The effect of human capital and innovation on China’s economic development University of Groningen Faculty of Economics and Business Msc Economic Development and Globalization Master

(1)

The effect of human capital and innovation on China’s economic development

University of Groningen

Faculty of Economics and Business

Msc Economic Development and Globalization

Master’s Thesis 2019-2020

Name: Xiaotian Dai

Student number: S3813517

Student email: x.dai.4@student.rug.nl

(2)

1 Abstract

This thesis aims to study the influence of human capital on China’s economic development in the context of crisis and examine the innovation channel through which human capital contributes to China’s economic development. We use a fixed effect model to examine if human capital has different effects in different conditions. Besides, we test if human capital contributes to China’s economic development by working as a facilitator for innovation capacity through an interaction term study. The results show that the effect of human capital is larger in big cities and before the crisis. Besides, innovation ability is proved to be the channel through which human capital contributes to China’s economic development.

(3)

Acknowledgements

I would like to thank my thesis supervisor Dr. Abdul Azeez Erumban sincerely for his patience and kindness during the supervision of the thesis. He always gives me detailed feedback and comments that are really valuable and I can always get inspiration. Also, I learn from him how to do research in a rigorous way. Research is not coming from nowhere but should be based on the previous solid literature ground. Just as Newton (1675) has said, ’if I have seen further it is by standing on the

shoulders of Giants’. I would also like to thank all the lecturers during my master’s study for

teaching me knowledge. I really learned a lot from their lectures and tutorials. And I would also thank my parents for their financial support during my study in the Netherlands especially during this special global coronavirus pandemic. Finally, I would thank my friends: when I was anxious and depressed during the process, they chatted with me and relieved my anxiety and always gave me great spiritual encouragement. Studying in RUG may be one of the most correct decisions I have ever made in my life. I am sure that the scientific and rigorous learning attitude I have learned can benefit me for my lifetime.

(4)

3

I. Introduction

China's economy has maintained a sustained and rapid growth in the past 40 years ever since the economic reform in 1978. Since 2000, China’s economy has grown at an average rate of 9.26 percent, arguably driven by its investment-driven manufacturing and export-led growth strategy. The 2008 subprime mortgage crisis has made policymakers in China realize the importance of independent innovation. With the launch of the 2006 National Plan (2006-2020), "independent innovation strategy", and "the goal of building an innovative country" are put forward by the Chinese government. And China's R&D expenditure increases constantly year by year: it increases from 1029.84 billion Yuan in 2012 to 1422 billion Yuan in 2015. And the proportion of R&D expenditure to GDP increases from 1.98% in 2012 to 2.1% in 2015. In 2008, a plan called "introducing high-level talents from overseas" was implemented, the aim of which is to invite outstanding scientific research scholars back to China to help change the current domestic scientific and research situation. It can be seen that China is consciously and systematically highlighting the engine role of human capital and innovation in economic development, showing the ever-increasing awareness of scientific and technological innovation.

This thesis is an attempt to understand the role of human capital and innovation in China's recent growth experience. The role of human capital in driving economic development is widely discussed both in the theoretical literature (Romer. 1986, 1990, Lucas. 1988) and empirical literature (Norman. 1996, Bils and Klenow. 2000, Engelbrecht. 2003) in general and also in China’s context (Buckley et al. 2002, Kawakami. 2004, Heckman &Yi. 2012). For instance, Romer (1986, 1990) and Lucas (1988) propose endogenous growth theory and propose that human capital can contribute to economic development from two main channels: productivity and innovation. Other literature in general has no consistent evidence on the effect of human capital on economic development: some find a positive and significant effect of human capital (Norman. 1996; Voon. 2001; Engelbrecht. 2003) while others find a negative effect or they find the results are sensitive to different measures of human capital (Kormendi. & Meguire. 1985; Dessus. 1999，Bils and Klenow. 2000). In addition, researchers have focused on human capital and innovation ability both from a macro level and micro level (Riddel. & Keith. 2003; Teixeira. & Fortuna. 2004; Dakhli & De 2004). For the Chinese literature, a large number of researchers have provided strong evidence for the positive effect of human capital on China’s economic development under the condition of a stable economic environment (Kawakami. 2004, Fleisher et al. 2010). Besides, plenty of Chinese literature focuses on the relationship between innovation and China’s economic development (Wu. 2010; Hu & Jefferson. 2004; Schaaper. 2009). However, seldom literature in China has combined human capital and innovation ability within the framework of regional economic development. Besides, the effect of human capital on China’s economic development in the context of the economic crisis has not been studied yet.

(6)

5

In the previous literature, researchers use the number of patents as a proxy of innovation capacity (Akçomak & Ter. 2009). But in the current study, we use city innovation index to measure innovation capacity and use the number of patents as an alternative measure of innovation capacity in the robustness test. It is important to use this new measurement of innovation capacity because the number of patents only looks at the quantity but cannot reflect the quality of heterogeneity patents. As the city innovation index is based on registration data of new enterprises and micro patent value data, we can get more insights from this proxy such as innovation output, entrepreneurship, and patent value instead of patent quantity only. So this index is a more comprehensive measure of innovation capacity compared with the number of patents. Second, most studies in the previous literature on China perform the analysis for the period of economic reform from 1978 to 2000. Although some recent studies have covered the time period after 2008, the effect of human capital on China’s economic development under the context of the economic crisis has not been studied yet. And we try to identify if human capital plays an important role in overcoming the crisis after 2008. We expect human capital to play different roles before and after the crisis according to the results provided by Sultanova & Chechina (2016) showing that human capital is a key driving force of Russia’s growth after the 2008 economic crisis. Third, most Chinese studies have examined the relationship between human capital and China’s economic development, or the relationship between innovation and China’s economic development. But we combine human capital and innovation ability within the framework of regional economic development by performing an interaction term analysis to examine if the relationship between human capital and China’s economic development is mediated by innovation capacity.

Specifically, we ask the following questions: the first research question is whether human capital plays different roles before and after the financial crisis in China. And this research question is important because by answering this we can identify if human capital plays an important role in overcoming the 2008 financial crisis in China. According to Sultanova & Chechina (2016), human capital is a key driving force of growth after the 2008 economic crisis and an important factor in overcoming economic crisis using the example of Russia. And we aim to address this question by using the example of Chinese cities. To address the above research question, we apply a fixed effect model using prefecture-level data (including both urban and rural areas in each city) from 2004 to 2013. By dividing the full sample into two non-overlapping sub time periods: 2004 to 2008 and 2009 to 2013, we can examine whether human capital plays different roles in China’s economic development before and after the 2008 crisis.

(7)

(8)

7

II. Literature review and hypothesis

This section provides a review of the literature related to the relationship between human capital and economic development. First, we discuss the theoretical literature on the relationship between human capital and economic development and major models. Subsequently, we will discuss empirical evidence on the relationship between human capital and economic development globally from three parts: (1) studies that link human capital directly to economic development; (2) studies that examine the impact of human capital on productivity – labor productivity and total factor productivity; (3) studies that look at the relationship between human capital and innovation. Then, we will focus on the existing empirical literature on the impact of human capital on economic development in China. Since our research is based on China, combining this part and previous analysis will help us identify the research gap we will address in this paper, and postulate the hypotheses. Finally, the contribution of this thesis to the existing literature will be discussed.

2.1 Theory literature and models

Adam Smith sets off the earliest wave of growth theory considering human capital, proposing that labor’s ability is restricted by the proficiency of labor skills and the strength of judgment ability and should be improved by education. He also proposes that economic growth is mainly expressed by wealth growth and that one of the important ways to contribute to wealth growth is to increase the number and quality of workers. The modern economic growth theory considering human capital starts with Theodore Schultz, Gary Becker, and Jacob Mincer. Schultz (1961) proposes that increasing the stock of human capital in a country or region can significantly promote economic growth. Schultz clearly defines the concept and nature of human capital, namely, knowledge, labor experience, and proficiency condensed on people. Human capital investment will increase the productivity of spending capital, thus generating positive returns. It is pointed out that human capital investment is an important source of economic growth. Schultz makes a quantitative study on the impact of American education investment on economic growth, and conclude that the return rate of human capital investment is the highest among all kinds of investments. While Schultz's contribution to the theory is more analyzed from a macro perspective, Gary Becker (1964) studies from a micro level. He broadens the scope of application of human capital from the field of education to the field of population and families. Becker establishes a model of human capital investment, proposing that human capital, similar to fixed capital, can be acquired through investment and is closely related to an individual's future income. Mincer (1974) links investment in human capital with income distribution, which determines the income model of human capital.

Most empirical research on human capital and economic growth is based on the following models and theories from Mankiw, Romer, Weil (1992), Uzawa (1965), Nelson and Phelps (1966), Romer (1986, 1990), and Lucas (1988).

Mankiw, Romer, and Weil (1992) use an augmented Solow model1_{by introducing human capital} into the Solow model. They propose that human capital can promote economic growth by slowing

1_{Solow model is an exogenous growth model which is mainly used to explain the impact of the accumulation of}

(9)

down the diminishing marginal returns to material capital. Their model provides a good starting point for empirical research based on the augmented Solow model. Uzawa (1965) put forward an economic growth model of two sectors: human capital and tangible input capital, both of which jointly contribute to growth. According to their model, education is the source of human capital. And human capital contributes to economic output by inducing technological progress and improving labor productivity through education. The important contribution of the Uzawa-Lucas model is to make exogenous technological progress endogenous, providing the basis for the subsequent endogenous growth theory. Nelson and Phelps (1966) construct a model to analyze the relationship between educational level and technological progress. They assume that exists a frontier technology, and the diffusion of technology will lead to the rapid introduction of this frontier technology innovation to various countries after its emergence. The dissemination of technology requires highly skilled personnel. After the technology is implanted in their own countries, they also need a certain number of skilled workers for operation and maintenance. And they apply a differential equation to express the role of human capital in the process of technological diffusion. Their growth theory mainly emphasizes the impact of human capital as an input factor on total factor productivity in economic activities and then discusses the impact on economic growth. This transmission path specifically means that the higher the human capital level of a country or region, the more it will have a key impact on technological innovation and technological absorption, which are the main factors affecting economic growth.

(10)

9

while manual labor has no effect on economic growth. In other words, productive human capital in R&D is the source of economic growth. Lucas (1988) combines Schultz's human capital theory with Romer's model. Based on this, he highlights the importance of human capital and proposes to improve the growth model. He proposes that the formation of human capital can be divided into two aspects: internal effect and external effect. The internal effect is the formation of human capital through school education and training. The external effect is to accumulate human capital through learning by doing to gain experience in actual production. Lucas (1988) proposes that the external effect can promote the increase of marginal return to production and overcome the marginal diminishing effect of other production factors. Human capital becomes an important production factor to promote economic growth.

2.2 Mechanism of human capital influencing economic development

Human capital accumulation will contribute to economic development through different channels. According to endogenous growth theory, human capital indirectly promotes economic development in the region by affecting technological diffusion and technological innovation.

Human capital is the carrier of knowledge. The improvement of the level of human capital can improve the ability of learning and absorbing frontier technologies and knowledge. Research by Mincer (1984) and Lucas (1988) both show that increasing investment in human capital can improve the marginal productivity of other input factors used in the production process, thus increasing the output level of the whole production process per unit time. This phenomenon is called the external effect of human capital by Lucas. Barro (2000) also points out that with the increase of human capital investment, the declining trend of marginal productivity of other input factors used in the production process can be slowed down, and the increase of human capital level will make workers improve the efficiency of the use of other input factors including physical capital. Besides, human capital has the ability of independent innovation. The independent innovation ability of human capital can be explained by the theory of learning by doing. Progress in knowledge must go through a process of accumulation. In order to ensure the rapid growth of the national and regional economy, the role of knowledge accumulation, technological progress, human capital, and the retraining of human capital should be emphasized. In the model proposed by Lucas (1988), the progress of knowledge in society is the result of human capital investment. And the accumulation of human capital helps to realize sustained economic growth.

So the two main channels through which economic development is impacted by human capital are productivity and innovation. And in this thesis, we will mainly focus on the innovation channel.

2.3 Empirical research on the impact of human capital and economic development globally 2.3.1 Empirical research that links human capital directly to economic development

(11)

the conclusion that human capital has positive effects on income per capita, and they predict that income per capita of countries with similar technological levels and population growth rates will converge, though more slowly than the textbook Solow model suggests. Norman (1996) classifies human capital into high-level, middle-level, and low-level by different qualifications and concludes that the high-level human capital makes a higher contribution to economic growth in developed countries, while the middle-level and low-level human capital have a greater impact on economic growth in less developed countries. Voon (2001) observes that the more the government invests in education, the richer the returns. In other words, the more highly educated talents, the stronger the effect on economic growth. Engelbrecht. (2003) test empirically the relationship between human capital stock and economic growth in 25 OECD countries. The results show that the stock of human capital has a significant positive impact on the country's economic development. Lucas (2015) draws the conclusion that education human capital can promote economic growth by simulating the relationship between human capital and economic growth.

On the other hand, some scholars find a negative relationship between human capital and economic growth or they find the results are not robust. Bils and Klenow (2000) use the Mincerian equation to estimate the impact of schooling on growth in human capital and find that the years of schooling enrollment as a proxy of human capital have a negative effect on economic growth with a coefficient of -0.1. They also find that the impact of schooling on growth explains less than one-third of the empirical cross-country relationship. Levine and Renelt (1992) test the robustness of the partial correlation between per capita income growth rates and a wide range of economic indicators including using the sensitivity analysis provided by Kormendi & Meguire (1985) and they find that the results for almost all the variables including secondary-school enrollment rate as a proxy of human capital are very fragile and not robust. Dessus (1999) tries to find reasons for the results that the accumulation of human capital is negatively correlated with economic growth and suggests that we should not ignore the international differences which are due to educational infrastructures and the initial endowment in human capital in the quality of schooling systems.

Since this study focusses on China, an emerging economy, it is important to review the literature in emerging markets, which are of great significance when do research on China since they may follow a similar logic path. Self, & Grabowski. (2004) use time serious estimator to examine the role of heterogeneous human capital in India’s economic growth and find that while primary education has a strong causal impact on growth, there is limited evidence for such a strong impact for secondary education. They also highlight that female education at all levels has the potential for generating economic growth. Maksymenko & Rabbani. (2008) use a multivariate time series model to make a comparative analysis in South Korea and India. They find that human capital has a both significant and positive effect on these two emerging economies. Din & Knight (2010) use China’s provincial data and find that high school and higher education have a strong impetus to the current economic growth while the primary education level does not contribute to the growth anymore.

(12)

11

the heterogeneity of human capital, ignoring the international differences in the quality of schooling systems, or choosing different econometric models.

2.3.2 Empirical research on human capital, productivity, and economic development

Plenty of researchers have confirmed that human capital has positive relationships with total factor productivity (TFP). Romer (1990) proposes that public education investment can improve the overall quality of workers, thus driving technological progress and enhancing the technological innovation capability of the region. Benhabib & Spiegel. (1994b) use the Cobb-Douglas production function to carry out regression analysis and find that the level of a country's human capital stock directly affects the growth rate of total factor productivity. Kumar et al., (2012) use empirical analysis to conclude that the higher the quality of education, the more it can improve total factor productivity. Hsieh & Klenow. (2010) find that TFP not only has a powerful direct impact on output but also has indirect effects through physical and human capital accumulation.

Though few scholars such as Fischer et al. (2009) find that the positive direct impact between human capital and labor productivity is offset by a significant and negative indirect impact, leaving the overall effect not significantly different from zero, a large number of researchers have proved the positive relationship between human capital and labor productivity. Mason and Finegold (1997) implement research in the UK, supporting the positive relationship between human capital and the firm’s performance. They find that education and training are more important than physical capital in determining productivity. Aggrey & Joseph. (2010) use firm-level panel data to examine the relationship between human capital and labor productivity in East African manufacturing firms. And they conclude that average education and training in Kenya have a positive impact on workers’ labor productivity in firms. Baharin et al. (2019) use ARDL analysis to deal with data for Indonesia and find that in the long run, primary and secondary educations are positively correlated with labor productivity while tertiary education has an opposite effect on labor productivity, which implies a low quality of human capital in Indonesia.

2.3.3 Empirical research on human capital, innovation, and economic development

(13)

Other scholars do research on human capital and innovation from a micro level, which also confirms the positive effect of human capital on innovation ability. Gallié & Legros. (2011) use the production data of France's enterprises from 1986 to 1992 to build a model and conclude that the improvement of France's human capital level obviously promotes the country's technological innovation capability. Abdul et al. (2014) take Malaysia's small and medium-sized enterprises as the research object, and the results show that high-level human capital helps to enhance the organizational culture of enterprises and improve innovation performance. Eriksson & Forslund. (2013) try to study whether the existence of university institutions would affect the employment rate in the places where they are located. They divide Sweden into groups according to whether there are university institutions and compare the employment rate data of the two regions from 2002 to 2008. The results show that the university itself will not increase the employment rate, but as a gathering place of talents, the university has cultivated high-level human capital with rich knowledge and professional skills for enterprises and has spillover effects, thus promoting technological progress and knowledge appreciation in the region.

Overall, a large number of scholars prove that human has a positive effect on innovation, therefore promoting economic growth indirectly both from macro and micro level. Besides, scholars have conducted in-depth research on human capital, not only focusing on the positive effect of human capital as a whole element on innovation capability but also studying the level and structure of human capital. They conclude that different levels of regional human capital have different impacts on technological innovation capability in the region.

2.4 Empirical research on the impact of human capital on economic development in China Plenty of literature has provided strong evidence for the positive effect of human capital on China’s economic development. Kawakami (2004) discovers the positive and significant impact of human capital on the income growth of China's provinces. Similarly, Fleisher & Zhao (2010) conclude that human capital has both direct and indirect impacts on productivity growth using provincial-level data. Heckman &Yi. (2012) propose that human capital accumulation and physical capital accumulation are important reasons for China's rapid economic development in recent years.

(14)

13

regions, which implies a pessimistic expectation for China that without government intervention, future regional inequality in China will increase: inland provinces cannot catch up with the coastal provinces in the near future.

Some scholars have calculated the contribution of human capital to China’s economic growth using provincial-level data. Borensztein and Ostry (1996) propose that a higher level of total factor productivity is very important to maintain a high economic growth rate in China. Wang & Yao. (1999) use a perpetual inventory method to estimate the provincial-level data from 1952 to 1999 to examine whether the remarkable growth performance in China during the reform period is mainly driven by productivity or factor accumulation. And they find that the accumulation speed of human capital in China is very fast and has important effects on growth.

Overall, a great deal of empirical literature has provided strong evidence for the positive effect of human capital in China. Though some scholars find that the effect of human capital is insignificant or even negative in China, they provide some possible reasons: inappropriate proxy of human capital, different model specifications, different econometric models, the existence of FDI that partly capturing the spillover effects and externalities related to human capital.

The city-level data is rarely seen in the research on China. And most research on China uses provincial-level or state-level data and they come to conflicting conclusions. However, it is not sufficient to use provincial-level data. This is because provincial-level data might hide several heterogeneities within each province because within a province there are highly developed cities and least developed cities. And if the research does not take these cities into account, the research might be biased. For example, within Guangdong Province, per capita GDP of Guangzhou and Zhuhai are 22,524 USD and 24,092 USD respectively in 2018, an equivalent of developed countries. In contrast, the per capita GDP of Meizhou and Shanwei is only 3,833 USD and 4,658 USD respectively in 2018. So it is more appropriate to use the city-level data to examine the effect of human capital on economic growth. According to Duranton and Puga (2014), studying urban growth is important because cities provide a channel to study the ways and reasons for economic growth. They stress the importance of direct interactions that happen among individuals while studying economic growth. Therefore it is important to use city-level data to carry studies. Also, the occurrence of spillover effects related to human capital is most relevant to the geographical unit of cities (Jacobs 1969, Glaeser, & Resseger 2010). So in this thesis, we will use the city-level data. To be exact, the city data captures all the regions including both urban and rural areas in each city.

Under the condition of a stable economic environment, human capital has been proved to be a key driving force of China’s economic development (Kawakami.2004, Fleisher et al.2010). However, the effect of human capital on China’s economic development under the context of the economic crisis has not been studied yet. According to Sultanova & Chechina (2016), human capital is verified to be an important factor of economic growth after the 2008 economic crisis and an important factor in overcoming economic crisis using the example of Russia. And we aim to identify this issue, so based on the previous analysis, we have the following hypothesis:

(15)

is an important driving force for China’s economic development after the economic crisis. In addition, according to the influence mechanism mentioned before, economic development is impacted by human capital through the channel of innovation capacity. So we have the following hypothesis:

H2: Human capital contributes to China’s economic development by working as a facilitator for innovation capacity.

This paper tries to contribute to the current literature of human capital and economic development in China from three aspects. First, traditionally researchers use the number of patents to represent innovation capacity (Akçomak & Ter. 2009). But we adopt a new measure of innovation capacity (city innovation index) provided by Kou and Liu (2017), which can reflect innovation output, entrepreneurship, and patent value instead of patent quantity only. In this way, we can measure city innovation capacity more comprehensively. Second, the effect of human capital on China’s economic development in the context of the economic crisis has not been studied yet, and we will test whether the effects of human capital on China’s economic development before and after the 2008 crisis vary using city-level data which is rarely seen in the literature. Third, seldom Chinese literature has combined human capital and innovation within the framework of economic development. But we apply an interaction term analysis to examine if human capital contributes to China’s economic development by working as a facilitator for innovation capacity.

(16)

15

III. Data and Methodology

3.1 Data

In order to detect the effect of human capital on China’s economic development, we collect the prefecture-level city data from 2004 to 2013. The prefecture-level cities are the second level of administrative division. The data of some cities have been dropped because of data unavailability and we also drop the outliers. And 277 cities are kept in the final sample for 10 years. For the prefecture-level city data, it captures both urban and rural areas of each city. Besides, the total number of population of these cities in the sample accounts for 92.97% of China’s total population. And the total GDP of these cities in the sample accounts for 94.83% of China’s total GDP. So the prefecture-level city data is very representative of the whole country.

For the data source, price data are collected from Wind Economic Database, patent data, and trade data are collected from National Intellectual Property Administration of China and China Statistical Yearbook for Regional Economy respectively. The data for the city innovation index is collected from Find Report on City and Industrial Innovation in China published by Fudan University Industrial Development Research Center. And the data for other variables are collected from City Statistical Yearbook of China which is published by the National Bureau of Statistics of China. And we integrate and process the data through Excel and Stata packages.

Dependent variable:

Real GDP per capita. As proposed by Pelinescu (2015), GDP per capita can be used to denote local

economic development, so we use GDP per capita as the dependent variable. GDP per capita takes away the population effect, which directly reflects the average income per person in the region. And we calculate the real GDP per capita of each city manually: we first collect GDP prices of each city with 2003 as the base year from Wind Economic Database which is an authoritative and widely used database for Chinese data in the area of economics and finance. Then, the nominal GDP of each city which is collected from City Statistical Yearbook of China is deflated by the GDP prices of each city with 2003 as the base year to get the real GDP of each city. In this way, we can take the city level price differences into account. Finally, we divide the real GDP of each city by the population to get the real GDP per capita of each city. For the prices, poorer cities experienced lower price growth from 2004 to 2013 compared with richer cities. Most prices vary from 1.002 to 1.37, which implies that most Chinese cities have experienced inflation during these time periods. However, Qitaihe, Hechi, and Hegang have prices smaller than 1 in 2012 and 2013, which means that these cities experienced deflation in 2012 and 2013. And they also experienced the lowest price growth during this time period. These three cities are relatively poorer cities in China. The average per capita income of all the cities in China is 45569 Yuan in 2013, while the average per capita income of these three cities is only 25081 Yuan in 2013, which is half of the average level.

Independent variables:

Human capital. According to Baro et al (2004), human capital plays a vital role in economic

(17)

Teixeira & Fortuna.2004). However, these data are not available at the city level for China. Alternatively, following Ouyang and Fu (2012), we use the percentage of enrolled university students in the city population to represent human capital. However, this measurement has limitations: university enrollment is not always proportional to graduates and this human capital measurement is zero for the cities without universities which is not true in reality. The expected effect of human capital on economic development is positive. And the data is collected from City Statistical Yearbook of China.

City innovation index. Technological innovation is a commercial innovation behavior to realize the

research and development of new ideas, new processes, new products, and new technologies and put them into production. Plenty of researchers have confirmed that technological innovation contributes to economic growth both theoretically (Romer, 1986) and empirically (Teixeira & Fortuna. 2004, Teixeira & Fortuna. 2004). Kou and Liu (2017) calculate the City Innovation Index for each city. This index is based on registration data of new enterprises and micro patent value data in each city with the help of econometric and statistical analysis methods. This index is measured by innovation output, entrepreneurship, and patent value instead of patent quantity, so it is a better proxy of innovation capacity. The expected effect of city innovation index on economic development is positive. And this data is collected from FIND Report on City and Industrial Innovation in China (2017) published by Fudan Institute of Industrial Development School of Economics, Fudan University.

Besides, according to Akçomak & Ter (2009), the number of patents is a good measure of technological innovation. So we use the number of patents as an alternative measurement of innovation capacity in the robustness test to examine if the variable (city innovation index) we use as a proxy of innovation capacity is reliable. And there are limitations to use patent as a proxy of innovation capacity: the quality and technology content of different patents is not taken into consideration. And the data is collected from National Intellectual Property Administration of China. The expected effect of patent on economic development is positive.

Interaction between human capital and innovation capacity:

Following the influence mechanism mentioned in the literature review part, innovation capacity is the channel through which human capital has an indirect effect on economic development. And we measure the innovation capacity of each city by city innovation index.

Human capital_ City innovation index. In this paper, we are especially interested in the linkage

(18)

17

this interaction term on China’s economic development is positive.

Besides, we also include an interaction term of human capital and the number of patents to represent the interaction between human capital and the alternative measurement of innovation capacity in the robustness test to examine if city innovation index is a good proxy of innovation capacity. And the expected effects of this interaction term on China’s economic development is also positive.

Control variables:

Lagged GDP per capita. Following Pelinescu (2015) and according to economic theory, the current

economic development is affected by past economic development. So we use the one-stage lagged value of GDP per capita to capture this effect. The expected effect of lagged per capita GDP on economic development is positive. And the data is collected from City Statistical Yearbook of China and Wind Economic Database.

Physical capital investment. As an important factor of economic development, the physical capital

investment rate can reflect the speed of capital accumulation. Belton et al. (2018) and Chen & Fang. (2018) both take physical capital investment into consideration while assessing the effect of human capital on economic development. So we use the percentage of physical asset investment in GDP of each city to represent the physical capital investment of each city. The expected effect of physical capital investment on economic development is positive. And the data is collected from City Statistical Yearbook of China.

Trade openness. The increase in import and export trade can improve the efficiency of resource

allocation, stimulate domestic demand, and have a strong pull on economic development. According to Harrison (1995), trade openness has a correlation with economic growth. So we use the percentage of total import and export volume of each city in GDP to represent the trade openness of each city. The expected effect of trade openness on economic development is positive. And the data is collected from China Statistical Yearbook for Regional Economy published by the National Bureau of Statistics of China.

FDI. Foreign direct investment can promote regional economic development, alleviate the funding

(19)

3.2 Methodology

Following the approach of Pelinescu (2015), we use a panel model based on the following function:

𝑌𝑖𝑡= 𝛽0+ 𝛽1𝐻𝑖𝑡+ 𝛽2𝐼𝑖𝑡+ 𝛽3𝑋𝑖𝑡+ 𝜃𝑖+ 𝛾𝑡+ 𝜀𝑖𝑡 (1)

In the above equation, subscripts i and t represent city and year respectively, 𝑌𝑖𝑡 is economic

development proxied by per capita GDP of city i in year t, 𝐻𝑖𝑡 is human capital proxied by the

percentage of enrolled university students in the city population in year t, 𝐼𝑖𝑡 is innovation capacity

proxied by city innovation index. 𝑋𝑖𝑡 are other control variables that have impacts on economic

development mentioned in the earlier part, 𝜃𝑖 and 𝛾𝑡 are two variables that capture the time and

city fixed effects where 𝜃𝑖 is used to control for the individual effect of cities that does not change

with time and 𝛾𝑡 is used to control for various macroeconomic factors that affect cities in different

years, 𝜀𝑖𝑡 is the error term, and the coefficients 𝛽 with subscripts are the parameters to be

estimated.

Though our model is mainly based on Pelinescu (2015), we optimize their model by adding more control variables such as fixed capital investment and FDI to add to the explanatory power of the model. Besides, we remove the control variable of public expenditure on education as proposed by Pelinescu (2015) because expenditure on education is highly correlated with the human capital measure we use: if we spend more on education, it is more likely that we will have more human capital which might cause the multicollinearity problem. Instead, we use the public expenditure on education as an alternative measurement of human capital in the later robustness test to test if our main results are sensitive to the choice of different measures of human capital. While the model provided by Pelinescu (2015) only examines the direct effect of human capital on economic development, we further apply an interaction term analysis to check the channel through which human capital has an indirect effect on China’s economic development.

While performing the econometric analysis, we take the natural logarithm of data, the advantage of which is to provide us with the intuitive elasticity of independent variables by using a log-log model. Besides, the logarithm of data can make our transformed variable more normally distributed. At the same time, we use robust standard errors to correct the possible heteroscedasticity.

After implementing the basic regression to examine the overall effect of human capital on China’s economic development, we group the full sample in two ways: first, the full sample is further divided into two sub-samples to check whether the magnitude of the effect of human capital is different in big cities and small cities. The sample of big cities consists of 4 super big cities and 15 sub-provincial level cities. These cities are the most developed cities in China. And the remaining prefecture-level cities make up the sample of small cities. The time periods for these two sub-samples are both from 2004 to 2013. Then, we divide the full sample into two non-overlapping five-year sub-periods (2004-2008 and 2009-2013) for all the cities. In this way, we want to check whether human capital plays different roles before and after the economic crisis.

(20)

19

including the interaction term between human capital and innovation capacity. When studying the interaction effect between human capital and FDI, Su & Liu (2016) include human capital, FDI, and the interaction term between human capital and FDI in their regression. Su & Liu (2016) assume that the relationship runs from human capital to FDI. And they conclude that human capital contributes to growth by working as a facilitator for FDI. As we are interested in the interaction between human capital and innovation capacity, we follow the approach provided by Su & Liu (2016) and replace their interaction term by the interaction term between human capital and innovation capacity (city innovation index). The function of this method is as follows:

𝑌𝑖𝑡= 𝛽0+ 𝛽1𝐻𝑖𝑡+ 𝛽2𝐼𝑖𝑡+ 𝛽3𝑋𝑖𝑡+ 𝛽4𝐻_ 𝐼𝑖𝑡+ 𝜃𝑖+ 𝛾𝑡+ 𝜀𝑖𝑡 (2)

In the above equation, subscripts i and t represent city and year respectively, 𝑌𝑖𝑡 is economic

development proxied by per capita GDP of city i in year t, 𝐻𝑖𝑡 is human capital proxied by the

percentage of enrolled university students in the city population in year t, 𝐼𝑖𝑡 is innovation capacity

proxied by the number of patents and city innovation index, 𝐻_ 𝐼𝑖𝑡 is the interaction term between

human capital and innovation capacity of each city. 𝑋𝑖𝑡 are other control variables that have

impacts on economic development mentioned in the earlier part, 𝜃𝑖 and 𝛾𝑡 are two variables that

capture the time and city fixed effects where 𝜃𝑖 is used to control for the individual effect of cities

that does not change with time and 𝛾𝑡 is used to control for various macroeconomic factors that

affect cities in different years, 𝜀𝑖𝑡 is the error term, and the coefficients 𝛽 with subscripts are the

(21)

IV. Empirical results

Before implementing the analysis, we do a multicollinearity test, the result of which can be seen from Appendix: Table A. The result proves that there does not exist multicollinearity problems. Besides, to check if it is appropriate to use the fixed effect model, we carry out the Hausman test, the result of which can be seen from Appendix: Table B. And the result shows that the p-value is 0, implying that we prefer the fixed effect model.

4.1 Data description and summary statistics

Table 1: Summary statistics

N Mean St.Dev min max

Per capita GDP (in 10000 Yuan) 2770 2.543 2.139 .195 21.021

Human capital (%) 2744 1.519 2.068 .024 12.704

Patent (in 10000 units) 2766 .182 .562 .0001 9.0771

City innovation index 2767 4.321 22.794 .005 543.08

Physical capital investment (%) 2769 59.657 23.163 10.223 216.905

Trade openness (%) 2751 21.446 38.668 .002 330.447

FDI (in 1 billion USD) Human capital_ patent Human capital_ index

2667 2740 2741 6.805 .631 19.632 18.304 2.417 106.273 .002 .00003 .0008 295.807 31.352 2431.068

Source: Own calculations

Summary statistics of the data are listed in Table 1. The units of GDP per capita, Patent, and FDI are 10000 Yuan, 10000 units, and 1 billion USD respectively. The City innovation index is an index and does not have units. And the units for all the other variables are percentages. It can be seen that this is not a strongly balanced panel data. The economic development gap of Chinese cities is very large: the economic development of the most developed city is more than 100 times that of the least developed city. Also, the distribution of human capital is very uneven.

Per capita GDP varies a lot across cities: there is a high concentration of cities with higher per capita in the eastern region of China. For instance, the average per capita income of all the cities in China is 42631 Yuan in 2013. And the average per capita income in the coastal cities of Guangzhou, Nanjing, and Hangzhou2_{is 95435 Yuan in 2013, which is twice as high as the average level of} China. The average per capita income of cities Hefei, Luoyang, and Datong3_{in the middle region} of China is 45569 Yuan in 2013. However, the per capita income of the western cities of Dingxi, Hechi, and Zhaotong4_{is only 9711 Yuan in 2013, which is far below the average level of China.} Even within a given region or a given province, there is substantial variation in per capita income across cities. Figure 1 shows that per capita income of cities can vary a lot even within the same

2_{Guangzhou, Nanjing, and Hangzhou belong to Guangdong Province, Jiangsu Province, and Zhejiang Province}

respectively, which are all in the coastal area.

3_{Hefei, Luoyang, and Datong belong to Anhui Province, Henan Province, and Shanxi Province respectively,}

which are all in the middle region of China.

4_{Dingxi, Hechi, and Zhaotong belong to Gansu Province, Guangxi Province, and Yunnan Province respectively,}

(22)

21

province. We take Guangdong Province as an example. Guangdong Province is one of the most developed provinces in China, however, there are huge disparities in per capita income across cities even within Guangdong province: we can see from Figure 1 that per capita income of Guangzhou and Zhuhai is 1076208 and 94760 Yuan respectively in 2013, which is an equivalent of developed countries. However, per capita income of Jieyang and Shantou is 23541 and 26129 Yuan respectively in 2013, which is much lower than the average level of Guangdong Province.

Source: Own calculations

Similar disparities are also observed in human capital level: there is a high concentration of cities with higher human capital levels in the eastern region of China. For instance, the average human capital level of all the cities in China is 1.85% in 2013. The average human capital level in the coastal cities of Guangzhou, Nanjing, and Hangzhou is 10.35% in 2013, which is more than five times the average level of China. The average human capital level of cities Hefei, Luoyang, and Datong in the middle region is 2.94% in 2013. However, the average human capital level in the western cities of Dingxi, Hechi, and Zhaotong is 0.21% in 2013, which is far below the average level of China. Also, we can see from Figure 1 that the human capital level distribution follows the same pattern of per capita income within the same province: human capital level of Guangzhou and Zhuhai is 11.81% and 11.7% respectively in 2013. However, human capital level of Jieyang and Shantou is only 0.173% and 0.178% respectively in 2013, which is not only lower than the average level of Guangdong Province but also lower than the average level of all the cities in China.

(23)

Then we come back to the summary statistics table 1. For other variables, the mean of physical capital investment is 59.657 %, but it varies from 10.223% to 216.905%. The mean of the number of patents and city innovation index is 1820 units and 4.321 respectively, but they both vary a lot. The mean of trade openness is 21.446%, but it varies a lot from 0.002% to 330.447%. The mean of FDI is 6.805 billion dollars, but it varies a lot from 0.002 to 295.807 billion. These facts mean that there exists a huge disparity in physical capital investment, city innovation capacity, trade openness, and FDI across cities: cities in coastal areas tend to have higher innovation capacity, have higher trade openness, and receive more FDI compared with inland cities; the disparity of physical capital investment is larger in inland cities than in coastal cities.

4.2 Empirical results of the basic regression

Table 2 lists the fixed effects estimation results for equation (1): Column (1) of Table 2 lists the result for the full sample from 2004 to 2013; Column (2) and (3) list the results for big cities and small cities from 2004 to 2013 respectively; Column (4) and (5) list the results for two non-overlapping five-year sub-periods (2004-2008 and 2009-2013) for all the cities including big cities and small cities.

In terms of human capital, it enters positively in all the five regressions in Table 2, though the magnitude varies, which is consistent with the expectation. So we can conclude from column (1) that human capital has a positive effect on China’s economic development. And when human capital is increased by 1 percent, regional per capita income will increase by 0.0324 percent. It can be seen from Column (1) that all of the coefficients except the constant are significant at a significance level of 5%. And the effect directions of all the coefficients are as expected. Besides, The R-squared is 0.961, which proves that the goodness-of-fit of the model is good. However, as we have included lagged per capita GDP in the model, it is possible that the high R-squared is mainly due to this control variable. And when we remove the lagged per capita GDP in the model, the R-squared fells to 0.829. So even if we remove the lagged per capita GDP from the model, the R-squared remains reasonably high at 0.829, which shows that our model still has good explanatory power.

After performing the basic regression, we further divide the full sample into two sub-samples: big cities and small cities. The sample of big cities consists of 4 municipalities and 15 sub-provincial level cities. These cities are the most developed cities in China. And the remaining prefecture-level cities make up the sub-sample of small cities. And Table 3 provides a comparison of the average values of all the variables between big cities and small cities. We can see from Table 3 that there is a large gap between the average level of human capital in big cities and small cities: the average value of human capital in big cities is almost five times that in small cities.

(24)

23

environment, transportation (especially aviation industry and high-speed railway), and industrial structure…It is likely that human capital has fewer impacts on big cities’ economic development compared with these factors. And the reason why human capital plays a more important role in big cities than in small cities might be that big cities have developed economic levels, favorable geographical location, and abundant employment opportunities. These advantages have attracted a large amount of high-level human capital, laying a good foundation for upgrading the regional industrial structure, and realizing the sound development of the local economy. Also, big cities have a greater need for a quality labor force, which may be due to a large number of relatively high-end industries gathered. In contrast, the technology level and urbanization rate in small cities are relatively low, so small cities have fewer job opportunities especially for high-skilled jobs. Therefore, the need for high-quality human capital is relatively small, leading to the fact that the scale of the high-quality labor force has little effect on regional economic development in small cities.

Table 2: Fixed effects regression results

Dependent variable: (1) (2) (3) (4) (5)

Log per capita GDP Full sample 2004-2013 Big cities 2004-2013 Small cities 2004-2013 Sub-periods 2004-2008 Sub-periods 2009-2013

Log human capital 0.0324*** 0.0426 0.0320*** 0.0549** 0.00993

(0.0105) (0.0328) (0.0107) (0.0278) (0.00823)

Log per capita GDP (-1) 0.852*** 0.799*** 0.852*** 0.746*** 0.611***

(0.0160) (0.0641) (0.0165) (0.0436) (0.0307)

Log physical capital 0.0553*** 0.124** 0.0547*** 0.123*** -0.0874***

(0.0105) (0.0542) (0.0108) (0.0230) (0.0280)

Log trade openness 0.0137** 0.0633 0.0129** 0.0143 -0.00569

(0.00559) (0.0429) (0.00564) (0.0114) (0.0129)

Log FDI 0.00634*** -0.00621 0.00766*** 0.00458* 0.00345

(0.00210) (0.00521) (0.00214) (0.00264) (0.00342)

Log innovation index 0.0273*** 0.0605** 0.0255*** 0.163*** 0.0631***

(0.00768) (0.0212) (0.00825) (0.0226) (0.0174) Constant -0.00874 -0.559* -0.000114 -0.112 0.875*** (0.0424) (0.286) (0.0448) (0.0969) (0.134) Observations 2,381 171 2,210 1,049 1,066 R-squared 0.961 0.962 0.961 0.901 0.810 Number of city 274 19 255 274 274

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

(Note: Big cities include four municipalities (super big city): Beijing, Shanghai, Tianjin, Chongqing; and fifteen sub-provincial cities: Guangzhou, Shenzhen, Hangzhou, Ningbo, Nanjing, Xiamen, Wuhan, Xi' an, Chengdu, Jinan, Qingdao, Shenyang, Dalian, Changchun and Harbin. And the remaining prefecture-level cities make up the sub-sample of small cities.)

(25)

Table 3: Comparison of average values of variables between big cities and small cities Variable Average value for big cities Average value for small cities Per capita GDP (in 10000 Yuan) 5.095 2.355

Human capital (%) 5.551 1.219

Physical capital investment (%) 54.278 60.053

Trade openness (%) 75.842 17.41

FDI (in 1 billion USD) 38.173 4.399

City innovation index 43.288 1.448

Human capital_ index 222.32 4.535

Source: own calculations

However, we should acknowledge that there might exist endogeneity issues: it is possible that big cities with higher levels of economic development also have more means to invest in education so that there will be more human capital accumulation in highly-developed cities (reverse causality). And finding appropriate instrumental variables is an ideal way to solve this possible endogeneity problem.

(26)

25

4.3 The role of innovation and human capital – empirical results of the interaction effect It is expected that the impact mechanism of human capital on economic development includes both direct and indirect effects. In addition to the direct promotion effect of human capital on economic development through labor output, human capital can also have an impact on economic development through indirect effects such as externalities. According to Su & Liu (2016), we apply the method provided by them to examine if human capital complements the innovation capacity of each city in promoting China’s economic development by including the interaction term between human capital and innovation capacity of each city. The results of equation (2) are listed in Table 4.

(27)

Table 4: Fixed effects results for the interplay of human capital and innovation capacity

Dependent variable: (1) (2) (3)

Log per capita GDP Full sample Big cities Small cities

Log human capital 0.0370*** -0.00499 0.0365***

(0.0102) (0.0561) (0.0105) Log per capita GDP (-1) 0.859*** 0.790*** 0.858*** (0.0162) (0.0634) (0.0165)

Log physical capital 0.0575*** 0.128** 0.0557***

(0.0104) (0.0543) (0.0106)

Log trade openness 0.0161*** 0.0746 0.0149***

(0.00549) (0.0451) (0.00551)

Log FDI 0.00638*** -0.00616 0.00774***

(0.00210) (0.00527) (0.00213)

Log innovation index 0.0231*** 0.0450* 0.0228***

(0.00803) (0.0223) (0.00839) Log human capital _ innovation index 0.00601*** 0.0134 0.00586** (0.00212) (0.0102) (0.00249) Constant -0.0350 -0.547* -0.0173 (0.0438) (0.261) (0.0454) Observations 2,381 171 2,210 R-squared 0.961 0.962 0.961 Number of city 274 19 255

Given that our measure of human capital is the percentage of university enrolled students in the city population, it is likely to show a lagged effect rather than an instantaneous effect on GDP. This is because, the fresh graduates constitute only a small fraction of the workforce, and their impact may come with a lag. Therefore, we also run the regression using 1, and 3 lags5_{. The results of the lagged} effect of human capital are listed in Table 5. The results in Table 5 show that the coefficients of human capital remain significant and positive at both 1 and 3 lags, although the magnitude of the coefficients is slightly lower than no lags. The main conclusion that comes out of it is regardless of whether we use the measure of education with or without lags, the impact is positive. We have to acknowledge that there might be statistical ways to find ideal lags, but in this analysis, we arbitrarily

5_{Since our time series is only 10 years, and going beyond 3 lags may lower the degrees of freedom, we do not try}

(28)

27

choose two lags to detect the lagged effect of human capital on China’s economic development.

Table 5: Fixed effects results with lags of human capital

Log per capita GDP Full sample

No lag

Full sample One lag

Full sample Three lags

Log human capital 0.0370***

Log human capital (-1)

Log human capital (-3)

(0.0102)

0.0320*** (0.00694)

0.0280*** (0.00752) Log per capita GDP (-1) 0.859*** 0.857*** 0.852*** (0.0162) (0.0154) (0.0140) Log physical capital 0.0575*** 0.0557*** 0.0112

(0.0104) (0.0101) (0.0124) Log trade openness 0.0161*** 0.0163*** 0.0221***

(0.00549) (0.00529) (0.00676)

Log FDI 0.00638*** 0.00649*** 0.00417*

(0.00210) (0.00210) (0.00221) Log innovation index 0.0231*** 0.0228*** 0.00709

(0.00803) (0.00798) (0.00688) Log human capital _ innovation index 0.00601*** 0.00570*** 0.0118*** (0.00212) (0.00211) (0.00229) Constant -0.0350 -0.0254 0.154*** (0.0438) (0.0418) (0.0570) Observations 2,381 2,369 1,843 R-squared 0.961 0.961 0.941 Number of city 274 274 274

(29)

V. Robustness test

Table 6: Fixed effects results for the alternative measurement of human capital

Dependent variable: (1) (2) (3) (4) (5)

Log per capita GDP Full sample 2004-2013 Big cities 2004-2013 Small cities 2004-2013 Sub-periods 2004-2008 Sub-periods 2008-2013

Log expenditure on education 0.0250*** 0.0201** 0.0249*** 0.0333*** -0.0227*** (0.00514) (0.00858) (0.00565) (0.00500) (0.00672)

Log per capita GDP (-1) 0.844*** 0.789*** 0.845*** 0.697*** 0.623***

(0.0154) (0.0647) (0.0161) (0.0449) (0.0312)

Log physical capital 0.0473*** 0.123** 0.0462*** 0.102*** -0.0858***

(0.0105) (0.0524) (0.0109) (0.0224) (0.0277)

Log trade openness 0.0117** 0.0713* 0.0105* 0.0149 -0.00452

(0.00533) (0.0398) (0.00540) (0.0109) (0.0127)

Log FDI 0.00634*** -0.00258 0.00725*** 0.00556** 0.00329

(0.00208) (0.00600) (0.00218) (0.00261) (0.00337)

Log innovation index 0.0186** 0.0503* 0.0178** 0.129*** 0.0647***

(0.00772) (0.0242) (0.00826) (0.0224) (0.0171) Constant 0.0817* -0.450 0.0875* 0.0241 0.814*** (0.0457) (0.302) (0.0487) (0.0923) (0.138) Observations 2,391 171 2,220 1,059 1,066 R-squared 0.962 0.963 0.962 0.907 0.812 Number of city 274 19 255 274 274

As mentioned earlier, there might exist endogeneity issues: there might exist reverse causality between human capital and economic development due to the fact that big cities with higher levels of economic development also have more means to invest in education so that there will be more human capital accumulation in highly-developed cities. The consequence of this endogeneity issue is that we may mistake the cause (GDP per capita) as the result (human capital). One way to solve this endogeneity problem is to identify appropriate instrumental variables. However, we do not find appropriate instrumental variables. So we have to acknowledge that we do not address this problem and that remains the weakness of this thesis.

(30)

29

measurement errors might occur, because this proxy is appropriate under the assumption that the percentage of enrolled university students is proportional to the share of graduates. Yao & Zhang (2001) use public expenditures on education as a proxy of human capital. And we also use public expenditure on education as an alternative measurement of human capital to examine if our main findings are robust when we use different measures of human capital. The results for the alternative measurement of human capital for equation (1) are listed in Table 6. It can be seen from column (1) that the overall impact of expenditure on education on economic development is still positive and statistically significant, which proves the robustness of our previous findings on the overall effect of human capital on China’s economic development. Then, we can see from column (2) and column (3) that the coefficients of human capital are similar in big cities and small cities. This is reasonable as public expenditure on education is likely to show a lagged effect rather than an instantaneous effect on GDP. So we also run the regression using a couple of lags of expenditure on education and find that when we use 3-year, 4-year, and 5-year lagged public expenditure on education, the coefficients are larger in big cities than in small cities (not reported in the table), which is consistent with our expectations that human capital contributes more to China’s economic development in big cities. Finally, we look at column (4) and column (5) and find that the coefficient of human capital before the crisis is larger than that after the crisis. This result is also consistent with our previous findings. Therefore, according to the results of alternative measures of human capital regressions, our main findings are not sensitive to different measures of human capital.

(31)

Table 7: Fixed effects results for the interplay of human capital and alternative measurement of innovation capacity (patent)

Log per capita GDP Full sample Big cities Small cities

Log human capital 0.0575*** 0.0411 0.0558***

(0.0131) (0.0309) (0.0143)

Log per capita GDP (-1) 0.861*** 0.816*** 0.861***

(0.0129) (0.0526) (0.0130)

Log physical capital 0.0544*** 0.0869** 0.0537***

(0.0104) (0.0385) (0.0107)

Log trade openness 0.0149*** 0.0529 0.0140**

(0.00538) (0.0353) (0.00542)

Log FDI 0.00668*** -0.00467 0.00794***

(0.00200) (0.00491) (0.00203)

Log patent 0.0219*** 0.0507 0.0209***

(0.00633) (0.0327) (0.00654)

Log human capital _ patent 0.00728*** 0.00969 0.00685***

(0.00202) (0.0134) (0.00239) Constant 0.0388 -0.193 0.0476 (0.0507) (0.194) (0.0532) Observations 2,381 171 2,210 R-squared 0.961 0.962 0.961 Number of city 274 19 255

(32)

31

VI. Conclusion

6.1 Conclusions and recommendations

This thesis is aimed to examine the effect of human capital on China’s economic development in the context of economic crisis. Besides, this thesis is aimed to examine if the relationship between human capital and China’s economic development is mediated by the innovation capacity of each city.

In order to answer the first research question, we use a fixed effect model collecting prefecture-level city data from 2004 to 2013, and then we divide the full sample into two non-overlapping sub-periods: 2004 to 2008 and 2009 to 2013 to identify if the effect of human capital varies. Besides, to answer the second research question, we apply an interaction term study to examine the channel through which human capital influences China’s economic development. After the analysis, we come up with several main findings: first, human capital has a positive and significant effect on China’s economic development, and this effect is larger in big cities with better economic level; second, human capital has a larger effect on China’s economic development before the crisis; third, human capital can contribute to China’s economic development by working as a facilitator of city innovation ability.

As is seen from the previous analysis, there is a large gap in human capital between big cities and small cities. And as we do not take migration effect (human capital move from one city to another city) into account, the gap in human capital between big cities and small cities is even larger in reality.

Based on the previous findings, we have the following recommendations: governments of small cities should implement more favorable policies to attract talents by more material rewards. For example, governments in small cities subsidize technological enterprises to make these enterprises more attractive to high-skilled labor. In this way, they can narrow the gap of human capital between them and big cities. And they should also increase the public expenditure on education, as education is the main way that can formulate human capital. Small cities will benefit from attracting human capital by contributing to the city’s innovation ability and they will benefit from investing more in education by cultivating more talents.

To check if our main findings are sensitive to the choice of measurement of human capital, we use public expenditure on education as an alternative measurement of human capital to perform the same analysis of equation (1), and we use the number of patents to perform the same analysis of equation (2), the results of which show that our main findings are not sensitive to different measures of innovation capacity and the variable (city innovation index) we use as a proxy of innovation capacity is reliable.

6.2 Limitations and future research

(33)

high-developed cities might be higher than what we estimate in our model, same for small cities. So if future research can take the migration effect into account, the result will be more precise.

Second, our measure of human capital is appropriate under the assumption that the percentage of enrolled university students is proportional to the percentage of university graduates. And if future research can measure human capital in a more accurate way such as the percentage of university graduates or share of tertiary educated population in the workforce, the result will be more reliable.

The effect of human capital and innovation on China’s economic development University of Groningen Faculty of Economics and Business Msc Economic Development and Globalization Master