FDI and Innovation in China:
To What Extent Does Inward FDI Determine
Regional Innovativeness?
Master’s Thesis
University of Groningen
Faculty of Economics and BusinessAbstract
China has experienced rapid growth over the last three decades, primarily driven by factor accumulation. However, due to a declining working-age population and diminishing returns to capital, this growth model is no longer. As a result, China actively develops policies to transform its economy towards an innovative powerhouse. The goal of this thesis was to assess whether inward foreign direct investment could act as a driver for this transition by examining whether FDI impacts regional innovativeness. By using a panel data set of 15 Chinese provinces over the period 1997-2009, this study finds that FDI has a significant and negative impact on regional innovation. Furthermore, this study also identifies that total R&D expenditure figures neglect the role of different organizations, such as research and educational institutions, and private firms. By differentiating R&D expenditure, measured as science and technology expenditure, this study shows that only R&D efforts by educational institutions significantly impacts innovativeness.
3
Table of Content
1. Introduction ... 4
2. Current Changes and Provincial Disparities ... 7
3. Theoretical Background ... 11
3.1 Defining Innovation ... 11
3.2 The Importance of Innovation for Economic Growth ... 13
3.3 Systems of Innovation ... 14
3.4 FDI and Innovation ... 17
4. Literature Review ... 19
4.1 Hypotheses ... 23
5. Methodology and Data ... 24
5.1 Data Sources and Data Description ... 25
4
1. Introduction
Within a timespan of about three decades, China managed to rise as a major economic power. Between 1984 and 2014, real gross domestic product (GDP) grew on average by 10 percent annually, thereby lifting the average GDP per capita from $1,200 to $13,400 in just 30 years1. Even more remarkable is that since the
beginning of the economic reforms in 1979, total output has risen from $699 billion to $18,257 billion (on a purchasing power parity basis) in 2014, making China one of the largest global economies2. In fact, China’s economy now accounts for
approximately 16 percent of the total world output3. However, the pace of China’s
growth has recently started to slowdown. According to The World bank, growth has decelerated to 7.4 percent in 2014 and 7.1 percent in 2015, and is expected to
drop below seven percent by 2017 (World Bank, 2015a)4. Given China’s role in the
world economy and the current deceleration of China’s growth, the question of whether China will be able to sustain long-term growth is of importance for world economic activity (ECB, 2013).
In this thesis, I investigate whether inward foreign direct investment can act as a driver for innovativeness in regional China, and thereby indirectly contribute to sustained long-term growth. Innovation, or technological change, is important for long-term economic growth because of the law of diminishing returns. Investments in physical capital can lead to short-term output growth, but, keeping other factors constant, profits from capital will eventually diminish progressively causing growth to cease (Solow, 1957). Only technological progress can sustain long-term output growth by increasing productivity, thereby offsetting diminishing returns
1Based on data from The Conference Board. 2016. The Conference Board Total Economy Databasetm, May 2016,
http://www.conference-board.org/data/economydatase/. The data is presented in 2015 US$ with updated 2011 purchasing power parities.
2See Figure A1 in the Appendix.
3 Own calculation based on China’s GDP data from The Conference Board and world GDP estimates from The World
Factbook 2016. Washington, DC: Central Intelligence Agency, 2016, http://www.cia.gov/library/publications/the-world-factbook/. Both sources are in 2015 US$.
4Note that there is a debate on the accuracy on China’s official GDP data. Therefore, several databases present different
5 to capital (Ibid). Therefore, innovation can be seen as the single most important factor for long-term economic growth (Rosenberg, 2004).
In the context of China, growth over the last decades was primarily driven by factor accumulation, i.e. increases in physical capital and labor (OECD, 2015). Given the law of diminishing returns and a declining working-age population, China’s growth model is no longer sustainable (World Bank, 2015b). In fact, China’s growth pace regarding labor productivity has been decelerating since 20105. As a response,
China has developed policies to achieve sustainable long-term growth. These policies aim to transform China into an innovative society by 20206. Several goals
include: R&D as a percentage of GDP greater than 2.5 percent, a rank among the top five global countries in terms of patenting, and advances in science and technology (S&T) contributing at least 60 percent to economic growth (EC, 2015). A key component herein is to continue to attract foreign direct investment (FDI) (Higgins, 2015). Indeed, China has received $129 billion in 2014, making it the world’s largest recipient of FDI inflows in 2014 (UNCTAD, 2015). In theory, FDI can be seen as a bundle of technology, capital, and know-how that can produce spillover effects in the form of job creation, technology transfers, upgrading of local skills, and by fostering local competition (World Bank, 2010). Therefore, FDI can theoretically contribute to the government’s plans of an innovative society by transferring technology. Another important aspect of these plans is to achieve coordinated and balanced development among rural and urban areas. Currently, growth and innovation has been unevenly distributed among provinces as will be shown in the section following this introduction.
The goal of this thesis is to assess whether foreign direct investment can act as a driver for regional innovation in China, and thereby contribute to a more coordinated and balanced innovative economy. This study complements to the existing literature in two ways. First, there is relative scarce empirical literature available on the drivers of provincial innovation in China. The literature that
5Based on data from The Conference Board. 2016. The Conference Board Total Economy Databasetm, May 2016,
http://www.conference-board.org/data/economydatase/.
6Based on the Medium and Long Term S&T Development Plan 2006-2020, State Council,
6 exists has found either, positive (Cheung & Lin, 2004; Fu, 2008; Roelfsema & Zhang, 2012), negative (Zhou, 2012; Hale, 2007), and inconclusive (Chen, 2007) results. With this study I complement to the existing literature by examining whether inward FDI acts as a driver for regional innovativeness once more. Second, besides having inward FDI as an independent variable, I use control variables that have not been tested before in the innovation systems (IS) approach. Previous studies used total R&D expenditure figures per region, whereas I differentiate between R&D expenditure contributed by different organizations. As such, the data allows me to ascertain which type of organization contributes more (or less) to innovativeness. These control variables are more detailed and appropriate in the context of provincial China.
To answer my research question, I employ a panel data analysis with fixed effects, covering 15 provinces over the period 1997-2009. The time period is chosen due to data limitations, but still covers 13 years. China can be divided into three regions, namely: eastern, central, and western China. Therefore, 15 provinces are equally split per region, meaning that each region is represented by five provinces. The empirical analysis is based on Regional Innovation System theory (RIS), complemented by the external factor of inward FDI.
7
2. Current Changes and Provincial Disparities
The Chinese government is well aware of the need to restructure the economy towards a sustainable long-term growth path (OECD, 2007). Faced with a decelerating growth pace in both GDP and productivity, the Ministry of Science and Technology (MOST) has developed the ‘Medium and Long Term S&T Development Plan 2006-20207 to transform China into an innovative society (Ibid).In order to attain sustainable long-term growth, the plan emphasizes a growing importance of indigenous innovation by stating:
‘At present, China is an economic superpower, but it not an economic powerhouse, a fundamental reason lies in weak innovation capacity’.8
In accordance with the governments 13th Five-Year Plan, emphasis is placed on
industrial upgrading towards the more innovative tertiary sector and stresses domestic consumption rather than exports. As shown in Appendix Figure A2, exports as a percentage of GDP is already declining, whereas household expenditure is slightly increasing in 2014. In addition, Appendix Figure A3 shows that the share of GDP of the secondary industry sector is declining and the share of the tertiary sector is rising9. In fact, a preliminary report for the first quarter of
2016 shows that the tertiary almost accounts for a 60% share in GDP10. Besides
these reforms, the 13th Five-Year Plan also focuses on coordination to achieve
balanced development among rural and urban areas. Currently, growth and innovation has been unevenly distributed among provinces, as illustrated in Figure 1. GDP per capita for total China in 2014 is almost five times higher than GDP per capita in 1995, approximating $13,500. However, eastern China almost reached a GDP per capita of $19,000 in 2014, whereas Central and Western China
7Medium and Long Term S&T Development Plan, 2006-2020, State Council,
http://www.gov.cn/jrzg/2006-02/09/content_183787.html
8Ibid, p.8.
9 According to the Industrial classification for National Economic Activities, China’s secondary industry refers to mining,
manufacturing, production and supply of electricity, heat, gas and water, and construction. The tertiary sector refers to the services sector.
10Preliminary Accounting Results of GDP for the Fourth Quarter and the Whole Year of 2015, NBS, 21 January 2016,
http://www.stats.gov.cn/english/PressRelease/201601/t20160121_1307717.html; (3) Preliminary Accounting Results of GDP for the First Quarter of 2016, NBS, 18 April 2016,
8 are far behind with respectively $11,100 and $10,600 GDP per capita. When total GDP is broken down per region, one can see that the largest share of GDP is contributed by eastern China, see Figure 2. Since 2000, the contribution of eastern China to total GDP has been rising, whereas the contribution of central and western China has been falling. According to these figures, China’s economic development has not been even across regions. The same applies for innovativeness when measured as the number of granted patents, see Figure 3. The top left corner graph presents the total amount of patents per type11, the other
three graph present the amount of each type of patent broken down per region.
Figure 1: GDP per capita (in 2014 US$)
Sources: (1) National Bureau of Statistics of China, NBS, GDP, http://data.stats.gov.cn/english/easyquery.htm?cn=C01; (2) The World Bank, World Development Indicators 2016, PPP conversion factor, http://data.worldbank.org/country/china; (3) The World bank, World Development Indicators 2016, US$ GDP deflator, http://data.worldbank.org/country/united-states
Notes: Nominal GDP in Yuan was transformed to GDP in current Intl $ using the PPP conversion factor. The World Bank GDP deflator with base year 2000 was rescaled to base year 2014 and used to construct GDP in constant 2014 US$.
9
Figure 2: Total GDP broken down per region (in 2014 US$)
Source: see Figure 1
Notes: Contrary to the earlier figure of $18,257 billion of GDP in 2015 US$, figure 2 shows a total GDP of $19,383 billion. This is caused by the use of different deflators and different base years. In the above figure, the World Bank deflators are used rescaled to 2014 as base year, because a 2015 deflator was not available.
Figure 3: Amount of patents per region
Source: National Bureau of Statistics of China, NBS, http://data.stats.gov.cn/english/.
482 125 84 0 100 200 300 400 500 600 Pa ten ts in Tho u san d s
Utility model patents
East Central West
276 32 35 0 100 200 300 400 1995 1998 2001 2004 2007 2010 2013 Pa ten ts in Tho u san d s
Design patents
East Central West
56.6% 56.4% 56.6% 56.9% 57.5% 25.0% 25.3% 25.2% 25.0% 24.6% 18.4% 18.3% 18.2% 18.1% 17.9% 0 2 4 6 8 10 12 14 16 18 20 1995 2000 2005 2010 2014 G DP in b ill ion s U S$
East Central West
114 24 20 0 20 40 60 80 100 120 1995 1998 2001 2004 2007 2010 2013 Pa ten ts in Tho u san d s
Invention patents
East Central West
1.192 158 690 344 0 200 400 600 800 1.000 1.200 1.400 1995 1998 2001 2004 2007 2010 2013 Pa ten ts in Tho u san d s
Total China
10 Eastern China produced 114 invention patents out of the national total of 158, i.e. 72 percent. A similar large contribution is observed for utility model and design patents, with respectively 70 and 80 percent produced by eastern China. Overall, eastern China has contributed 73 percent of all patents in China.
Bo do these disparities matter? According to the OECD (2012), “economic activity tends to concentrate in large metropolitan areas” (p.16). China’s eastern region contains mostly coastal provinces and it is the region with the densest metropolitan areas. Therefore, it should be no surprise that most of China’s economic activity is concentrated in this region. However, less developed regions can greatly contribute to national aggregate growth (OECD, 2012). In fact, less developed regions accounted for 43% of aggregate OECD growth (Ibid).
Efforts are made to address these disparities. For example, the Ninth Five-Year Plan of 1996-2000 initiated the beginning of encouraging economic development in central and western China (Lai, 2002). The ‘Great Western Development Strategy’, as it is called, aims to channel state investments, foreign loans and private capital into these regions (Ögütçü & Taube, 2002). In addition, efforts are also made regarding the level of human capital. Under ‘A Plan for Human Resource Development in the West’, initiated in 2000, the government aimed to attract talent in science, technology, and management and relocate teachers from the coastal regions to the west (Lai, 2002). Surely these policies have had a positive effect, as shown by Yu et al. (2008), but did not close the gap in terms of innovation and GDP between the coastal, central and western regions as shown by the previous figures. Consequently, the government is still pursuing coordinated and balanced growth among regions, according to the latest 13th Five-Year Plan.
Given the importance of the potential in less developed regions and the goals the government is aiming for, the question of how to achieve balanced growth and innovativeness among regions remains unanswered. Therefore, as discussed in the introduction, I am interested to investigate whether inward foreign direct investment can be a vehicle for greater regional innovativeness.
11
3. Theoretical Background
This section presents the theoretical background by first discussing the definition of innovation. Second, the importance of innovation for economic growth is discussed using Solow’s (1957) and Romer’s (1994) theories of economic growth, followed by the national- and regional innovation system (NIS/RIS) framework. Lastly, the role of foreign direct investment (FDI) in relation to innovation is discussed.
3.1 Defining Innovation
Innovation is a rather abstract concept, and as such, many definitions exist along with different perspectives and levels of analysis (Damanpour & Schneider, 2006). According to Schumpeter (1949), one of the first economists to define ‘innovation’, the entrepreneur creates innovation by creative response and, thereby, makes something new for commercial application outside of existing practice. The entrepreneur is not limited to an individual, instead whoever initiates the creative response fills the role of entrepreneur (Ibid).
12 The feature that is common to all definitions is that innovation implies something new and novelty (Pérez-Luño et al., 2010). Therefore, ‘how innovation is defined
and what it includes depends on the purpose of the research’ (Zhou, 2012, p.39). In
13 3.2 The Importance of Innovation for Economic Growth
In a rudimentary sense, economic growth can happen in two ways. Firstly, by increasing the factors of production such as capital and labor (Rosenberg, 2004). In fact, since the industrial revolution – over more than two hundred years ago- capital investments have continued (Grossman & Helpman, 1993). However, as Adam Smith notes:
‘As capitals increase in any country, the profits which can be made by employing them necessarily diminish’ (Smith, 1776, p. 276).
Therefore, growth driven by capital accumulation is not sustainable due to the law of diminishing returns. The second way of increasing growth is by increasing productivity, i.e. create more output for the same number of factors of production (Rosenberg, 2004). To accomplish this, Solow (1957) states that growth is ultimately driven by exogenous technological progress, measured, in a statistical
sense, by the Solow residual12. In fact, numerous studies have confirmed that a
part of economic growth cannot be accounted for by increases in factors of production, but is accounted for by the residual. For example, Nelson & Pack (1999) conclude that the rapid growth of the so-called Asian miracle economies could not have been the result of ‘marshalling of inputs’ (Ibid, p. 434). Instead, rapid growth was made possible by learning about and coming to master, technologies that were
new to their economies (Ibid)13. Another theory that emphasizes economy growth
through technological progress is endogenous growth theory. Contrary to Solow (1957), endogenous growth theory emphasis that technological progress is the endogenous outcome of an economic system, specifically driven by investments in human capital and knowledge. (Romer, 1994). Both theories stress the importance of innovation, but regard each other as ‘mutual exclusive drivers’ of economic growth (Wang & Wu, 2015). However, they not need to be. By looking into the different ways of developing innovation in an economy, both theories can be
12The Solow residual is also called total factor productivity. It is the part of a growth accounting procedure that cannot be
explained through capital accumulation or an increase in labor force participation.
13On the contrary, Young (1994) finds that growth of productivity (TFP) was not the cause of this growth. Instead,
14 complementary. One way of doing so is by developing systems of innovation, which will be discussed in the next section.
3.3 Systems of Innovation
Due to increased recognition of the importance of knowledge in economies, the increased usage of systems approaches, and the growing amount of knowledge generating institutions; the systems of innovation (SI) approach has gained importance in economics (OECD, 1997a). Instead of assuming a linear model of innovation14, the SI approach stresses that innovativeness and technological
development are the result of a set of actors in the economic system and relationships among them (Ibid). Founded by Freeman (1987), and conceptualized by several others (Lundvall, 1992; Nelson, 1993; Metcalfe & Georgiou, 1997; Cooke et al., 1997), a system of innovation can be defined as:
‘…elements and relationships which interact in the production, diffusion, and use of new, and economically useful, knowledge…either located within or rooted inside the borders of a nation state.’ (Lundvall, 1992, p.2)
Because of the geographical consideration in Lundvall’s definition, his version is commonly called a ‘national innovation system’ (NIS) (Godin, 2010). However, an innovation system can also be approached regionally (Cooke et al., 1997). The regional innovation system (RIS) approach is more suitable for large countries, because the RIS approach does not assume homogeneity within nations (Schremph et al., 2013; Edquist, 2004). Regions, defined as territories that are less in size than its sovereign state and have distinctive supralocal administrative, cultural, political or economic power (Cooke et al., 1998), are often not homogenous. In fact, as shown in the introductory figures, China’s regions and provinces are by no means homogeneous. Besides economic differences, Asheim & Gertler (2004) argue that regional cultures, face-to-face interactions and trust based relationships are also important factors for the way that firms and institutions interact with each other. Therefore, innovative capacity can also be thought of as embedded in social relationships that are developed over time along culturally and
14The linear model of innovation postulates that innovation starts with science and an increase in the amount of scientific
15 regionally determined lines (Doloreux & Parto, 2004). Consequently, there is growing recognition that important elements of the process of innovation become regionalized (Ibid). Given the characteristics of China’s heterogeneity across regions and the aforementioned regionalism of innovation, the RIS theory seems an appropriate empirical approach for this thesis. The main elements and relationships are illustrated in Figure 4. The knowledge generation and diffusion subsystem consists of organizations that produce and diffuse knowledge. As such, it consists of research organizations, educational institutions, technology mediating organizations, and the people that work in them (Trippl, 2006); OECD, 199a7).
16
Figure 4: Framework Regional Innovation System
Source: Trippl 2006. Public research organizations Educational organizations Technology mediating organizations, incubators, etc.
Knowledge generation & diffusion subsystem
Regional policy subsystem
Policy institutions Regional development agencies Industrial companies Contractors Customers Competitors Collaborators
Knowledge application & exploitation subsystem
Finance - Subsidies – Innovation And Cluster Policies
17 3.4 FDI and Innovation
19
4. Literature Review
In theory, as described earlier, the innovative output of a system depends on its inputs, actors, and relationships both internal and external to the system. In this section a review of the empirical literature is provided. This literature is focused on the determinants of innovativeness, along with literature on FDI spillovers.
The idea of university research that affects innovation is explored by Jaffe (1989). By using a Cobb-Douglas model he explored whether patents as a proxy for new knowledge, is determined by R&D of industries and university research, controlled for population and economic activity, his findings indicate that both independent variables are significant and positively affect patent production. The role of universities with regard to innovation is also found in Veugelers and Del Rey (2014).
20 Using the proposed model of Furman et al. (2002), a country level production function in which innovation is a function of innovation capacity and the resources employed for creating new ideas, Riddel and Schwer (2003) conducted a similar study on regional innovative capacity in the U.S. They too use patents as a measure for technological innovations for several reasons. First, patents issued for ‘genuinely novel ideas’ are a good indicator for the production rate of new technological concepts. Second, because patenting is costly, they argue that inventors or firms will only issue patents if here is some economic return on their investment. And finally, patents offer the inventor exclusive rights to revenues. Therefore, the economic returns can be reflected in the state in which it is granted. Independent variables include the stock of labor and capital measured as the number of high-tech workers in the state, expenditure on R&D by both universities and industry, state population, and the number of university degrees issued in the state. They find that human capital (university degrees) has a significant relationship with patents. A one percent increase in degrees reveals a 0.26 percent increase in patents. In addition, both R&D expenditures and the number of high-tech employees (stock of labor) are positive and significant.
Although the previous studies found evidence of R&D expenditure on patents, Rodriquez-Pose and Crescenzi (2006) warn for measurement problems to this variable. First, they warn that the returns for public and private R&D investments may vary. Second, not all innovative activities are classified as R&D.
21 Zhou (2012) investigates regional innovation capacity in China over two periods, 1991 to 1998 and 1999 to 2005, using a panel data model estimated by fixed effects GLS regression. The time span was separated, in order to measure if the impact of the independent variables would change over time. According to Zhou, this would measure the policy changes. Using the NIS and NIC approach, she uses the following variables. For innovation output the number of domestic patents was used, distinction them between inventions and utility patents. Innovation actors are measured as the number of large and medium-sized enterprises and the number of higher education institutions. Innovation input is not measured as R&D expenditure as in previous studies, but as funding for science and technology activities divided by regional GDP. In addition, human resources were included as the number of scientists and engineers employed full time per million people. GDP per capita was included to represent the economic infrastructure and the economic value of its existing knowledge stock. Interactions between actors were measured as the value of domestic technology contracts. Zhou (2012) also includes external interactions by using inward FDI and the sum if imports and exports. The results reveal that higher education institutions, GDP per capita, S&T funding, full time employed scientists and engineers, the sum of imports and exports and the value of domestic technology contracts are positively significant. In contrast, the effects of large and medium-size enterprises and FDI were negative and significant. One of the explanations of the negative FDI value is that foreign invested enterprises do not favor patenting in China.
Besides the aforementioned studies on national and regional innovative systems and capabilities, studies on FDI spillovers are reviewed. Since the literature is large on this subject, I focus on the studies that are more relevant for China, i.e. studies that investigate spillover effects of FDI on China.15
Cheung and Lin (2004) study spillover effects of FDI on regional innovation from 1995 to 2000. They use patents as innovation output and five independent variables, namely: the number of S&T personnel, expenditure on S&T
15 Interested may want to look into Blomström et al (1999) for a more extensive review on the determinants of FDI on host
22 development, the share of FIE export to gross output, and per capita GDP. Both pooled ordinary least squares and fixed and random effects panel data approach are used. High collinearity is observed between S&T personnel and S&T expenditure. Therefore, they estimate the equation using only one of these variables at a time. Their OLS results show that FDI affects patents positively and significantly. Fist, random effects show that FDI is again positive and significantly affecting invention patents and utility model patents, whereas the fixed effects does not. They find positive and significant effects for external design patents in both fixed and random effect models. Overall, FDI seems to have a positive effect on patents, but the OLS seems to overestimate the effect. S&T personnel and S&T expenditure are both positive and significant, as was the FIE share of export in the panel data analysis. Finally, GDP was also find significantly positive for all models.
A similar research is done by Fu (2008) over the period 1998 to 2004, but he introduces human capital and FDI intensity in the model, measured as he ratio of new fixed assets of FIE to total industrial net fixed assets in the region. Human capital is measured as the percentage of population with university degrees. Overall, Fu finds that the magnitude FDI significantly contributes to overall regional innovation capacity. These same magnitudes are found for R&D expenditure and for the human capital variable. Shang, Poon, Yue (2012) find similar results over the period 2001 to 2008, but measure human capital as the skilled personnel of research institutes and universities. This form of human capital is also found to be positively and significantly related to innovation.
23 Overall, all studies on innovative capacity in China utilize the number of patents as innovation output, using either the total amount of patents, or a split between invention, utility and design patents. The main determinants of innovative capacity seem to be educational systems, industrial characteristics such as firm size, GDP per capita, R&D expenditures, and various measures of human capital. In addition, the review also shows that FDI can have a significant effect on the production of patents.
4.1 Hypotheses
Four hypotheses are postulated to examine China’s regional innovativeness and the determinants thereof. Although the results of the empirical literature show positive and negative signs of inward FDI on innovativeness, the theory predicts a positive effect due to spillovers. To examine this once more, the first hypotheses is: H1. Inward FDI has a positive impact upon regional innovation performance. Secondly, human capital, often referred to as absorptive capacity (Li et al., 2015), is found to be positively related with innovation performance. Therefore, the second hypothesis is:
H2. Human capital has a positive impact upon regional innovation performance. R&D competencies (see Figure 4) have also been found as a driver for innovativeness in the literature. In addition, theory predicts that R&D is an important factor that facilitates interactions in the RIS. Therefore, the third hypothesis is:
H3. Research & Development competencies have a positive impact upon regional innovation performance.
Lastly, RIS theory emphasizes the importance of the regional policy subsystem. Therefore, the fourth hypothesis is:
24
5. Methodology and Data
To test the above hypotheses, I conduct a statistical analysis for a panel data set consisting of 15 Chinese provinces over the time period of 1997 until 2009. The empirical model is based on a knowledge production function in which innovative performance depends on the level of capital and labor resources devoted to the ideas sector of the economy (Furman et at., 2002). By including the factors identified in RIS theory and literature review, I derive at the following log-log models to estimate the spillover effects of inward FDI on innovation in China: 𝐼𝑃𝑖𝑡 = 𝛼 + 𝛽1𝐹𝐷𝐼𝑖𝑡+ 𝛽2𝑅𝐼𝑅𝐼𝑖𝑡+ 𝛽3𝑅𝐼𝐿𝑀𝐸𝑖𝑡 + 𝛽4𝑅𝐼𝐻𝐸𝑖𝑡+ 𝛽5𝐻𝐶𝑅𝐼𝑖𝑡+ 𝛽6𝐻𝐶𝐿𝑀𝐸𝑖𝑡 + 𝛽7𝐻𝐶𝐻𝐸𝑖𝑡+ 𝛽8𝑆𝐸𝑍𝑖𝑡 + 𝛽9𝐺𝐷𝑃𝑃𝐶𝑖𝑡 + 𝛽10𝑃𝑂𝑃𝑖𝑡 + 𝛽11𝐸𝑋𝑃𝑂𝑅𝑇𝑖𝑡+ 𝜀𝑖𝑡 (1) 𝑈𝑃𝑖𝑡 = 𝛼 + 𝛽1𝐹𝐷𝐼𝑖𝑡+ 𝛽2𝑅𝐼𝑅𝐼𝑖𝑡+ 𝛽3𝑅𝐼𝐿𝑀𝐸𝑖𝑡+ 𝛽4𝑅𝐼𝐻𝐸𝑖𝑡+ 𝛽5𝐻𝐶𝑅𝐼𝑖𝑡+ 𝛽6𝐻𝐶𝐿𝑀𝐸𝑖𝑡 + 𝛽7𝐻𝐶𝐻𝐸𝑖𝑡+ 𝛽8𝑆𝐸𝑍𝑖𝑡 + 𝛽9𝐺𝐷𝑃𝑃𝐶𝑖𝑡 + 𝛽10𝑃𝑂𝑃𝑖𝑡 + 𝛽11𝐸𝑋𝑃𝑂𝑅𝑇𝑖𝑡+ 𝜀𝑖𝑡 (2) 𝐷𝑃𝑖𝑡 = 𝛼 + 𝛽1𝐹𝐷𝐼𝑖𝑡+ 𝛽2𝑅𝐼𝑅𝐼𝑖𝑡+ 𝛽3𝑅𝐼𝐿𝑀𝐸𝑖𝑡+ 𝛽4𝑅𝐼𝐻𝐸𝑖𝑡+ 𝛽5𝐻𝐶𝑅𝐼𝑖𝑡+ 𝛽6𝐻𝐶𝐿𝑀𝐸𝑖𝑡 + 𝛽7𝐻𝐶𝐻𝐸𝑖𝑡+ 𝛽8𝑆𝐸𝑍𝑖𝑡 + 𝛽9𝐺𝐷𝑃𝑃𝐶𝑖𝑡 + 𝛽10𝑃𝑂𝑃𝑖𝑡 + 𝛽11𝐸𝑋𝑃𝑂𝑅𝑇𝑖𝑡+ 𝜀𝑖𝑡 (3)
25 5.1 Data Sources and Data Description
The data is collected from the National Bureau of Statistics of China, various issues of China´s Statistical Yearbook, and various issues of China´s Statistical Yearbook on Science and Technology. Because of a lack of data availability of several explanatory variables, the starting year for the analysis is 1997. Unfortunately, the yearbooks only provide provincial data on FDI until the year 2003. Consequently, if all Chinese provinces were used in this study, the data would only cover seven years (1997-2003). To overcome this, a compromise is made by using 15 provinces, five from each region. The additional FDI data for these 15 provinces is taken from Hongtian (2011) who reports this data up to 200916.
Therefore, the final sample includes 15 Chinese provinces for the period 1997-2009. See Appendix Figure A4 for their geographical location.
Dependent Variables
The dependent variable is the regional innovation performance, which can be measured in several ways, namely: patent citations, new product announcements, new product sales, and patents. One can view patent citations as being similar to academic references, since the citation indicates which part of the knowledge described in the patent document, has been claimed by earlier patents (Maurseth & Verspagen, 2002). These citations then indicate that the knowledge in the cited patent was useful for developing new kinds of knowledge in the citing patent (Jaffe et al. 1993). New product announcements are used as an indicator for innovative performance of companies. The announcements are based on press releases by these companies themselves with little screening by an independent database operator (Hagedoorn & Cloodt, 2003). As such, the data can be biased (Ibid). New product sales as a measure of innovative performance also has two major limitations according to Cheung & Lin (2004). First, new product sales capture product innovation but not process innovation. Therefore, using this indicator would not capture the whole concept of innovation (Ibid). Secondly, Chinese
16In Hongtian (2011) these regions are specifically chosen for their geographical location. According to the economic
26 enterprises receive tax credits by the government for new products sales in order to stimulate R&D (Ibid). Therefore, these enterprises tend to misreport information. An alternative way is by using patents, which is common in the literature on innovation. Patents capture both product and process innovation (Cheung & Lin, 2004), thereby complying to the definition of innovation used in this study, and patents seems to offer the best available indicator (Freeman, 2004). In addition, patent statistics are available on provincial level in China and Chinese patent laws are uniform (Yueh, 2006). Therefore, these statistics are an appropriate measure across provinces. The patents are reported in three types: invention, utility model, and external design. Invention patents refer to new technical proposals to a product, process, or improvement thereof17. Utility patents
refer to the practical and new technical proposal on the shape and structure of the product, or both, but is not directly related to its aesthetic properties (Fu, 2008). Lastly, external design patents refer to the aesthetics and industrially applicable new designs for the shape, patterns, and color of the product or their combinations. These distinctions give the opportunity to investigate the effect of FDI on each type of patent. Therefore, all three types of patents are treated as dependent variables, namely: invention patents granted (IP), utility models patents granted(UP), and design patents granted (DP). All three types are measured in the actual amount of patents.
Independent Variables
Inward foreign direct investment (FDI), which is considered a driver for innovativeness, is measured as the actual amount utilized per region. The data is taken from the Chinese statistical yearbooks as well as from Hongtian (2011), and is reported in current US dollars. The other monetary independent variables, which will be described subsequently, are reported in Chinese renminbi (RMB). In order to be consistent, it is useful to deflate the current values into constant values and to convert these constant values into RMB as well. According to Wei et al. (2006), the appropriate deflator is the US consumer price index. Therefore, the
17The definitions of invention, utility model, and design patents are taken from the National Bureau of Statistics of
27 current US dollar FDI values are deflated by the US consumer price index based in 2000. The deflated values are converted to values in RMB by using the official exchange rate in 200018.
Research and development expenditure (R&D) is also identified as an important driver for innovativeness. In the Chinese context, many papers use the total R&D expenditure per region (see among other, Fu, 2008; Shang et al., 2012). Instead of using total R&D expenditure figures, I include expenditures on ‘science and technology’ (S&T). These expenditures include operating expenses, including labor costs, scientific research, investment in the acquisition or construction of fixed assets, and R&D activities. As such, Zhou (2012) argues that it is a better variable than R&D expenditure to represent the effort put into innovation development. In fact, Zhou (2012) and Cheung & Lin (2004) found that S&T expenditure positively influences innovation performance. However, these are the only two papers incorporating this value and both papers utilize the total amount of S&T expenditure per province. In this paper, S&T expenditure is broken down in three figures, namely: S&T expenditure by independent research institutes, S&T expenditure by large & medium sized enterprises, and S&T expenditure by higher educational institutions. This allows me to investigate deeper the impact of research efforts of these institutions on innovative performance in the context of China. This data is reported in current RMB values and only available in China’s Yearbooks on Science and Technology in pdf format. These files are not free of charge. As such, all files had to be bought and all values had to be manually converted into excel format. The data is deflated by the Chinese GDP deflator to the price level of the year 200019. Similar to Li et al. (2015), who use R&D
expenditure figures, I divide these S&T expenditures by provincial GDP to take into account the relative sizes of the provincial economies and label them as ‘research intensity’ figures. Therefore, the final variables are: research intensity independent research institutes (RIRI), research intensity large to medium sized
18 Both the CPI index and the official exchange rate are taken from the World Bank Database, see:
http://data.worldbank.org/
28 enterprises (RILME), and research intensity higher educational institutions (RIHE).
Human capital has also been identified as important for innovative performance. Some measure this as the percentage of the population with a university degree (Fu, 2008), as the number of skilled personnel of research institutes and universities (Shang et al., 2012), or as the number of scientists and engineers employed full time per million people (Zhou, 2012). Similar to Cheung & Lin (2004), I measure human capital as the number of S&T personnel. However, instead of using the total number of S&T personnel per province, I break it down to the number of S&T personnel employed by the three types of institutions mentioned earlier. Again, as with S&T expenditures, this has not been done before and it allows me to investigate deeper what influence each type of human capital has on innovative performance. To control for provincial size, I divide these human capital amounts by the total amount of people employed per province in the corresponding year and label it ‘human capital intensity’. Therefore, the final variables are: human capital intensity independent research institutes (HCRI), human capital intensity large to medium sized enterprises (HCLME), and human capital intensity higher educational institutions (HCHE).
Regional policy regarding innovation (RP) is also identified as an important factor for innovative performance. To the best of my knowledge, there is not a single paper that incorporates this in the context of China’s regional innovative performance, most likely due to the difficulty of measuring these policies. However, Sun (2013) constructed a weighting scheme that incorporates China’s Open Door Policy and its Western Development Campaign by using the types of economic
zones per province20. The development of these so-called ‘economic and
technological development zones’ started in the early 1980s and were meant to encourage economic growth and development by attracting FDI. In the late 1980s, the ‘Torch Program’ initiated the construction of ‘high tech industrial development zones’ with the aim of utilizing the technological capacity of research institutes, educational organizations and large to medium sized enterprises to develop new
29 high tech products and expedite the commercialization of innovations21. Sun (2013)
uses these, and other types of zones, in a study to explain China’s regional disparities in economic growth from 1978 to 2011. By assigning weights to different types of zones, a policy index was created to account for these zones that were the product of governmental policies. However, that index is not applicable anymore since the total amount, and different types of zones per province, has increased. Therefore, this weighting scheme would give each province the same weight if used for this study. Instead, I use the actual amount of economic zones per province as a proxy for governmental policy to attract FDI and to develop technological capabilities. In other words, the variable SEZ (special economic zones) is the sum of the amount of ‘economic and technological development zones’ and ‘high-tech industrial development zones’ per province. To control for the fact that different provinces are at a different stage of economic development and are of different size, I include provincial GDP per capita (GDPPC) and the total population size per province (POP). The GDP per capita figures are deflated to 2000 RMB price levels. Lastly, to test whether exposure to foreign markets can lead Chinese firms to acquire technology or knowledge abroad, and thereby affect domestic innovative performance, I include the share of exports to provincial GDP deflated to 2000 RMB price level (EXPORT). For a brief overview of these variables and their measurements, see Appendix Table A1.
5.2 Estimation Methodology
Before empirically testing the model, the data is examined to check the assumptions of linear regression.
First, all independent variables were individually examined against the dependent variables using scatterplots. This visual examination confirmed that the variables approximate a straight line pattern, confirming linearity between the independent and the dependent variables. Second, the assumption of normality is tested using histograms. The result show that all variables are positively skewed. The logarithm transformation of all variables is used to ensure an approximately normal distribution, hence the log-log models. Third, a high degree of correlation
30 among several independent variables may cause multicollinearity issues. If the variables are perfect correlated with each other (singularity), i.e. if these variables have a linear relationship, the assumption of no multicollinearity would be violated. To test this, a correlation matrix of the independent variables was made (see Appendix Table A4).
According to the table, FDI is moderately correlated with the variables SEZ, GDPPC, and EXPORT. In addition, EXPORT is moderately correlated with SEZ and RIRI is highly correlated with HCRI (0.8437). To further investigate this issue, the variance inflation factors (VIF) are examined to measure the inflation in the variances of the parameter estimates as a result of collinearities (Mina, 2007). Appendix Table A5 reports the obtained VIF results. According to Bowerman et al. (2005), multicollinearity is ‘severe’ if a VIF is greater than 10. The results show that the variables RIRI and HCRI have VIF’s larger than 10, thereby causing concern for multicollinearity issues. In other words, the amount of S&T employees employed in independent research institutes, highly correlated with the amount of S&T expenditure by these independent research institutes. In an effort to solve this, I estimate the VIF’s again, thereby excluding the variable RIRI. The results show that all VIF’s are well below 10 by doing so. In addition, I estimate the VIF’s once more, but this time exclude the variable HCRI. Again, this results in all VIF’s being well below 10. Therefore, these two variables cause severe multicollinearity issues and I decide to estimate them separately in equations (1), (2), and (3). Consequently, the amount of models to be estimated is increased from three to six models, in which (1a) will exclude RIRI and (1b) will exclude HCRI etc.
32
6. Results
The descriptive statistics for all independent and dependent variables, both before and after the log transformation, are given in Appendix Tables A2 and A3. The disparities between the Chinese regions in terms of patent production are substantial. Invention patents range from an amount of 11 to 11,355 patents, whereas the mean is only 573. The same variation is present in the number of utility model and design patents. The GDP per capita differences are also notable and range from 3,827 US$ to 62,252 US$, whereas the average country has a GDP per capita of 15,910 US$. FDI is also not evenly spread across China. In fact, the maximum observed amount of FDI is 16.8 billion US$, whereas the minimum amount is 523 million US$. The log transformation significantly reduced this variation as shown by the standard deviations. For example, invention patents had a standard deviation of almost 1200 before the transformation, whereas a standard deviation of 1.28 is observed after the transformation.
Table 1 reports the regression results for all three models. Model (1) is regressed using invention patents (IP) as dependent variable, in model (2) utility model patents (UP) is the dependent variable, and model (3) is regressed using design patents (DP) as the dependent variable. Because of detected multicollinearity, each model is estimated twice: once by dropping the variable RIRI and once by dropping the variable HCRI.
33 Table 1: Fixed Effects Regression Results for IP, UP, and DP
Model (1) Model (2) Model (3)
Dependent variable Invention patents Utility model patents Design patents 1A 1B 2A 2B 3A 3B FDI -0.243** (0.107) -0.219** (0.111) -0.132 (0.105) -0.115 (0.102) -0.015 (0.119) 0.013 (0.119) RIRI -0.175 (0.263) 0.100 (0.169) -0.122 (0.247) RILME 0.009 (0.126) 0.077 (0.128) -0.153 (0.115) -0.147 (0.117) -0.021 (0.202) 0.044 (0.205) RIHE 0.396*** (0.106) 0.561*** (0.120) 0.045 (0.081) 0.114 (0.094) -0.099 (0.143) 0.083 (0.165) HCRI -0.742** (0.277) -0.202 (0.150) -0.770** (0.301) HCLME 0.215 (0.260) 0.046 (0.285) 0.377* (0.177) 0.329* (0.161) 0.183 (0.290) 0.006 (0.258) HCHE -0.231 (0.227) -0.236 (0.255) -0.002 (0.170) -0.011 (0.171) -0.203 (0.229) -0.211 (0.260) SEZ 0.126 (0.373) 0.153 (0.381) -0.255 (0.191) -0.253 (0.198) 0.127 (0.262) 0.152 (0.262) GDPPC 1.346*** (0.144) 1.291*** (0.152) 0.731*** (0.129) 0.747*** (0.129) 0.861*** (0.175) 0.817*** (0.194) POP 4.705*** (0.941) 5.721*** (1.049) 2.701*** (0.838) 3.081*** (0.839) 0.160 (1.362) 1.257 (1.241) EXPORTS 0.285* (0.158) 0.306* (0.167) 0.128 (0.159) 0.110 (0.156) 0.435* (0.224) 0.447* (0.229) R2 0.9180 0.9151 0.8451 0.8447 0.7495 0.7401 N 195 195 195 195 195 195 Notes:
(1): In model (1) the DV is IP, in model (2), the DV is UP, in model (3), the DV is DP.
(2): The models 1A, 2A, and 3A exclude the variable RIRI. The models 1B, 2B, and 3B exclude the variable HCRI. (3): Robust standard errors are reported in parentheses.
34 In turn, the amount of S&T personnel employed by independent research institutes bears a large negative sign of -0.742 and is statistically significant. The negative sign is also found for S&T personnel employed by higher educational institutions, although not significant. The amount of S&T personnel employed by large to medium sized enterprises and the amount of special economic zones are both insignificantly contributing to invention patent production. GDP per capita and the amount of inhabitants (POP) both show a positive effect on patent production. In fact, a one percent increase in population raises patent production by 4.7 percent. The share of exports to GDP has a significant coefficient of 0.285 at the 90 percent significance level. Adding the variable HCRI (model 1B), shows that the expenditure on science and technology by independent research institutes as a share of GDP negatively impacts invention patent production, albeit insignificant. Analyzing the results of models 2A and 2B reveals that inward FDI bears a negative effect on utility model patents. Contrary to invention patents, this effect is statistically insignificant. The variable RIRI now has a positive effect, whereas the amount of special economic zones and RILME now negatively impacts patent production. All three observations are however insignificant. The impacts of RIHE and HCRI are smaller and both lost their significance. The amount of S&T personnel employed by large to medium sized enterprises however has turned significant at the 90 percent significance level, whereas EXPORTS lost significance. The impacts of GDP per capita and the amount of population is somewhat smaller, but still highly significant (p=0.000).
35
7. Discussion
Based on the regression results, I can conclude that inward foreign direct investments (FDI) has no independent positive effect on regional innovativeness in China. Therefore, hypothesis one is rejected. Compared to other literature, this finding is consistent with Zhou (2012) and Hale (2007), but contrary to Cheung & Lin (2004), Fu (2008), and Roelfsema & Zhang (2012) who find that FDI is positively related to patent production. Several explanations can be found in the literature for this negative effect. First of all, foreign firms who commit to FDI might not favor patenting in China. A study by Zhou (2006) shows that 91 per cent of foreign invested enterprises do not file applications for patents in China. Even though China’s patent system corresponds to the WTO organization’s standards, infringement of property rights and trademarks are still a concern in China (OECD, 2007). In addition, foreign firms pay, on average, higher wages than domestic firms (Fosfuri et al, 2001). This may lead foreign firms to crowd out local firms in terms of human capital availability by preventing labor turnover. In turn, this might prevent knowledge spillovers by reducing the innovative capacity of local firms. Lastly, local organizations might simply be not able to absorb and learn from the technologies and knowledge brought by foreign firms.
The results also indicate that FDI has a strong and significant impact on invention patents, whereas this effect is less negative and insignificant for utility model patents, and even positive but insignificant for design patents. A possible explanation could be that design patents are less sophisticated than utility model patents, which are in turn are less sophisticated than invention patents. According to Dunning (2001), foreign firms tent to prevent ‘leaking’ of technologies to host countries. As such, more sophisticated technologies that could lead to invention patents might be kept more protected than less sophisticated technologies.
36 research organizations, educational institutions, and private firms play important roles in the process of innovation. Therefore, when total R&D or S&T figures are used the separate research efforts of these organizations are neglected. The results indicate that the research efforts of research institutions and large to medium sized enterprises mostly reduce innovativeness, whereas the efforts of higher educational institutions strongly favor invention patent production. Therefore, the positive results found by Erdal & Göcer (2015), Li et al., (2015), Fu (2008), Zhou (2012), and Cheung & Lin (2004), might be the result of the positive effect of S&T expenditure by higher educational institutions offsetting the mostly negative effects of S&T expenditure by research institutes and large to medium sized enterprises. As
The efforts of the Chinese government to develop innovativeness has apparently focused mostly on expanding the higher educational system. Plans such as the ‘A Plan for Human Resource Development in the West’ have apparently mostly contributed to develop innovative capabilities in higher educational institutions, whereas the amount of independent research institutes have been downsized (OECD, 1997). This might also explain the large negative impact of S&T personnel employed by these research institutes (HCRI) on patent production.
37 Hypotheses four, in which regional policy regarding innovation has a positive impact upon regional innovation performance, is also inconclusive. None of the models produce statistically significant results. Therefore, I can’t reject the null hypothesis and do not accept the null hypothesis of no difference either. I attribute the inconclusive results to the way regional policy is measured. First of all, the amount of special economic zones is a proxy for regional policy based on Sun (2013). However, the methodology used by Sun (2013) was not applicable is this situation as shown in the methodology section. Therefore, I had to count to amount of zones per region. Some regions had only one special economic zone. Consequently, the log transformation turned these values into a zero, thereby reduce the number of observations for this variable.
Moving away from the main objective of this study, I found that GDP per capita and the total inhabitants per region have a large and statistically significant impact on innovativeness. This indicates that the provinces that are more economically developed, e.g. eastern coastal regions, tend to be more productive in innovative activities. This is not surprising, because a better developed economic infrastructure is in general positively correlated with more innovation (Cheung & Lin, 2004).
Finally, I want to point out that several limitations in this thesis should not be overlooked. First of all, China’s national output figures are not as reliable as western data sources. As pointed out earlier, there is a debate on whether China exaggerates GDP figures. Although alternative estimates are available, such as from the The Conference Board, regional GDP is not. Therefore, I had to rely on the official statistics from the National Bureau of Statistics of China.
39
8. Conclusion
China has experienced rapid growth over the last three decades, primarily driven by factor accumulation (OECD, 2015). However, due to a declining working-age population and diminishing returns to capital, this growth model is no longer sustainable (World Bank, 2015b). As a result, China actively develops policies to transform its economy towards an innovative powerhouse. The goal of this thesis was to assess whether inward foreign direct investment could act as a driver for this transition, by examining whether FDI impacts regional innovativeness. Based on innovation systems (IS) theory, I employed a panel data analysis which included 15 Chinese provinces over the period 1997-2009. By using different types of patents as a proxy for innovativeness, I find that FDI does not contribute significantly to the overall regional innovativeness. In fact, FDI impacts regional innovative capabilities negatively. The result both confirm (Zhou, 2012; Hale, 2007) and challenge (Cheung & Lin, 2004; Fu, 2008; Roelfsema & Zhang, 2012) previous researches. I interpret this as foreign firms not favoring filing patent applications in China. In addition, local firms might not be able to absorb and learn from the technologies and knowledge brought by FDI due to a lack of competent human capital within research institutions and local firms.
Furthermore, I used different measures for science and technology expenditure and human capital. These measures made it possible to investigate the impact of research institutions, educational institutions and private firms. Although
40 provinces of the total of 33 were used in this study due to restrictions on data availability for FDI. Efforts were made to obtain additional data, such as contacting the NBS, but with no success. Even though these 15 provinces were spread across China, large variations exist within the data which probably caused heterogeneity problems. Lastly, a panel data set consisting of all 33 provinces could mitigate this problem and would also enable future research to estimate the effects of separate S&T expenditure and human capital figures between eastern, central, and western regions. Therefore, if more regional FDI statistics for China would be available in the future, other researchers may want to investigate the impact of FDI on regional innovativeness across eastern,
41
9. Appendix
Figure A1: Total GDP comparison in 2015 US$
Sources: (1) The Conference Board. 2016. The Conference Board Total Economy DatabaseTM, May 2016,
http://www.conference-board.org/data/economydatabase/; (2) The World Factbook 2016-17. Washington, DC: Central Intelligence Agency, 2016. https://www.cia.gov/library/publications/the-world-factbook/.
Figure A2: Exports and Household consumption expenditure (% of GDP)
Source: The World Bank, World Development Indicators (2016), Exports of goods and services (% of GDP) and Household finale consumption expenditure, etc. (% of GDP), http://data.worldbank.org/country/china.
0 2 4 6 8 10 12 14 16 18 20 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013 Trilli o n s U S$
China United States European Union Japan India
22,6% 36,5% 15 20 25 30 35 40 45 50 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Sh ar e o f G DP (% )
42
Figure A3: Secondary vs Tertiary industry (% of GDP)
Sources: (1) National Bureau of Statistics of China, NBS, http://data.stats.gov.cn/english/easyquery.htm?cn=C01; (2) Preliminary Accounting Results of GDP for the Fourth Quarter and the Whole Year of 2015, NBS, 21 January 2016,
http://www.stats.gov.cn/english/PressRelease/201601/t20160121_1307717.html; (3) Preliminary Accounting Results of GDP for the First Quarter of 2016, NBS, 18 April 2016, http://www.stats.gov.cn/english/PressRelease/201604/t20160418_1345160.html.
37,5% 59,9% 36 39 42 45 48 51 54 57 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 Q1 Sh ar e o f G DP (% )
43
Western provinces Central provinces Eastern provinces Other
Figure A4: Geography Chinese provinces
Western provinces: Shaanxi, Yunnan, Sichuan, Guangxi, Inner Mongolia. Central Provinces: Hubei, Hunan, Jilin, Shanxi, Henan.
Eastern provinces: Hebei, Guangdong, Jiangsu, Zhejiang, Fuijan.
Other provinces: Beijing, Tianjin, Laoning, Shanghai, Shandong, Hainan, Heilongjiang, Anhui, Jiangxi, Chongqing, Guizhou, Ganxu, Qinghai, Ningxia, Xinjiang, Tibet.
Table A1: Variables, definitions, and data sources
Variable (label) Definition Source
IP (invention patents) Log of number of invention patents National Bureau of Statistics of China UP (utility model patents) Log of number of utility model patents National Bureau of Statistics of China DP (design patents) Log of number of design patents National Bureau of Statistics of China FDI (FDI) Log of FDI inflows in constant 2000 RMB Author’s calculation based on data from
1997-2003 China Statistical Yearbook 2004-2009 Hongtian (2011) RIRI (exp. S&T research institutes) Log of expenditure on S&T in independent
research institutes as a share of GDP in 2000RMB
Author’s calculation based on data from China Statistical Yearbook on Science and Technology
RILME (exp. S&T L&M sized enterprises) Log of expenditure on S&T in large to medium sized enterprises as a share of GDP in 2000RMB
Author’s calculation based on data from China Statistical Yearbook on Science and Technology
RIHE (exp. S&T higher education) Log of expenditure on S&T in higher educational institutes as a share of GDP in 2000RMB
Author’s calculation based on data from China Statistical Yearbook on Science and Technology
HCRI (pers. S&T research institutes) Log of S&T personnel in independent research institutes as a share of total employed population
Author’s calculation based on data from China Statistical Yearbook on Science and Technology
HCLME (pers. S&T L&M sized enterprises) Log of S&T personnel in large to medium sized enterprises as a share of total employed population
Author’s calculation based on data from China Statistical Yearbook on Science and Technology
HCHE (pers. S&T higher education) Log of S&T personnel in higher educational institutes as a share of total employed population
Author’s calculation based on data from China Statistical Yearbook on Science and Technology
SEZ (special economic zones) Log of amount of special economic zones China Knowledge Online GDPPC (GDP per capita) Log of gdp per capita in constant 2000
RMB
Author’s calculation based on data from National Bureau of Statistics of China data POP (population) Log of total population National Bureau of Statistics of China EXPORT (exports) Log of exports as a share of GDP in 2000
RMB
44 Table A2: Descriptive statistics (before log transformation)
Variable Obs Mean Std. Dev. Min Max
IP 195 573.0256 1196.406 11 11355
UP 195 3119.621 4548.78 264 27438
DP 195 3749.313 8408.603 68 60025
FDI 195 2.29e+10 3.39e+10 5.23e+08 1.68e+11
RIRI 195 0.0039644 0.0041165 0.0006651 0.0210751 RILME 195 0.0086349 0.0048673 0.0014312 0.0248285 RIHE 195 0.0013379 0.0013028 0.0001047 0.0079322 HCRI 195 0.0003156 0.0002359 0.0001155 0.0013143 HCLME 195 0.0012 0.0007318 0.0002389 0.0040363 HCHE 195 0.000311 0.0001604 0.0000844 0.0010389 SEZ 195 3.630769 2.336858 1 11 GDPPC 195 15910.95 12989.81 3827 62252
POP 195 5.50e+07 2.22e+07 2.33e+07 1.01e+08
EXPORTS 195 0.1363903 0.1778092 0.0111951 0.8498778
Table A3: Descriptive statistics (after log transformation)
Variable Obs Mean Std. Dev. Min Max
45 Table A4: Correlation Matrix (pairwise correlation)
46 Table A5: Variance inflation factors (VIF)
(1) (2) (3) RIRI 18.90 3.64 HCRI 18.84 3.62 HCLME 7.54 4.95 3.68 FDI 7.17 6.29 6.54 POP 6.77 3.64 3.70 RILME 5.67 5.31 5.37 HCHE 5.45 5.37 5.45 GDPPC 5.36 5.18 5.35 SEZ 5.27 5.26 5.27 RIHE 4.75 4.25 4.45 EXPORT 4.25 4.12 4.03 Mean VIF 8.18 4.80 4.75
Notes: (1) includes all independent variables, (2) excludes the variable RIRI but includes HCRI, (3) excludes HCRI but includes RIRI.
Figure A5: Residuals against fitted values (excluding RIRI)
47
Figure A6: Residuals against fitted values (excluding HCRI)
48
Table A6: Modified Wald, Wooldridge and Hausman Test
Modified Wald Test Excluding RIRI Excluding HCRI
49
10. References
Alfaro, L., Chanda, A., Kalemli-Ozean, S., Sayek, S. (2004). FDI and Economic Growth: The Role of Local Financial Markets. Journal of International Economics, 64 (1), 89-112.
Athreye, S., Cantwell, J. (2007). Globalization and the emergence of new technology producers. Research Policy, 36 (2), 209-226.
Asheim, B., Gertler, M. (2004). Understanding regional innovation systems. In J. Fagerberg, D.C. Mowery. R.R. Nelson (Eds.), Oxford Handbook of Innovation and
Policy. Oxford, UK: Oxford University Press.
Baragheh, A., Rowley, J., Sambrook, S. (2009). Towards a multidisciplinary definition of innovation. Management Decision, 47 (8), 1323-1339.
Barra, C., Zotti, R. (2015). Regional innovation system (in) efficiency and its determinants: an empirical evidence from Italian regions. MPRA Paper No. 67067, University library of Munich, Germany.
Blomström, M., Globerman, S., Kokko, A. (1999). The Determinants of Host Country Spillovers from Foreign Direct Investment: A Review and Synthesis of the Literature. In N. Pain (Ed.), Inward Investment, Technological Change, and
Growth (pp. 34-65). Basingstoke, U.K.: Palgrave.
Bloom, M. (1992). Technological change in the Korean electronics industry. OECD Publishing, Paris.
Bosworth, B., Collins, S.M. (1996). Economic Growth in East Asia: Accumulation versus Assimilation. Brookings Papers on Economic Activity, 1996 (2), 135-191. Bosworth, B., Collins, S.M. (2008). Accounting for Growth: comparing China and India. Journal of Economic Perspectives, 22 (1), 45-66.
Bowerman, B., O’Connell, R.T., Koehler, A.B. (2005). Forecasting, Time Series, And
Regression. California: Thomson Brooks/Cole.
Chen, Y. (2007). Impact of Foreign Direct Investment on Regional Innovation Capability: A Case of China. Journal of Data Science, 5, 577-596.
Cheung, K.Y., Lin, P. (2004). Spillover effects of FDI on innovation in China: Evidence from the provincial data. China Economic Review, 15 (1), 25-44.
Cooke, P., Memedovic, O. (2003). Strategies for regional innovation systems:
Learning transfer and Application Policy Papers. Vienna: UNIDO.
