- Email: [email protected]
Sources of External Technology, Absorptive Capacity, and Innovation Capability in Chinese State-Owned High-Tech Enterprises
World Development Vol. 39, No. 7, pp. 1240–1248, 2011 Ó 2011 Elsevier Ltd. All rights reserved 0305-750X/$ - see front matter www.elsevier.com/locate/worlddev
doi:10.1016/j.worlddev.2010.05.011
Sources of External Technology, Absorptive Capacity, and Innovation Capability in Chinese State-Owned High-Tech Enterprises XIBAO LI * Tsinghua University, Beijing, PR China Summary. — This paper examines the pattern of innovation and learning among state-owned enterprises in Chinese high-tech sectors and empirically estimates the impact of three types of investment for acquiring technological knowledge—in-house R&D, importing foreign technology, and purchasing domestic technology—on the innovation capabilities of firms. Based on a panel dataset consisting of 21 high-tech sectors during the period 1995–2004, an augmented knowledge production function is estimated. The results show that importing foreign technology alone does not facilitate innovation in Chinese state-owned high-tech enterprises, unless in-house R&D is also conducted. Domestic technology purchases, however, are found to have a favorable direct impact on innovation, suggesting that firms have less difficulty in absorbing domestic technological knowledge than utilizing foreign technology and that absorptive capacity is contingent upon the source or nature of the external knowledge. Ó 2011 Elsevier Ltd. All rights reserved. Key words — R&D, foreign technology import, domestic technology purchase, absorptive capacity, innovation capability
1. INTRODUCTION
basis of which technology and innovation are developed are difficult to transfer across firms. Cohen and Levinthal (1990) argued that the importance of external technology on technological competence and innovation capability depends on whether recipient firms have related prior knowledge or absorptive capacity to understand and exploit technological opportunities. They further noted that in-house R&D has a function of learning and helps recipient firms build up their absorptive capacity (Cohen & Levinthal, 1989). Since R&D is also crucial for firms to innovate and generate new knowledge, given the dual functions of in-house R&D, what is the pattern of innovation and learning among non-frontier firms in developing countries? How and under what conditions can these firms make the best use of the three different sources of knowledge—in-house R&D, imported foreign technology, and acquired domestic technology—to promote innovation? Does the pattern of learning and innovation in firms vary with the sources of external knowledge? To address the above issues, this study examines the pattern of learning and innovation among state-owned and stateholding enterprises (SOEs) in Chinese high-tech sectors, and empirically examines the impact on innovation capability of both internally developed knowledge via in-house R&D and externally acquired knowledge through the technology market. In particular, the study considers two forms of external disembodied technological knowledge which are originally generated in different contexts: technology imported from foreign countries and domestic technology licensed by universities, research institutes, or other domestic firms. Following the definition of the National Bureau of Statistics of China
The role of in-house R&D in a firm’s innovation has been recognized since Schumpeter’s (1942) work, and the relationship between R&D and innovation outcomes has been extensively studied within the framework of knowledge production functions in previous empirical studies (see e.g., Cincera, 1997; Furman, Porter, & Stern, 2002; Hausman, Hall, & Griliches, 1984; Jaffe, 1986). Beside in-house R&D, a firm can also obtain useful technological knowledge from technology markets (Arora, Fosfuri, & Gambardella, 2001), which in turn contribute to the firm’s knowledge generation or innovation. For firms in late-comer economies, exploiting technological knowledge created in developed countries is particularly common and often ranks among the most important strategies for catching-up (Amsden, 1989). Previous studies in the research stream of international R&D spillover emphasized the importance of knowledge spillovers embodied in foreign technological products (Coe & Helpman, 1995; Grossman & Helpman, 1991; MacGarvie, 2006). By contrast, inadequate attention has been paid to the role of disembodied knowledge, such as designs, formulae, drawings, processes, patents, and knowhow. Such knowledge is often licensed or acquired by recipient firms through a formal channel and, thus, may be directly exploited. Firms can obtain licenses for disembodied knowledge from both developed countries and domestic innovators, such as universities, research institutes, and other innovating firms. In China, for example, during the early stage of economic transition, indigenous firms not only spent a large amount of money on importing foreign technology (Kang & Yang, 1991; Xu, 1996), but were also advised to acquire useful knowledge from the domestic technology market (Liu & White, 2001). As a result, a complicated pattern of learning and innovation has emerged in Chinese industries or sectors. The complexity of learning and innovation among catch-up firms lies in the differences in the nature and sources of knowledge. Due to the tacit and context-specific nature of technological knowledge, it is difficult, if not impossible, for recipient firms to acquire innovation capacity through the mere license or purchase of external technology. For instance, Nelson and Winter (1982) pointed out that organizational routines on the
* This paper benefits greatly from the constructive comments and suggestions of Dr. Xiaolan Fu, two anonymous reviewers, and participants in the Oxford conference on “Confronting the Challenge of Technology for Development: Experiences from the BRICS”, May 30, 2008. The author also gratefully acknowledges the financial support provided by the National Natural Science Foundation of China (Project No.: 70602005) and The United Nations Industrial Development Organization (UNIDO). Final revision accepted: May 20, 2010. 1240
SOURCES OF EXTERNAL TECHNOLOGY, ABSORPTIVE CAPACITY, AND INNOVATION CAPABILITY
(NBSC), foreign technology import (FTI) expenditures are defined as including purchases of design knowledge, formulae, drawings, processes, patents and know-how, and key equipment closely related to new product development. They do not include those used directly for production, such as production lines, complete knock-down kits, and turn-key facilities (Liu & White, 1997). In this sense, a large portion of FTI expenditures is spent on disembodied knowledge. FTI mainly takes the form of technological licensing and is substantially different from the import of foreign machinery and equipment, which represents a form of embodied knowledge. Similarly, domestic technology purchase (DTP) refers to the licensing of outside technologies developed by universities, research institutes, or other firms, representing a form of knowledge transfer from domestic sources. Hence, both FTI and DTP reflect a firm’s efforts to acquire technological capabilities from external sources. They can be transacted in markets for technology, as defined in Arora et al. (2001). The major difference between FTI and DTP lies in the source of technological knowledge. This analysis is based on a panel dataset constructed from publicly available official statistics on five Chinese high-tech industry categories: medical and pharmaceutical products, aircraft and spacecraft, electronic and telecommunication equipment, computer and office equipment, and medical instruments and meters. The categories of Chinese high-tech industries are similar to and comparable with those used in OECD countries, although they are not as intensive in R&D as their counterparts (Xu, 2000). With the four-digit Chinese standard industrial classification (SIC) sector as the unit of analysis, a panel of 21 sectors was constructed and sector-level data for the period from 1995 to 2004 analyzed. Table A1, in the appendix, gives a list of all sectors included. Drawing upon the previous studies (Acs, Anselin, & Varga, 2002; Furman et al., 2002; Li & Wu, 2010), domestic patent application counts were used to measure innovation capability. A firm’s innovation capability is thought to be determined by three sources of technological knowledge that the firm obtained from inhouse R&D, foreign technology import, and domestic technology transfer. The contribution to the innovation capability of the three sources of knowledge is empirically estimated within a framework of knowledge production functions (Basant & Fikkert, 1996; Hall & Mairesse, 1995; Hu, Jefferson, & Qian, 2005). This paper contributes to the literature on absorptive capacity and the role of in-house R&D for catch-up firms. The estimated results show that SOEs in the Chinese high-tech industries can gain leverage by investing in both in-house R&D and FTI, while technology imports alone do not contribute to the rate of patenting. By contrast, investing in domestic technology alone facilitates a firm’s innovation. The firm can easily assimilate and utilize domestic technological knowledge, independent of its absorptive capacity. That is, absorptive capacity is crucial for assimilating foreign technology, but not for taking advantage of domestic knowledge. The finding suggests that the argument for absorptive capacity is contingent upon the source or nature of external knowledge and that the diffusion of domestic technology is also conducive to fostering innovation. The remainder of this paper is organized as follows. Section 2 presents a brief review of the relevant literature. Section 3 introduces the estimating strategy employed in the analysis, followed by a description of the data construction and summary statistics. The results and main findings are reported in Section 4. Section 5 concludes with some caveats and areas for further research.
1241
2. LITERATURE REVIEW In consideration of their weak technological capability, firms in developing countries often regard the use of technology developed abroad as an expedient way to develop and expand their manufacturing capacity (Amsden, 1989), instead of investing intensively in R&D. In China, for example, the presence of foreign capital as well as the use of imported technology from developed countries has contributed considerably to the country’s rapid economic growth since the 1980s (Hu & Jefferson, 2002; Liu & Wang, 2003; Madariaga & Poncet, 2007; Yao, 2006; Zhang, 2001). Besides foreign technology, the Chinese firms also look for appropriate technologies from domestic sources, such as university labs, government-run research institutes, or even local competitors (Li & Wu, 2010; Liu & White, 2001). Irrespective of whether the external knowledge is developed abroad or domestically, what really matters is whether recipient firms can assimilate and effectively exploit such external knowledge to build up their own innovation capability. In reality, however, it proves rather difficult for recipient firms to effectively utilize knowledge from external sources to boost their own innovation. During the early catch-up stage in China, despite the government’s wishful policy goal of “trading market for technology”, it was often found that Chinese indigenous firms could not even acquire or have access to state-of-the-art technology through direct foreign investment inflow and import of foreign technology (Wang & Gao, 2006). One situation that occurs frequently among Chinese indigenous firms is that the very weak technological capability of some domestic firms forces them to rely continuously on foreign technology. Such examples have been noted in the Chinese automobile and civilian aircraft industries, which are discussed in Lu (2005) and Lu and Feng (2005), respectively. One explanation for this phenomenon relates to the nature of the knowledge to be assimilated. Technology acquired through FTI and DTP contains both codified (e.g., patents, formulae, and drawings) and tacit (e.g., know-how) knowledge. Although the transmission of codified knowledge usually ends with the acquisition of documents and manuals, tacit knowledge associated with use of external technology does not easily transfer with the purchase of a technology. Whether the licensees can effectively take advantage of such licensed knowledge is closely contingent on the level of their prior related knowledge or absorptive capacity. According to Cohen and Levinthal (1990, p128), absorptive capacity refers to “the ability of a firm to recognize the value of new, external information, assimilate it and apply it to commercial ends”. It strongly hinges on the level of the firm’s prior related knowledge, which can, in turn, be developed through its own R&D efforts. From this perspective, in-house R&D not only helps a firm generate new knowledge (i.e., innovation effect) but also contributes to its absorptive capacity (i.e., learning effect) as well. It thus plays a dual role (Cohen & Levinthal, 1989; Griffith, Redding, & Van Reenen, 2004). Forbes and Wield (2000) contended that the learning function of R&D is particularly relevant and important in developing countries. In view of the two functions of in-house R&D and the two sources of external knowledge, it seems an intricate matter for licensee firms to make the best use of both internally developed and externally licensed knowledge for their own innovation. This reasoning suggests that the knowledge portfolio has an important influence on a firm’s innovation performance. Although recent studies have confirmed the contribution of FTI to productivity and growth in developing countries (Basant & Fikkert, 1996; Hu et al., 2005), the impact of foreign
1242
WORLD DEVELOPMENT
technology on the innovation capacity of local firms has not been adequately discussed. The limited empirical findings in previous work are at best mixed or inconclusive. Liu and White (1997) found that firms can gain leverage when investing in both R&D personnel and foreign technology in highly innovative industries, while imported technology alone has no direct impact on a firm’s innovation measured by its share of new product sales. By contrast, Liu and Buck (2007) found that learning-by-importing -technology can directly accelerate a local firm’s introduction of new products, independent of absorptive capacity. Two empirical studies thus provide contrasting results on the importance of absorptive capacity, and few studies have ever considered the impact of DTP, despite its importance as an additional source of external knowledge. Furthermore, it remains unclear how important the leverage effect of R&D is for a recipient firm to exploit domestic technology. An exception is the recent work by Li and Wu (2010). These authors addressed both issues and found that the contribution of foreign knowledge to the innovation of Chinese firms is strongly dependent on the investment of the firms in R&D, while R&D does not leverage the effect of domestic technology. Based on region-industry level data, their study makes no distinction between domestic firms and foreign-investment joint ventures. This is problematic given the fact that foreign firms rarely purchase domestic technology in China and have a quite different pattern of learning and innovation from domestic firms. To fill this gap, the current analysis adds to Li and Wu (2010) by focusing on an important type of domestic firm, that is, only the SOE. To understand how essential and different the learning effect of R&D is for assimilating knowledge from different sources, it is wise to review the basic argument of absorptive capacity. Cohen and Levinthal (1990) posited that the importance of absorptive capacity is determined by both the quantity of knowledge to be assimilated and utilized and the difficulty of learning. The difficulty or ease of learning is largely determined by the characteristics of the underlying technological knowledge, which include the complexity of knowledge to be assimilated and degree to which outside knowledge is targeted to the needs and concerns of a firm. Following this reasoning, it may be expected that a systematic difference exists between technological knowledge imported from foreign countries and that licensed from domestic sources. First, since the technological level in many Chinese industries is far from technology frontiers, foreign imported technology is normally more complex and sophisticated than the best domestic technology. Secondly, since the accumulation process of tacit knowledge is influenced by both internal (e.g., organizational routines, structures, coordination mechanism) and external (e.g., market structure, competition, regulation) context-specific factors, it is rather difficult for recipient firms in one context to fully understand and assimilate tacit knowledge created in another context. According to Nelson and Winter (1982), the more tacit the relevant knowledge the more difficult it is to assimilate and exploit. Differences in contextual factors are presumably much larger across countries than between firms. As a result, it will be more challenging to draw on foreign technology than domestic knowledge. Furthermore, labor mobility is much easier within a country than across national borders. Related technological knowledge may not be essential for a recipient firm to exploit external knowledge generated by another local firm, especially when it can acquire such capacity by hiring those with the requisite knowledge from other local firms. Taking all these points together, it is reasonable to argue that exploiting foreign technology is a more demanding task than utilizing domestic technology for Chinese firms.
In another stream of research, empirical studies on the technology gap provide ample evidence for the argument that complexity and sophistication of external knowledge have an influential impact on the incentives of firms to develop absorptive capacity, and the learning effect of R&D in developing absorptive capacity is contingent on the level of the technology gap. For instance, Griffith et al. (2004) and Kneller (2005) demonstrated that the further a country is behind the world technological frontier, the greater the importance of its R&D investment in creating absorptive capacity. Researchers found that only firms above a certain technological level were likely to benefit from external technology spillovers (Borensztein, Gregorio, & Lee, 1998; Girma, 2005; Glass & Saggi, 2002), implying that the larger the technology gap, the more crucial the absorptive capacity. In China, the technology gap between a firm’s own technology and imported foreign technology is much larger than the gap between its own technology and licensed domestic technology. Thus, it is expected that the leverage effect of absorptive capacity on the effect of external knowledge will be different between the two sources of outside knowledge. 3. EMPIRICAL STRATEGY (a) Methodology To compare the impact of knowledge developed from different sources on innovation, and to empirically examine the mode of learning and innovation in Chinese indigenous firms, an augmented knowledge production function is proposed as the basis of economic analysis and uses the number of domestic patent applications as the measure of innovation performance. Taking into account the specific nature of patent counts in the dataset (Cameron & Trivedi, 1998), it is assumed that the number of sector level patents can be specified in a count panel data model as follows: 9 8 dt þ b1 RDit þ b2 FTI it þ b3 DTP it > > > > > = < þb ðRD FTI Þ þ b ðRD DTP Þ > it it it it 4 5 ; EðPatentit Þ ¼ exp > > þb6 JVPCT it þ b7 INTEN it > > > > ; : þb8 CAPINT it þ b9 LGSIZEit ð1Þ where Patentit is the patent counts for sector i in year t. RDit, FTIit and DTPit represent log-scaled knowledge stocks KRit, KFit, and KDit, respectively. KRit is the stock of technical knowledge generated by the in-house R&D investment of firms in sector i in year t; KFit is the stock of foreign technical knowledge imported by sector i; and KDit is the stock of domestic technology acquired by firms in sector i. dt represents year dummy and b’s are coefficients to be estimated from the model. In light of the fact that absorptive capacity can be created through in-house R&D, two interaction terms (RDFTI) and (RDDTP) are included to account for the leverage effect of absorptive capacity on assimilating and utilizing knowledge generated through FTI and DTP, respectively. Thus, the model specification (1) incorporates the dual role of in-house R&D. It implicitly posits that a firm is unable to assimilate externally generated knowledge passively. In order to exploit such knowledge, the firm must invest in absorptive capacity, which can, in turn, be created by investment in R&D. Along with the three variables of knowledge capital, four control variables are incorporated in the model. The first
SOURCES OF EXTERNAL TECHNOLOGY, ABSORPTIVE CAPACITY, AND INNOVATION CAPABILITY
variable, JVPCT, is the sales share from overseas owned or invested firms in each sector (including Hong Kong Special Administrative Region of China, Macao Special Administrative Region of China and Taiwan Province of China). It is intended to capture the implicit knowledge spillover from these oversea firms, and account for the potential effect of FDI which is regarded as another important channel through which foreign technologies influence accumulation of domestic technological capabilities (Cheung & Lin, 2004). Because the unit of this analysis is the high-tech sector, two sector level variables (INTEN and CAPINT) are added to allow for sector heterogeneity. The variable INTEN reflects sector-level R&D intensity and is defined as R&D expenditure per unit of sales. CAPINT is sectoral capital intensity, measured as the ratio of the book value of fixed assets to total outputs. Finally, sector size is controlled by incorporating the variable LGSIZE which is measured by the logged sector-level sales. By including year dummies in the model, time-related sector-wide effects are subsumed in the coefficient of the dummy years. It should be noted that there are several other channels through which technology diffuses, such as technical information provided by original equipment manufacturer (OEM) contracts, reverse engineering efforts, and labor mobility (Liu & Wang, 2003). Since most of these leave no paper trail from which to analyze the flow of technology and its impact, in the proposed model, only the three sources of technological knowledge are focused on, that is, in-house R&D, FTI, and DTP. (b) Data and measurement Given the importance of high-tech industries to the national economy, the NBSC has been collecting information on science and technology activities from high-tech firms. For the purpose of this analysis, this official source of information was used to investigate innovation activities across high-tech sectors. Specifically, the data used all came from office statistics published by the NBSC, in the series China Statistics Yearbook on High Technology Industry (2002–2005), which covers the period from 1995 to 2004. The Yearbook contains about 30 major indicators of innovation activities in five Chinese high-tech industrial categories: pharmaceuticals, aircraft and spacecraft, telecom, computers, and instruments. Firm-level data were originally reported to the NBSC by all large and medium-sized enterprises (LME). To maintain the confidentiality of the information from the firms, the NBSC aggregated firm-level data according to a combination of firm ownership and four-digit SIC sectors. To highlight the impact of different sources of external technology, in this study, only SOEs were considered for two reasons. First, these enterprises accounted for a dominant share of the FTI and R&D during the sample period. In 1995, for example, the SOEs account for 76.2% and 81.4% of R&D investments and FTI expenditures in Chinese high-tech industries. Despite continuous ownership reform in China, in 2004, over 40% of R&D expenditures in Chinese high-tech industries still came from the SOEs. Second, from a practical standpoint, data information on FTI and DTP was unavailable for firms under other ownership categories. Therefore, each sector in this analysis consists only of SOEs which are occasionally referred to as indigenous firms. In previous studies, both Liu and White (1997) and Liu and Wang (2003) used new product sales as a measure of innovation. In this paper, however, the domestic patent applications count was employed as a proxy indicator of endogenous innovation output. In terms of measuring the endogenous or indigenous features of innovation, these two measures are largely different. In China, if a product is designated by the
1243
government as new, the firm can obtain a tax subsidy from the provincial or national government. For this reason, firms have a strong incentive to over-record the sales of new products. In addition, the procedures for new product approval are neither completely standardized nor comparable between regions. The newness of products is a relatively arbitrary and geographically-bounded concept. According to the Chinese statistical convention, a product designated as new can simply be new to a local market, such as a county, a city, or a province, whether or not it has been on the market in other places. Alternatively, a product can be regarded as new as long as it is new to a firm and has not been produced for more than 1 year. Thus, the measure of new product sales will inevitably bring in some measurement bias. In innovation literature, the use of patent statistics also raises important measurement issues (Archambault, 2002; Archibugi, 1992; Griliches, 1990; Pavitt, 1988). It is widely recognized that patent information can neither include all important technological innovations nor reflect the importance of different innovations. There are pitfalls associated with equating patenting with the level of innovative activity in both “quantity” and “quality” perspectives. Nonetheless, patents usually contain technological improvements and/or innovative ideas that have, at least, to be new to the country. More importantly, in contrast to new product sales, the procedures for filing patents are uniform across all sectors and industries. In this sense, patents can be regarded as a good reflection of endogenous or indigenous innovation efforts. In effect, it is the best indicator available in the NBSC aggregated level dataset to reflect the indigenous nature of an innovation. In the empirical estimation of R&D-patenting knowledge production functions, two measures of R&D, or knowledge, are usually considered. For example, Hausman et al. (1984), Acs and Audretsch (1988), Cincera (1997) and Blundell, Griffith, and Windmeijer (2002) used the current and past flow of R&D expenditure as inputs, and assumed that all past R&D investments are substitutes, with a unit elasticity of substitution. Due to the high persistence of R&D expenditure, this specification often leads to multicollinearity problems between lagged regressors. Drawing on empirical literature on R&D and productivity (Hall & Mairesse, 1995), Cre´pon and Duguet (1997), instead, employed estimated R&D capital stock. According to these studies, the annual flow of R&D expenditure is taken to be investment adding to a firm’s knowledge capital. Knowledge capital depreciates over time, so that the contribution of past R&D becomes less valuable as time passes (Griliches, 1979). One advantage of this specification is that it allows for both complementarity and depreciation in past R&D expenditures. This paper adopts the second approach, to construct stock variables for the three different sources of knowledge. Following common practice in the previous work, the variable of knowledge capital stocks generated from R&D can be estimated with a perpetual inventory model (Hall & Mairesse, 1995). The initial knowledge stock SR1 and the knowledge stock at the beginning of year t (SRt) are computed from annual R&D investment (FRt), as follows: FR1 ; ð2Þ SR1 ¼ ðg þ dÞ SRt ¼ ð1 dÞSRt1 þ FRt1 ;
t P 2:
ð3Þ
Here, g denotes the pre-sample growth rate of annual R&D flow FRt, and d is the annual depreciation rate of R&D investment. One drawback associated with using knowledge stock in (3) as technological input is that it ignores the contribution of
1244
WORLD DEVELOPMENT
current R&D investment in patenting. Hall, Griliches, and Hausman (1986) showed that a strong contemporaneous relationship exists between in-house R&D investment and patenting. To incorporate the contemporaneous effect, the in-house R&D stocks at the middle of year t were used, which are computed as: KRt ¼ ðSRt þ SRtþ1 Þ=2;
aggregated at sector level, the patent applications filed by Chinese high-tech firms are rather limited. Before 2001, more than half of the sectors had annual total patent applications of no more than ten, although since then firms in more regions took out more than that number annually. This indicates that innovation capability in Chinese indigenous firms is at a very low level in terms of patenting.
ð4Þ
1 6 t 6 T:
Both SRt and SRtþ1 are obtained from (2) and (3). The effect of contemporaneous R&D investment is thus reflected in KRt through SRtþ1 . To compute the initial R&D stock SR1, two parameters, g and d, have to be determined (Griliches & Mairesse, 1984). Since the specification of d and g does not affect the results significantly, as confirmed in the literature (Hall & Mairesse, 1995), a pre-sample growth rate and an annual depreciation rate were assumed, both at 15% as in previous studies (Hu & Jefferson, 2004). The same approach was adopted to construct the knowledge stocks generated from FTI and DTP. In the computation, in-house R&D, FTI, and DTP expenditures are all adjusted by GDP deflators to their 1995 constant values. Since the book value of fixed assets used in constructing the variable CAPINT is not available for 1995, the empirical analysis is based on data from 1996 to 2004, which amounts to 189 observations. Table 1 provides a description of the correlation matrix and summary statistics of the constructed variables. The correlation matrix shows that there are moderate correlations between knowledge stock variables and sector size, which is not surprising given that the sector size affects the amount of investment in acquiring knowledge from all sources. A multicollinearity diagnostic test tells that the highest Variance Inflation Factor (VIF) is 5.29, below the critical point of 10 (Belsley, Kuh, & Welsch, 1980), suggesting that multicollinearity is not severe and no remedial action should be considered. As is usually found in patent data, the number of domestic patent applications in this analysis has more weight on the right tail than expected from the usual Poisson distributions, which results in a much larger variance than their means. It indicates that patent counts are over-dispersed. To account for the over-dispersion of patent counts in estimating, the fixed effect negative binomial regression proposed by Hausman et al. (1984) was used. Table 2 further describes the distribution of the patent application counts in different years. While
4. RESULTS AND DISCUSSIONS Table 3 reports the results from negative binomial regressions, based on various model specifications. The estimation proceeds with the usual maximum likelihood techniques. First, a simple case was considered where only two sources of knowledge stocks (RD and FTI) were included (column 1). It can clearly be seen that only in-house R&D contributes significantly to the introduction of patents. Although one might expect FTI to have a similar effect on innovation, the coefficient of FTI is insignificant although positive, suggesting that the impact of FTI on innovation is not as important as expected. When the term DTP is included, its estimated coefficient is found to be significantly positive (column 2), indicating that the impact of acquired domestic knowledge is important to patenting. Furthermore, the model including the interaction term (RDFTI), the result of which is reported in column 3, is found to be preferred to the one listed in column 2. In this case, the estimated coefficient of the interaction term is significantly positive, indicating that positive leverage can be gained by firms investing in both in-house R&D and FTI. It suggests that, when coupled with in-house R&D, FTI has a significant positive impact on patent applications, while the purchasing of foreign technology alone does not. To examine whether a similar leverage effect exists between in-house R&D and DTP, one more interaction term RDDTP was added to the model specification (1) (column 4). In this specification, the estimated coefficients of both RD and RDFTI still remain significantly positive, while the coefficient of RDDTP is not significant statistically. Surprisingly, the coefficient of R&D knowledge stock is now insignificant. This is probably because of a severe multicollinearity among the explanatory variables. In this case, the VIF statistics for the variable RD is 11.67, larger than the common cut-off value, indicating a serious multicollinearity. The likelihood ratio test
Table 1. Correlation matrix and descriptive statistics Variable
2 3 4 5 6 7 8 9 10 Mean Std. Dev. Min Max
Patent
RD
FTI
DTP
RDFTI
RDDTP
JVPCT
INTEN
CAPINT
LGSIZE
1
2
3
4
5
6
7
8
9
10
0.515 0.379** 0.341** 0.306** 0.202** 0.024 0.001 0.105 0.511**
0.658** 0.665** 0.202** 0.653** 0.462** 0.354** 0.064 0.777**
0.778** 0.253** 0.418** 0.285** 0.075 0.295** 0.609**
0.100 0.460** 0.480** 0.179* 0.182* 0.615**
0.516** 0.044 0.118 0.072 0.121
0.336** 0.358** 0.060 0.489**
0.417** 0.327** 0.159*
0.253** 0.027
0.286**
40.989 74.220 0 621
0.731 1.317 3.277 3.690
0.176 1.827 5.353 3.876
2.114 1.966 7.945 2.191
1.448 2.743 2.558 10.845
0.166 3.288 5.079 16.096
0.488 0.290 0 0.951
1.974 1.838 0.041 13.825
1.128 0.761 0.099 3.714
**
Total 189 observations. * p < 0.05. ** p < 0.01.
3.861 1.323 0.915 6.323
SOURCES OF EXTERNAL TECHNOLOGY, ABSORPTIVE CAPACITY, AND INNOVATION CAPABILITY
1245
Table 2. Distribution of patent counts Year
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
Number of patent applications 0
1–10
11–50
50–100
101–250
>250
2 2 2 2 4 2 3 4 4 0
11 11 9 10 8 9 8 5 6 4
5 7 6 6 5 5 6 6 6 5
2 1 4 1 2 3 3 3 2 1
1 0 0 2 2 2 1 3 2 8
0 0 0 0 0 0 0 0 1 3
Table 3. Results from negative binomial regressions Coefficient
RD FTI
Fixed effect (1)
(2)
(3)
(4)
(5)
(6)
0.461** (0.136) 0.116 (0.079)
0.502** (0.133) 0.084 (0.101) 0.272** (0.095)
0.337** (0.143) 0.124 (0.099) 0.341** (0.101) 0.124** (0.047)
0.345** (0.130) 0.123 (0.099) 0.342** (0.100) 0.124** (0.046)
0.394** (0.126) 0.074 (0.091) 0.239** (0.092) 0.088** (0.039)
0.484 (0.459) 0.209** (0.050) 0.133 (0.144) 0.039 (0.137)
0.209 (0.496) 0.217** (0.049) 0.222* (0.127) 0.030 (0.134)
0.227 (0.499) 0.189** (0.050) 0.220* (0.119) 0.017 (0.134)
0.157 (0.207) 0.160 (0.106) 0.362** (0.107) 0.192** (0.073) 0.091 (0.076) 0.174 (0.511) 0.185** (0.050) 0.224* (0.127) 0.024 (0.135)
0.225 (0.498) 0.189** (0.050) 0.214** (0.109)
0.308 (0.412) 0.199** (0.049) 0.173 (0.118) 0.188 (0.123)
189 21
189 21
189 21
189 21
189 21
189 21
596.23 260.74
591.94 280.37
588.53 321.56
587.78 315.97
588.54 320.04
724.16 387.42
DTP RDFTI RDDTP JVPCT INTEN CAPINT LGSIZE No. of observations No. of sectors Log-likelihood Wald Statistics
Random effect
Standard errors are in parentheses; coefficients for constant terms and year dummies are not reported for brevity. * p < 0.1. ** p < 0.05.
for nested models suggests that this model is not preferred to the model reported in column 3. When the interaction term RDDTP is added to the model reported in columns 1 and 2, the likelihood ratio tests always favor the model without the interaction term. Both tests suggest that no important leverage effect can be found between R&D and DTP. Thus, the model specification (3) is the most preferred of all the specifications. To alleviate the concern of multicollinearity, a model specification without LGSIZE is estimated. In this case, the mean and maximum VIF values are 2.27 and 3.83, indicating that multicollinearity is quite trivial. The results reported in column 5 are found to be almost the same as those obtained from the preferred specification. As an additional check, the model in column 3 is estimated from a random effect negative binomial regression, the results of which are similar to those displayed in column 3. From these findings, one can conclude that absorptive capacity is more crucial for assimilating foreign technology than for exploiting domestic technology.
It may be wondered whether the way that the three knowledge stock variables are constructed has any detectable impact on the estimated results, since they were computed from an assumed growth rate g and depreciation rate d. To check for the robustness of these results, Table 4 lists four groups of results estimated from the preferred model with knowledge stock variables constructed from different combinations of g and d. These combinations reflect a possible range of g and d, that is, from 5% to 35%. These four groups of results are reasonably similar to those obtained in the preferred model in column 3 of Table 3, although the magnitude of estimated coefficients varies somewhat. Where both g and d are set at 5%, the direct effect of FTI knowledge is even significantly negative, which further strengthens the argument that FTI knowledge itself, does not promote innovation in domestic firms. From the econometric analysis, it is evident that investing in in-house R&D and DTP can directly facilitate the introduc-
1246
WORLD DEVELOPMENT Table 4. Robustness check
(g, d)
RD FTI DTP RDFTI JVPCT INTEN CAPINT LGSIZE No. of observations No. of sectors Log-likelihood Wald statistics
(5%, 5%)
(35%, 5%)
(5%, 35%)
(35%, 35%)
(1)
(2)
(3)
(4)
**
**
0.132 (0.170) 0.336** (0.108) 0.465** (0.105) 0.135** (0.045) 0.541 (0.489) 0.185** (0.049) 0.245** (0.109) 0.073 (0.128)
0.332 (0.151) 0.129 (0.103) 0.343** (0.102) 0.091** (0.038) 0.287 (0.499) 0.199** (0.050) 0.208* (0.122) 0.008 (0.134)
0.510 (0.130) 0.002 (0.090) 0.201** (0.083) 0.088* (0.046) 0.032 (0.514) 0.196** (0.051) 0.198 (0.130) 0.022 (0.135)
0.503** (0.134) 0.032 (0.091) 0.171** (0.081) 0.083* (0.046) 0.115 (0.517) 0.196** (0.052) 0.182 (0.133) 0.012 (0.136)
189 21
189 21
189 21
189 21
585.27 334.77
589.70 313.88
590.37 306.25
591.11 305.22
Standard errors are in parentheses; coefficients for constant terms and year dummies are not reported for brevity. * p < 0.1. ** p < 0.05.
tion of innovations by an indigenous firm. Although this does not hold true for knowledge obtained through FTI, the findings do reveal that firms can take advantage of the leverage effect through in-house R&D. This result provides evidence supporting Cohen and Levinthal’s (1990) argument of internal absorptive capacity. It is consistent with the results reported in Liu and White (1997) and Li and Wu (2010), but contrasts those of Liu and Buck (2007) which found that the significance of the FTI is independent of the absorptive capacity of domestic firms. This is probably because the latter study used new product sales as a measure of innovation, which overestimates the rate of innovation, as previously discussed. Surprisingly, in all model specifications, the coefficients of JVPCT are statistically insignificant, showing the presence of foreign firms has no detectable impact on the indigenous innovation of domestic firms. One can claim that the presence of foreign firms in local markets has two opposing effects on innovation in domestic firms. On the one hand, competitive pressure from market penetration by foreign firms encourages domestic firms to innovate more actively and build their own competencies. On the other hand, the relative technological advantage of foreign firms may be a barrier for indigenous firms to patent or patent around. The existence of a large technological gap may also induce Chinese firms to devote more attention to learning and less to innovation. Given the weak technological capability in most Chinese indigenous firms in the early period, this conjecture does not seem unreasonable. However, solid empirical evidence is still needed. Among all the model specifications in Tables 3 and 4, the estimated coefficients of INTEN are significantly negative, indicating that, in rapidly evolving industries requiring a large amount of R&D commitment, indigenous firms have a relative disadvantage in terms of patenting. The underlying reason might be that firms in R&D intensive sectors are confronted with more fierce competition from foreign frontier companies and thus it is more difficult to patent or patent around. Although a careful analysis of this issue is not possible in this
study due to the limitation of data availability, a comparison of R&D expenditure and patent applications across sectors provides further support for this finding. In 1995, among all 21 sectors, the seven most R&D-intensive sectors applied for only 27.2% of the patents, while they invested 62.9% on R&D expenditure. In comparison, the seven sectors with moderate R&D intensity accounted for 62.1% of patent applications, although their share of R&D was less than 27%. During the sample period, the general pattern has actually not changed much. In 2004, for example, about a half of R&D expenditure was invested in the seven high-R&D sectors. Their contribution to patent applications, however, accounts for less than 20%. It seems to suggest that different sectors have different innovation patterns. Furthermore, from a learning perspective, greater technological opportunity represents a greater amount of external knowledge, which, in turn, increases the importance of learning. Cohen and Levinthal (1990) postulated that the faster the pace of knowledge generation in a field the more necessary the R&D efforts for developing absorptive capacity or learning. As a result, firms in R&D intensive sectors are probably more involved in learning or catch-up than innovation. If this is the case, one would not be surprised to find the negative impact of INTEN. In terms of the estimated coefficient CAPINT, it is significantly positive in the preferred model, suggesting that the capital intensity of a sector is favorable to the innovation of domestic firms in Chinese high-tech industries. Higher capital intensity means a higher entry barrier and less competitive market structure. Firms in such a sector will be able to appropriate their innovation benefits more effectively and, therefore, have a stronger incentive to innovate. In this sense, the result is intuitive and justifiable. Finally, the sector size is found to be insignificant across all model specifications. One reason may be that the sector-level investment in R&D and other channels for obtaining external knowledge are closely related to sector size, as demonstrated in the correlation matrix. The effect of sector size cannot
SOURCES OF EXTERNAL TECHNOLOGY, ABSORPTIVE CAPACITY, AND INNOVATION CAPABILITY
properly be identified and estimated. One way to circumvent this problem is to conduct a firm-level analysis, which is clearly beyond the scope of this research. 5. CONCLUSIONS In this paper, the effect of three types of investment in acquiring technological knowledge (in-house R&D, FTI and DTP) on Chinese SOEs’ innovation capacity in 21 high-tech sectors was empirically examined. Taking domestic patent applications as a measure of innovation output and the four-digit SIC sector as the unit of analysis, a sector-level knowledge production function with the fixed effect negative binomial count model technique was estimated. The data were constructed from a panel of 21 sectors from 1995 to 2004, which were drawn from recent sector level statistics published by the NBSC. Interesting findings revealed themselves from the estimated results. It was found that FTI knowledge alone does not effectively promote the innovation of domestic firms in terms of patents, although its contribution is prominent in the early development stage of Chinese high-tech industries. A significant leverage effect between in-house R&D and FTI was apparent, which is consistent with the finding reported in Liu and White (1997) and provides empirical evidence for Cohen and Levinthal’s argument for absorptive capacity. This implies that the purchase of foreign technology alone is not conducive to a firm’s innovation performance unless it is coupled with investment in R&D. From a strategic perspective, firms can better use foreign technology by first enhancing their absorptive capacity, which can be gained through their own R&D activities. By contrast, the assimilation and utilization of domestic technology is less demanding for absorptive capacity, as shown in Li and Wu (2010). Besides the relatively low sophistication of domestic technology in terms of knowledge complexity, the transferability of tacit knowledge between firms with similar contextual knowledge may also play a role. Therefore, when acquiring external technology, firms are advised to pay particular attention to contextual factors
1247
that might have influenced the accumulation of tacit knowledge associated with the technology. On the other hand, it also indicates that firms can start with appropriate domestic technology for competence building when they have a low level of absorptive capacity. This research contributes to the literature of innovation in two aspects. First, it validates the importance of learning to innovation in developing countries, and lends further support for the argument of absorptive capacity. Second, it reveals that the role of absorptive capacity is contingent upon the source and nature of external technology, which is related to the contextual factors influencing the creation and accumulation of the associated tacit knowledge. In terms of policy and research implications, although the findings in this analysis are drawn from the Chinese situation, they can also shed light on catch-up strategies for non-frontier firms in a broad spectrum of other developing economies, because firms in these economies share similarities with Chinese indigenous firms. In particular, they rely heavily on knowledge developed externally for technological capability and have multiple sources of external knowledge to draw upon. The learning effect of R&D is also crucial for them to effectively assimilate and utilize external technology. How to strike a balance between investing in R&D and purchasing external technology developed abroad and domestically is always an important strategic decision for these firms to make. This study will hopefully serve as a starting point to inspire future research interests in a larger and broader context. Finally, in interpreting findings reported in this analysis, there are several caveats. One limitation is associated with measurement of innovation output and the construction of knowledge stocks. In particular, knowledge generated from different sources may depreciate at different rates, which will inevitably complicate analysis. Use of a sector as the unit of analysis also leads to a rather small sample, making small sample bias a possible concern, and endogeneity is difficult to deal with. In addition, issues examined in this analysis are better investigated at the firm level. All of these warrant further research.
REFERENCES Acs, Z. J., & Audretsch, D. B. (1988). Innovation in large and small firms: An empirical analysis. American Economic Review, 78(4), 678–690. Acs, Z. J., Anselin, L., & Varga, A. (2002). Patents and innovation counts as measures of regional production of new knowledge. Research Policy, 31, 1069–1085. Amsden, A. (1989). Asia’s next giant: South Korea and late industrialization. New York: Oxford University Press. ´ . (2002). Methods for using patents in cross-country Archambault, E comparisons. Scientometrics, 54(1), 15–30. Archibugi, D. (1992). Patenting as an indicator of technological innovation: A review. Science and Public Policy, 19(6), 357–368. Arora, A., Fosfuri, A., & Gambardella, A. (2001). Markets for technology and their implications for corporate strategy. Industrial and Corporate Change, 10(2), 419–451. Basant, R., & Fikkert, B. (1996). The effects of R&D, foreign technology purchase, and domestic and international spillovers on productivity in Indian firms. Review of Economics & Statistics, 78(2), 187–199. Belsley, D. A., Kuh, E., & Welsch, R. E. (1980). Regression diagnostics: Identifying influential data and sources of collinearity. New York: John Wiley and Sons. Blundell, R., Griffith, R., & Windmeijer, F. (2002). Individual effects and dynamics in count data model. Journal of Econometrics, 108, 113–131. Borensztein, E., Gregorio, J., & Lee, J. (1998). How does foreign direct investment affect economic growth. Journal of International Economics, 45, 115–135.
Cameron, A.C. & Trivedi, P.K. (1998). Regression analysis of count data. Cambridge University Press. Cheung, K., & Lin, P. (2004). Spillover effects of FDI on innovation in China: Evidence from the provincial data. China Economic Review, 15(1), 25–44. Cincera, M. (1997). Patents, R&D, and technological spillovers at the firm level: Some evidences from econometric count models for patent data. Journal of Applied Econometrics, 12(12), 265–280. Coe, D., & Helpman, E. (1995). International R&D spillovers. European Economic Review, 39, 859–887. Cohen, W. M., & Levinthal, D. A. (1989). Innovation and Learning: Two faces of R&D. Economic Journal, 99, 569–596. Cohen, W. M., & Levinthal, D. A. (1990). Absorptive capacity: A new perspective on learning and innovation. Administrative Science Quarterly, 35, 128–152. Cre´pon, B., & Duguet, E. (1997). Estimating the innovation function from patent numbers: GMM on count panel data. Journal of Applied Econometrics, 12, 243–263. Forbes, N., & Wield, D. (2000). Managing R&D in technology-followers. Research Policy, 29, 1095–1109. Furman, J. L., Porter, M. E., & Stern, S. (2002). The determinants of national innovative capacity. Research Policy, 31, 899–933. Girma, S. (2005). Absorptive capacity and productivity spillovers from FDI: A threshold regression analysis. Oxford Bulletin of Economics and Statistics, 67(3), 281–306.
1248
WORLD DEVELOPMENT
Glass, A., & Saggi, K. (2002). Multinational firms and technology transfer. Scandinavian Journal of Economics, 104, 1153–1191. Griffith, R., Redding, S., & Van Reenen, J. (2004). Mapping the two faces of R&D: Productivity growth in a panel of OECD industries. Review of Economics and Statistics, 86(4), 883–895. Griliches, Z., & Mairesse, J. (1984). Productivity and R&D at the firm level. In Z. Griliches (Ed.), R&D, patents, and productivity (pp. 339–374). Chicago: University of Chicago Press. Griliches, Z. (1979). Issues in assessing the contribution of research and development to productivity growth. Bell Journal of Economics, 10(1), 92–116. Griliches, Z. (1990). Patent statistics as economic indicators: A survey. Journal of Economic Literature, 28(4), 1661–1707. Grossman, G., & Helpman, E. (1991). Innovation and growth in the global economy. Cambridge, MA: MIT Press. Hall, B. H., & Mairesse, J. (1995). Exploring the relationship between R&D and productivity in French manufacturing firms. Journal of Econometrics, 65, 263–293. Hall, B. H., Griliches, Z., & Hausman, J. (1986). Patents and R and D: Is there a lag?. International Economic Review, 27(2), 265–283. Hausman, J., Hall, B. H., & Griliches, Z. (1984). Economic-models for count data with an application to the patents R and D relationship. Econometrica, 52(4), 909–938. Hu, A. G., & Jefferson, G. H. (2002). FDI impact of spillover: Evidence from China’s electronic and textile industries. The World Economy, 25(8), 1063–1076. Hu, A. G., & Jefferson, G. H. (2004). Returns to research and development in Chinese industry: Evidence from state-owned enterprises in Beijing. China Economic Review, 15, 86–107. Hu, A. G., Jefferson, G. H., & Qian, J. (2005). R&D and technology transfer: Firm-level evidence from Chinese industry. Review of Economics and Statistics, 87(4), 780–786. Jaffe, A. B. (1986). Technological opportunity and spillover of R&D: Evidence from firms’ patents, profits, and market value. American Economic Review, 76(5), 984–1001. Kang, R., & Yang, Y. (1991). A historical review of foreign technology import during the past 40 years in China. Management World (in Chinese), 6, 168–174. Kneller, R. (2005). Frontier technology, absorptive capacity and distance. Oxford Bulletin of Economics and Statistics, 67(1), 1–23. Li, X., & Wu, G. (2010). In-house R&D, technology purchase and innovation: Empirical evidences from Chinese hi-tech industries, 1995– 2004. International Journal of Technology Management, 51(2/3/4), 217–238. Liu, X., & Buck, T. (2007). Innovation performance and channels for international technology spillovers: Evidence from Chinese high-tech industries. Research Policy, 36, 355–366. Liu, X., & Wang, C. (2003). Does foreign direct investment facilitate technological progress? Evidence from Chinese industries. Research Policy, 32(6), 945–953. Liu, X., & White, S. (1997). The relative contributions of foreign technology and domestic inputs to innovation in Chinese manufacturing industries. Technovation, 17(3), 119–125. Liu, X., & White, S. (2001). Comparing innovation systems: A framework and application to China’s transitional context. Research Policy, 30, 1091–1114. Lu, F., & Feng, K. (2005). The policy choice to develop our state’s automobile industry with independent intellectual property rights. Beijing: Peking University Press. Lu, F. (2005). Thinking about the strategy of developing China’s civilian aircraft industry. Chinese Soft Science (In Chinese), 4, 10–16. MacGarvie, M. (2006). Do firms learn from international trade? Review of Economics and Statistics, 88, 46–60. Madariaga, N., & Poncet, S. (2007). FDI in Chinese cities: Spillovers and impact on growth. World Economy, 30(5), 837–862. National Bureau of Statistics of China (2002–2005). China statistical yearbook on high technology industry. Beijing, China: China Statistics Press. Nelson, R. R., & Winter, S. G. (1982). An evolutionary theory of economic change. Cambridge, MA: Harvard University Press.
Pavitt, K. (1988). Uses and abuses of patents statistics. In A. F. J. van Raan (Ed.), Handbook of quantitative studies of science and technology (pp. 509–536). Amsterdam: Elsevier Science Publishers. Schumpeter, J. (1942). Capitalism, socialism and democracy. New York: Harper and Row. Wang, Q., & Gao, J. (2006). A retrospective analysis of the foreign capital attracting practice based on the “exchange-market-for-technology” strategy in China. Journal of Hubei University (Philosophy and Social Science), 33(3), 261–264. Xu, J. (1996). Technology import in China: Status, problems and prospect. Quantitative and Technology Economics Research (in Chinese), 3, 24–32. Xu, Y. (2000). Research on the definition of Chinese hi-tech industries (In Chinese). Beijing, China: National Research Center of Science and Technology Development. Yao, S. (2006). On economic growth, FDI and exports in China. Applied Economics, 38, 339–351. Zhang, K. (2001). How does foreign direct investment affect economic growth in China? Economics of Transition, 9(3), 679–693.
APPENDIX A
Table A1. 4-Digit SIC sectors in the Chinese high-tech industries Industries and sectors
R&D intensity in 2004 (%)
Medical and pharmaceutical products Chemical pharmaceutical products Processing of traditional chinese medicine Biology products
1.40 1.41 2.39
Aircraft and spacecraft Aircraft Spacecraft
4.85 13.82
Electronics and telecommunication equipment Telecommunication transmission unit Telecommunication exchange unit Telecommunication terminal unit Radar & peripheral equipment Broadcast & television equipment Electronic vacuum apparatus Semiconductor separated parts Integrated circuits Electronic components Household audiovisual equipment Other electronic equipment
2.92 8.01 0.63 3.69 1.93 1.57 1.73 1.23 0.76 1.25 0.54
Computers and office equipment Computers Peripheral equipment of computers Office equipment
0.46 0.47 0.49
Medical equipment and meters Medical equipment & instruments Instruments & meters
1.39 1.86
Note: R&D Intensity is calculated for Large and Medium-Sized Enterprises of all ownership types. Data are drawn from the National Bureau of Statistics of China (2002–2005).
Available online at www.sciencedirect.com
doi:10.1016/j.worlddev.2010.05.011
Sources of External Technology, Absorptive Capacity, and Innovation Capability in Chinese State-Owned High-Tech Enterprises XIBAO LI * Tsinghua University, Beijing, PR China Summary. — This paper examines the pattern of innovation and learning among state-owned enterprises in Chinese high-tech sectors and empirically estimates the impact of three types of investment for acquiring technological knowledge—in-house R&D, importing foreign technology, and purchasing domestic technology—on the innovation capabilities of firms. Based on a panel dataset consisting of 21 high-tech sectors during the period 1995–2004, an augmented knowledge production function is estimated. The results show that importing foreign technology alone does not facilitate innovation in Chinese state-owned high-tech enterprises, unless in-house R&D is also conducted. Domestic technology purchases, however, are found to have a favorable direct impact on innovation, suggesting that firms have less difficulty in absorbing domestic technological knowledge than utilizing foreign technology and that absorptive capacity is contingent upon the source or nature of the external knowledge. Ó 2011 Elsevier Ltd. All rights reserved. Key words — R&D, foreign technology import, domestic technology purchase, absorptive capacity, innovation capability
1. INTRODUCTION
basis of which technology and innovation are developed are difficult to transfer across firms. Cohen and Levinthal (1990) argued that the importance of external technology on technological competence and innovation capability depends on whether recipient firms have related prior knowledge or absorptive capacity to understand and exploit technological opportunities. They further noted that in-house R&D has a function of learning and helps recipient firms build up their absorptive capacity (Cohen & Levinthal, 1989). Since R&D is also crucial for firms to innovate and generate new knowledge, given the dual functions of in-house R&D, what is the pattern of innovation and learning among non-frontier firms in developing countries? How and under what conditions can these firms make the best use of the three different sources of knowledge—in-house R&D, imported foreign technology, and acquired domestic technology—to promote innovation? Does the pattern of learning and innovation in firms vary with the sources of external knowledge? To address the above issues, this study examines the pattern of learning and innovation among state-owned and stateholding enterprises (SOEs) in Chinese high-tech sectors, and empirically examines the impact on innovation capability of both internally developed knowledge via in-house R&D and externally acquired knowledge through the technology market. In particular, the study considers two forms of external disembodied technological knowledge which are originally generated in different contexts: technology imported from foreign countries and domestic technology licensed by universities, research institutes, or other domestic firms. Following the definition of the National Bureau of Statistics of China
The role of in-house R&D in a firm’s innovation has been recognized since Schumpeter’s (1942) work, and the relationship between R&D and innovation outcomes has been extensively studied within the framework of knowledge production functions in previous empirical studies (see e.g., Cincera, 1997; Furman, Porter, & Stern, 2002; Hausman, Hall, & Griliches, 1984; Jaffe, 1986). Beside in-house R&D, a firm can also obtain useful technological knowledge from technology markets (Arora, Fosfuri, & Gambardella, 2001), which in turn contribute to the firm’s knowledge generation or innovation. For firms in late-comer economies, exploiting technological knowledge created in developed countries is particularly common and often ranks among the most important strategies for catching-up (Amsden, 1989). Previous studies in the research stream of international R&D spillover emphasized the importance of knowledge spillovers embodied in foreign technological products (Coe & Helpman, 1995; Grossman & Helpman, 1991; MacGarvie, 2006). By contrast, inadequate attention has been paid to the role of disembodied knowledge, such as designs, formulae, drawings, processes, patents, and knowhow. Such knowledge is often licensed or acquired by recipient firms through a formal channel and, thus, may be directly exploited. Firms can obtain licenses for disembodied knowledge from both developed countries and domestic innovators, such as universities, research institutes, and other innovating firms. In China, for example, during the early stage of economic transition, indigenous firms not only spent a large amount of money on importing foreign technology (Kang & Yang, 1991; Xu, 1996), but were also advised to acquire useful knowledge from the domestic technology market (Liu & White, 2001). As a result, a complicated pattern of learning and innovation has emerged in Chinese industries or sectors. The complexity of learning and innovation among catch-up firms lies in the differences in the nature and sources of knowledge. Due to the tacit and context-specific nature of technological knowledge, it is difficult, if not impossible, for recipient firms to acquire innovation capacity through the mere license or purchase of external technology. For instance, Nelson and Winter (1982) pointed out that organizational routines on the
* This paper benefits greatly from the constructive comments and suggestions of Dr. Xiaolan Fu, two anonymous reviewers, and participants in the Oxford conference on “Confronting the Challenge of Technology for Development: Experiences from the BRICS”, May 30, 2008. The author also gratefully acknowledges the financial support provided by the National Natural Science Foundation of China (Project No.: 70602005) and The United Nations Industrial Development Organization (UNIDO). Final revision accepted: May 20, 2010. 1240
SOURCES OF EXTERNAL TECHNOLOGY, ABSORPTIVE CAPACITY, AND INNOVATION CAPABILITY
(NBSC), foreign technology import (FTI) expenditures are defined as including purchases of design knowledge, formulae, drawings, processes, patents and know-how, and key equipment closely related to new product development. They do not include those used directly for production, such as production lines, complete knock-down kits, and turn-key facilities (Liu & White, 1997). In this sense, a large portion of FTI expenditures is spent on disembodied knowledge. FTI mainly takes the form of technological licensing and is substantially different from the import of foreign machinery and equipment, which represents a form of embodied knowledge. Similarly, domestic technology purchase (DTP) refers to the licensing of outside technologies developed by universities, research institutes, or other firms, representing a form of knowledge transfer from domestic sources. Hence, both FTI and DTP reflect a firm’s efforts to acquire technological capabilities from external sources. They can be transacted in markets for technology, as defined in Arora et al. (2001). The major difference between FTI and DTP lies in the source of technological knowledge. This analysis is based on a panel dataset constructed from publicly available official statistics on five Chinese high-tech industry categories: medical and pharmaceutical products, aircraft and spacecraft, electronic and telecommunication equipment, computer and office equipment, and medical instruments and meters. The categories of Chinese high-tech industries are similar to and comparable with those used in OECD countries, although they are not as intensive in R&D as their counterparts (Xu, 2000). With the four-digit Chinese standard industrial classification (SIC) sector as the unit of analysis, a panel of 21 sectors was constructed and sector-level data for the period from 1995 to 2004 analyzed. Table A1, in the appendix, gives a list of all sectors included. Drawing upon the previous studies (Acs, Anselin, & Varga, 2002; Furman et al., 2002; Li & Wu, 2010), domestic patent application counts were used to measure innovation capability. A firm’s innovation capability is thought to be determined by three sources of technological knowledge that the firm obtained from inhouse R&D, foreign technology import, and domestic technology transfer. The contribution to the innovation capability of the three sources of knowledge is empirically estimated within a framework of knowledge production functions (Basant & Fikkert, 1996; Hall & Mairesse, 1995; Hu, Jefferson, & Qian, 2005). This paper contributes to the literature on absorptive capacity and the role of in-house R&D for catch-up firms. The estimated results show that SOEs in the Chinese high-tech industries can gain leverage by investing in both in-house R&D and FTI, while technology imports alone do not contribute to the rate of patenting. By contrast, investing in domestic technology alone facilitates a firm’s innovation. The firm can easily assimilate and utilize domestic technological knowledge, independent of its absorptive capacity. That is, absorptive capacity is crucial for assimilating foreign technology, but not for taking advantage of domestic knowledge. The finding suggests that the argument for absorptive capacity is contingent upon the source or nature of external knowledge and that the diffusion of domestic technology is also conducive to fostering innovation. The remainder of this paper is organized as follows. Section 2 presents a brief review of the relevant literature. Section 3 introduces the estimating strategy employed in the analysis, followed by a description of the data construction and summary statistics. The results and main findings are reported in Section 4. Section 5 concludes with some caveats and areas for further research.
1241
2. LITERATURE REVIEW In consideration of their weak technological capability, firms in developing countries often regard the use of technology developed abroad as an expedient way to develop and expand their manufacturing capacity (Amsden, 1989), instead of investing intensively in R&D. In China, for example, the presence of foreign capital as well as the use of imported technology from developed countries has contributed considerably to the country’s rapid economic growth since the 1980s (Hu & Jefferson, 2002; Liu & Wang, 2003; Madariaga & Poncet, 2007; Yao, 2006; Zhang, 2001). Besides foreign technology, the Chinese firms also look for appropriate technologies from domestic sources, such as university labs, government-run research institutes, or even local competitors (Li & Wu, 2010; Liu & White, 2001). Irrespective of whether the external knowledge is developed abroad or domestically, what really matters is whether recipient firms can assimilate and effectively exploit such external knowledge to build up their own innovation capability. In reality, however, it proves rather difficult for recipient firms to effectively utilize knowledge from external sources to boost their own innovation. During the early catch-up stage in China, despite the government’s wishful policy goal of “trading market for technology”, it was often found that Chinese indigenous firms could not even acquire or have access to state-of-the-art technology through direct foreign investment inflow and import of foreign technology (Wang & Gao, 2006). One situation that occurs frequently among Chinese indigenous firms is that the very weak technological capability of some domestic firms forces them to rely continuously on foreign technology. Such examples have been noted in the Chinese automobile and civilian aircraft industries, which are discussed in Lu (2005) and Lu and Feng (2005), respectively. One explanation for this phenomenon relates to the nature of the knowledge to be assimilated. Technology acquired through FTI and DTP contains both codified (e.g., patents, formulae, and drawings) and tacit (e.g., know-how) knowledge. Although the transmission of codified knowledge usually ends with the acquisition of documents and manuals, tacit knowledge associated with use of external technology does not easily transfer with the purchase of a technology. Whether the licensees can effectively take advantage of such licensed knowledge is closely contingent on the level of their prior related knowledge or absorptive capacity. According to Cohen and Levinthal (1990, p128), absorptive capacity refers to “the ability of a firm to recognize the value of new, external information, assimilate it and apply it to commercial ends”. It strongly hinges on the level of the firm’s prior related knowledge, which can, in turn, be developed through its own R&D efforts. From this perspective, in-house R&D not only helps a firm generate new knowledge (i.e., innovation effect) but also contributes to its absorptive capacity (i.e., learning effect) as well. It thus plays a dual role (Cohen & Levinthal, 1989; Griffith, Redding, & Van Reenen, 2004). Forbes and Wield (2000) contended that the learning function of R&D is particularly relevant and important in developing countries. In view of the two functions of in-house R&D and the two sources of external knowledge, it seems an intricate matter for licensee firms to make the best use of both internally developed and externally licensed knowledge for their own innovation. This reasoning suggests that the knowledge portfolio has an important influence on a firm’s innovation performance. Although recent studies have confirmed the contribution of FTI to productivity and growth in developing countries (Basant & Fikkert, 1996; Hu et al., 2005), the impact of foreign
1242
WORLD DEVELOPMENT
technology on the innovation capacity of local firms has not been adequately discussed. The limited empirical findings in previous work are at best mixed or inconclusive. Liu and White (1997) found that firms can gain leverage when investing in both R&D personnel and foreign technology in highly innovative industries, while imported technology alone has no direct impact on a firm’s innovation measured by its share of new product sales. By contrast, Liu and Buck (2007) found that learning-by-importing -technology can directly accelerate a local firm’s introduction of new products, independent of absorptive capacity. Two empirical studies thus provide contrasting results on the importance of absorptive capacity, and few studies have ever considered the impact of DTP, despite its importance as an additional source of external knowledge. Furthermore, it remains unclear how important the leverage effect of R&D is for a recipient firm to exploit domestic technology. An exception is the recent work by Li and Wu (2010). These authors addressed both issues and found that the contribution of foreign knowledge to the innovation of Chinese firms is strongly dependent on the investment of the firms in R&D, while R&D does not leverage the effect of domestic technology. Based on region-industry level data, their study makes no distinction between domestic firms and foreign-investment joint ventures. This is problematic given the fact that foreign firms rarely purchase domestic technology in China and have a quite different pattern of learning and innovation from domestic firms. To fill this gap, the current analysis adds to Li and Wu (2010) by focusing on an important type of domestic firm, that is, only the SOE. To understand how essential and different the learning effect of R&D is for assimilating knowledge from different sources, it is wise to review the basic argument of absorptive capacity. Cohen and Levinthal (1990) posited that the importance of absorptive capacity is determined by both the quantity of knowledge to be assimilated and utilized and the difficulty of learning. The difficulty or ease of learning is largely determined by the characteristics of the underlying technological knowledge, which include the complexity of knowledge to be assimilated and degree to which outside knowledge is targeted to the needs and concerns of a firm. Following this reasoning, it may be expected that a systematic difference exists between technological knowledge imported from foreign countries and that licensed from domestic sources. First, since the technological level in many Chinese industries is far from technology frontiers, foreign imported technology is normally more complex and sophisticated than the best domestic technology. Secondly, since the accumulation process of tacit knowledge is influenced by both internal (e.g., organizational routines, structures, coordination mechanism) and external (e.g., market structure, competition, regulation) context-specific factors, it is rather difficult for recipient firms in one context to fully understand and assimilate tacit knowledge created in another context. According to Nelson and Winter (1982), the more tacit the relevant knowledge the more difficult it is to assimilate and exploit. Differences in contextual factors are presumably much larger across countries than between firms. As a result, it will be more challenging to draw on foreign technology than domestic knowledge. Furthermore, labor mobility is much easier within a country than across national borders. Related technological knowledge may not be essential for a recipient firm to exploit external knowledge generated by another local firm, especially when it can acquire such capacity by hiring those with the requisite knowledge from other local firms. Taking all these points together, it is reasonable to argue that exploiting foreign technology is a more demanding task than utilizing domestic technology for Chinese firms.
In another stream of research, empirical studies on the technology gap provide ample evidence for the argument that complexity and sophistication of external knowledge have an influential impact on the incentives of firms to develop absorptive capacity, and the learning effect of R&D in developing absorptive capacity is contingent on the level of the technology gap. For instance, Griffith et al. (2004) and Kneller (2005) demonstrated that the further a country is behind the world technological frontier, the greater the importance of its R&D investment in creating absorptive capacity. Researchers found that only firms above a certain technological level were likely to benefit from external technology spillovers (Borensztein, Gregorio, & Lee, 1998; Girma, 2005; Glass & Saggi, 2002), implying that the larger the technology gap, the more crucial the absorptive capacity. In China, the technology gap between a firm’s own technology and imported foreign technology is much larger than the gap between its own technology and licensed domestic technology. Thus, it is expected that the leverage effect of absorptive capacity on the effect of external knowledge will be different between the two sources of outside knowledge. 3. EMPIRICAL STRATEGY (a) Methodology To compare the impact of knowledge developed from different sources on innovation, and to empirically examine the mode of learning and innovation in Chinese indigenous firms, an augmented knowledge production function is proposed as the basis of economic analysis and uses the number of domestic patent applications as the measure of innovation performance. Taking into account the specific nature of patent counts in the dataset (Cameron & Trivedi, 1998), it is assumed that the number of sector level patents can be specified in a count panel data model as follows: 9 8 dt þ b1 RDit þ b2 FTI it þ b3 DTP it > > > > > = < þb ðRD FTI Þ þ b ðRD DTP Þ > it it it it 4 5 ; EðPatentit Þ ¼ exp > > þb6 JVPCT it þ b7 INTEN it > > > > ; : þb8 CAPINT it þ b9 LGSIZEit ð1Þ where Patentit is the patent counts for sector i in year t. RDit, FTIit and DTPit represent log-scaled knowledge stocks KRit, KFit, and KDit, respectively. KRit is the stock of technical knowledge generated by the in-house R&D investment of firms in sector i in year t; KFit is the stock of foreign technical knowledge imported by sector i; and KDit is the stock of domestic technology acquired by firms in sector i. dt represents year dummy and b’s are coefficients to be estimated from the model. In light of the fact that absorptive capacity can be created through in-house R&D, two interaction terms (RDFTI) and (RDDTP) are included to account for the leverage effect of absorptive capacity on assimilating and utilizing knowledge generated through FTI and DTP, respectively. Thus, the model specification (1) incorporates the dual role of in-house R&D. It implicitly posits that a firm is unable to assimilate externally generated knowledge passively. In order to exploit such knowledge, the firm must invest in absorptive capacity, which can, in turn, be created by investment in R&D. Along with the three variables of knowledge capital, four control variables are incorporated in the model. The first
SOURCES OF EXTERNAL TECHNOLOGY, ABSORPTIVE CAPACITY, AND INNOVATION CAPABILITY
variable, JVPCT, is the sales share from overseas owned or invested firms in each sector (including Hong Kong Special Administrative Region of China, Macao Special Administrative Region of China and Taiwan Province of China). It is intended to capture the implicit knowledge spillover from these oversea firms, and account for the potential effect of FDI which is regarded as another important channel through which foreign technologies influence accumulation of domestic technological capabilities (Cheung & Lin, 2004). Because the unit of this analysis is the high-tech sector, two sector level variables (INTEN and CAPINT) are added to allow for sector heterogeneity. The variable INTEN reflects sector-level R&D intensity and is defined as R&D expenditure per unit of sales. CAPINT is sectoral capital intensity, measured as the ratio of the book value of fixed assets to total outputs. Finally, sector size is controlled by incorporating the variable LGSIZE which is measured by the logged sector-level sales. By including year dummies in the model, time-related sector-wide effects are subsumed in the coefficient of the dummy years. It should be noted that there are several other channels through which technology diffuses, such as technical information provided by original equipment manufacturer (OEM) contracts, reverse engineering efforts, and labor mobility (Liu & Wang, 2003). Since most of these leave no paper trail from which to analyze the flow of technology and its impact, in the proposed model, only the three sources of technological knowledge are focused on, that is, in-house R&D, FTI, and DTP. (b) Data and measurement Given the importance of high-tech industries to the national economy, the NBSC has been collecting information on science and technology activities from high-tech firms. For the purpose of this analysis, this official source of information was used to investigate innovation activities across high-tech sectors. Specifically, the data used all came from office statistics published by the NBSC, in the series China Statistics Yearbook on High Technology Industry (2002–2005), which covers the period from 1995 to 2004. The Yearbook contains about 30 major indicators of innovation activities in five Chinese high-tech industrial categories: pharmaceuticals, aircraft and spacecraft, telecom, computers, and instruments. Firm-level data were originally reported to the NBSC by all large and medium-sized enterprises (LME). To maintain the confidentiality of the information from the firms, the NBSC aggregated firm-level data according to a combination of firm ownership and four-digit SIC sectors. To highlight the impact of different sources of external technology, in this study, only SOEs were considered for two reasons. First, these enterprises accounted for a dominant share of the FTI and R&D during the sample period. In 1995, for example, the SOEs account for 76.2% and 81.4% of R&D investments and FTI expenditures in Chinese high-tech industries. Despite continuous ownership reform in China, in 2004, over 40% of R&D expenditures in Chinese high-tech industries still came from the SOEs. Second, from a practical standpoint, data information on FTI and DTP was unavailable for firms under other ownership categories. Therefore, each sector in this analysis consists only of SOEs which are occasionally referred to as indigenous firms. In previous studies, both Liu and White (1997) and Liu and Wang (2003) used new product sales as a measure of innovation. In this paper, however, the domestic patent applications count was employed as a proxy indicator of endogenous innovation output. In terms of measuring the endogenous or indigenous features of innovation, these two measures are largely different. In China, if a product is designated by the
1243
government as new, the firm can obtain a tax subsidy from the provincial or national government. For this reason, firms have a strong incentive to over-record the sales of new products. In addition, the procedures for new product approval are neither completely standardized nor comparable between regions. The newness of products is a relatively arbitrary and geographically-bounded concept. According to the Chinese statistical convention, a product designated as new can simply be new to a local market, such as a county, a city, or a province, whether or not it has been on the market in other places. Alternatively, a product can be regarded as new as long as it is new to a firm and has not been produced for more than 1 year. Thus, the measure of new product sales will inevitably bring in some measurement bias. In innovation literature, the use of patent statistics also raises important measurement issues (Archambault, 2002; Archibugi, 1992; Griliches, 1990; Pavitt, 1988). It is widely recognized that patent information can neither include all important technological innovations nor reflect the importance of different innovations. There are pitfalls associated with equating patenting with the level of innovative activity in both “quantity” and “quality” perspectives. Nonetheless, patents usually contain technological improvements and/or innovative ideas that have, at least, to be new to the country. More importantly, in contrast to new product sales, the procedures for filing patents are uniform across all sectors and industries. In this sense, patents can be regarded as a good reflection of endogenous or indigenous innovation efforts. In effect, it is the best indicator available in the NBSC aggregated level dataset to reflect the indigenous nature of an innovation. In the empirical estimation of R&D-patenting knowledge production functions, two measures of R&D, or knowledge, are usually considered. For example, Hausman et al. (1984), Acs and Audretsch (1988), Cincera (1997) and Blundell, Griffith, and Windmeijer (2002) used the current and past flow of R&D expenditure as inputs, and assumed that all past R&D investments are substitutes, with a unit elasticity of substitution. Due to the high persistence of R&D expenditure, this specification often leads to multicollinearity problems between lagged regressors. Drawing on empirical literature on R&D and productivity (Hall & Mairesse, 1995), Cre´pon and Duguet (1997), instead, employed estimated R&D capital stock. According to these studies, the annual flow of R&D expenditure is taken to be investment adding to a firm’s knowledge capital. Knowledge capital depreciates over time, so that the contribution of past R&D becomes less valuable as time passes (Griliches, 1979). One advantage of this specification is that it allows for both complementarity and depreciation in past R&D expenditures. This paper adopts the second approach, to construct stock variables for the three different sources of knowledge. Following common practice in the previous work, the variable of knowledge capital stocks generated from R&D can be estimated with a perpetual inventory model (Hall & Mairesse, 1995). The initial knowledge stock SR1 and the knowledge stock at the beginning of year t (SRt) are computed from annual R&D investment (FRt), as follows: FR1 ; ð2Þ SR1 ¼ ðg þ dÞ SRt ¼ ð1 dÞSRt1 þ FRt1 ;
t P 2:
ð3Þ
Here, g denotes the pre-sample growth rate of annual R&D flow FRt, and d is the annual depreciation rate of R&D investment. One drawback associated with using knowledge stock in (3) as technological input is that it ignores the contribution of
1244
WORLD DEVELOPMENT
current R&D investment in patenting. Hall, Griliches, and Hausman (1986) showed that a strong contemporaneous relationship exists between in-house R&D investment and patenting. To incorporate the contemporaneous effect, the in-house R&D stocks at the middle of year t were used, which are computed as: KRt ¼ ðSRt þ SRtþ1 Þ=2;
aggregated at sector level, the patent applications filed by Chinese high-tech firms are rather limited. Before 2001, more than half of the sectors had annual total patent applications of no more than ten, although since then firms in more regions took out more than that number annually. This indicates that innovation capability in Chinese indigenous firms is at a very low level in terms of patenting.
ð4Þ
1 6 t 6 T:
Both SRt and SRtþ1 are obtained from (2) and (3). The effect of contemporaneous R&D investment is thus reflected in KRt through SRtþ1 . To compute the initial R&D stock SR1, two parameters, g and d, have to be determined (Griliches & Mairesse, 1984). Since the specification of d and g does not affect the results significantly, as confirmed in the literature (Hall & Mairesse, 1995), a pre-sample growth rate and an annual depreciation rate were assumed, both at 15% as in previous studies (Hu & Jefferson, 2004). The same approach was adopted to construct the knowledge stocks generated from FTI and DTP. In the computation, in-house R&D, FTI, and DTP expenditures are all adjusted by GDP deflators to their 1995 constant values. Since the book value of fixed assets used in constructing the variable CAPINT is not available for 1995, the empirical analysis is based on data from 1996 to 2004, which amounts to 189 observations. Table 1 provides a description of the correlation matrix and summary statistics of the constructed variables. The correlation matrix shows that there are moderate correlations between knowledge stock variables and sector size, which is not surprising given that the sector size affects the amount of investment in acquiring knowledge from all sources. A multicollinearity diagnostic test tells that the highest Variance Inflation Factor (VIF) is 5.29, below the critical point of 10 (Belsley, Kuh, & Welsch, 1980), suggesting that multicollinearity is not severe and no remedial action should be considered. As is usually found in patent data, the number of domestic patent applications in this analysis has more weight on the right tail than expected from the usual Poisson distributions, which results in a much larger variance than their means. It indicates that patent counts are over-dispersed. To account for the over-dispersion of patent counts in estimating, the fixed effect negative binomial regression proposed by Hausman et al. (1984) was used. Table 2 further describes the distribution of the patent application counts in different years. While
4. RESULTS AND DISCUSSIONS Table 3 reports the results from negative binomial regressions, based on various model specifications. The estimation proceeds with the usual maximum likelihood techniques. First, a simple case was considered where only two sources of knowledge stocks (RD and FTI) were included (column 1). It can clearly be seen that only in-house R&D contributes significantly to the introduction of patents. Although one might expect FTI to have a similar effect on innovation, the coefficient of FTI is insignificant although positive, suggesting that the impact of FTI on innovation is not as important as expected. When the term DTP is included, its estimated coefficient is found to be significantly positive (column 2), indicating that the impact of acquired domestic knowledge is important to patenting. Furthermore, the model including the interaction term (RDFTI), the result of which is reported in column 3, is found to be preferred to the one listed in column 2. In this case, the estimated coefficient of the interaction term is significantly positive, indicating that positive leverage can be gained by firms investing in both in-house R&D and FTI. It suggests that, when coupled with in-house R&D, FTI has a significant positive impact on patent applications, while the purchasing of foreign technology alone does not. To examine whether a similar leverage effect exists between in-house R&D and DTP, one more interaction term RDDTP was added to the model specification (1) (column 4). In this specification, the estimated coefficients of both RD and RDFTI still remain significantly positive, while the coefficient of RDDTP is not significant statistically. Surprisingly, the coefficient of R&D knowledge stock is now insignificant. This is probably because of a severe multicollinearity among the explanatory variables. In this case, the VIF statistics for the variable RD is 11.67, larger than the common cut-off value, indicating a serious multicollinearity. The likelihood ratio test
Table 1. Correlation matrix and descriptive statistics Variable
2 3 4 5 6 7 8 9 10 Mean Std. Dev. Min Max
Patent
RD
FTI
DTP
RDFTI
RDDTP
JVPCT
INTEN
CAPINT
LGSIZE
1
2
3
4
5
6
7
8
9
10
0.515 0.379** 0.341** 0.306** 0.202** 0.024 0.001 0.105 0.511**
0.658** 0.665** 0.202** 0.653** 0.462** 0.354** 0.064 0.777**
0.778** 0.253** 0.418** 0.285** 0.075 0.295** 0.609**
0.100 0.460** 0.480** 0.179* 0.182* 0.615**
0.516** 0.044 0.118 0.072 0.121
0.336** 0.358** 0.060 0.489**
0.417** 0.327** 0.159*
0.253** 0.027
0.286**
40.989 74.220 0 621
0.731 1.317 3.277 3.690
0.176 1.827 5.353 3.876
2.114 1.966 7.945 2.191
1.448 2.743 2.558 10.845
0.166 3.288 5.079 16.096
0.488 0.290 0 0.951
1.974 1.838 0.041 13.825
1.128 0.761 0.099 3.714
**
Total 189 observations. * p < 0.05. ** p < 0.01.
3.861 1.323 0.915 6.323
SOURCES OF EXTERNAL TECHNOLOGY, ABSORPTIVE CAPACITY, AND INNOVATION CAPABILITY
1245
Table 2. Distribution of patent counts Year
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
Number of patent applications 0
1–10
11–50
50–100
101–250
>250
2 2 2 2 4 2 3 4 4 0
11 11 9 10 8 9 8 5 6 4
5 7 6 6 5 5 6 6 6 5
2 1 4 1 2 3 3 3 2 1
1 0 0 2 2 2 1 3 2 8
0 0 0 0 0 0 0 0 1 3
Table 3. Results from negative binomial regressions Coefficient
RD FTI
Fixed effect (1)
(2)
(3)
(4)
(5)
(6)
0.461** (0.136) 0.116 (0.079)
0.502** (0.133) 0.084 (0.101) 0.272** (0.095)
0.337** (0.143) 0.124 (0.099) 0.341** (0.101) 0.124** (0.047)
0.345** (0.130) 0.123 (0.099) 0.342** (0.100) 0.124** (0.046)
0.394** (0.126) 0.074 (0.091) 0.239** (0.092) 0.088** (0.039)
0.484 (0.459) 0.209** (0.050) 0.133 (0.144) 0.039 (0.137)
0.209 (0.496) 0.217** (0.049) 0.222* (0.127) 0.030 (0.134)
0.227 (0.499) 0.189** (0.050) 0.220* (0.119) 0.017 (0.134)
0.157 (0.207) 0.160 (0.106) 0.362** (0.107) 0.192** (0.073) 0.091 (0.076) 0.174 (0.511) 0.185** (0.050) 0.224* (0.127) 0.024 (0.135)
0.225 (0.498) 0.189** (0.050) 0.214** (0.109)
0.308 (0.412) 0.199** (0.049) 0.173 (0.118) 0.188 (0.123)
189 21
189 21
189 21
189 21
189 21
189 21
596.23 260.74
591.94 280.37
588.53 321.56
587.78 315.97
588.54 320.04
724.16 387.42
DTP RDFTI RDDTP JVPCT INTEN CAPINT LGSIZE No. of observations No. of sectors Log-likelihood Wald Statistics
Random effect
Standard errors are in parentheses; coefficients for constant terms and year dummies are not reported for brevity. * p < 0.1. ** p < 0.05.
for nested models suggests that this model is not preferred to the model reported in column 3. When the interaction term RDDTP is added to the model reported in columns 1 and 2, the likelihood ratio tests always favor the model without the interaction term. Both tests suggest that no important leverage effect can be found between R&D and DTP. Thus, the model specification (3) is the most preferred of all the specifications. To alleviate the concern of multicollinearity, a model specification without LGSIZE is estimated. In this case, the mean and maximum VIF values are 2.27 and 3.83, indicating that multicollinearity is quite trivial. The results reported in column 5 are found to be almost the same as those obtained from the preferred specification. As an additional check, the model in column 3 is estimated from a random effect negative binomial regression, the results of which are similar to those displayed in column 3. From these findings, one can conclude that absorptive capacity is more crucial for assimilating foreign technology than for exploiting domestic technology.
It may be wondered whether the way that the three knowledge stock variables are constructed has any detectable impact on the estimated results, since they were computed from an assumed growth rate g and depreciation rate d. To check for the robustness of these results, Table 4 lists four groups of results estimated from the preferred model with knowledge stock variables constructed from different combinations of g and d. These combinations reflect a possible range of g and d, that is, from 5% to 35%. These four groups of results are reasonably similar to those obtained in the preferred model in column 3 of Table 3, although the magnitude of estimated coefficients varies somewhat. Where both g and d are set at 5%, the direct effect of FTI knowledge is even significantly negative, which further strengthens the argument that FTI knowledge itself, does not promote innovation in domestic firms. From the econometric analysis, it is evident that investing in in-house R&D and DTP can directly facilitate the introduc-
1246
WORLD DEVELOPMENT Table 4. Robustness check
(g, d)
RD FTI DTP RDFTI JVPCT INTEN CAPINT LGSIZE No. of observations No. of sectors Log-likelihood Wald statistics
(5%, 5%)
(35%, 5%)
(5%, 35%)
(35%, 35%)
(1)
(2)
(3)
(4)
**
**
0.132 (0.170) 0.336** (0.108) 0.465** (0.105) 0.135** (0.045) 0.541 (0.489) 0.185** (0.049) 0.245** (0.109) 0.073 (0.128)
0.332 (0.151) 0.129 (0.103) 0.343** (0.102) 0.091** (0.038) 0.287 (0.499) 0.199** (0.050) 0.208* (0.122) 0.008 (0.134)
0.510 (0.130) 0.002 (0.090) 0.201** (0.083) 0.088* (0.046) 0.032 (0.514) 0.196** (0.051) 0.198 (0.130) 0.022 (0.135)
0.503** (0.134) 0.032 (0.091) 0.171** (0.081) 0.083* (0.046) 0.115 (0.517) 0.196** (0.052) 0.182 (0.133) 0.012 (0.136)
189 21
189 21
189 21
189 21
585.27 334.77
589.70 313.88
590.37 306.25
591.11 305.22
Standard errors are in parentheses; coefficients for constant terms and year dummies are not reported for brevity. * p < 0.1. ** p < 0.05.
tion of innovations by an indigenous firm. Although this does not hold true for knowledge obtained through FTI, the findings do reveal that firms can take advantage of the leverage effect through in-house R&D. This result provides evidence supporting Cohen and Levinthal’s (1990) argument of internal absorptive capacity. It is consistent with the results reported in Liu and White (1997) and Li and Wu (2010), but contrasts those of Liu and Buck (2007) which found that the significance of the FTI is independent of the absorptive capacity of domestic firms. This is probably because the latter study used new product sales as a measure of innovation, which overestimates the rate of innovation, as previously discussed. Surprisingly, in all model specifications, the coefficients of JVPCT are statistically insignificant, showing the presence of foreign firms has no detectable impact on the indigenous innovation of domestic firms. One can claim that the presence of foreign firms in local markets has two opposing effects on innovation in domestic firms. On the one hand, competitive pressure from market penetration by foreign firms encourages domestic firms to innovate more actively and build their own competencies. On the other hand, the relative technological advantage of foreign firms may be a barrier for indigenous firms to patent or patent around. The existence of a large technological gap may also induce Chinese firms to devote more attention to learning and less to innovation. Given the weak technological capability in most Chinese indigenous firms in the early period, this conjecture does not seem unreasonable. However, solid empirical evidence is still needed. Among all the model specifications in Tables 3 and 4, the estimated coefficients of INTEN are significantly negative, indicating that, in rapidly evolving industries requiring a large amount of R&D commitment, indigenous firms have a relative disadvantage in terms of patenting. The underlying reason might be that firms in R&D intensive sectors are confronted with more fierce competition from foreign frontier companies and thus it is more difficult to patent or patent around. Although a careful analysis of this issue is not possible in this
study due to the limitation of data availability, a comparison of R&D expenditure and patent applications across sectors provides further support for this finding. In 1995, among all 21 sectors, the seven most R&D-intensive sectors applied for only 27.2% of the patents, while they invested 62.9% on R&D expenditure. In comparison, the seven sectors with moderate R&D intensity accounted for 62.1% of patent applications, although their share of R&D was less than 27%. During the sample period, the general pattern has actually not changed much. In 2004, for example, about a half of R&D expenditure was invested in the seven high-R&D sectors. Their contribution to patent applications, however, accounts for less than 20%. It seems to suggest that different sectors have different innovation patterns. Furthermore, from a learning perspective, greater technological opportunity represents a greater amount of external knowledge, which, in turn, increases the importance of learning. Cohen and Levinthal (1990) postulated that the faster the pace of knowledge generation in a field the more necessary the R&D efforts for developing absorptive capacity or learning. As a result, firms in R&D intensive sectors are probably more involved in learning or catch-up than innovation. If this is the case, one would not be surprised to find the negative impact of INTEN. In terms of the estimated coefficient CAPINT, it is significantly positive in the preferred model, suggesting that the capital intensity of a sector is favorable to the innovation of domestic firms in Chinese high-tech industries. Higher capital intensity means a higher entry barrier and less competitive market structure. Firms in such a sector will be able to appropriate their innovation benefits more effectively and, therefore, have a stronger incentive to innovate. In this sense, the result is intuitive and justifiable. Finally, the sector size is found to be insignificant across all model specifications. One reason may be that the sector-level investment in R&D and other channels for obtaining external knowledge are closely related to sector size, as demonstrated in the correlation matrix. The effect of sector size cannot
SOURCES OF EXTERNAL TECHNOLOGY, ABSORPTIVE CAPACITY, AND INNOVATION CAPABILITY
properly be identified and estimated. One way to circumvent this problem is to conduct a firm-level analysis, which is clearly beyond the scope of this research. 5. CONCLUSIONS In this paper, the effect of three types of investment in acquiring technological knowledge (in-house R&D, FTI and DTP) on Chinese SOEs’ innovation capacity in 21 high-tech sectors was empirically examined. Taking domestic patent applications as a measure of innovation output and the four-digit SIC sector as the unit of analysis, a sector-level knowledge production function with the fixed effect negative binomial count model technique was estimated. The data were constructed from a panel of 21 sectors from 1995 to 2004, which were drawn from recent sector level statistics published by the NBSC. Interesting findings revealed themselves from the estimated results. It was found that FTI knowledge alone does not effectively promote the innovation of domestic firms in terms of patents, although its contribution is prominent in the early development stage of Chinese high-tech industries. A significant leverage effect between in-house R&D and FTI was apparent, which is consistent with the finding reported in Liu and White (1997) and provides empirical evidence for Cohen and Levinthal’s argument for absorptive capacity. This implies that the purchase of foreign technology alone is not conducive to a firm’s innovation performance unless it is coupled with investment in R&D. From a strategic perspective, firms can better use foreign technology by first enhancing their absorptive capacity, which can be gained through their own R&D activities. By contrast, the assimilation and utilization of domestic technology is less demanding for absorptive capacity, as shown in Li and Wu (2010). Besides the relatively low sophistication of domestic technology in terms of knowledge complexity, the transferability of tacit knowledge between firms with similar contextual knowledge may also play a role. Therefore, when acquiring external technology, firms are advised to pay particular attention to contextual factors
1247
that might have influenced the accumulation of tacit knowledge associated with the technology. On the other hand, it also indicates that firms can start with appropriate domestic technology for competence building when they have a low level of absorptive capacity. This research contributes to the literature of innovation in two aspects. First, it validates the importance of learning to innovation in developing countries, and lends further support for the argument of absorptive capacity. Second, it reveals that the role of absorptive capacity is contingent upon the source and nature of external technology, which is related to the contextual factors influencing the creation and accumulation of the associated tacit knowledge. In terms of policy and research implications, although the findings in this analysis are drawn from the Chinese situation, they can also shed light on catch-up strategies for non-frontier firms in a broad spectrum of other developing economies, because firms in these economies share similarities with Chinese indigenous firms. In particular, they rely heavily on knowledge developed externally for technological capability and have multiple sources of external knowledge to draw upon. The learning effect of R&D is also crucial for them to effectively assimilate and utilize external technology. How to strike a balance between investing in R&D and purchasing external technology developed abroad and domestically is always an important strategic decision for these firms to make. This study will hopefully serve as a starting point to inspire future research interests in a larger and broader context. Finally, in interpreting findings reported in this analysis, there are several caveats. One limitation is associated with measurement of innovation output and the construction of knowledge stocks. In particular, knowledge generated from different sources may depreciate at different rates, which will inevitably complicate analysis. Use of a sector as the unit of analysis also leads to a rather small sample, making small sample bias a possible concern, and endogeneity is difficult to deal with. In addition, issues examined in this analysis are better investigated at the firm level. All of these warrant further research.
REFERENCES Acs, Z. J., & Audretsch, D. B. (1988). Innovation in large and small firms: An empirical analysis. American Economic Review, 78(4), 678–690. Acs, Z. J., Anselin, L., & Varga, A. (2002). Patents and innovation counts as measures of regional production of new knowledge. Research Policy, 31, 1069–1085. Amsden, A. (1989). Asia’s next giant: South Korea and late industrialization. New York: Oxford University Press. ´ . (2002). Methods for using patents in cross-country Archambault, E comparisons. Scientometrics, 54(1), 15–30. Archibugi, D. (1992). Patenting as an indicator of technological innovation: A review. Science and Public Policy, 19(6), 357–368. Arora, A., Fosfuri, A., & Gambardella, A. (2001). Markets for technology and their implications for corporate strategy. Industrial and Corporate Change, 10(2), 419–451. Basant, R., & Fikkert, B. (1996). The effects of R&D, foreign technology purchase, and domestic and international spillovers on productivity in Indian firms. Review of Economics & Statistics, 78(2), 187–199. Belsley, D. A., Kuh, E., & Welsch, R. E. (1980). Regression diagnostics: Identifying influential data and sources of collinearity. New York: John Wiley and Sons. Blundell, R., Griffith, R., & Windmeijer, F. (2002). Individual effects and dynamics in count data model. Journal of Econometrics, 108, 113–131. Borensztein, E., Gregorio, J., & Lee, J. (1998). How does foreign direct investment affect economic growth. Journal of International Economics, 45, 115–135.
Cameron, A.C. & Trivedi, P.K. (1998). Regression analysis of count data. Cambridge University Press. Cheung, K., & Lin, P. (2004). Spillover effects of FDI on innovation in China: Evidence from the provincial data. China Economic Review, 15(1), 25–44. Cincera, M. (1997). Patents, R&D, and technological spillovers at the firm level: Some evidences from econometric count models for patent data. Journal of Applied Econometrics, 12(12), 265–280. Coe, D., & Helpman, E. (1995). International R&D spillovers. European Economic Review, 39, 859–887. Cohen, W. M., & Levinthal, D. A. (1989). Innovation and Learning: Two faces of R&D. Economic Journal, 99, 569–596. Cohen, W. M., & Levinthal, D. A. (1990). Absorptive capacity: A new perspective on learning and innovation. Administrative Science Quarterly, 35, 128–152. Cre´pon, B., & Duguet, E. (1997). Estimating the innovation function from patent numbers: GMM on count panel data. Journal of Applied Econometrics, 12, 243–263. Forbes, N., & Wield, D. (2000). Managing R&D in technology-followers. Research Policy, 29, 1095–1109. Furman, J. L., Porter, M. E., & Stern, S. (2002). The determinants of national innovative capacity. Research Policy, 31, 899–933. Girma, S. (2005). Absorptive capacity and productivity spillovers from FDI: A threshold regression analysis. Oxford Bulletin of Economics and Statistics, 67(3), 281–306.
1248
WORLD DEVELOPMENT
Glass, A., & Saggi, K. (2002). Multinational firms and technology transfer. Scandinavian Journal of Economics, 104, 1153–1191. Griffith, R., Redding, S., & Van Reenen, J. (2004). Mapping the two faces of R&D: Productivity growth in a panel of OECD industries. Review of Economics and Statistics, 86(4), 883–895. Griliches, Z., & Mairesse, J. (1984). Productivity and R&D at the firm level. In Z. Griliches (Ed.), R&D, patents, and productivity (pp. 339–374). Chicago: University of Chicago Press. Griliches, Z. (1979). Issues in assessing the contribution of research and development to productivity growth. Bell Journal of Economics, 10(1), 92–116. Griliches, Z. (1990). Patent statistics as economic indicators: A survey. Journal of Economic Literature, 28(4), 1661–1707. Grossman, G., & Helpman, E. (1991). Innovation and growth in the global economy. Cambridge, MA: MIT Press. Hall, B. H., & Mairesse, J. (1995). Exploring the relationship between R&D and productivity in French manufacturing firms. Journal of Econometrics, 65, 263–293. Hall, B. H., Griliches, Z., & Hausman, J. (1986). Patents and R and D: Is there a lag?. International Economic Review, 27(2), 265–283. Hausman, J., Hall, B. H., & Griliches, Z. (1984). Economic-models for count data with an application to the patents R and D relationship. Econometrica, 52(4), 909–938. Hu, A. G., & Jefferson, G. H. (2002). FDI impact of spillover: Evidence from China’s electronic and textile industries. The World Economy, 25(8), 1063–1076. Hu, A. G., & Jefferson, G. H. (2004). Returns to research and development in Chinese industry: Evidence from state-owned enterprises in Beijing. China Economic Review, 15, 86–107. Hu, A. G., Jefferson, G. H., & Qian, J. (2005). R&D and technology transfer: Firm-level evidence from Chinese industry. Review of Economics and Statistics, 87(4), 780–786. Jaffe, A. B. (1986). Technological opportunity and spillover of R&D: Evidence from firms’ patents, profits, and market value. American Economic Review, 76(5), 984–1001. Kang, R., & Yang, Y. (1991). A historical review of foreign technology import during the past 40 years in China. Management World (in Chinese), 6, 168–174. Kneller, R. (2005). Frontier technology, absorptive capacity and distance. Oxford Bulletin of Economics and Statistics, 67(1), 1–23. Li, X., & Wu, G. (2010). In-house R&D, technology purchase and innovation: Empirical evidences from Chinese hi-tech industries, 1995– 2004. International Journal of Technology Management, 51(2/3/4), 217–238. Liu, X., & Buck, T. (2007). Innovation performance and channels for international technology spillovers: Evidence from Chinese high-tech industries. Research Policy, 36, 355–366. Liu, X., & Wang, C. (2003). Does foreign direct investment facilitate technological progress? Evidence from Chinese industries. Research Policy, 32(6), 945–953. Liu, X., & White, S. (1997). The relative contributions of foreign technology and domestic inputs to innovation in Chinese manufacturing industries. Technovation, 17(3), 119–125. Liu, X., & White, S. (2001). Comparing innovation systems: A framework and application to China’s transitional context. Research Policy, 30, 1091–1114. Lu, F., & Feng, K. (2005). The policy choice to develop our state’s automobile industry with independent intellectual property rights. Beijing: Peking University Press. Lu, F. (2005). Thinking about the strategy of developing China’s civilian aircraft industry. Chinese Soft Science (In Chinese), 4, 10–16. MacGarvie, M. (2006). Do firms learn from international trade? Review of Economics and Statistics, 88, 46–60. Madariaga, N., & Poncet, S. (2007). FDI in Chinese cities: Spillovers and impact on growth. World Economy, 30(5), 837–862. National Bureau of Statistics of China (2002–2005). China statistical yearbook on high technology industry. Beijing, China: China Statistics Press. Nelson, R. R., & Winter, S. G. (1982). An evolutionary theory of economic change. Cambridge, MA: Harvard University Press.
Pavitt, K. (1988). Uses and abuses of patents statistics. In A. F. J. van Raan (Ed.), Handbook of quantitative studies of science and technology (pp. 509–536). Amsterdam: Elsevier Science Publishers. Schumpeter, J. (1942). Capitalism, socialism and democracy. New York: Harper and Row. Wang, Q., & Gao, J. (2006). A retrospective analysis of the foreign capital attracting practice based on the “exchange-market-for-technology” strategy in China. Journal of Hubei University (Philosophy and Social Science), 33(3), 261–264. Xu, J. (1996). Technology import in China: Status, problems and prospect. Quantitative and Technology Economics Research (in Chinese), 3, 24–32. Xu, Y. (2000). Research on the definition of Chinese hi-tech industries (In Chinese). Beijing, China: National Research Center of Science and Technology Development. Yao, S. (2006). On economic growth, FDI and exports in China. Applied Economics, 38, 339–351. Zhang, K. (2001). How does foreign direct investment affect economic growth in China? Economics of Transition, 9(3), 679–693.
APPENDIX A
Table A1. 4-Digit SIC sectors in the Chinese high-tech industries Industries and sectors
R&D intensity in 2004 (%)
Medical and pharmaceutical products Chemical pharmaceutical products Processing of traditional chinese medicine Biology products
1.40 1.41 2.39
Aircraft and spacecraft Aircraft Spacecraft
4.85 13.82
Electronics and telecommunication equipment Telecommunication transmission unit Telecommunication exchange unit Telecommunication terminal unit Radar & peripheral equipment Broadcast & television equipment Electronic vacuum apparatus Semiconductor separated parts Integrated circuits Electronic components Household audiovisual equipment Other electronic equipment
2.92 8.01 0.63 3.69 1.93 1.57 1.73 1.23 0.76 1.25 0.54
Computers and office equipment Computers Peripheral equipment of computers Office equipment
0.46 0.47 0.49
Medical equipment and meters Medical equipment & instruments Instruments & meters
1.39 1.86
Note: R&D Intensity is calculated for Large and Medium-Sized Enterprises of all ownership types. Data are drawn from the National Bureau of Statistics of China (2002–2005).
Available online at www.sciencedirect.com