Intra-product Specialisation and the Innovation Ability of the High-technology Industry in China

Abstract

The purpose of this paper is to analyse the influence of intra-product specialisation on the innovation ability of the high-technology industry in China. To achieve this research goal, we use the generalised method of moments (GMM) with a dynamic panel data model in our empirical research. Total factor productivity (TFP) and its decomposition variables are used to measure the innovation ability. The Malmquist index is employed to calculate TFP and its decomposition variables, namely, technology progress, technical efficiency and scale efficiency. Previous studies show that intra-product specialisation can raise TFP, but our empirical results reveal that the impact of the intra-product specialisation on the TFP is uncertain: it decreases TFP in the initial period but increases it in the lag period. Results also indicate that an increase in the proportion of state-owned property rights exerts a positive influence on TFP and the change in pure technical efficiency. Technical efficiency is greatly influenced by intra-product specialisation, whereas intra-product specialisation exerts a slight influence on technological progress. The capital investment in research and development (R&D) has a positive effect on TFP, and the contribution rate of R&D labour input to TFP is relatively low. The main conclusion is that China need to improve the degree of intra-product specialisation, accelerate the evolution of industry technical standards and reduce the coordination cost of intra-product specialisation to enhance the innovation ability of the high-technology industry.

Introduction

After decades of rapid growth, the economic development of China is now facing new challenges. As the speed of economic growth shifts from high to medium-high, the industrial structure is optimised continuously. Moreover, the economic development has transformed from the cheap-labour-driven and investment-driven modes to the innovation-driven mode. Thus, implementing an innovation-driven development strategy is a promising endeavour. The sustainable economic development of China relies not only on technological innovation and product innovation but also on organisational innovation, which fully exploits global production factors efficiently. Economic organisation is used as a key factor of a country’s economic growth in classical economics, but it is not strongly emphasised in neoclassical economics (Yang & Shi, 1995). The division of labour and specialisation level in the information economy era are higher than that in the industrial economy era. Economic organisation plays a crucial role. The promotion of independent innovation ability through efficient economic organisation is fundamental for China. The country should occupy the high end of global value chains to enhance its global competitiveness. The industries of different countries have adapted to the active promotion of information technology since the beginning of the twenty-first century. The international division of labour has developed from intra-industry division to intra-product specialisation, which is highly complex. The different processes or modules of a product are completed by the production networks of inter-enterprises (Wang & Hsu, 2014). Intra-product specialisation is the essence of industrial structure in the new economic era. The high-technology (high-tech) industry in China also presents a trend of intra-product specialisation. Total factor productivity (TFP) can deeply reflect the influence generated by scientific and technological innovation and progress, scale economy and organisational innovation, and other advanced production factors (Zhang, 2015). Therefore, TFP can represent the innovation ability of an industry. How does intra-product specialisation affect the technological progress and efficiency of the high-tech industry? The present study empirically investigates the effect of intra-product specialisation on the TFP of the high-tech industry in current and lag periods. The TFP of the high-tech industry may be affected by other industrial environment factors, in addition to the degree of intra-product specialisation (Segarra-Ona & Peiro-Signes, 2014). After considering the importance and availability of data, we adopt the property right structure, industrial opening degree and industrial research and development (R&D) investment of enterprises as control variables. We calculate TFP and its decomposition variables (i.e., technology progress, technical efficiency and scale efficiency) with the Malmquist index on the basis of the data envelopment analysis (DEA) method. We also use the generalised method of moments (GMM) and dynamic panel data models to study intra-product specialisation and the effect of industrial environment factors on TFP and its decomposition variables. In the empirical study, the Data Envelopment Analysis Program (DEAP) software is used to calculate TFP and its decomposition variables, and the GMM dynamic panel data model is analysed with the STATA software. The following basic assumptions are derived on the basis of the analysis: Intra-product specialisation can improve the TFP of the high-tech industry in China in the long term, and industry environment factors exert a significant influence on the TFP of this particular industry.

Literature Review

The global division of labour has developed from the product level to the process or module level. The process of producing a product is divided into different steps that are distributed to different countries; such process is called intra-product specialisation (Gereffi, Humphrey & Sturgeon, 2005; Song, 2011; Vogiatzogloua, 2012). A number of scholars believe that the theory of comparative advantage is the theoretical basis of intra-product specialisation (Lu, 2007). Intra-product specialisation is also called vertical specialisation (VS) (Hummels, Ishii & Yi, 2001; Hummels, Rapoport & Yi, 1998; Kohler, 2003), global outsourcing, slicing up of the value chain or modularity (Brusoni & Prencipe, 2011; Cheng, 2011; Lau, Yam & Tang, 2011). Falk (2012) investigated the impact of international outsourcing to low and high income countries on TFP growth based on manufacturing industry data of fourteen Organisation for Economic Co-operation and Development (OECD) countries from 1995 to 2000. Furthermore, Falk made a distinction between the outsourcing of materials from developed countries and that from the emerging nations, and also made a distinction between outsourcing of materials and outsourcing of services. Falk’s research gives us much inspiration.

Intra-product specialisation is closely related to VS. The vertical specialisation index (VSI) is often used to measure the degree of intra-product specialisation and outsourcing (Grossman & Helpman, 2005). Scholars mainly use three methods to measure VS. The first method employs the amount of parts trade data or the processing trade data to calculate VS degree (Feenstra & Hanson, 1996; Görg, Strobl & Hanley, 2008; Hanson, Mataloni & Slaughter, 2005). In this method, imported products are divided into intermediate and final products, but such division is difficult in practice. The second method involves the use of input and output tables and export trade data to measure VS. The vertical specialisation share (VSS) proposed by Hummels et al. (2001) is the most representative. The VSS index is the proportion of imported intermediate products in a country’s export commodities, and it has been improved by many scholars, including Dean, Fung and Wang (2011). In the latest development of the input–output method, multinational customs data and transnational input–output tables are combined to solve the problem of repeated calculation caused by multiple cross-borders of intermediate products (Koopman, Wang & Wei, 2014). Several Chinese research institutes and scholars have used this method to conduct empirical studies in China. The measurement methods of Hummels require input–output tables and an inter-industry consumption coefficient matrix. However, the National Statistics Bureau of China releases limited annual input–output table data, thus increasing the difficulty of empirical research in China. In addition, the implicit assumptions of this method can hardly meet the real situation in many countries. The third method is the use of the VSI developed from VSS. The VSI measures VS or outsourcing level through the proportion of intermediate goods in the total output. The empirical studies of Egger and Egger (2006) used VSI to measure outsourcing. The VSI also needs input–output table, but the high-technology industry input–output table in China has not been announced by the government. As for whether the VS can be calculated without the input–output table, Liu and Wu (2006) and Dai (2012) believed that the simplified method of value-added service (VAS) can also measure the degree of VS and the disintegration of production, which provides important implications for this paper. According to the VAS method, total output = intermediate input + value added, VAS = 1 – value added/gross output. So VAS can avoid the use of input–output table, and make the research object in a continuous time.

Some scholars have studied extensively on intra-product specialisation or VS from the perspective of global value chain. However, VS and intra-product specialisation differ in terms of the study objects. Vertical specialisation focuses on the international division of the value chain, whereas intra-product specialisation emphasises the segmentation of the value chain at home and abroad. It is inadequate to solely focus on international trade data while studying intra-product specialisation. The VAS method essentially involves the segmentation of the value chain at home and abroad. Therefore, we suggest using the VAS method to measure the degree of intra-product specialisation. The influence of intra-product specialisation on innovation ability in the current and lag periods is also studied in this paper.

Moreover, some scholars have studied the TFP of China’s manufacturing industry and high-tech industry. Zhang, Sun, Delgado and Kumbhakar (2012) analysed the impact of R&D on the productivity of China’s high-technology industry by using a novel semiparametric approach that allows heterogeneities across provinces and time. Zhao and Boasson (2015) measured the productivity spillover effects of offshore outsourcing in the Chinese manufacturing industry by using the Cobb–Douglas (C–D) production function and VSS. These studies give us a lot of inspirations.

Research Methods and Analytical Framework

Empirical Model

The GMM dynamic panel data model can overcome endogeneity and the industry effects of section data, especially in a research using large cross sections and short-term data. To measure the effect of intra-product specialisation on TFP accurately, we use the GMM dynamic panel data model, which is written as follows:

Y_{i t} = α X_{i t} + β W_{i t} + v_{i} + ε_{i t} i = 1, 2, ..., N; t = 1, 2, ..., T

(1)

In Equation (1), i represents the high-tech industry; t represents time; v_i represents the industry effect, and the value of different industries are not identical; ε_it denote random disturbance terms, which exhibit a normal distribution. X_it is strictly an exogenous variable in the model. X_it is not related to the random disturbance term of the current or previous period. W_it represents the predetermined and endogenous variables in the model. α and ß are the parameters to be estimated. Two assumptions of the model are as follows: (i) zero expectation and mutual independence and (ii) no correlation between groups of error terms. First, we identify a difference in the model to solve the problem of omitted variables. However, the predetermined variables after identifying the difference are not strictly exogenous; we must use the instrumental variable for the estimation. The GMM method uses the lagged term of the explanatory variable as the instrumental variable. We validate the model with the Arellano–Bond test. If the residual term in the GMM regression system indicates autocorrelation, we can infer that the variable is not strictly exogenous (Dai, 2012). The Sargan test is used to examine the problem of over-identification. The STATA software is used in the econometric analysis.

Measurement of Intra-product Specialisation

Proposed by Hummels et al. (2001), VSS is the most recognised method for calculating VS. According to Hummels’ model (Hu, 2006),

V S S_{z} = \frac{V S_{z}}{E X_{z}}

(2)

where VS_z represents the total VS trade volume of all industries in country Z and EX_z represents the total export volume of country Z. The VS of industry i can be defined as:

V S_{i} = \frac{M_{i}^{I}}{Y_{i}} \cdot X_{i} = \frac{X_{i}}{Y_{i}} M_{i}^{I}

(3)

In Equation (3), $M_{i}^{I}$ is the intermediate goods imported of industry i, Y_i represents the total output of industry i and X_i represents the industry’s total exports. Assuming that n industries exist in country Z, then the total VS degree of all industries in country Z is computed as:

V S_{z} = \sum_{i = 1}^{n} V S_{i}

(4)

Thus, we can define the proportion of VS_z in the export of country Z.

V S_{z} = \frac{V S_{z}}{X_{z}} = \frac{\sum_{i = 1}^{n} V S_{i}}{\sum_{i = 1}^{n} X_{i}}

(5)

In Equation (5), X_z represents the total export value of country Z. The input–output tables are used to calculate the VS degree, from which we can obtain the data of industrial input, output and export. Thus, Equation (5) can be written in the following matrix form:

\frac{V S_{z}}{X_{z}} = \frac{1}{X_{z}} u A^{M} X^{V}

(6)

In Equation (6), u is the 1 × n vector consisting of 1, A™ is the n × n dependence coefficient matrix of imported intermediate products and X^V is the n × 1 export vector. If the full coefficient matrix is used and the imported intermediate products indirectly included in the export are considered, we have:

\frac{V S_{z}^{i}}{X_{z}} = \frac{1}{X_{z}} u A^{M} {(I - A^{D})}^{- 1} X^{V}

(7)

In Equation (7), $\frac{V S_{z}^{i}}{X_{z}}$ is the total VS trade volume of all industries in country Z, including the imported intermediate products indirectly counted in the export. I is the unit matrix, and A^D is the n × n domestic consumption coefficient matrix. A is the consumption coefficient matrix of input–output tables (Hu, 2006).

A = A^{D} + A^{M}

(8)

The measurement of intra-product specialisation is different from that of VS. When building the index system of VS, we must distinguish between intermediate and final products or obtain the inter-industry consumption coefficient matrix to study the proportion of international industrial division of a certain country. In practice, intermediate products are difficult to define, and obtaining inter-industry consumption coefficient matrix statistics is challenging. Furthermore, the National Bureau of Statistics of China only releases a few years’ worth of input–output table data. Intra-product specialisation requires the study of the value-added proportion of an industry or enterprise in the product value chain and the measurement of the division of the value chain among the related industries or enterprises. When studying the division of the value chain among related industries or enterprises, we do not need to obtain inter-industry trade or consumption coefficient matrix data. Hence, we will use an approximate method to estimate the degree of VS of China’s high-tech industries. Because:

S V = I V + A V

(9)

In Equation (9), AV represents the added value of industrial value chain, IV represents the inputs value of industrial intermediate and SV represents the sum value of the whole industrial chain. Therefore, the proportion of the intermediate inputs value in the total industrial value chain (i.e., the intra-product specialisation degree) can be simply represented by the proportion of the added value in the total industrial value chain. The VAS (or the intra-product specialisation degree) can be stated as Equation (10):

V A S = \frac{I V}{S V} = 1 - \frac{A V}{S V}

(10)

The VAS has been used by some scholars, such as Dai (2012), but the scholars have not realised that the VAS based on the added value method is not suitable to measure the degree of international division of labour. The VAS method is more suitable for studying the degree of separation between industries or industrial intra-product specialisation than for reflecting the degree of international industrial division or VS. If a value chain decomposes or restructures among different companies in an industry, the value of VAS can effectively reflect this change.

Measurement of TFP

Total factor productivity shows the economic growth caused by intangible production factors, including scientific and technological progress, organisational innovation, production innovation, specialisation and scale economy. Thus, TFP can reflect the efficiency of technology development and the quality of economic growth.

The traditional TFP calculation methods can be divided into three kinds: growth accounting method, index method and production frontier method. Growth accounting method is used to measure the contribution of different production factors to economic growth. As a residual of traditional production function, it was introduced by Solow (1957). The translog production function is a generalisation of the C–D production function. The index method calculates TFP by the ratio of total output and total input of a production unit. A lot of scholars use growth accounting method and index method to calculate TFP (Balakrishnan, 2004; Balakrishnan, Pushpangadan & Babu, 2000; Goldar, 2004; Kathuria, Rajesh & Sen, 2013). The translog (transcendental logarithmic) index method is a development of the C–D production function (Berndt & Christensen, 1973). Translog index method can be used to distinguish the contribution of the quantity and quality of factors to output. Tornqvist index is also a commonly used index method. The use of Tornqvist index is relatively simple, and it suitable for microeconomic analysis. But Tornqvist index and translog index cannot provide more information about technical progress and technical efficiency. The production frontier method estimates TFP through the comparison of the actual value and the optimal value of production. The production frontier methods can be divided into parametric and non-parametric methods. When a parametric method is applied to estimate TFP, assumptions such as a perfectly competitive market and scale benefit can be relaxed. However, priori assumptions must be given to the estimated parameters, and such requirement may lead to the instability of the estimated parameters. The non-parametric estimation method is suitable for the regular statistical research of productivity. This paper uses the non-parametric method (Malmquist index) to estimate the TFP of the high-tech industry in China. Some economists have used Malmquist index to calculate the productivity of India (Deb & Ray, 2014; Mitra, Sharma & Veganzones-Varoudakis, 2012; Mitra, Varoudakis & Veganzones-Varoudakis, 2002), which has given a lot of inspiration to this paper. The Malmquist index is based on DEA. It was initially introduced by Sten Malmquist. Then Caves, Christensen and Diewert (1982) applied the method to conduct a productivity analysis. They constructed a productivity index with the ratio of the distance function and named it the Malmquist productivity index. Thereafter, Färe, Grosskopf, Norris and Zhang (1994) further decomposed the Malmquist index into technical efficiency change, technological progress and scale efficiency change. The formula is as follows:

M_{0} (x^{t + 1}, y^{t + 1}; x^{t}, y^{t}) = [\frac{D_{0}^{t} (x^{t + 1}, y^{t + 1})}{D_{0}^{t} (x^{t}, y^{t})} * \frac{D_{0}^{t + 1} (x^{t + 1}, y^{t + 1})}{D_{0}^{t + 1} (x^{t}, y^{t})}]^{\frac{1}{2}}

(11)

= \frac{D_{0}^{t + 1} (x^{t + 1}, y^{t + 1}}{D_{0}^{t} (x^{t}, y^{t})} [\frac{D_{0}^{t} (x^{t + 1}, y^{t + 1})}{D_{0}^{t + 1} (x^{t + 1}, y^{t + 1})} * \frac{D_{0}^{t} (x^{t}, y^{t + 1}}{D_{0}^{t + 1} (x^{t}, y^{t})}]^{\frac{1}{2}}

(12)

= e f f * t e c h

(13)

= p e c h \times \sec ​ h \times t e c h

(14)

In Equations (11)–(14), x^t and y^t are the numbers of input and output in t period, respectively. x^t⁺¹ and y^t⁺¹ represent the total input and output in t + 1 period, respectively. $D_{0}^{t}$ and $D_{0}^{t + 1}$ represent the distance function in t and t + 1 period, respectively.

Equation (13) implies that TFP can be decomposed into technical efficiency change (eff ) and technological progress change (tech). Under the conditions in which the scales return is constant and the investment has no limit, eff is the technical efficiency change. eff represents the distance between the industry’s actual production surface and the best frontier during t to t + 1 period. tech indicates that the technological frontier moves from t to t + 1 period. Such movement is called technological progress. In Equation (14), the efficiency change index can be decomposed into pure technical efficiency change (pech) and scale efficiency change (sech). Therefore, TFP can be expressed as: pech×sech×tech.

The DEAP software is used to calculate TFP. We use the following data to calculate the TFP: the output data is the added value; capital investment is equal to the amount of capital stock plus the new capital per year; the weighted average number of employees is used to measure labour input. In calculating TFP, the calculation method of capital input is involved in many arguments. This work replaces capital input with capital stock to simplify the analysis. This approach is commonly practised in China. Only capital and human input are considered when solving the Malmquist index. We use the added value to indicate the output and employ 1995 as the base period. The fixed asset is deflated by the price index of fixed assets investment.

Control Variables

In addition to the degree of intra-product specialisation, TFP may be affected by other variables. After considering the importance of variables and the availability of data, we use property right structure, export performance and R&D manpower and capital input as control variables (Dai, 2012). But the measurement indexes of the control variables (property right structure and export performance) are different from that of Dai’s control variables.

Property right structure. When measuring property right structure, we adopt the method by Jefferson, Bai, Guan and Yu (2006), which involves the use of the proportion of state-owned enterprises accounted for the proportion of the whole industry. We use ‘prop’ to represent property right structure.

Export performance. Import penetration and export performance can reflect ratio of dependence on foreign trade. Considering the research object and availability of research data, we utilise export performance to measure the degree of involvement of a country in international trade. Export performance is used to measure the ratio of export of a particular industry. The equation for this index is as follows: Export performance = X_i/Q_i. X_i represents the number of exports of the country’s i industry and Q_i stands for the total amount of domestic production. In this paper, ‘export’ is used to show ratio of dependence on foreign trade.

R&D input comprising capital and manpower input. The R&D capital investment is deflated with the ‘consumer price index’ and the weighted average of the fixed assets investment price index (Li, 2007). ‘Log l’ and ‘log k’ are used to represent the logarithmic values of the R&D labour input and capital input, respectively.

Empirical Data and Analysis

The Chinese government began releasing high-tech industry data in 1995, but it stopped publishing the key data (i.e., added value) for measuring intra-product specialisation degree after 2011. The data used in the present study are obtained from the Statistical Yearbook of China’s High-tech Industry, the China Statistical Yearbook and the website of the high-tech industry division of the National Development and Reform Commission in China. We screen the data of 18 sub-industries, including the chemical pharmaceutical manufacturing industry, on the basis of The Classified Catalog of the High-tech Industry published in 2012 (see Table 1). The added value data of the high-tech sub-industries from 2008 to 2011 are calculated according to the data of cumulative increase published on the website of the high-tech industry division of the National Development and Reform Commission in China. The price index data are derived from China Statistical Yearbook, and other data are generated from the Statistical Yearbook of China’s High-tech Industry. Given the lack of information on the export delivery volume in 1996 and 1997, we assume no changes in the growth rate of the export delivery volume in 1996, 1997 and 1998. The data of the communication equipment manufacturing industry in 1995 were not published; thus, we use the data in 1996 to re-estimate the data in 1995. We assume that the growth rate of fixed assets in 1994, 1995 and 1996 is the average growth rate from 1996 to 2000. Given the changes in the statistical calibre, the fixed assets from 2009 to 2011 are shown in the section of total assets rather than in the year-end original value of fixed assets. Moreover, the ‘new increase in fixed assets’ is the sum of the ‘update, transformation, and new fixed assets’ and ‘basic construction newly added fixed assets’ in 1997 and 1998. The ‘total assets’ in 2009 showed an abnormal change relative to the changes in the remaining years, which did not meet the normal economic laws. To ensure the reasonableness of the data, we assume that the ‘total assets’ growth rate did not change in 2010 and 2011, and then we re-estimate the value of total assets in 2009. Added value and capital stock are closely related to the fixed assets investment price index; therefore, the added value and capital stock are deflated by fixed assets investment price index. Export delivery value of industrial goods is based on producer price index of industrial goods, and all of the 18 high-tech industries belong to the manufacturing sector, so export delivery value of high-tech industries has been deflated by producer price index. The R&D expenditure is affected by internal and external factors; hence, R&D expenditure is deflated by the weighted average of the consumer price index and the fixed asset investment price index. The fixed asset is deflated by the fixed asset investment price index.

Table 1

High-tech Sub-industries in the Empirical Study

Two-digit Code Industry	Three-digit Code Industry	Four-digit Code Industry
Pharmaceutical manufacturing industry	Chemical pharmaceutical manufacturing Chinese traditional medicine manufacturing Biological and biochemical products manufacturing
Aircraft and spacecraft manufacturing industry	Aircraft manufacturing and repairSpacecraft manufacturing
Electronic and communication equipment manufacturing industry	Communication equipment manufacturing	Communications transmission equipment manufacturingCommunication switching equipment manufacturingCommunication terminal equipment manufacturing
	Radar and ancillary equipment manufacturingBroadcasting and television equipment manufacturing
	Electronic device manufacturing	Electronic vacuum device manufacturingSemiconductor discrete device manufacturingIntegrated circuit manufacturing
	Electronic components manufacturing
Electronic computer and office equipment manufacturing industry	Whole computer manufacturingComputer peripheral equipment manufacturing
Medical equipment and instruments and meters manufacturing industry	Medical equipment and instruments manufacturing, measurement instruments and meters manufacturing

Source: Chinese Statistical Yearbook of High-tech Industry.

The TFP of the high-tech sub-industries in China undergo a fluctuating process (see Figure 1). The progress of science and technology and the advancement of economic reform led to the continuous growth of China’s TFP from 1995 to 2006. The Chinese government increased its investments to stimulate the economy and overcome the devastating effects of the international financial crisis in 2008. The average annual growth rate of the fixed assets investment of China is up to 26.1 per cent from 2007 to 2010. Meanwhile, TFP began to decrease in 2008. The growth rate of the fixed assets investment of China declined, and its TFP showed a slow rising trend after 2010. Among the five high-tech sub-industries, the aircraft and spacecraft manufacturing industry showed an obvious increase in TFP, whereas the computer and office equipment manufacturing industry exhibited small fluctuation amplitude. The TFP values of the aircraft and spacecraft manufacturing, pharmaceutical manufacturing, medical equipment and instruments and meters manufacturing industries showed a sharp decline after 2008 and achieved stability after 2010. We use the data from 1995 to 2011 in the regression model of this paper because of the lack of official added value data at the industry level.

Figure 1

Source: Authors’ calculation.

The pech (the change of pure technical efficiency) development trends of high-tech industries present different development trends (see Figure 2). The pech of pharmaceutical manufacturing industry, aircraft and spacecraft manufacturing industry and medical equipment and instruments and meters manufacturing industry all have the process of falling at first and then rising. The pech of electronic computer and office equipment manufacturing industry is at a stable level. The pech of electronic and communication equipment manufacturing industry always maintains a relatively low level.

Figure 2

Source: Authors’ calculation.

The sech (the change of scale efficiency) development trends of high-tech industries are shown in Figure 3. The trends of sech in pharmaceutical manufacturing industry, medical equipment and instruments and meters manufacturing industry and electronic and communication equipment manufacturing industry all have an upward trend. The trends of sech in aircraft and spacecraft manufacturing industry and electronic computer and office equipment manufacturing industry all maintain a relatively low development level.

Figure 3

Source: Authors’ calculation.

The tech (technological progress) development trends of five high-tech sub-industries are all showing an upward fluctuation trend. But the downward trend of technological progress in each sub-industry is particularly evident after 2008. Since the reform and opening up, China’s economic growth has been mainly relying on export-oriented policies, as well as investment-driven policies. China’s economic development has been largely driven by the government, which is very different from most Western countries. After the 2008 international financial crisis, China’s export growth fell sharply while the domestic demand went relatively weak. Then the Chinese government launched an economic stimulus plan of four trillion yuan. In addition, the local governments of China also launched trillions of yuan of economic stimulus plans. As a result, a large part of these funds flowed into the real estate market through various channels since real estate was considered a high-profit industry after the global financial crisis. As a result, China’s economic growth has been driven by investment after the global financial crisis rather than by technological progress. In order to avoid the disadvantages of the development strategy driven by investment and low-quality labour force, the Chinese government put forward the ‘innovation-driven’ development strategy in 2012. It is obvious that technological progress has a lower contribution to economic growth than the investment, and the government is trying to change this unfavourable situation (Figure 4).

Figure 4

Source: Authors’ calculation.

The VAS (intra-product specialisation index) of five high-tech sub-industries shows a narrow range of shock development trend. In addition to the pharmaceutical manufacturing industry, the indexes of the other four industries are significantly improved (see Figure 5).

Figure 5

Source: Authors’ calculation.

Results and Discussion

In this study, the impact of the current period of intra-product specialisation on TFP is investigated by examining the relationship between the VAS and the TFP in the current period. The effect of lag period of intra-product specialisation on TFP is inspected by analysing the relationship between the VAS and the TFP in the lag period. The regression results indicate the rationality of the model setting and the effectiveness of the instrument variables (see Table 2). The p value of AR(1) significantly neglects the primary hypothesis of the non-existence of auto-regression. The p value of AR(2) verifies that auto-regression does not exist. Thus, the model setting is rational. The p value of the Sargan test is 0.91, which ensures the reliability of the instrument variables. Except for the t value of ‘export’, the t values of the other parameters are rejected by the null hypothesis at the 5 per cent significance level. The result of intra-product specialisation degree in the current period is negatively related to the TFP. Meanwhile, the degree of intra-product specialisation has a positive effect on TFP in the lag period. Other variables indicate that when the proportion of state-owned ownership increases, it can promote TFP. This outcome may be due to the government strongly supporting certain high-tech sub-industries, which play a certain role in promoting TFP. Nevertheless, high state ownership does not necessarily result in enhanced TFP. Export performance can promote TFP in certain aspects but not dramatically. The R&D capital investment has a positive influence on TFP, whereas R&D labour investment has a negative effect.

Table 2

Influence of Intra-product Specialisation on TFP

Variable	TFP_t–1	VAS	VAS_t–1	Prop	Log l	Log k	Export	AR(1)	AR(2)	Sargan
Result	0.93***(39.11)	–6.93***(–12.40)	6.07***(10.47)	0.23**(2.15)	–0.13***(–2.88)	0.08**(2.40)	0.01(0.03)	0.00	0.64	0.91

Source: Authors’ calculation.

Notes: Figures in brackets are t statistics. ***, ** and *, respectively, denote the statistical significance levels of 1%, 5% and 10%. Subscript ‘t – 1’ indicates one lag period. The values of AR(1), AR(2) and Sargan test are p values.

TFP can be decomposed as the product of pech (pure technical efficiency change), sech (scale efficiency change) and tech (technological change). We carry out the regression of the three decomposition variables on TFP in this work (see Table 3). The results show that intra-product specialisation exerts a negative effect on the decomposition variables in the current period but a positive effect in the lag period. Therefore, the long-term impact of the intra-product specialisation on the TFP is uncertain. According to the significance of coefficient, intra-product specialisation has the greatest effect on technological efficiency, followed by scale efficiency and technological change. Intra-product specialisation achieves a minimal effect on technological progress.

Table 3

Influence of Intra-product Specialisation on the Decomposition Variables of TFP

Variable	VAS	VAS_t–1	Prop	Log l	Log k	Export	AR(1)	AR(2)	Sargan
pech	–2.00***(–6.54)	1.76***(5.67)	0.10*(1.62)	–0.20(0.81)	0.02(1.23)	0.02(0.21)	0.00	0.32	0.96
sech	–1.13***(–4.47)	1.09***(4.29)	–0.15***(–2.74)	–0.01(–0.83)	0.019(1.28)	–0.12(–1.57)	0.00	0.04	0.72
tech	–0.59***(–0.88)	0.57(0.86)	0.33***(2.46)	–0.08(–1.49)	0.02(0.7)	0.10(0.48)	0.00	0.01	0.94

Source: Authors’ calculation.

To explore the differences among the industries, we analyse the influences of five industries (i.e., pharmaceutical, aircraft and spacecraft, electronics and communication equipment, electronic computer and office equipment, medical equipment and instruments and meters manufacturing industries) on TFP. The result shows that intra-product specialisation has a significantly negative effect on TFP in the current period. The effect of the intra-product specialisation of the aircraft and spacecraft manufacturing industry on TFP is insignificant in the lag period. The intra-product specialisation of the other four industries improves productivity efficiency in the lag period. In terms of coefficient size, the influence of the intra-product specialisation of the selected industries on TFP is listed in a descending order: medical equipment and instruments and meters, electronic computer and office equipment, pharmaceutical, electronic communication and equipment and aircraft and spacecraft manufacturing industries (Table 4).

Table 4

Influence of Intra-product Specialisation on the TFP of High-tech Sub-industries

Variable	TFP_t–1	VAS	VAS_t–1	Prop	Log l	Log k	Export	AR(1)	AR(2)	Sargan
Industry 1	0.90***(12.52)	–6.90***(–6.99)	5.32***(4.92)	–0.02(–0.04)	0.03(0.18)	0.05(0.43)	2.98**(2.11)	0.00	0.48	1.00
Industry 2	0.62***(5.69)	–13.34***(–4.87)	0.095(0.03)	0.61(0.62)	–0.26(–1.50)	0.35*(2.00)	3.16(1.04)	0.01	0.15	1.00
Industry 3	0.76***(16.16)	–7.87***(–8.93)	3.93***(4.21)	–0.52**(–2.32)	–0.27***(–3.84)	0.18***(3.60)	–0.33(–1.24)	0.00	0.52	0.98
Industry 4	0.93***(10.37)	–8.27***(–4.75)	7.01***(5.82)	–0.45(–0.72)	–0.19*(–1.85)	0.15*(1.83)	–1.66***(–2.93)	0.05	0.90	1.00
Industry 5	0.93***(10.24)	–9.78***(–3.96)	11.05***(3.89)	2.87**(2.09)	0.02(0.09)	0.09(0.40)	4.43**(2.31)	0.35	0.29	1.00

Source: Authors’ calculation.

Notes: Industry 1 is the pharmaceutical manufacturing industry, Industry 2 is the aircraft and spacecraft manufacturing industry, Industry 3 is the electronic equipment and communication equipment manufacturing industry, Industry 4 is the electronic computer and office equipment manufacturing industry, and Industry 5 is the medical equipment and instruments and meters manufacturing industry. The figures in brackets are t statistics. ***, ** and *, respectively, denote the statistical significance levels of 1%, 5% and 10%. Subscript ‘t – 1’ indicates one lag period. The values of AR(1), AR(2) and Sargan test are p values.

Intra-product specialisation can improve TFP in the lag period, but it can slow down TFP in the current period for four reasons. First, decomposing complicated products and setting unified rules require considerable time and technology. Second, at the beginning of intra-product specialisation, the complex product system of association rules cannot immediately adapt to the market environment. As a result, adjustment and error correction costs increase. When the intra-product specialisation enters maturity, the running efficiency of a product system increases, and the cost declines. Third, in the early development period of intra-product specialisation, technology innovation remains at the mutual imitation and small breakthrough stages. Afterwards, intra-product specialisation acquires an increasing number of innovative branches and gains important technological breakthroughs. Fourth, innovation resources are not widely available at the beginning of intra-product specialisation, but they become abundant and mature at a later stage.

The influence of intra-product specialisation on technology efficiency is significant, but its effect on technological change is relatively small. This finding is closely related to the mechanism of intra-product specialisation. Intra-product specialisation can increase human capital accumulation through a specialised division of labour and thus promote technical efficiency. Although technological changes are influenced by intra-product specialisation, it is not the main factor of technological changes; R&D inputs occupy a more important position. However, the influence of the intra-product specialisation of the aircraft and spacecraft manufacturing industry is insignificant in the lag period. This industry is highly monopolised because of national security; thus, making a judgement on the basis of market rules is difficult.

The R&D input has a negative effect on TFP. The possible reason is that a low-level labour input does not lead to the TFP growth of the high-tech industry in China. Only a high level of human capital investment can result in TFP growth. State-owned property rights increase the TFP of high-tech industries, which may be related to the selected industries. In the empirical study, we select aircraft and spacecraft, electronic communication equipment and other industries. In these industries, many large state-owned enterprises control the industrial chains. The export performance that can promote the improvement of TFP is not significant because intra-product specialisation can improve the TFP of high-tech industries through international trade in the early period. When intra-product specialisation develops to a certain degree, the local production and R&D of intermediate products gradually increase. The perspective of international trade can only partly explain intra-product specialisation. The configurations of domestic value chains are crucial in a large manufacturing country such as China.

Some scholars consider the influences of VS and intra-product specialisation the same in their researches. Dai (2012) found that VS enhanced TFP, but our empirical results support the conclusion that the impact of intra-product specialisation on the TFP of high-tech industry is uncertain. And there are reasons for the paradox. First, China implemented a policy to stimulate the economy after the 2008 international financial crisis, resulting in the decline of the TFP, which makes the empirical results appear different. Second, Because of the length of the article, Dai (2012) did not list all the data sources and the empirical results. So the data we selected may not be the same. Third, the control variables we selected are not the same as that of Dai’s. In addition, we studied the different impacts of intra-product specialisation on the TFP of the high-tech sub-industries, which makes our empirical results more detailed.

Conclusions

The impact of intra-product specialisation on TFP is different in the current period and in the lag period. So the basic assumptions of the article have not been fully confirmed. In addition, the industrial environment exerts a significant effect on TFP. Intra-product specialisation is found to have a negative effect on the TFP of the high-tech industry in China in the current period and a positive effect on TFP and its decomposition variables in the lag period. In other words, the impact of intra-product specialisation on the TFP of high-tech industry is uncertain. This empirical result is an important contribution of this work, which is different from the empirical results of the previous studies, such as Dai’s research. We also used Hansen test to examine the problem of excessive recognition, and the empirical results are significantly different from that of Dai’s study. The degree of intra-product specialisation in the high-tech industry has shown an overall upward trend since 1995. The TFP of the high-tech industry showed a downward trend from 2007 to 2010. Improving the ratio of state ownership has a positive effect on TFP and pure technical efficiency. This outcome differs from previous findings. The R&D capital investment has a positive effect on TFP, whereas R&D human capital investment slightly contributes to technical efficiency. The degree of intra-product specialisation has the most significant influence on technical efficiency and a relatively minimal effect on technological change. Intra-product specialisation has significant effects on the TFP of all high-tech sub-industries. The main contribution of this study is the empirical investigation into the effects of intra-product specialisation during the current and lag periods by constructing a dynamic panel model of the GMM. Given the difficulties in data acquisition, this work also has limitations that should be addressed. Certain important industrial environmental factors, such as industrial technology standards and other factors, are not considered as variables in the empirical model. Other complex factors should be used in future empirical research.

According to the empirical conclusions, we put forward four policy recommendations. First, China should take advantage of intra-product specialisation to promote the development of its high-tech and traditional manufacturing industries. The large high-tech enterprises of China should strive to become integrators of global intra-product specialisation networks. The high-tech enterprises with technological advantages should attempt to become pioneers of international industrial standards. Second, to reduce the integration costs of intra-product specialisation, China should accelerate the evolution of industrial standards and reduce trial-and-error costs. Third, China should deepen the reform of state-owned enterprises and use the spillover effects of technological knowledge to promote the TFP growth of the high-tech industry. At the same time, China should encourage the participation of private enterprises in global intra-product specialisation networks and promote their cooperation with state-owned enterprises to occupy a favourable position in global value chains. Finally, China should use advantageous policies that can encourage enterprises to increase R&D capital investment. Human capital structure should also be optimised to improve its contribution to technical efficiency.

Footnotes

Acknowledgements

This work is supported by National Social Science Foundation of China (No. 13CJY057) and Social Science Foundation of Hunan Province (No. 14ZDB013).

References

Balakrishnan

(2004). Measuring productivity in manufacturing sector. Economic and Political Weekly, 39(14), 1465–1471.

Balakrishnan

, Pushpangadan

, & Babu

M.S.

(2000). Trade liberalisation and productivity growth in manufacturing: Evidence from firm-level panel data. Economic and Political Weekly, 35(41), 3679–3682.

Berndt

E.R.

, & Christensen

L.R.

(1973). The translog function and the substitution of equipment, structures, and labor in U.S. manufacturing 1929–68. Journal of Econometrics, 1(1), 81–113.

Brusoni

, & Prencipe

(2011). Patterns of modularization: The dynamics of product architecture in complex systems. European Management Review, 8(2), 67–80.

Caves

D.W.

, Christensen

L.R.

, & Diewert

W.E.

(1982). The economic theory of index numbers and the measurement of input, output, and productivity. Econometrica, 50(6), 1393–1414.

Cheng

L.C.

(2011). Assessing performance of utilizing organizational modularity to manage supply chains: Evidence in the US manufacturing sector. International Journal of Production Economics, 131(2), 736–746.

Dai

(2012). Productivity effects of high-tech industry vertical specialization in China. Statistical Research, 29(1), 55–60.

Dean

J.M.

, Fung

K.C.

, & Wang

(2011). Measuring vertical specialization: The case of China. Review of International Economics, 19(4), 609–625.

Deb

A.K.

, & Ray

S.C.

(2014). Total factor productivity growth in Indian manufacturing: A biennial Malmquist analysis of inter-state data. Indian Economic Review, 49(1), 1–25.

10.

Egger

, & Egger

(2006). International outsourcing and the productivity of low-skilled labor in the EU. Economic Inquiry, 44(1), 98–108.

11.

Falk

(2012). International outsourcing and productivity growth. Review of Economics and Institutions, 3(1), 1–19.

12.

Färe

, Grosskopf

, Norris

, & Zhang

(1994). Productivity growth, technical progress, and efficiency change in industrialized countries. American Economic Review, 84(1), 66–83.

13.

Feenstra

R.C.

, & Hanson

G.H.

(1996). Globalization, outsourcing, and wage inequality. American Economic Review, 86(2), 240–245.

14.

Gereffi

, Humphrey

, & Sturgeon

(2005). The governance of global value chains. Review of International Political Economy, 12(1), 78–104.

15.

Goldar

(2004). Indian manufacturing: Productivity trends in pre-and post-reform periods. Economic and Political Weekly, 39(46), 5033–5043.

16.

Görg

, Hanley

, & Strobl

(2008). Productivity effects of international outsourcing: Evidence from plant level data. Canadian Journal of Economics, 41(2), 670–688.

17.

Grossman

G.M.

, & Helpman

(2005). Outsourcing in a global economy. Review of Economic Studies, 72(1), 135–159.

18.

Hanson

G.H.

, Mataloni

R.J.

, & Slaughter

M.J.

(2005). Vertical production networks in multinational firms. Review of Economics and Statistics, 87(4), 664–678.

19.

(2006). International vertical specialization and trade: A survey. Nankai Economic Studies, 22(5), 12–26.

20.

Hummels

, Ishii

, & Yi

(2001). The nature and growth of vertical specialization in world trade. Journal of International Economics, 54(1), 75–96.

21.

Hummels

, Rapoport

, & Yi

(1998). Vertical specialization and the changing nature of world trade. FRBNY Economic Policy Review, 4(2), 79–99.

22.

Jefferson

G.H.

, Bai

, Guan

, & Yu

(2006). R&D performance in Chinese industry. Economics of Innovation and New Technology, 15(4–5), 345–366.

23.

Kathuria

, Rajesh

R.S.N.

, & Sen

(2013). Productivity measurement in Indian manufacturing: A comparison of alternative methods. Journal of Quantitative Economics, 11(1–2), 148–179.

24.

Kohler

(2003). The distributional effects of international fragmentation. Germen Economic Review, 4(1), 89–120.

25.

Koopman

, Wang

, & Wei

(2014). Tracing value-added and double counting in gross exports. American Economic Review, 104(2), 459–494.

26.

Lau

A.K.W.

, Yam

R.C.M.

, & Tang

(2011). The impact of product modularity on new product performance: Mediation by product innovativeness. Journal of Product Innovation Management, 28(2), 270–284.

27.

(2007). Own R&D technology purchased and productivity development. Journal of Quantitative and Technical Economics Research, 24(7), 15–24.

28.

Liu

, & Wu

(2006). Integration of trade and disintegration of production: An empirical study based on two important hypotheses of economic globalization. Social Science in China, 27(2), 80–92.

29.

(2007). Economics of service outsourcing: From the perspective of intraindustrial trade. Journal of World Economy, 30(8), 22–35.

30.

Mitra

, Sharma

, & Veganzones-Varoudakis

(2012). Estimating impact of infrastructure on productivity and efficiency of Indian manufacturing. Applied Economics Letters, 19(8), 779–783.

31.

Mitra

, Varoudakis

, & Veganzones-Varoudakis

(2002). Productivity and technical efficiency in Indian states’ manufacturing: The role of infrastructure. Economic Development and Cultural Change, 50(2), 395–426.

32.

Segarra-Ona

M.D.

, & Peiro-Signes

(2014). Environmental eco-orientation determinants at the Spanish tile industry. Dyna, 89(2), 220–228.

33.

Solow

(1957). Technical change and the aggregate production function. Review of Economics and Statistics, 39(3), 312–320.

34.

Song

E.Y.

(2011). On gravity, specialization and intra-industry trade. Review of International Economics, 19(3), 494–508.

35.

Vogiatzogloua

(2012). Vertical specialization and export performance in electronics: A cross-country analysis. International Economic Journal, 26(1), 109–139.

36.

Wang

J.H.

, & Hsu

J.L.

(2014). Effects of service innovation on oral spreading and consumption intention in catering industry. Revista de Cercetare si Interventie Sociala, 47(1), 220–232.

37.

Yang

, & Shi

(1995). A new theory of industrialization. Journal of Comparative Economics, 20(2), 171–189.

38.

Zhang

H.P.

(2015). An agent-based simulation model for supply chain collaborative technological innovation diffusion. International Journal of Simulation Modelling, 14(2), 313–324.

39.

Zhang

, Sun

, Delgado

M.S.

, & Kumbhakar

S.C.

(2012). Productivity in China’s high technology industry: Regional heterogeneity and R&D. Technological Forecasting and Social Change, 79(1), 127–141.

40.

Zhao

, & Boasson

(2015). Productivity gains from offshore outsourcing: Evidence from the Chinese manufacturing industry. China and World Economy, 23(5), 104–122.