Assessing industrial policy to promote SME cooperative associations: motivations, activities, and impacts

Nobuya Fukugawa Tohoku University

6-6-11-804 Aramaki Aoba-ku Sendai 980-8579 Japan Graduate School of Engineering, Tohoku University [email protected]

Abstract The establishment and promotion of cooperative associations was one of the most prevalent SME (Small- and Medium-sized Enterprise) policies in postwar Japan. This study quantitatively evaluated industrial policy to promote cooperative associations in the period before the fundamental reform of SME policy in 1999. Estimation results of fixed-effect models using an industry-level panel (1966-1987) show that SMEs in industries where the dependence on government-affiliated financial institutions was high were engaged in finance-driven activities in SME cooperatives. SMEs in industries where the minimum efficient size was small joined production-driven activities in SME cooperatives. Estimation results of treatment regression models using a firm-level dataset (1992-1995) show that the participation in cooperative associations was negatively correlated with the SMEs' total factor productivity growth, even after controlling for unobservable firm-specific factors influential both in productivity and the participation in cooperative associations. The results imply adverse selection or the mismanagement of SME cooperative associations.

Keywords Industrial policy, cooperative associations, small firms, Japan

JEL D78, L52, O38

0 / 25

INTRODUCTION A cooperative of small and medium-sized enterprises (SMEs) was a intra-industry organization uniquely developed through industrial policy in postwar Japan. The economic aim of the policy was to improve the bargaining power of SMEs in the product market, to create economies of scale, and to mitigate the constraints encountered by SMEs in the capital market by making policy loans more accessible for SMEs. It is also said that the political aim underlying this policy was to establish a stable base for the conservative government, which faced serious conflict with the socialist party in the beginning of the Cold War. The recognition that SMEs were socially vulnerable and should be supported as such has been widely accepted among the public and policymakers, resulting in the passage of many laws aiming to support SMEs, including the promotion of cooperative associations, without much political controversy. Such recognition has also led to less attention to the quantitative evaluation of the SME policy. Although SME cooperatives were prevalent at least until the seventies, few quantitative evaluation of this industrial policy had so far been conducted.1

This study quantitatively evaluates industrial policy to promote SME cooperatives in the period before the fundamental reform of SME policy. Specifically, using an industry-level dataset of SME cooperatives (1966-1987), this study examines the determinants and types of collective activities that SME cooperatives were engaged, and using a firm-level dataset of SMEs (1992-1995), this study also examines the impacts of SME cooperatives on productivity growth of member firms. Foreshadowing the results, the collective activities in SME cooperatives were characterized by finance-driven and production-driven activities, and the participation in SME cooperatives, even after controlling for selection bias, negatively affected productivity growth of member firms. The remainder of the paper is organized as follows. Section 2 overviews the industrial policy to promote SME cooperative associations in Japan. Section 3 describes the empirical method and Section 4 shows empirical results. Section 5 discusses implications of the findings and refers to issues for future research.

SME COOPERATIVES IN JAPAN There were cartels among small-sized exporters in the Meiji era (1868-1912) that controlled price, prevented low-quality goods from being exported, and enforced sanctions against a member that broke 1

One exception is Okamuro (2009). 1 / 25

the cartel. Although initiated as voluntary associations, intraindustry associations among small firms were gradually institutionalized by the relevant laws before WWII. These cartels were dismantled during the occupation era (1945-1951) by the enactment of the antitrust law in 1947. However, they were restored under the increasingly intense Cold War, reflected in domestic political conflict between the conservative party and the socialist party. The establishment and promotion of SME cooperatives was put forward in 1947 by the conservative government, which was seriously concerned about winning the imminent general election. The conservative government aimed to attract small firm managers who were likely to support a progressive party by offering them financial advantage in the form of policy loans and debt guarantee through government-affiliated financial institutions; this practice has made SME policy symbolic of a compensation scheme in the face of a political crisis (Calder, 1988). Although the progressive party won the general election, policy instruments to support SMEs proposed by the conservative party were inherited by the centrist government.

As for the economic foundation of the policy, it was assumed that individual SMEs were vulnerable to their larger counterparts' abuse of their dominant bargaining position, and the policy intervention was necessary to support such SMEs. The SME Cooperative Association Law, enacted in 1949, defined its goal as “to secure opportunities for the fair economic activities and improve the economic position of SMEs” (Article 1). This was why joint activities of SME cooperative associations were exempted from the application of the antitrust law (Article 7). The weak position of SMEs was assumed not only in the product market but also in the financial market. The SME Cooperative Association Law designated cooperatives as receivers of long-term capital from government-affiliated financial institutions, such as the Central Cooperative Bank for Commerce and Industry established in 1936, since SMEs were considered to have difficulty in accessing capital from the commercial banks because of serious asymmetric information. Another assumption was that SMEs were too small and thus inefficient in cost structure. Therefore, cooperatives were expected to bundle SMEs in the same industry to the point where economies of scale prevailed.

The SME Cooperative Association Law defined several types of SME cooperatives (Article 3). Figure 1 shows the number of SME cooperatives and business cooperative associations in all sectors. Information was collected from a comprehensive survey conducted by the National Federation of Small 2 / 25

Business Associations (NFSBA, 2000). Approximately 80 percent of SME cooperatives are business cooperative associations, implying that examining business cooperative associations is equivalent to assessing industrial policy to promote SME cooperatives. The number of business cooperative associations in all sectors peaked in 1981, and has since gradually declined, stagnating in the nineties.

-----Figure 1-----

Figure 2 shows the ratio of SMEs that join business cooperative associations to SMEs in the manufacturing sector. Information was collected from the “Basic Survey on Manufacturing Structure and Activities” by the Ministry of Economy, Trade, and Industry. With respect to time-series variation, the ratio drops sharply during 1976 and 1987, which holds true for all industries. This suggests that industrial policy to promote SME cooperatives became less attractive for SMEs, presumably because the implicit assumptions of the policy that SMEs were the socially vulnerable became less relevant after the structural changes in the economy and in technology. In other words, the policy environment had significantly changed from the high-growth era in the sixties, where economies of scale worked best, to the low-growth era in the seventies, where innovations symbolized by the diffusion of micro-electronics, numerically controlled machine tools, and factory automation would be of more importance. With respect to cross-sectional variation, the ratio of SMEs joining business cooperative associations is higher in traditional industries, such as textile and wood, and lower in high-tech and globally competitive industries, such as electronics. This suggests that industrial policy to promote SME cooperatives was more important in industries where government protection was likely to be adopted and less relevant to industries where innovation and global competition mattered.

-----Figure 2-----

The activities of business cooperative associations are defined as follows by the SME Cooperative Association Law (Article 9-2). Variables in parentheses, used in factor analysis introduced in the next section, denote the ratio of SMEs engaged in a particular type of activity in business cooperative associations.

3 / 25

The business cooperative association borrows long- and short-term capital from government-affiliated financial institutions and loans it to members (FINA). To expand members’ business, the business cooperative association guarantees the member’s debt when the member engages in transactions with customers and suppliers (GUAR). The business cooperative association purchases new equipments that an individual member cannot afford, in order to produce, process, and supply what the members need (PROD). The business cooperative association inspects the quality of member firms’ raw materials, facilities, and final products (INSP). The business cooperative association purchases the raw materials and consumable goods required by member firms in bulk (PROC). The business cooperative association sells the products of member firms all together (SALE). The business cooperative association receives orders, passes them on to member firms, and delivers products (MARK). The business cooperative association stores and transports member firms’ raw materials, intermediate goods, and final products jointly (STOR). The business cooperative association establishes research facilities or outsources research to local public technology centers to solve member firms’ technological problems (RES).

METHOD Determinants of SME cooperative's activities: an industry-level analysis Model The first purpose of this study is to examine determinants of SME cooperative association. I address this issue using an industry-level data. Specifically, following models address the issue that which type of business cooperative association's activity is more prevalent under which type of business environment. Equations 1 and 2 show regression models and suffixes i and t denote industry and time. Definitions of variables are as follows.

F1it=a+b1MESit+b2PCMCVit+b3SHGRit+b4SUBit+b5IPRit+b6GAFIit+μi+λt+νit

(1)

F2it=a+b1MESit+b2PCMCVit+b3SHGRit+b4SUBit+b5IPRit+b6GAFIit+μi+λt+νit

(2)

SME cooperative's activities (F1 and F2) Dependent variables are the factor scores based on a factor analysis of various types of activities in which member firms are engaged. Since there were many variables representing activities of business 4 / 25

cooperative associations, as shown in the last paragraph of Section 2, I reduced the data using factor analysis to identify a small number of factors that characterized activities in which SMEs engaged as members of business cooperative associations (the results of factor loadings in Figure 3). Appendix Figures 1 and 2 show sectoral mean of F1 and F2.

-----Figure 3-----

Figure 3 shows that information on activities in business cooperative associations can be reduced to two factors. Factor 1, which is highly correlated with FINA and GUAR both of which are uncorrelated with Factor 2, can be considered to represent finance-driven activities. Factor 2, which is highly correlated with PROD and INSP both of which are uncorrelated with Factor 1, can be considered to represent production-driven activities. Then, I compute the factor scores of the two factors identified, F1 for finance-driven activities and F2 for production-driven activities.

Minimum efficient size (MES) One assumption lying behind the industrial policy to promote SME cooperative associations was that SMEs were too small to exert economies of scale. It is, however, meaningless to argue whether SMEs as a whole are too small because the minimum efficient size (i.e., establishment size at which quantity expansion reduces average cost) significantly differs according to industrial characteristics. Thus, it is predicted that the ratio of SMEs joining production-driven activities (F2) is higher in industries where the minimum efficient size is smaller. The minimum efficient size is defined as the average size of establishments in the industry divided by 0.75 (Lyons, 1980).

Vulnerability of profitability (PCMCV) The other assumption of the policy was that SMEs were vulnerable to larger firms' abuse of their dominant bargaining position in the product market. Thus, it is predicted that the ratio of SMEs joining finance-driven (F1) and production-driven activities (F2) is higher in industries where the profit rate of SMEs is lower and/or less stable. The price-cost margins are used to represent profitability at an industry-level. The coefficient of variation (PCMCV), that is, standard deviation divided by the average of price-cost margins, which is higher when price-cost margins are lower and/or less stable, is 5 / 25

introduced to the regression model.

Output growth (SHGR) When business prospects look poor, SMEs may want to join business cooperative associations for financial assistance and solutions to technological problems. Thus, the growth rate of industry shipments (SHGR) was added to the regression model. It is predicted that the ratio of SMEs joining finance-driven activities (F1) is higher in industries with a lower growth rate.

Subcontracting (SUB) This variable represents the ratio of subcontractors to SMEs. In a discrete process industry, where the entire production process can be divided into several processes in technological terms, it may be economically rational for large, integrated firms to outsource a particular part of the entire production process to smaller-sized subcontractors. In such case, there will be an increased need to organize intraindustry networks among subcontractors. The more active subcontracting was in the industry, the more the need to organize intraindustry supply chain networks among subcontractors would arise. Thus, it is predicted that the ratio of SMEs joining production-driven activities (F2) is higher in industries where subcontracting is more prevalent.

Government support (GAFI) The SMEs’ debt ratio to government-affiliated financial institutions for SMEs (i.e., the Central Cooperative Bank for Commerce and Industry, the People’s Finance Corporation, and the Japan Finance Corporation for Small Business) is expected to represent the SMEs' dependence on government protection. It is predicted that the ratio of SMEs joining finance-driven activities (F1) is higher in industries with higher dependence on government support, such as policy loans.

Innovation (IPR) The ratio of SMEs possessing intellectual property rights, such as patents, developed on their own is used to represent SMEs’ technological capability at an industry-level. As shown in Figure 2, it is predicted that the ratio of SMEs joining finance-driven (F1) and production-driven activities (F2) is lower in industries where SMEs are more active in developing new technology independently. 6 / 25

Data The only source of information of business cooperative association's activities was an industry-level data, that is, the Ministry of Economy, Trade, and Industry's “Basic Survey on Manufacturing Structure and Activities” in 1971, 1976, 1981, and 1987. The dataset of 1966 was not used since GUAR appeared in the survey after 1966. Information of independent variables were collected from the Ministry of Economy, Trade, and Industry's “Census of Manufactures” in 1971, 1976, 1981, and 1987. PCMCV and SHGR were computed using the data for the five preceding years of each survey year. Two-digit industries examined in this study are foods, textiles, clothing, woods, furniture, pulp, publishing, chemicals, petroleum, rubber, leather, ceramics, iron, nonferrous, metal, machinery, electronics, transportation, and precision. Due to the change in standard industrial classification in the empirical period, information of beverage and plastics could not be used in regression analysis. The number of observations of a balanced panel is 76 (19 two-digit manufacturing industries and 4 points of time).

Impacts on TFP growth: a firm-level analysis Model The second question is the impact of business cooperative associations on members' performance. A treatment effect analysis based on a firm-level dataset is used for evaluation since it enables one to control for endogeneity stemming from the fact that the selection of a policy target can significantly correlate to a policy evaluation indicator, which makes it difficult to distinguish actual policy impacts from the contributions of firms’ inherent characteristics (Deaton, 1995). Regression models and variables are as follows.

TFP  a  bBCA  cX  u (3) P   Z  v

(4)

 D  1 if P  0    D  0 if P  0 

(5)

In the first stage, a probit model is ran on Equation 4 to estimate the determinants of the probability of SMEs participating in a business cooperative association. Based on a probit model, selection correction 7 / 25

variable is added to Equation 3. In the second stage, an ordinary least square is ran on Equation 3 to estimate the impact of a business cooperative association on total factor productivity growth. A significant estimated coefficient of the selection correction variable, rho in the result table, indicates a significant correlation between unobservable factors that affect SMEs' being a member of a business cooperative association, v in Equation 4, and unobservable factors influential in TFP growth, u in Equation 3, implying the presence of endogeneity.

Total factor productivity growth (TFP) A dependent variable of the main equation is the total factor productivity growth (annual average between 1992 and 1995) of SMEs with 50 employees or more. TFP was measured by using the Cobb-Douglas production function where labor and capital are inputs, based on the assumption of a competitive market. To deflate capital and value added, the net fixed capital formation index and the GDP deflator in the System of National Account by the Economic and Social Research Institute were used.

The participation in business cooperative associations (BCA) As shown in Figure 1, most of the SME cooperative associations are business cooperative associations. Therefore, a dummy variables taking a value of one if the firm joins business cooperative associations (BCA) is introduced to examine the policy effect of SME cooperative associations. Any policy effect stemming from activities in business cooperative associations, such as improved bargaining position in a product market, mitigated constraints in a financial market, reputation establishment in a business community and brand royalty entailed by that, and cost reduction through scale economies, would result in the output growth of member firms that cannot be explained by the growth of labor and capital. Thus if SME cooperatives contribute to improving productivity of member firms, even after controlling for unobservable firm-specific factors influencing productivity growth, there will be a positive correlation between TFP and BCA.

Determinants of TFP growth (X) The most important source of total factor productivity growth is accumulated technological knowledge of the firm. The following variables are considered to represent technological resources of the firm and 8 / 25

are expected to have positive impacts on total factor productivity growth of the firm. R&D expenditure is considered as representative of a not-codified knowledge resource (Arora and Gambardella, 1990) embodied in scientists. R&D intensity is defined as R&D expenditure divided by value added (RDI). Similarly, the codified knowledge resource is presented by the number of patents developed and held by the firm (PATENT). The royalty revenue from licensing-out agreements in domestic and overseas technology transactions (LICENSEOUT), and the expenditure for licensing-in agreements with domestic or foreign partners (LICENSEIN) are also included. Control variables included in X are as follows. TFP92 is the level of total factor productivity in 1992. This variable controls for a very high growth between 1992 and 1995 caused by 1992's TPF level's being very low. Two-digit industrial dummies are included in the regression analysis but their results are not reported.

Latent variable for program participation (P*) P* denotes a latent variable representing the propensity of SMEs to join a business cooperative association. An SME is a member of a business cooperative association if P* exceeds zero, and a non-member otherwise.

Determinants of the participation in business cooperative associations (Z) As resource-based view of the firm suggests, internal resources as well as market structure affect firm behavior. If firms with less internal resources are highly motivated to join business cooperative associations to augment internal resources, internal resources are expected to negatively correlate with the probability of participation in business cooperative associations. As shown in the definitions of X, RDI and PATENT are proxy variables for technological resources. Advertisement-sales ratio (ADI) and export-sales ratio (EXPORT) represent market resources that enable the firm to introduce new products to the market successfully. The presence of the investment in information communication technology (INTRANET, INTERNET) represents organizational resources since the investment of information communication technology implies organizational changes, such as empowerment and less hierarchical structure, that enable the investment to contribute to performance more efficiently. In capital intensive industries, policy loans accessible through business cooperative associations may seem attractive to SMEs. Thus, tangible-fixed-asset-labor ratio (CAPLAB) of the firm is expected to positively correlate to the probability of participation in business cooperative associations. A subsidiary (SUBSIDIARY) may 9 / 25

receive financial assistance from a parent firm and are unlikely to joint SME cooperatives due to less financial constraints in a capital market. Ordinary profit divided by sales (ORDINARYP) represents financial resources, such as cash flow, and is considered to be negatively correlated with P* since business cooperative association was establish to mitigate SMEs' constraints in a financial market. A dummy for being subcontractor (SUBCONTRACTOR) represents SMEs' being vulnerable to their larger counterparts' abuse of their dominant bargaining position, thus predicted to be positively correlated with P*. Control variables included in Z are as follows. Younger firms are less likely to establish legitimacy in the business community. If younger firms are motivated to join business cooperative associations to exploit reputation effect of being a member, firm age (AGE) is expected to negatively correlate with the probability of participation in business cooperative associations. Another control variable is firm size as measured by log of the number of employees (SIZE).

Data So far the only information source for SME cooperative associations based on a firm-level dataset is the Ministry of International Trade and Industry's "Basic Survey of Japanese Business Structure and Activities" (1992 and 1995). Among firms with establishments in mining, manufacturing, wholesale and retail, and restaurants, this survey covers establishments with 50 employees or more, or firms with capitalization of 30 million JPY or more. This survey was tentatively initiated in 1992 and annually conducted after 1995. The 1992 survey had collected information on business cooperative associations and so far this is the last. Thus, independent variables and treatment variables were generated from the 1992 survey and a dependent variable, annual average of TFP growth, was generated from surveys in 1992 and 1995. Thus, this study assumes that impacts of SME cooperatives on productivity become visible within three years. The Small and Medium-sized Enterprise Basic Law defines SMEs as, in the manufacturing industry, as firms with 300 or fewer employees, or firms with a capitalization of 100 million JPY or less. The threshold applied varies according to industry. Although the official definition was altered after the amendment of SME Basic Law in 1999, since this study used dataset in 1992-1995, the previous definition was applied. Observations are 9049 SMEs in the manufacturing sector.

RESULTS Determinants of SME cooperative's activities: an industry-level analysis 10 / 25

Table 1 shows estimation results of fixed-effect models. The results of time dummies are not reported. The F-test rejects the null hypothesis of no individual effects (p<.000 for Equation 1 and p<.000 for Equation 2). The null hypothesis of random-effect is rejected by the Hausman test (p<.000 for Equation 1 and p<.022 for Equation 2). The number of observations is 76 (19 industries and 4 points of time) since information of GUAR could not be obtained in 1966.

The results show that SMEs dependent on government support through government-affiliated financial institutions were engaged in finance-driven activities (F1) in SME cooperatives. This suggests the characteristics of SME cooperatives as a receiver of policy loans. The more SMEs were familiar with policy loans, the more likely they were involved in finance-driven activities in SME cooperatives. In addition, the minimum efficient size (MES) is negatively correlated with the ratio of SMEs joining production-driven activities (F2), which is consistent with the prediction. It is reasonable that joint activities that foster scale economies are important in industries where a larger amount of capital is needed for more efficient operation of establishments. IPR shows negative signs in both models, as predicted, but the results are not statistically significant, which does not support that business cooperative associations are likely to be organized in less innovative industries. Although the results are not significant, SUB shows a positive sign in Equation 1 and a negative sign in Equation 2, suggesting that SMEs attempt to join business cooperative associations so that the benefits and functions offered by the cooperatives will not overlap and be complementary. In other words, core firms in keiretsu (note the empirical period is 1970s and 1980s when subcontracting network represented as keiretsu system was working in discrete process industries) may offer small subcontractors inspection services for materials and final products and technical assistance in the production process. If small subcontractors were to join business cooperative associations, they may have preferred to be provided with financial assistance that might not have been obtained in subcontracting networks. The level and stability of an industry-level profit rate and output growth had no impact on SMEs’ propensity to join business cooperative associations, implying that member firms were not motivated by the improvement of the bargaining position in the product market. Impacts on TFP growth: a firm-level analysis Table 2 shows the results of treatment regression models without and with industry dummies. The inclusion of industry dummies does not affect the main results and both type of regression models show 11 / 25

that the participation in business cooperative association negatively affects SMEs' total factor productivity growth, even after controlling for unobservable firm-specific factors affecting both the participation in SME cooperatives and productivity growth. This is in line with another study on the Japanese industrial policy, although not exclusively focusing on SME policy, by Beason and Weinstein (1996) showing that policy instruments on tax, tariff, subsidy, and loans from the Japan Development Bank yielded no significant effect on an industry-level TFP growth. However, the results are more striking than those of the previous study in that the promotion of business cooperative associations had a negative effect on the TFP growth of SMEs. One interpretation is adverse selection, whereby assistance related to finance and production offered by business cooperative associations attracted unmotivated small firm managers (Urata and Kawai, 2002). Alternatively, activities in business cooperative associations might have been inappropriately designed so that participants might deteriorate their technological and/or managerial capabilities.

The coefficients of RDI represent the rate of return to R&D, that is, R&D productivity. The results show that the rate of return to R&D is 5.6% in the model without industry dummies and 6.6% in the model with industry dummies. Considering the rate of return to R&D for all SMEs in the manufacturing sector is about 7.2% (not reported in the table), the inclusion of cooperative association dummy does not change R&D productivity greatly. It is reasonable that the royalty revenue from licensing out improves total factor productivity since it represents the quality of explicit or codified knowledge resources. However, the number of patents has a significantly negative impact on technological progress, which is opposite to the prediction. Technologies introduced from domestic and foreign firms via patent licensing do not affect productivity growth of SMEs.

The results for the selection equation show that R&D-inactive firms are likely to join business cooperative associations, which is consistent with the results in Table 1 where IPR is (though insignificant) negatively correlated with F1 and F2. Subcontractors tend to join business cooperative associations. Taking account of the results in Table 1 where SUB is positively correlated with finance-driven activities while negatively correlated with production-driven activities, subcontractors might have joined cooperative associations in order to secure financial access through government affiliated financial institutions. Contrary to predictions, tangible-asset-labor ratio is negatively correlated 12 / 25

with the probability of SMEs to join business cooperative associations. Subsidiaries find it unnecessary to joint SME cooperatives since they are able to obtain financial and technological support from parent firms. Lastly, smaller and older SMEs tend to participate in SME cooperatives.

Although uniquely developed in Japan, the case of SME cooperatives sheds light on the target, method, and dynamics of industrial policy. As shown in the estimation results of Equation 4, the program chiefly targeted less innovative SMEs. It is not surprising that this type of targeting made no contribution to the improvement of productivity. The results contrast with the findings that recent SME policies focusing on innovative firms show a positive program impact in terms of capital investment and firm growth (Motohashi, 2001; Harada and Honjo, 2005). Furthermore, the estimation results of Equations 1 and 2 show that the policy implementation was through generating scale economies and mitigating financial constraint in the capital market, both of which were found to be detrimental to productivity growth. As discussed in Section 2, this policy instrument was originally aimed at the creation of a political base of the conservative government at that time, by offering small firm managers easier access to policy loans that could be used for bridge financing. With this in mind, it is reasonable that finance-driven activities explained a significant proportion of the variation in activities of business cooperative associations. Considering that they did not contribute to productivity growth, a financial advantage offered through SME cooperatives seemed to be exploited by member firms for operation, and not for innovation. The absence of positive program effects may have resulted from a failure of the policy to adapt to changes in business environments in the empirical period. In the seventies, many microelectronics innovations, such as numerically controlled machine tools, diffused among SMEs, and the rapid appreciation of the Yen in the mid-eighties accelerated foreign direct investments by large firms, which forced SMEs to follow core firms or to expand market globally on their own. These changes in business environments forced SMEs to recognize the importance of innovation and globalization, with which SME cooperatives were not supposed to deal. The fundamental reform of the SME Basic Law in 1999, which shifted the purpose of SME policy from supporting all SMEs as the socially vulnerable to the creation and promotion of innovative SMEs, can be regarded as the reflection of such dynamics of the policy environments.

Another important implication of this policy instrument is with regard to spillover, which seems to be 13 / 25

very important when considering small firms’ innovative activities. Pioneering works in the economics of innovation suggest a positive correlation between firm size and innovation (Schumpeter, 1942; Galbraith, 1956). However, stylized facts from empirical studies indicate that scale economies do not prevail in the relationship between R&D input and output (Cohen, 1995). Rather, smaller firms tend to show higher R&D productivity (Griliches, 1980; Acs and Audretsch, 1990; Tsai and Wang, 2005). The absence of scale economies in innovation partially stems from the fact that interorganizational knowledge networks act as a significant source of small firms’ innovation (Freel, 2000; Rogers, 2004; Motohashi, 2005; Fukugawa, 2006; Okamuro, 2007; Nieto and Santamaria, 2010). This implies that by leveraging knowledge networks, small firms can have advantages in innovation against their larger counterparts. Though established as interfirm organizations, as defined by the SME Cooperative Association Law (Article 9-2), SME cooperatives were not originally designed to promote mutual learning among member firms. Instead, they virtually acted as a receiver of policy loans. This seemed to be a serious problem when innovation became more important for SMEs after the high-growth era in the sixties. This factor may also relate to a failure of the industrial policy to adapt to changes in business environments in the empirical period.

CONCLUSION Although the establishment and promotion of SME cooperatives was one of the most prevalent SME policies in postwar Japan and SME cooperatives were prevalent at least until the seventies, few quantitative evaluation of this industrial policy had so far been conducted. Using an industry-level and a firm level datasets, this study quantitatively evaluated SMEs' motivations to participate in SME cooperatives, activities that members were engaged in SME cooperatives, and the impacts of SME cooperatives on total factor productivity growth of member firms. The key findings are as follows. SMEs in industries where the dependence on government-affiliated financial institutions was high tended to join SME cooperatives for finance-driven activities. SMEs in industries where the minimum efficient size was small tended to join production-driven activities. The participation in SME cooperatives was negatively correlated with the SMEs' total factor productivity growth, even after controlling for unobservable firm-specific factors influential both in productivity and the participation in SME cooperatives. As the negative effects of SME cooperatives on productivity growth imply the mismanagement of SME cooperatives, further empirical research, based on a cooperative-level dataset, 14 / 25

will need to examine how they were managed. In addition, because of the availability of the relevant dataset, the empirical period of this study was limited to that before the fundamental reform of SME policy. The fundamental reform of the SME Basic Law in 1999 shifted the purpose of SME policy from supporting all SMEs as the socially vulnerable to the creation and promotion of innovative SMEs. Future study also should take account of the impact of such shifts on how the way SME cooperatives were managed.

15 / 25

References Acs, Z. and Audretsch, D. (1990) Innovation and Small Firms, MIT Press: Boston. Amemiya, T. (1985) Advanced Econometrics, Harvard University Press. Anand, B. and Khanna, T. (2000) Do firms learn to create value? the case of alliances, Strategic Management Journal, 21(3), 295–315. Arvanitis, S., Sydow, N., and Woerter, M. (2008) Is there any impact of university-industry knowledge transfer on innovation and productivity? an empirical analysis based on Swiss firm data, Review of Industrial Organization, 32(2), 77–94. Beason, R. and Weinstein, D. (1996) Growth, Economies of Scale, and Targeting in Japan (1955–1990), Review of Economics and Statistics, 78(2), pp.286–295. Calder, K. (1988) Crisis and Compensation: Public Policy and Political Stability in Japan, 1949–1986, Princeton University Press: Princeton. Cohen, W. (1995) Empirical Studies of Innovative Activity, in Stoneman, P. (ed.), Handbook of the Economics of Innovation and Technological Change, Blackwell: Oxford, p.182–264. Deaton, A. (1995) Data and Econometric Tools for Development Analysis, in Behrman, J. and Srinivasan, T. N. (eds.), Handbook of Development Economics Volume 3A, North Holland: Amsterdam. Freel, M. (2000) External Linkages and Product Innovation in Small Manufacturing Firms, Entrepreneurship and Regional Development, 12(3), pp.245–266. Fukugawa, N. (2006) Determining Factors in Innovation of Small Firm Networks: A Case of CROSSdustry GROUPs in Japan, Small Business Economics, 27(2–3), pp.181–193. Galbraith, J. (1956) American Capitalism: the Concept of Countervailing Power, Houghton Mifflin: Boston. Griliches, Z. (1980) Returns to Research and Development Expenditure in the Private Sector. In: Kendrick, J. and Vaccara, B. (Eds.), New developments in productivity measurement and analysis, University of Chicago Press: Chicago. Harada, N. and Honjo, Y. (2005) Does the Creative Business Promotion Law Enhance SMEs’ Capital Investments? Evidence from a Panel Dataset of Unlisted SMEs in Japan, Japan and the World Economy, 17(4), pp395–406. Irwin, D. and Klenow, P. (1996) High-tech R&D subsidies: estimating the effects of Sematech, Journal of International Economics, 40(3-4), 323–344. Link, A. and Bauer, L. (1989) Cooperative research in U.S. manufacturing: assessing policy initiatives and corporate strategies, Lexington Books: Lexington. Link, A. and Rees, J. (1990) Firm size, university based research, and the returns to R&D, Small Business Economics, 2(1), 25–31. Maddala, G. S. (1983) Limited-dependent and qualitative variables in econometrics, Cambridge University Press. March, J. (1991) Exploration and Exploitation in Organizational Learning, Organization Science, 2(1), 1-13. Motohashi, K. (2001) Use of Plant Level Micro-data for SME Innovation Policy Evaluation in Japan, RIETI Discussion Paper, 01E006. Motohashi, K. (2005) University-Industry Collaborations in Japan: The Role of New Technology-Based Firms in Transforming the National Innovation System, Research Policy, 34(5), pp.583–594. 16 / 25

National Federation of Small Business Associations (2000) White Paper on Small Business Associations, National Federation of Small Business Associations: Tokyo. National Federation of Small Business Associations (2001) Survey on Business Cooperative Associations, National Federation of Small Business Associations: Tokyo. Nieto, M. and Santamaria, L. (2010) Technological Collaboration: Bridging the Innovation Gap between Small and Large Firms, Journal of Small Business Management, 48(1), pp.44–69. Odagiri, H. and Kato, Y. (1996) University-industry research collaboration in the biotechnology industry: an empirical analysis, Business Review, 45, 62–80. Okamuro, H. (2009) Economic analysis of technological alliances, Doyukan: Tokyo (In Japanese). Okamuro, H. (2007) Determinants of Successful R&D Cooperation in Japanese Small Businesses: The Impact of Organizational and Contractual Characteristics, Research Policy, 36(10), pp.1529–1544. Organization for Economic Co-operation and Development (2000) OECD Small and Medium sized Enterprise Outlook, OECD: Paris. Rogers, M. (2004) Networks, Firm Size and Innovation, Small Business Economics, 22(2), pp.141–153. Schumpeter, J. (1942) Capitalism, Socialism, and Democracy, G. Allen and Unwin: London. SMEA (1981) White Paper on Small- and Medium-sized Enterprises, Small- and Medium-sized Enterprise Agency: Tokyo. Tsai, K. and Wang, J. (2005) Does R&D Performance Decline with Firm Size? A Re-examination in Terms of Elasticity, Research Policy, 34(6), pp.966-976. Urata, S. and Kawai, H. (2002) Technological Progress by Small and Medium Enterprises in Japan, Small Business Economics, 18(1-3), pp.53-67.

17 / 25

Figure 1 The number of SME cooperatives

Notes Author's elaboration based on NFSBA (2000). BCA denotes business cooperative associations. Total denotes BCA and other forms of SME cooperatives.

18 / 25

Figure 2 The ratio of SMEs that join business cooperative associations

Notes Author's elaboration based on the Ministry of Economy, Trade, and Industry's “Basic Survey on Manufacturing Structure and Activities”.

19 / 25

Figure 3 Factor loadings

1

Factor loadings

stor3

.5

prod3

Factor 2

insp3

sale3 guar3

0

proc3 res3

fina3

-.5

mark3

-.5

0

.5

1

Factor 1 Rotation: orthogonal varimax Method: principal factors

Notes the ratio of SMEs engaged in a particular type of activity in business cooperative associations See Section 2 for the definitions of variables.

20 / 25

Table 1 Estimated fixed-effect models

N Equation Dependent variable MES PCMCV SHGR SUB IPR GAFI _cons F test for pooled OLS Hausman test for random-effect sigma_u sigma_e Rho

76 1 F1 Coef. -.010 .067 .034 1.351 -.943 5.186 -.735 .000 .000 1.182 .377 .907

Std. Err. .011 .347 .066 1.079 4.041 2.665 .722

Significance

* *** ***

76 2 F2 Coef. -.030 .248 .088 -1.930 -2.475 5.446 1.971 .000 .022 1.297 .499 .871

Std. Err. .015 .459 .087 1.425 5.339 3.521 .954

Significance *

** *** **

Note *** p<.01; ** p<.05; * p>.1

21 / 25

Table 2 Estimated treatment regression models N 9049 Equation 3 Coef. Std. Err. Dependent variable TFP RDI 0.056 0.016 PATENT -3.6E-05 1.17E-05 LICENSEOUT 0.025 0.006 LICENSEIN 3.54E-06 3.11E-05 BCA -0.032 0.003 TFP92 -0.183 0.003 _cons 0.207 0.004 Industry dummy No Equation 4 Dependent variable BCA RDI -1.785 0.323 PATENT -0.0003 0.0004 ADI -1.502 1.158 EXPORT -0.277 0.181 INTERNET 0.255 0.043 INTRANET 0.319 0.032 ORDINARYP -0.039 0.128 CAPLAB -0.005 0.001 SUBSIDIARY -0.488 0.035 SUBCONTRACTOR 0.131 0.029 SIZE -0.081 0.023 AGE 0.014 0.001 _cons -0.514 0.123 Industry dummy No /athrho 0.275 0.034 /lnsigma -2.682 0.020 Rho Sigma Lambda Wald test of rho = 0

0.268 0.068 0.018

Significance *** *** *** *** *** ***

***

*** *** *** *** *** *** *** *** *** ***

0.032 0.001 0.002

9049 3 Coef. TFP 0.066 -3.1E-05 0.025 3.25E-06 -0.030 -0.182 0.193 Yes 4 BCA -0.855 -0.0002 -0.292 -0.058 0.253 0.327 0.029 -0.006 -0.480 0.193 -0.077 0.013 -0.414 Yes 0.255 -2.691 0.249 0.067 0.016

***

Std. Err.

Significance

0.017 0.00001 0.006 3.04E-05 0.003 0.003 0.012

*** *** ***

0.316 0.0004 1.220 0.183 0.044 0.032 0.139 0.001 0.035 0.031 0.024 0.001 0.406

***

0.034 0.019

*** ***

*** *** ***

*** *** *** *** *** *** ***

0.032 0.001 0.002 ***

Note *** p<.01; ** p<.05; * p>.1

22 / 25

Appendix Figure 1 Finance-driven activities in SME cooperatives

ceramic chemical clothing electronics food furniture iron leather machinery metal nonferrous petroleum precision publishing pulp rubber textiles transportation wood -1

0

1

2

mean of f1

23 / 25

Appendix Figure 2 Production-driven activities in SME cooperatives

ceramic chemical clothing electronics food furniture iron leather machinery metal nonferrous petroleum precision publishing pulp rubber textiles transportation wood -1

0

1

2

mean of f2

24 / 25