX 22XX, 23XX 24XX, 25XX 26XX 281, 2, 6 283X 284, 5, 7, 9 291X 30XX 32XX 331X, 332X 333, 5, 6 34XX 357X 35XX 365X 366X 367X 36XX 371X 372X, 376X 37XX 381 X, 382X 383X-7X 27, 31, 39

COPART

SHARE

STAND

RDCST

BASIC

IMPORT

EXPORT

6.18 1.46 10.38 7.45 54.55 25.64 23.03 95.72 0 25.45 19.94 21.05 6.91 75 39.11 68.97 11.8 73.12 20.65 66.25 13.63 0 58.58 19.42 5.97 30.01 27.98

23.63 3 37.33 11 40.88 55.5 28.71 70.51 0 59.05 92.86 53.25 51.88 35.97 40.59 7.25 21.25 28.62 17.23 67.82 24.34 0 13 6.94 7.25 31.91 24.85

3.92 3.95 4.7 3.38 3.35 3.29 3.42 4.56 3.35 4 3.32 4.07 4.1 4.88 4.09 4.8 4.47 4.58 4.19 4.21 4.05 4.18 4.15 4.21 3.82 4.04 0.48

104.7 63.8 93.9 79 108.6 96.7 86.4 139.3 70.1 83.1 79.2 90.7 61.8 84.4 85.9 60.9 77.9 58.9 60.4 146.2 36.8 70 66.4 106.7 83.8 83.82 24.51

4.67 3.88 3.57 4.24 4.26 5.55 4.5 4.16 4.41 3.88 3.73 3.92 3.49 3.93 3.25 3.74 3.75 4.75 3.87 3.45 3.54 3.49 4.23 4.29 3.74 4.01 0.51

5.49 11.41 9.04 7.35 6.59 4.81 2.49 8.72 5.77 1.61 15.74 21.56 6.05 8.29 9.86 52.40 6.44 17.91 7.7 16.66 5.68 3.34 8.51 9.49 8.76 10.47 9.96

2.87 3.2 6.27 5.41 15.77 9.48 8.01 2.24 4.96 1.19 2.95 6.66 5.46 26.17 21.61 6.88 6.95 16.57 9.06 7.9 23.12 8.61 24.18 12.54 2.73 9.63 7.39

*Includes printing and leathers. SIC: Standard Industrial Classification. COPART: the number of cooperative R&Ds per 1000 firms. SHARE: the extent of firms from the same industry joining a cooperative R&D (points). STAND: industry's emphasis on standardization R&D (scale 1-7). RDCST: annual R&D spending per R&D personnel (thousand dollars). BASIC: industry's emphasis on basic science research (scale 1-7). IMPORT: import/output ratio (%). EXPORT: export/output ratio (%).

TABLE 2.

Regression results

Variable

Expected sign

constant dummy SHARE STAND RDCST BASIC IMPORT EXPORT Adjusted R2

Regression (1)

-198.35 (-4.888) (-) (+) (+) (+) (+) (+)

0.264 (2.091) 20.582 (2.742)*** 0.364 (2.438)** 19.605 (3.223)*** 1.116 (4.116)*** 1.646 (4.234)*** 0.762

Regression (2)

-219.06 (-3.768) -5.094 (-1.607) 0.423 (2.651) 24.360 (2.590)*** 0.234 (1.436)* 23.149 (2.756)*** 1.017 (2.656)*** 1.552 (3.061)*** 0.603

t-ratios in parentheses. *one-tailed 10% level significance. **one-tailed 5% level significance. ***one-tailed 1% level significance. In regression (1), both sides of the equation were divided by RDCST to fix heteroscedasticity. In regression (2), both sides of the equation were divided by BASIC to fix heteroscedasticity. Federal funds were included in RDCST of communication equipment, electric devices and aerospace industries with dummy variable one.

In summary, the hypotheses are supported, with the single exception of hypothesis HI.

128 Tedunova~onVol.16 No.3

6. SUPPORTIVEEVIDENCEFROMTHECASESTUDYOF SEMATECH In the early 1980s, the US semiconductor industry lost its competitiveness against the Japanese, particularly in the memory chip market. In 1987, 14 semiconductor firms organized a cooperative R&D project for semiconductor processing and manufacturing technology. 4 SEMATECH (SEmiconductor MAnufacturing TECHnology) started operation in 1988 in Austin, Texas, while receiving subsidies from the Department of Defense (DoD) which expressed its concern that the decline of the US semiconductor industry would cause a serious national security problem. By 1994, SEMATECH achieved technological parity with Japan [6, 7] and contributed to the revival

4 The fourteen members were: Advanced Micro Devices Inc., A T & T Co., Digital Equipment Corp., Harris Corp., Hewlett Packard Co., Intel Corp., IBM Corp., LSI Logic Corp., Micron Technology Inc., Motorola Inc., National Semiconductor Corp., NCR Corp., Rockwell International Corp, and Texas Instruments, Inc. The members included the captive makers, which produced semiconductors for their own use, as well as the merchant makers, which produced semiconductors for sales.

An analysisof cooperaUveR&D in the UnitedStates

of the US semiconductor manufacturing equipment industry. According to an estimate by VLSI Research Inc. [8], without SEMATECH, US equipment manufacturers would have lost another 25 points of their world market share. As the complement to crossindustry regression analysis, this section analyzes whether or not SEMATECH has satisfied the conditions for the initiation of cooperative R&D by using the same hypotheses as the regression analysis. Although SEMATECH consists of semiconductor manufacturers and does not intend to diversify in R&D, its members have cooperated with semiconductor manufacturing equipment makers through contract research. SEMATECH was expected to provide the opportunity for cooperation and information exchange between equipment makers and users (semiconductor chip manufacturers). Exploitation of the complementarity of R&D expertise through vertical cooperation was an important reason for firms to participate in SEMATECH. The annual budget of SEMATECH was $200M, half of which was a subsidy from the DoD. Each member contributed $1M or 1% of its annual semiconductor sales, whichever was greater. Because SEMATECH research, which reduced production costs of semiconductor manufacturing, would benefit most a firm with large semiconductor sales, it was reasonable that contributions were proportional to sales. A minimum contribution was set because SEMATECH expected that as long as investment was significant, members had to be serious about SEMATECH research [9]. Because participant firms spent 8.75% of their sales on R&D in 1988 [10], the contribution to SEMATECH would account for about 11% of a firm's R&D. Investment in SEMATECH was large enough to force each participant to pay attention to the performance of SEMATECH research. SEMATECH also restricted a single firm's contribution so as not to exceed 15% of the total members' contribution ($100M) so that one big contributor would not control SEMATECH. Even if a large firm such as IBM or Texas Instruments contributed this upper limit, it could still access the research which would have cost over twelve times the amount of its contribution. In addition, most of the participants' in-house R&D was for product, not for processing technology. It was estimated that, in the mid-1980s, semiconductor manufacturers annually spent $200-300M on processing technology R&D and semiconductor equipment manufacturers spent another $500M likewise [ 11]. Therefore, the SEMATECH budget ($200M/year including $100M subsidies) would be quantitatively significant relative to the entire US funding for processing technology R&D. Cost sharing, as well as receiving subsidies, was an important factor for even large firms to participate in SEMATECH. SEMATECH research was not basic research but

development research; it involved processing technology for semiconductor chips, although SEMATECH did not include actual joint production of the chips. Because the R&D topic was development research, the technological goal and strategy were clear: the use of short-wavelength (ultraviolet) light for lithography in order to achieve a 0.35 ~m line width by the end of 1992. The new processing technology was critical for the 14 semiconductor manufacturers but was not sensitive to their marketing strategies or their factory-level know-how. It was difficult for semiconductor manufacturers to keep a new processing technology secret because such a technology could be diffused by equipment manufacturers [12]. Even though SEMATECH research was not pure basic research, it had low appropriability; SEMATECH members were, therefore, willing to conduct such research in cooperative R&D rather than as in-house R&D. In addition, SEMATECH had an aspect of R&D for standardization which also had a spillover problem. The fourteen semiconductor manufacturers of SEMATECH, which accounted for 80% of US semiconductor manufacturing, actually specified standardized qualification of equipment so that an equipment maker did not have to adjust equipment for each user [8]. SEMATECH members had incentives to support standardization R&D which they were reluctant to conduct individually. Finally, the existence of foreign (Japanese) competitors was a strong motive for SEMATECH. As well as the US semiconductor industry, the world market share of the US semiconductor equipment industry declined in the mid-1980s. In particular, the Japanese manufacturers Nikon and Canon came to dominate as manufacturers of the lithography machine, the most critical part of semiconductor processing. US semiconductor manufacturers grew concerned that Japanese semiconductor manufacturing equipment makers would not supply state-of-the-art equipment to US semiconductor manufacturers as fast as to their Japanese counterparts, thus disadvantaging the US manufacturers,s Similarly, Japanese semiconductor manufacturers may not supply state-of-the-art semiconductor chips to US computer manufacturers as fast as to their Japanese customers. US firms which wanted to protect the 'electronics food chain' of semiconductor chip user-chip manufacturer-manufacturing equipment maker from Japanese competition were willing to join SEMATECH. SEMATECH satisfied the hypotheses of this paper. SEMATECH members were strongly interested in participating in it, although the DoD played an important role in its initiation. SEMATECH members

5 Tyson [19] points out that the delays in the shipment of Japanese state-of-the-art equipment to US users have actually happened.

Tedmovat~ Vol.16 No. 3

Y. Miyata

have been willing to send their high quality research personnel to the SEMATECH facility [13], achieving the original technical goal. As the US semiconductor industry and equipment manufacturing industry have regained their competitiveness against the Japanese, in October 1994 SEMATECH decided not to accept federal subsidies after fiscal 1996 [14]. SEMATECH research will then be half of what it is today. When the threat of foreign competitors declined, finns lost interest in industrywide cooperative R&D.

7. CONCLUSIONS This paper has studied cooperative R&D among US finns. The National Cooperative Research Act of 1984 deregulated antitrust enforcement for cooperative R&D. Since then, US firms have been able to use cooperative R&D as a business strategy with little concern of antitrust lawsuits. Using cooperative R&D data for the period after the NCRA of 1984, this paper has tested hypotheses regarding which industries are more willing to organize cooperative R&D. The results show that there is no strong relationship between cooperative R&D and R&D diversification. If a US firm needs to obtain a new technology, it can purchase a firm which has such a technology or, through the labor market, hire research personnel who have the knowledge. Diversification may, therefore, not be a strong motivation for cooperative R&D, thus not supporting my first hypothesis. Industries facing costly R&D resources, industries emphasizing basic research and standardization R&D, both of which have the spillover problem, and industries with a large export/output or import/output ratio tend to be more willing to organize cooperative R&D, supporting the rest of my hypotheses. The fourteen major US semiconductor firms organized a cooperative R&D project called SEMATECH. These finns had a strong motive for participating in SEMATECH, satisfying the hypotheses regarding willingness to participate in cooperative R&D. As the cross-industry regression results indicate, SEMATECH was not organized for participants' diversification. However, vertical cooperation between semiconductor manufacturers and manufacturing equipment makers exploited the complementarity of R&D expertise. Cost sharing was an important reason for the finns to participate in SEMATECH. This agreed with the cross-industry regression result. Although SEMATECH did not conduct pure basic research, SEMATECH research - - semiconductor processing technology - - had a spillover problem for semiconductor manufacturers. The SEMATECH participants wanted to conduct such research in cooperative R&D rather than as in-house R&D. The choice of R&D topics, which were far from pure basic research, actually

130 TechnevationVoi.16 No.3

made cooperation smooth because research goals and strategies became unambiguous. SEMATECH research also had an aspect of standardization R&D which had a low appropriability. Finally, the existence of foreign competitors was a strong motive for the fourteen finns to participate in SEMATECH; this agrees with the results of the cross-industry regression analysis. An interesting future research topic would be the relationship between basic research and cooperative R&D. The cross-industry regression analysis supports a positive relationship between the necessity of basic research and cooperative R&D, while the SEMATECH case indicates that the avoidance of basic research made cooperation easy. However, the Federal Register did not tell much whether cooperative R&D actually conducts basic research or development research. Industries which emphasize basic research may organize many cooperative R&D projects for applied or development research to save money for critical basic in-house R&D. More case studies would be necessary to investigate what kind of research topics cooperative R&D is conducting. The National Cooperative Research Act of 1984 did not exempt joint production from treble damage lawsuits, but the National Cooperative Production Amendment of 1993 has extended the deregulation to joint production. Commercialization of new technology is as costly as R&D, and feedback between the stages of basic research, applied research, development, production and marketing is critical for innovations [15]. In addition, if firms do not have to compete in commercializing the results of cooperative R&D, no firm will try to be a free-rider which does not disclose its expertise but tries to steal knowledge from other participants, and cooperation at the R&D stage would be smooth. However, finns do not want to cooperate with each other in a field which is too close to their marketing strategies or factory-level know-how. SEMATECH was successful because its R&D topic found a niche: it was critical for every participant, but it was not basic research, which often has vague research goals and strategies, and it was not close to factory-level know-how. The impact of the 1993 deregulation on cooperative R&D would be another interesting research topic.

ACKNOWLEDGMENTS I would like to thank Michele Blood of SEMATECH who patiently answered my questions about this project. I would like to thank Frederic Scherer at the Kennedy School of Government, Harvard University, for providing me with a portion of Yale University's survey. I would like to thank Arthur Denzau at the Department of Economics, Claremont Graduate School, and Mary Olson at the Department of Economics, Washington University in St Louis, for their comments. I also thank Claire Cuccio for editing my English.

An analysisof cooperaUveR&D in the UnitedStates

REFERENCES 1 US DoJ (Department of Justice). Antitrust Guide Concerning Research Joint Ventures. US Government Printing Office, Washington, DC, 1980. 2 A. Goto. Innovation and Industrial Organization in Japan (in Japanese). University of Tokyo Press, Tokyo, 1993, p. 102. 3 National Science Foundation. Research and

4

5

6 7

8

9

10 11

12

Development Resources in US Industry: 1987. NSF89-323. US Government Printing Office, Washington, DC, 1989. National Science Foundation. Science and Technology Resources in US Industry. NSF88-321. US Government Printing Office, Washington, DC, 1989. US DoC (Department of Commerce). US Industrial Outlook. US Government Printing Office, Washington, DC, various years. J. Carey and O. Port. One stepper forward for Sematech. Business Week (8 June 1992) 110-111. US GAO (General Accounting Office). SEMATECH's Technological Progress and Proposed R&D Program. GAO/RCED-92-223BR. US Government Printing Office, Washington, DC, 1992. W.J. Spencer and P. Grindley. SEMATECH after five years: High technology consortia and US competitiveness. California Management Review, 35(4) (September 1993) 9-32. US Congressional Hearing. Semiconductors: The Role of Consortia. Subcommittee on Technology and Competitiveness, Committee on Science, Space, and Technology, House of Representatives, 102nd Congress, First Session, 23 July 1991. US Government Printing Office, Washington, DC, 1991, p. 145. Business Week. Innovation in America. Business Week, Special Issue (16 June 1989) 180-232. US CBO (Congressional Budget Office). The Benefits and Risks of Federal Funding for Sematech. US Government Printing Office, Washington, DC, 1987, pp. 43-44. W.G. Ouchi. The new joint R&D. Proceedings of the IEEE, 77(9) (September 1989) 1318-1326.

13 P. Grindley, D.C. Mowery and B. Silverman.

14

15

16

17

18

19

Sematech and Collaborative Research: Lessons in the Design of High-Technology Consortia. CCC (Consortium on Competitiveness and Cooperation) Working Paper No. 93-21. Haas School of Business, University of California, Berkeley, CA, 1994. S. McCartney. Sematech to stop federal funding for its research. The Wall Street Journal (5 October 1994) B-5. T.M. Jorde and D.J. Teece. Competition and cooperation: striking the right balance. California Management Review, 31(3): (Spring 1989) 25-37. K.P. Ewing, Jr. Joint research, antitrust, and innovation. Research Management (March 1981) 2529. Commerce Clearing House. CCH Trade Cases. Commerce Clearing House, Chicago, IL, various years. MITI (Ministry of International Trade and Industry). The Comparison of Corporate Strategies in Japan and the United States (in Japanese). Nihon Noritsu Kyoukai, Tokyo, 1989, pp. 96-97. L.D. Tyson. Who's Bashing Whom?: Trade Conflict in High-Technology Industries. Institute for International Economics, Washington, DC, 1992, pp. 146, 275.

Yukio Miyata is a lecturer at the Department of Commerce and Economics,Osaka Universityof Commerce,HigashiOsaka-shi, Japan. Afterobtaininghis BA degreein Economics from Osaka University, Japan, in 1983 and BS degree in Materials Engineering from the University of Washington, Seattle, USA, in 1987, he obtained an MS degree in Engineering and Policy in 1989 and a PhD degree in Economics in 1994, both from Washington University in St Louis, USA. Dr Miyata'scurrentresearch interestis economicand policyanalysis of technological innovations. He is particularly interested in Japanese and US goverment policies on cooperative research and developmentamong firms.

Technova'donYd. 16 No. 3