Regional Innovation Systems Based on Stochastic Frontier Analysis: A Study on Thirty-One Provinces in China

Abstract

This article summarises the literature on regional innovation system (RIS) The study uses thirty-one Chinese provinces and cities as the unit of analysis. Subsequently, an empirical study is conducted using stochastic frontier analysis on unbalanced panel data covering 31 provinces. The empirical result shows that innovation performance of RISs differs greatly in the various provinces and cities analyzed. The average score of the national regional innovation performance is 0.4514, indicating that non-efficiency is very common in Chinese provinces and cities. There are various factors that should be considered to promote innovation performance of the RIS. These are openness of RISs, maturity of the technology market, collaboration of RISs and intellectual property protection. Interestingly the unique circumstances of the Chinese market show that enterprise technology input and government investments have a negative impact on the regional innovation performance.

Introduction

Economic Competition in the current economic era is mainly manifested as the competition of science and technology. Social economic structures, economic relations and customs have changed dramatically during the era of the knowledge economy. A regional innovation system (RIS) is an effective tool for economic development (Li and Hu, 2014; Peng and Zheng, 2012; Shang, 2011). Unlike other tools, a RIS is a more flexible analytic tool (Li, 2009).

The ‘innovation system’ concept can be understood in both a narrow as well as a broad sense. A narrow definition of the innovation system primarily incorporates the R&D functions of universities, public and private research institutes and corporations, reflecting a top-down model of innovation as exemplified by the triple helix approach (Etzkowitz and Leydesdorff, 2000). The broad definition encompasses all interrelated institutional actors that create, diffuse and exploit innovations (Chung, 2002; Lundvall, 2010). In this article, the broad definition is followed. The concept of RIS originated from discussions about national innovative system (NIS) (Edquist, 1997; Freeman, 2000; Lundvall, 2010; Nelson and Green, 1996). We define a RIS as a complex of innovation actors and institutions in a region (administrative regions) that are related with the generation, diffusion and appropriation of technological innovation and an interrelationship between these innovation actors (Asheim and Isaksen, 1997; Cooke, Gomez Uranga and Etxebarria, 1997). The RIS can be thought of as the institutional infrastructure supporting innovation within the production structure of a region (Asheim and Vang, 2006). The precise distinction between an NIS and an RIS is difficult to ascertain. In fact, distinctions are not always made; some authors categorise these ideas as different concepts (Asheim and Isaksen, 1997; Autio, 1998) whereas others see regional systems as a subset of a national system (Archibugi et al., 1999; Aslesen, 1996). A regional approach is better to formulate and implement a competent NIS than a sectoral approach. It emphasises that a concept of RIS is a good tool to generate an effective national innovation system, as it can effectively create different sectoral innovation systems in different regions (Chung, 2002). A RIS is composed of five main innovation actors: firms, universities, public research institutions, financial sector and governments. Firms, universities, public research institutions are actual research producers who carry out R&D activities. In addition, governments play the role of coordinator among research producers in terms of their policy instruments, visions and perspectives for the future. Financial sector can provide funding for firms, universities, public research institutions. All 31 provinces contain these five main innovation actors. Openness of RIS, maturity of the technology market, collaboration of RISs, intellectual property protection, the extent of government funding and marketisation are the factors which influence the efficiency of RIS. They were used in many prior papers (Goto and Suzuki, 1989; Liu and Li, 2007).

Innovation performance varies not only between nations, but also between sub-national regions; such as states or provinces (Acs et al. 2002; Evangelista et al., 2001; Fritsch, 2002). For large countries, the national innovation system approach is probably less relevant (Edquist, 2010). Many researchers noticed that the innovation systems in developing economies and transitional economies had quite different systematic characteristics from those in developed countries (Gu and Lundvall, 2006; Hu and Mathews, 2005; Liu and White, 2001; Radosevic, 2002). In the case of China, this was particularly true (Li, 2009). In 1999, the RIS appeared in Chinese literature for the first time (Huang Lucheng, 2000; Mao Yanhua, 1999; Hu Zhijian, 1999). In 2008, Zhu Xiaoxia argued that provincial boundary was the best RIS, in the Chinese context. China actively built provincial RISs, and there were lots of researches about provincial RISs (Liu, Guan, 2002; Guan Jiancheng, 2005; Han Ying, 2006; Ren Shenggang, 2007; Sun Kai, 2006; Bai Junhong, 2009; Zhang Manyin et al., 2011; Chen Wei, 2012; Li Mingchao, 2013; Zhou Hongwen, 2014).

Each region constitutes an innovation system, which in turn constitutes an important part of the national innovation system in China (Chung, 2002). In addition, the use of administrative regions as the unit of analysis has at least three advantages (Li, 2009). First, from a practical perspective, comparable datasets covering innovation indicators are easily compiled for these systems; thus making an empirical comparison between regions possible and reliable. Second, using administrative regions as the unit of analysis provides an applicable and comparable framework. Furthermore, institutional factors determined the national level can be ignored when comparing regional variation.

Numerous researchers claim that the evaluation of a RISs performance can provide a scientific basis for the judgment, establishment and correction of regional policies (Chi et al., 2004; Cooke, 2008; Guan and Liu, 2003; Peng and Zheng, 2012; Wu et al. 2010; Xie et al., 2009; Zabala-Iturriagagoitia, Voigt et al., 2007). Guan Jiancheng (2003) used DEA to analyse the impact of institutions on innovation efficiency in RISs and found that there was no clear connection between regional innovation capability and the innovation efficiency. Moreover, Chi et al. (2004) investigated regional technological innovation efficiency in thirty provinces, municipal cities and autonomous regions in China using the DEA method (Note: In 1978, Charnes developed a new method named data envelopment analysis (DEA), which can evaluate the relative efficiencies achievable among a group of comparable operations or processes (Cooper et al., 2007; Zhu, 2003). DEA considers a particular set of DMUs or observations with each DMU possessing a set of input measures and output measures representing its multiple performance metrics, and it can be used to determine the efficiency of a group of decision-making units (DMUs) relative to an envelope (efficient frontier) by optimally weighting inputs and outputs.). The results revealed that regional technological innovation efficiency scores were higher in the Eastern regions compared the Western regions in China. Zabala analyzed 2002–2003 European innovational integration rankings using the DEA method, showing that innovation efficiency was relative to the level of regional cooperation Zabala (2007). Specifically, higher the level of technology areas, higher the need for collaboration systems. Lin Yun (2008) analysed panel data of thirty provinces in China and their cities’ input and output in technology innovation. The results revealed that Research and Development (R&D), technological opportunities and absorptive capacity had cumulative and circular influence on the performance of regional technology innovation. The cumulative cycle had become a regional technology innovational ‘track’, resulting in increasingly prominent regional differences in innovation. Xie (2009) evaluated the regional efficiency of thirty provinces in China’s provinces and their cities to obtain indirectly an environmental assessment of regional innovation. Jie Wu et al. (2010) divided the innovation process into technological sub-processes and economic sub-processes to measure the innovation efficiency of the 2001–2005 China’s RIS. He used an improved approach by using the DEA method to establish a multi-factor innovation performance evaluation system that provides a future summary of the successes and failures of planning and regulatory management.

There has been some progress in the literature on RISs, but the method mainly focused is on the DEA method. DEA can compare the innovation performance of the RISs, but it cannot do the detailed analysis of the impact factors (Li et al., 2009). Conversely, Stochastic Frontier Analysis (SFA) divides the real output into three parts. These are 1) production function, 2) random factors and 3) technical inefficiency. SFA allows to statistically test the results while assessing the confidence interval, which is more suitable for calculations of a large sample. Furthermore, SFA has been applied extensively in various areas, such as mathematics, science and engineering construction, automotive, forestry, sustainable development and macroeconomic management. SFA technology can not only measure the efficiency of each individual, but can also quantitatively analyze the specific impact of various factors on the efficiency differences between individuals (Battese and Coelli, 1995).

The remainder of the article is organised as follows. In section II the model is discussed. Section III contains data issues and variable measures. In section IV the empirical study is presented. Finally, section VII draws the main conclusions.

The Construction of the Model

This section introduces the modelling process of SFA, and then builds a SFA Model based on panel data, on the factors in Table 1, of RISs and innovation performance.

Stochastic Frontier Analysis was proposed by Meeusen and Broeck in the late 1970s, and the specific model is calculated as follows (Battese and Coelli, 1995; He and Chen, 2008):

H_{it} G_{it} (V_{it} U_{it}), i 1,, N; t 1,, T

(1)

m_{it} z_{it}

(2)

In Equation (1), H_it represents the output of period t, while region i represents the RIS in the form of logarithm. G_it represents the input of period t, region i represents the RIS. β is the parameter to be estimated. The random variable $V_{it} iid$ , the same to U_it . V_it is subject to $N (0, V^{2})$ , and is independent of U_it . U_it is non-negative random variable and subject to $N (m_{it}, U^{2})$ . U_it represents those random factors of RISs in period t affecting only region i.

${TE}_{it} \exp (U_{it})$ represents technology non-efficiency, measuring the efficiency relative to the efficient frontier (Note: SFA calculates the efficient frontier and efficiency scores. DMUs, which performance project directly onto this efficient frontier, are determined to be relatively efficient with scores of 1. That means no other DMU’s performance is strictly better in either its inputs or its outputs.). Technical validity refers to if the technical input-output vector is valid; specifically, considering prior technology and the situation of declining other outputs (or increase other inputs) and if it is technically impossible to reduce any investment (or any increase in output). The TE value is used as a measure of the effectiveness of technical indicators, reflecting the degree of effectiveness to resources among producers. Regional innovation activities are regarded as production activities, and the performance of the RIS can be seen as the development of productivity. TE is selected as a measure of performance indicators to measure regional innovation. ${TE}_{it} \exp (U_{it})$ represents the level of technical efficiency of RISs in period t, region i in the sample.

Equation (2), z_it is the vector affecting the performance of the system. δ is the parameter to be estimated $\frac{_{u}^{2}}{(_{v}^{2}_{u}^{2})}$ which determines whether SFA models should be used $(0, 1)$ . If $0$ , it indicates that V_it plays the dominant role, while U_it does not have an effect. Subsequently, the method of least squares analysis can be used. If $1$ , it indicates that U_it works. While V_it does not have an effect, in which case it is not necessary to use SFA. γ’ test statistics are progressively subject to mixed analysis.

This article uses the trans log production function, based on the model proposed by Battese and Coelli. The empirical framework used in this article is:

\begin{matrix} \ln (Y_{it})_{0}_{1} \ln K_{it}_{2} \ln L_{it}_{3} \ln {(K_{it})}^{2} \\ _{4} \ln {(L_{it})}^{2}_{5} \ln (K_{it}) \ln (L_{it}) (V_{it} U_{it}) \end{matrix}

(3)

\begin{matrix} m_{it}_{0}_{1} {open}_{it}_{2} {maturity}_{it}_{3} {collaboration}_{it}_{4} {input}_{it} \\ _{5} {awareness}_{it}_{6} {government}_{it}_{7} {market}_{it} \end{matrix}

(4)

Equation (3), Y represents the output of new products in the RIS. The Western literature indicates that most Western scholars use ‘the number of patent applications’ or ‘the number of patents granted’ as a performance indicator to measure the output (Doloreux and Parto, 2005; Radosevic, 2002). This article chooses the value of new products instead of the number of patents. This is because a key restrictor on the development of nation’s innovative ability is the low transformation rate of scientific and technological achievements, that is, the eminent divergence between R & D and links in the production chain. In 2008, China’s transformation rate of scientific and technological achievements was on average only 20 per cent. The proportion to achieve the industrialisation was less than 5 per cent, the rate of agricultural technology transformation was slightly higher at 30 per cent (65 per cent −85 per cent in developed countries) (Jiang and Cui, 2009; Jiang and Liu, 2013; Lan, 2008). The transformation rate of scientific and technological achievements in universities was under 5 per cent and the pharmaceutical and technological achievements conversion rate was less than 8 per cent, which is far lower than in developed countries. Therefore, to study regional innovation performance in the Chinese context, it is not advisable to use ‘the number of patent applications’ or ‘the number of patents granted’. It is suggested ‘the number of patent applications’ can only be measured as an awareness of a body for the regional innovation as intellectual property protection. Therefore, to study regional innovation performance in the Chinese context, it is not advisable to use ‘the number of patent applications’ or ‘the number of patents granted’. Moreover, one of the important functions of the RIS is to solve scientific and technological achievements conversions. Hence, it is suggested ‘the number of patent applications’ can only be measured as an awareness of a body for the regional innovation as intellectual property protection. For the aforementioned reasons it is chosen to use the output value of new products as a measure rather than the number of patents. Although the output value of new products grows rapidly, to a large extent, this result is not generated from the improvement of China’s regional science and technology innovation competitiveness, but from multinational companies with operation in China (Note: The multinational companies are in our sample.).

K is the capital stock of scientific and technological activities (Note: The science and technology activities are organised activities which arise from the natural sciences, agricultural sciences, medical sciences, engineering sciences, humanities and social sciences and technology (referred to scientific and technical area), relative to the emergence, development, dissemination and application of science and technology-tech knowledge.) of RISs. R & D inputs are not used as a measurement. R & D refers to the scientific fields, in order to increase the amount of knowledge and the knowledge used to create systematic and creative activities of new applications. This includes basic research, applied research and experimental development activities. However, the value of new products is used as an output measurement index, while input includes not only the (R & D), but also the application of research results and related IT services costs. According to UNESCO, the science and technology activities are organised activities which arise from the natural sciences, agricultural sciences, medical sciences, engineering sciences, humanities and social sciences and technology (referred to scientific and technical area), relative to the emergence, development, dissemination and application of science and technology-tech knowledge. It can be divided into three types; specifically 1) research and experimental development (R&D), 2) applied research and 3) experimental development results and related technology services. On the other hand, investment in science and technology activities is an investment flow. Its impact on output is largely dependent on accumulated prophase investment, in addition to the current expenditure of funds. This also includes the raising stock. Thus, R & D capital input is substituted by Capital Stock to measure capital investment, following Xi Jiangguo (Xi, 2012).

L represents the S&T personnel full-time equivalent of a RIS. Jaffe (Jaffe, 1986) expanded knowledge on the production function, introduced the notion about investment in human resources, and pointed out that the knowledge production function has a broad applicability that can be applied for regions and macroeconomic categories. This article uses all R & D staff of RISs as the equivalent of the measurement index of human resources. The most appropriate choice should be S&T personnel full-time equivalent, but there are no relevant statistics available in the statistical yearbook. Hence, this article selects RISs R & D personnel as an alternative measure (Peng et al., 2014).

Equation (4) is used to measure the factors (Note: The factors are open, maturity, collaboration, awareness, input, government and market.) that influence the performance of RISs from the two main perspectives; namely, the innovation system and the main body of innovation. open indicates the degree of openness of RISs. For a sustainable developmental system, openness is an essential characteristic. Some scholars use the proportion of total import and export trade in GDP to measure the openness of the region (Liu and Li, 2007), and it should be used to measure the openness in a broader sense. When measuring the innovation performance of RISs, it is more appropriate to choose openness technically. Therefore, the proportion of regional technical introduction in GDP is used as measure. Some scholars use the proportion of total import and export trade in GDP to measure the openness of the region (Liu and Li, 2007). This article argues that it is more appropriate to choose the opening degree measured through the amount and proportion of regional technical introduction of GDP, when measuring the RISs performance.

Maturity represents the maturity of the technology market, and is measured by the proportion of the turnover in the technology market in GDP. The more sophisticated the technology market, the more conducive to technology flows within the system. The mature technology market acts as an ‘invisible hand’ to promote the product forming and industrialisation of scientific and technological achievements consequently affect regional innovation performance. Therefore, the maturity of the technology market is one of the key factors of regional innovation performance.

Collaboration represents the level of collaboration among the main bodies, which are firms, universities, public research institutions, financial sector, governments, in internal innovation system. As a system, its coordination is an important manifestation of system functions. It is measured by the proportion of the amount of finance liquidity in GDP (Li, 2009). From a functional point of view, the RIS is expected to achieve optimal innovation output by integrating the key resources of the region effectively. Therefore, the level of inter-regional cooperation for the main body has an important impact on innovation performance (Li, 2009).

awareness represents the awareness of protecting property rights in the RIS. It is measured by the proportion of the number of patent applications to the amount of people who are engaged with technology. Conversely, as previously mentioned, the number of patent applications should not be used as a measure of performance output indicators. It is more reasonable to use it as a factor to study its relationship with innovation performance. New products can bring high profits to pioneering entrepreneurs, but imitators of innovation will quickly follow to take share of the profits in the market. Technical barriers are effective means of preventing knowledge spillovers and patent law is an effective law to establish technical barriers. High awareness of IPR protection within the region allows enterprises to retain a competitive advantage after introducing an effective innovation. On the other hand, the sense of innovation can’t be measured by the absolute amount of patent applications compared with the number of employed individuals in the population. Employed represents a person who is beyond the age of 16years and engaged in social labour, for which s/he receive remuneration or business income. This indicator, an important indicator of China’s basic national conditions and strength, reflects the actual utilisation of labour resources in a specific time frame (Li, 2009). Evidently, it is more reasonable to choose the number of employed rather than the total population of an area, because the relationship between intellectual property rights and the unemployed or younger than 16 is irrelative.

input represents the science and technology investment intensity of enterprises in the area, measured by the proportion of the amount of technology investment in the main business revenue. The figures can be found in the book ‘China Statistical Yearbook on Science and Technology’. In the RIS, the enterprise is undoubtedly the key to achieve the final product and industrialisation. Hence, the science and technology investment intensity of enterprises in the area has an important impact on the level of innovation performance.

Government represents the local government investment intensity of the RIS. RIS is an effective tool for the government to develop the regional economy, and the government’s financial support for regional innovation has a significant impact (Goto and Suzuki, 1989). Therefore, the government role in the RIS cannot be ignored. Under normal circumstances, it demonstrates a positive role.

Market represents marketisation degree, measured by the ratio of the total output value of state-owned or state-controlled enterprises in industrial enterprises. Market can provide inexhaustible innovative power, so the extent of the market has an important effect on innovation performance. The development between China’s state-owned enterprises and private enterprises is unjust. In terms of the number of enterprises, private enterprises show a promising future, for example, the proportion of private enterprises in the foremost 500 was 3.8 per cent in 2002, 13.8 per cent in 2003, 14.8 per cent in 2004 (Wang, 2005). However, if the total assets are used to measure, it will show that the year-end total assets of state-owned enterprises accounted for more than 95 per cent of the total assets. That means China’s state-owned economy is in a dominant position (Liu and Li, 2007).

Data and Variable

Data Sources

The data on openness, continuity and authority stems, mainly from the ‘China Statistical Yearbook’ (2000–2009), ‘Fifty-five years of new China Statistical Data Collection 1949–2004’ (2005), ‘China statistical Yearbook on Science and Technology’ (2000–2009). The technology model and the computing software SFA can solve problems with unbalanced panel data and the selected data is based on China’s provinces and cities from 1999 to 2008, resulting in a total of 3100 observations.

Variables Selected

The Main Function

The main function is Equation (3). Y is the value of innovative product output in the region (unit: million), new products include two indicators: (a) new product sales and (b) new product output value. Due to the impact of the consumer market, the economic cycle, corporate marketing strategy, and additional factors, new product sales have greater uncertainty. New product output value can reflect the actual level of innovation performance better than new product sales. Choosing 1999 as the base year, we use the producer price index to deflate the value of innovative product output.

K is the capital stock for scientific and technological activities in the area. This is calculated by summing up science and technology expenditures in the lag period and the current collections (Goto and Suzuki, 1989; Y. Wu, 2008).

K_{it}_{k 1}^{n}_{k} E_{i (t k)} R_{t}

(5)

$_{k}$ is the lag discount coefficient of science and technology expenditures. E is the expenditure for science and technology, while R is the collection for science and technology. Put θ as average lag phase, when $k$ , $_{k} 1$ , otherwise $_{k} 0$ . Based on previous research, two years are selected as the average lag period, namely: $K_{it} E_{i (t 2)} R_{t}$ . Choosing 1999 as the base year, we use the R&D price index to deflate the value of science and technology expenditures. The R&D price index equals to 0.55* the consumer price index plus 0.45 * the fixed asset investment price index.

L represents the whole equivalent of R & D personnel (Unit: person-years). The accurate calculation unit of labour investment should be the effective time used to measure the amount of labour input in developed market economies (Liu and Li, 2007).

Efficiency Function Section

When analysing the factors influencing regional innovation performance, various indexes are selected. These are the proportion of the technology introduction value in GDP (per cent), the proportion of technology market turnover in GDP (per cent), the proportion of financial liquidity among innovative main body in GDP (per cent), the proportion of technology input in the main business income (per cent), the number of patent applications and employment (item/million), the proportion of government IT spending in total government expenditure (per cent), the proportion of non-nationalisation total industrial output value in industrial enterprises output value (%) and others. They reflect the different factors of innovation performance from the aspects of the extent of system openness (open), the maturity of technology market (maturity), the system level of collaboration (collaboration), the intensity of enterprise technology input (input), the awareness of protecting property rights (awareness), the extent of government funding (government) and marketisation degree (market), respectively.

Statistical Description of Variables

Table 1 offers a description of the statistics based on a sample of 31 Chinese provinces from 1999 to 2008.

According to Table 1, the index difference in China’s 31 provinces and cities is high, displaying an obvious ‘polarization phenomenon’. Since the first three are the absolute amount of data, factors such as the size of the area are taken into account and excluded. Among the factors outlined, the gap per 10,000 employed persons who own patent applications is up to 3997-fold. The gap for the proportion of non-nationalisation total industrial output value in industrial enterprises output value is nearly 16 times.

Table 1

Statistics of Thirty-One Chinese Provinces 1999 to 2008

	Observed Value	Mean Value	Standard Deviation	Minimum	Maximum
New product output value (million)	302	1.10*10⁷	1.78*10⁷	26.6335	1.35*10⁸
Technology Activities capital stock (million)	310	1982300	2539463	6592.12	1.47*10⁷
The whole equivalent of R & D personnel (person-years)	310	39977.03	39856.87	204	238683.79
The proportion of the technology introduction value in GDP (%)	237	2.1926	7.5034	0.0013	76.0112
The proportion of technology market turnover in GDP (%)	298	0.7935	1.4504	0.0055	10.9824
The proportion of financial liquidity among innovative main body in GDP (%)	310	1.8137	5.5571	0.2153	85.9138
The proportion of technology input in the main business income (%)	308	3.2577	6.1396	0.0284	43.1118
The number of patent applications and employment (item/million)	310	5.3972	8.9640	0.0203	81.1440
The proportion of government IT spending in total government expenditure (%)	310	4.3130	5.7361	0.3728	37.9818
The proportion of non-nationalization total industrial output value in industrial enterprises output value (%)	310	44.9876	20.6028	5.6025	88.6445

Empirical Research

According to the model constructed earlier, empirical research was conducted on China’s 31 provinces from 1999 to 2008 through Froniter 4.1 (Note: Froniter 4.1 can use the computer program to provide maximum likelihood estimates of a wide variety of stochastic frontier production and cost functions.). The output is shown in Table 2.

General Evaluation

From the Table 2, by calculating the 3100 observations, the value of γ in other information is 0.7611, at a significance level of 1 per cent. After analyzing, the value γ it shows that it is neither close to 0 nor close to 1 (0.7611). This suggests that the inefficiency phenomenon does exist in China’s various RISs. It is necessary to use SFA to analyze innovation performance. The estimation of the production function is all significant at the 1 per cent level. The majority of the inefficiency function is also significant. Log-likelihood function is −547.7202, indicating that the maximum likelihood estimation algorithm is better. The LR statistic is 22.5617, significant at the 1 per cent level, indicating that the model is validated. The data shows that average technical efficiency of the RISs in China is 0.4514, suggesting that the non-efficiency problem in the RIS is serious. As long as there are effective ways to promote regional innovation performance, innovation output can grow two-fold under the existing conditions of input and output. Based on this finding, factors affecting regional innovation performance are analyzed and it is expected to find effective ways to enhance regional innovation performance through scientific methods.

Table 2

The Estimation Results of the Stochastic Frontier Function and Efficiency Equations

Variables	Estimated Coefficient	Standard Deviation	Value of T
Part of the production function
Constant term	−46.5629***	1.5183	30.6670
K	3.3610***	0.9735	3.4524
L	6.6355***	1.4576	4.5525
K²	0.7556***	0.1819	4.1532
L²	1.3100***	0.2454	5.3386
K*L	−2.3529***	0.4202	−5.5996
Inefficiency function section
Constant term	−0.2108	0.7251	−0.2907
The extent of system openness (open)	−9.7701***	3.6236	−2.6963
The maturity of the technology market (maturity)	−2.3346*	1.3232	−1.7643
The system level of collaboration (collaboration)	−11.7920***	4.4239	−2.6655
The intensity of enterprise technology input (input)	0.6911	1.0110	0.6835
The awareness of intellectual property protection (awareness)	−17.6681***	6.3998	−2.7607
Recognition level of the government	−1.1711*	1.0733	−1.9115
Marketisation degree	−0.8273	1.0486	−0.7889
Other Information
γ	0.7611***	0.0531	14.3438
Logarithm likelihood function value	−547.7202
LR statistic	22.5617***
The average technical efficiency	0.4514
Sample volume	310

Source: Authors’ own, computed through SFA.

Notes: *,**,***significant at 10, 5 and 1 per cent level respectively; LR is the likelihood ratio test, consistent with a mixture of chi-square distribution.

Factors Analysis

First, the empirical results show that there is a positive correlation between the extent of system openness and innovation performance. Thirty years of reform and opening mad China evolve enormously, and the extent of regional openness has led to remarkable achievements. Hence, a deepening reform, opening up and strengthening inter-regional cooperations are effective ways to promote system performance. However, some parts of China are still affected by Chinese traditional concepts and operate in a relatively closed stage. Both, the monopolistic thoughts derived from Chinese feudal dynasties and ‘never contact with each other’ philosophy are boycotting the Chinese traditional concept to openness. Therefore, the situation must be evaluated from inside to outside perspectives. From a regional perspective, the key to promote the extent of system openness is adhering to continuous learning, and introducing foreign advanced technology actively.

Second, the maturity of technology market (maturity) is an important medium to achieve commercialisation and industrialisation of scientific research, and the prosperity of technology market marks the smoothness of technology flows. The empirical results suggest that the technology market has indeed a positive influence on innovation performance. Therefore, the government can promote regional innovation performance by further standardising the technology market order, encouraging transactions in technology market positively and creating a better environment for technology innovation.

Third, the system level of collaboration (collaboration) is key as any system must have the ability to collaborate, which is the key to achieve 1 +1 > 2 through systems integration. The RIS uses the flow of funds among the main bodies to characterise the overall collaboration level of the system, as empirical studies have shown positive relationship shown between them. The operational nature of the RIS should aim to achieve effective integration among the main bodies and show a stronger competitive edge of the whole region by effective mechanism. Therefore, further strengthening cooperation among the main bodies is an effective and scientific way to enhance regional innovation performance and improve regional innovation ability.

Fourth, the intensity of enterprise technology input is an important factor. Throughout the world today, countries with an overall leading position as a nation or where the level of economic and social development is the highest, tend to invest heavily into science, technology and R&D activities. The formation of the US national innovation system started with enterprises, followed by universities, independent research departments and government agencies. Enterprises have become the main driver of American innovation activities, and a large number of important inventions are developed by corporate R & D institutions. Thus, companies should be the principal force in the area of innovation. Since the year 2000, China’s high-tech private enterprises such as Huawei and ZTE, ‘the successful development of Godson’ (Universal CPU), ‘Ark I, II’ (embedded CPU) are growing, suggesting that Chinese companies are adopting a leading role in R & D and innovation. However, Chinese companies do not have the ability to become the subject of scientific and technological innovation, as empirical studies have shown that there is a negative correlation between IT investment and regional scientific and technological innovation performance. This is a realistic portrayal of national conditions, as only through reforms can enterprises be the subject of technological innovation, and subsequently the increased investment in technology can enhance regional innovation performance. That’s why even though Chinese companies continue to increase their investment in science and technology the innovation capacity has not improved significantly (Jiang, Shen and Wang, 2009; Liu and Guan, 2002).

Fifth, the awareness of intellectual property protection (awareness) has been shown to have a positive relationship bon regional innovation performance. As mentioned earlier, intellectual property protection can protect innovation effectively. Since China has joined WTO, continuous crackdowns on the infringement of intellectual property are prevalent. In the process of injuring, China has constantly formulated relevant laws and regulations to enhance the awareness of IPR protection among all stakeholders. The awareness of intellectual property protection is an important reflection of a culture of innovation. The intensity of the awareness of intellectual property protection in an area reflects the recognition level of innovation within the inner-region from indirect sources, promoting innovation performance.

Sixth, the extent of government funding is another important factor. Chinese RISs are different from foreign ones. RISs in China are government-led, hence, it would be expected that there should be a positive relationship between government investment and regional innovation performance. However, the empirical studies show that the positive relationship between them is not significant. The main reason for this phenomenon is that our government often provide fund to science and technology in a way of ‘preferential policies’ to achieve support for RISs. The current study finds that, in Shanghai, for example, the amount can reach more than 40 million. Although it can meet the needs of early development of enterprises, it is very difficult to carry out technological innovation. This study argues that more governmental investment should be applied to create a good business and financing environment, rather than offering direct financial rewards. The government must change the direction of its investment, otherwise the government investment does not exert influence on innovation performance.

Seventh, studies show that the higher the degree of marketisation, the stronger the enhancement of regional innovation performance. For example, China’s fixed asset investment growth was around 30 per cent, but the government-based infrastructure investment (railways, highways, airports) grew up to 60 per cent, while in many fields of private enterprises (manufacturing industry), the investment growth rate is only equivalent to the national average. This means that a large proportion of the resources used for stimulating domestic demand enter directly into the state sector. Private enterprises play an important role in economic development. Zhang Weiying pointed that ‘the proportion of state-owned economy in areas with a large income gap, and vice versa,’ is so excessive in the development of the state economy that it will bring inequality and then affect the system of innovation performance. The value of final goods and services created by small and medium enterprises relying mainly on private enterprises is equivalent to about 60 per cent of the GDP. Tax amounted to about 50 per cent of total national tax revenue, providing nearly 80 per cent of urban jobs, which shows the important role of the private economy in regional development. Recently, the phenomenon of ‘the nation gets forward while people get backward’ shows that here are still some areas to improve for the socialist market economy in China. As pointed out by Professor Chen Zhiwu “there is no end for privatisation and marketization” and China needs to further reforms.

The Innovation Performance Analysis According to Geographical and Time Dimensions

Selecting the geographical dimension hierarchy, based on the output of the model, Table 3 shows the average performance of 31 provinces during 10 years.

Table 3 shows that SFA evaluates more objectively than DEA. The evaluation process shows that at first sight DEA appears to be the more valid unit (Liu and Guan, 2002), which cannot be compared with each other. Although some scholars have improved DEA, when dealing with a large sample, it still entails limitations. For example the value of efficiency is more than 1. Unlike DEA, SFA allows drawing an accurate evaluation of the efficiency in empirical research and offers more appropriate alternative for a contrastive analysis.

Figure 1 shows the volatility of China’s average innovation efficiency from 1999 to 2008. The data indicates that the innovation performance in all Chinese regions of China shows certain volatility. Overall, the entire waveform is comparatively regular and shows no large deviations.

Table 3 and Figure 1 show that the average performance of China’s RIS is about 0.42, while 1999 and 2008 demonstrate higher scores. This shows that China still performs at a low level and has large room for improvement. Meanwhile, Shanghai, which ranks first, reached 0.6918, which is 8.56 times as high as Tibet. This score illustrate the real conditions of Shanghai’s innovation performance rather than DEA=1. The differences in infrastructure, personnel training and other factors are directly related to geographical location, government policies and so on (Avelar-Sosa et al., 2014). Thus, from a geographical point of view, innovation performance is shaped just like a ‘pyramid’, as is shown in Figure 2. The Yangtze River Delta, the Southeast Coastal Areas and the Northwest Territories are in a leading position, while the average efficiency in other areas is lower than the national average. There are 14 areas above the average, where the efficiency score is above 0.69 (Shanghai, which are classified as ‘outstanding’). Areas where the efficiency is above 0.6 include Guangdong, Tianjin, Beijing, Jiangsu, Hainan are classified as ‘good’. Zhejiang, Shandong, Ningxia, Fujian, Chongqing, Liaoning, Jilin and Guangxi are classified as ‘mediocre’. More than half of the areas in China are still below the average. These include Hubei, Sichuan, Hunan, Hebei, Anhui, Inner Mongolia, Jiangxi, Qinghai, Guizhou, Shanxi, Yunnan, Xinjiang, Heilongjiang, Shaanxi, Gansu and Tibet. Of these areas the highest score is Inner Mongolia (0.4431), while the lowest score is obtained by Tibet (0.0808). These areas are classified as ‘low’.

Table 3

The Performance of Thirty-One Regional Innovation Systems

City	Average Performance	Ranking	City	Average Performance	Ranking	City	Average Performance	Ranking
Beijing	0.6331	4	Anhui	0.3968	19	Sichuan	0.4050	18
Tianjin	0.6371	3	Fujian	0.5256	10	Guizhou	0.3834	24
Hebei	0.3841	23	Jiangxi	0.3954	20	Yunnan	0.3186	27
Shanxi	0.3477	25	shandong	0.5761	8	Tibet	0.0808	31
InnerMongolia	0.4431	15	Henan	0.3891	22	Shanxi	0.2839	29
Liaoning	0.4905	12	Hubei	0.4092	17	Gansu	0.2781	30
Jilin	0.4820	13	Hunan	0.3915	21	Qinghai	0.4129	16
Heilongjiag	0.3356	26	Guangdong	0.6377	2	Ningxia	0.5426	9
Shanghai	0.6918	1	Guangxi	0.4802	14	Xinjiang	0.3100	28
Jiangsu	0.6203	5	Hainan	0.6105	6
Zhejiang	0.5870	7	Chongqing	0.5137	11

Source: Authors’ own computed via SFA.

Figure 1

Source: Based on Table 3.

Figure 2

Source: Based on Table 3.

Conclusions

The article summarised the literature on RISs and is based on the actual situation in China. China’s 31 provinces and cities were selected as units of analysis. Subsequently an empirical study was conducted using stochastic frontier analysis based on the 31 provinces and the cities’ unbalanced panel data. The findings show that innovation performances of RISs in different provinces and cities differ greatly and that there are various factors affecting regional innovation performance. The main conclusions drawn are as follows:

The national average regional innovation performance is only 0.4514, indicating that non-efficiency is very common in Chinese provinces and cities. Meanwhile, this shows large room for improvement on regional innovation performance. If reinvested amount is unchanged, Chinese innovation performance can improve about 2 times.

Various factors affect the performance of the RIS. These are the openness of RIS (−9.7701), maturity of the technology market (−2.3346), collaboration of RISs (−11.7920), intellectual property protection (−17.6681), the extent of government funding (−1.1711) and marketisation (−0.8273). The performance of Chinese RISs can be improved effectively by further allocating resources and improving the previously outlined factors. Note that the necessary changes with regard to the direction of government investment need to be considered. Instead of offering direct financial rewards, emphasis should be placed on creating a good business investment and finance environment.

Due to the unique nature of China’s development experience, ‘the intensity of enterprise technology input’ plays a reverse role in innovation performance. Therefore, it is recommended not to increase investment blindly, but rather to analyse the reasons for this phenomenon combined with the current situation in China, as the coefficient of the intensity of enterprise technology input is 0.6911 > 0. Otherwise the effect will only be counterproductive. Only if enterprises truly become the main innovation driver, can regional innovation performance be promoted by increasing technology investment.

On the one hand, the volatility of China’s overall innovation efficiency of RIS from 1999 to 2008 presents an appearance of stable volatility, as Figure 1 shows. On the other hand, the distribution of regional innovation performance is uneven (Tibet, 0.0808; Yunnan, 0.3186; Shanghai, 0.6918), and the differences between regions found in the data are in some cases above eight times (the highest is 0.6918, and the lowest is 0.0808). Therefore, China needs to enhance the learning curve in various regions, while popularising effective practices and promoting backward innovation performance.

Footnotes

Acknowledgements

We are grateful to Krishna V.V. (Editor in Chief) and the reviewers for their valuable comments and suggestions for improving the article. This work was supported by Doctoral Fund of Ministry of Education of China (20122304120021), Heilongjiang Province Postdoctoral Start Foundation (LBH-Q13050), the Special Foundation of Central Universities Basic Research Fee (HEUCF140907).

Appendix

Table A1

Science, Technology and Innovation Indicators for Thirty-One Provinces

	New Product Output Value (million yuan)	R&D Investment (million yuan)	Human Resources (person-years)	Publications (piece)	Patents (piece)
Beijing	30308746	9112536	189551	48076	17747
Tianjing	28187577	3044039	48348	7101	6790
Hebei	10572869	1832047	46155	4484	5496
Shanxi	6775804	1842181	43986	2365	2279
Inner Mongolia	3279831	741176	18264	439	1328
Liaoning	19824273	3347327	76673	11103	10665
Jilin	16388333	1084675	31731	6251	2984
Heilongjiag	4751014	1331604	50717	9137	4574
Shanghai	47255773	5802852	95129	24011	24468
Jiangsu	72901516	11249469	195333	20252	44438
Zhejiang	67532379	6142914	159589	12633	52953
Anhui	11295422	2420714	49465	6375	4346
Fujian	17775430	2095120	59270	3925	7937
Jiangxi	6238765	1007469	28241	1690	2295
shandong	56312044	7398101	160420	10842	26688
Henan	14497944	2457757	71494	4372	9133
Hubei	18301079	2834072	72751	14654	8374
Hunan	16420299	2065783	50253	8983	6133
Guangdong	78621730	8386674	238684	10044	62031
Guangxi	5536319	809441	23243	1255	2228
Hainan	639621	137275	1726	124	341
Chongqing	17265239	1364847	34421	3419	4820
Sichuan	19021875	3209439	86736	9554	13369
Guizhou	1884371	469418	11458	525	1728
Yunnan	2862040	785053	19754	1438	2021
Tibet	39106	52384	635	2	93
Shanxi	5278713	2638225	64752	12761	4392
Gansu	2446938	625520	20118	3658	1047
Qinghai	494442	147677	2501	110	228
Ningxia	847684	179524	5153	79	606
Xinjiang	1670297	490340	8810	510	1493

Source: The data is mainly from the ‘China statistical Yearbook on Science and Technology’ National Bureau of Statistics of the People’s Republic of China, 2009). Patents index means patents granted by region. Publications index means Chinese scientific papers taken by major foreign referencing system by region. Thirty-one provinces are except Macao, Taiwan, Hong Kong.

National Bureau of Statistics of the People’s Republic of China. (2009, 2010).

References

Acs

Zoltan J

Anselin

Luc

Varga

Attila

. (2002). Patents and innovation counts as measures of regional production of new knowledge. Research policy, 31(7), 1069–1085.

Avelar-Sosa

Liliana

García-Alcaraz

Jorge Luis

Cedillo-Campos

Miguel Gastón

Adarme-Jaimes

Wilson

. (2014). Effects of regional infrastructure and offered services in the supply chains performance: Case Ciudad Juarez. Dyna, 81(186), 208–217.

Archibugi

Daniele

Howells

Jeremy

Michie

Jonathan

. (1999). Innovation systems and policy in a global economy. Cambridge: Cambridge University Press.

Asheim

Bjørn T.

Isaksen

Arne

. (1997). Location, agglomeration and innovation: Towards regional innovation systems in Norway? European Planning Studies, 5(3), 299–330.

Asheim

Bjorn

Vang

Jan

(2006). Regional innovation systems in Asian countries: A new way of exploiting the benefits of transnational corporations. INNOVATION: Management, Policy and Practice, 8(1–2), 27–44.

Aslesen

Heidi Wiig

. (1996). An empirical study of the innovation system in Finnmark. Norway: STEP group.

Autio

Erkko

(1998). Evaluation of RTD in regional systems of innovation. European Planning Studies, 6(2), 131–140.

Battese

George Edward

Coelli

Tim J.

(1995). A model for technical inefficiency effects in a stochastic frontier production function for panel data. Empirical economics, 20(2), 325–332.

Bai Junhong

Jiang Kesheng

Li Jing

Li Jia

. (2009). Analysis of the Impact of Environment Factors on Regional Innovation Efficiency–Empirical Test Based on the DEA-Tobit Two-step Method. R&D Management, 21(2), 96–102.

10.

Chi

Renyong

Xiaofen

Zhengwei

. (2004). An analysis of differences in innovation efficiency between the eastern & the western regions in China and its causes. China Soft Science, 0(8), 661–672.

11.

Chung

Sunyang

. (2002). Building a national innovation system through regional innovation systems. Technovation, 22(8), 485–491.

12.

Cooke

Philip

. (2008). Regional innovation systems: Origin of the species. International Journal of Technological Learning, Innovation and Development, 1(3), 393–409.

13.

Cooke

Philip

Gomez Uranga

Mikel

Etxebarria

Goio

. (1997). Regional innovation systems: Institutional and organisational dimensions. Research policy, 26(4), 475–491.

14.

Cooper

William W

Seiford

Lawrence M

Tone

Kaoru

. (2007). Data envelopment analysis: A comprehensive text with models, applications, references and DEA-Solver Software (2nd ed.). New York: Springer.

15.

Doloreux

David

Parto

Saeed

. (2005). Regional innovation systems: Current discourse and unresolved issues. Technology in society, 27(2), 133–153.

16.

Edquist

Charles

. (1997). Systems of innovation: Technologies, institutions, and organizations. London: Psychology Press.

17.

Edquist

Charles

. (2010). Systems of innovation perspectives and challenges. African Journal of Science, Technology, Innovation and Development, 2(3), 14–45.

18.

Etzkowitz

Henry

Leydesdorff

Loet

. (2000). The dynamics of innovation: From National Systems and ‘Mode 2’ to a Triple Helix of university–industry–government relations. Research policy, 29(2), 109–123.

19.

Evangelista

Rinaldo

Iammarino

Simona

Mastrostefano

Valeria

Silvani

Alberto

. (2001). Measuring the regional dimension of innovation. Lessons from the Italian Innovation Survey. Technovation, 21(11), 733–745.

20.

Freeman

Chris

. (2000). The political economy of the long wave. Cheltenham, UK: Edward Elgar Press.

21.

Fritsch

Michael

. (2002). Measuring the quality of regional innovation systems: A knowledge production function approach. International Regional Science Review, 25(1), 86–101.

22.

Goto

Akira

Suzuki

Kazuyuki

. (1989). R&D capital, rate of return on R&D investment and spillover of R&D in Japanese manufacturing industries. The Review of Economics and Statistics, 71(4), 555–564.

23.

Shulin

Lundvall

Bengt- ke

. (2006). China’s innovation system and the move toward harmonious growth and endogenous innovation. Innovation: management, policy & practice, 8(1), 1–26.

24.

Guan

Jiancheng

Liu

Shunzhong

. (2003). The study on impact of institutions on innovation efficiency in regional innovation systems. Studies in Science of Science, 21(2), 210–214.

25.

Jiancheng

Guan

Ying

. (2005). The performance of Chinese regional innovation system evaluation based on data envelopment analysis. Studies in Science of Science, 23(2), 265–272.

26.

Ying

Han

Kai

Mei

Yanbo

Zhang

. (2006). Evaluation of the efficiency of regional innovation system in China. Technology Economics, 0(3), 13–16.

27.

Feng

Chen

Rong

. (2008). Corporate governance managerial incentive and firm efficiency—An empirical analysis of Chinese listed company in several industries. Journal of Management Sciences in China, 11(4), 142–152.

28.

Mei-Chih

Mathews

John A.

(2005). National innovative capacity in East Asia. Research Policy, 34(9), 1322–1349.

29.

Huang

Lucheng

. (2000). Approach to the main content of regional innovation system. Science Research Management, 21(2), 43–48.

30.

Zhijian

. (1999). The development of the theory of regional innovation system. Forum on Science and Technology in China, 0(6), 20–23.

31.

Jaffe

Adam B.

(1986). Technological opportunity and spillovers of R&D: Evidence from firms’ patents, profits and market value. USA: National Bureau of Economic Research Cambridge, MA.

32.

Jiang

Heping

Cui

Kai

. (2009). Accelerate agricultural scientific and technological achievements to promote food production and farmers’ income. China Development Observation, (9), 19–21.

33.

Jiang

Junfeng

Shen

Junfeng

Wang

Xiulai

. (2009). Modeling the techno-innovative cooperation between enterprises with asymmetric capability. Journal of Systems Engineering, (3), 335–342.

34.

Jiang

Kebing

Liu

Qida

. (2013). A preliminary analysis of university’s low scientific and technical achievements transformation rate. Technology and Innovation Management, 34(3), 201–204.

35.

Jie Wu

Zhou Zhixiang

Liang

. (2010). Measuring the performance of Chinese regional innovation systems with two-stage DEA-based model. International Journal of Sustainable Society, 2(1), 85–99.

36.

Lan

Ming

. (2008). CPPCC members cries technology investment efficiency needs to be improved. Retrieved from http://scitech.people.com.cn/GB/6981981.html

37.

Jing

Tan

Qingmei

Bai

Junhong

Yue

Liangyun

. (2009). Stochastic frontier analysis of China’s regional innovation efficiency. Systems Engineering, 27(8), 44–50.

38.

Xibao

. (2009). China’s regional innovation capacity in transition: An empirical approach. Research Policy, 38(2), 338–357.

39.

Xinwei

Dengfeng

. (2014). Research progress and future prospects of regional innovation system in the knowledge diffusion foreign. Research Progress and Future Prospects of Regional Innovation System in the Knowledge Diffusion Foreign (1), 33–39.

40.

Yun

Lin

Xiangrong

Jin

. (2008). Regional innovation performance of the “Matthew Effect” —30 Chinese provinces and cities based on panel data analysis. Economist, 3, 78–85.

41.

Liu

Lingli

Jianhua

. (2007). Empirical study on the efficiency of regional R&D resource allocation based on stochastic frontier analysis. Science of Science and Management of S&T, 12(16), 39–44.

42.

Liu

Shunzhong

Guan

Shunzhong

. (2002). The evaluation on the innovating performance of regional innovation systems. Journal of Management Sciences in China (1), 75–78.

43.

Liu

Xielin

White

Steven

. (2001). Comparing innovation systems: A framework and application to China’s transitional context. Research policy, 30(7), 1091–1114.

44.

Lundvall

Bengt-Åke

. (2010). National systems of innovation: Toward a theory of innovation and interactive learning (Vol. 2). London: Anthem Press.

45.

Yanhua

Mao

. (1999). The connotation and policy implication of regional innovation system. Economist, 0(2), 84–90

46.

National Bureau of Statistics of the People’s Republic of China. (2009, 2010). China statistical Yearbook on Science and Technology. Beijing, China: China Statistics Press.

47.

Nelson

Richard R

Green

Kenneth

. (1996). National innovation systems: A comparative analysis. R and D Management, 26(2), 191–191.

48.

Peng

Jianguo

Guo

Peng

Mingjie

. (2014). Research on relationship between innovation input and output of high—tech industry based on canonical correlation analysis. Science and Technology Management Research, 34(6), 101–104.

49.

Peng

Zheng

(2012). Regional innovation system: Theory and policy. Special Zone Economy, (10), 209–211.

50.

Radosevic

Slavo

(2002). Regional innovation systems in Central and Eastern Europe: Determinants, organizers and alignments. The Journal of Technology Transfer, 27(1), 87–96.

51.

Shenggang

Ren

. (2007). The demonstration research of dominant industry development selecting of Central China. Journal of Central South Huiversity: Social Science, 13(4), 446–451

52.

Shang

Qian

(2011). Regional innovation system: Dynamic positioning for policy. Science of Science and Management of S&T, 32(7), 81–85.

53.

Kai

Sun

. (2006). Study on the evaluation of innovation efficiency of regional innovation system based on DEA. Science and Technology Management Research, 28(3), 139–141.

54.

Wang

Yansen

. (2005). A study on Chinese private enterprises foreign direct investment. Wuhan University of Technology Press.

55.

Jie

Zhou

Zhixiang

Liang

. (2010). Measuring the performance of Chinese regional innovation systems with two-stage DEA-based model. International Journal of Sustainable Society, 2(1), 85–99.

56.

Yanbing

. (2008). Measurement on R&D output elasticity of Chinese industrial sector. China Economic Quarterly, (4), 869–890.

57.

Jianguo

. (2012). Analysis on the influencing factors of provincial factor allocation efficiency: Based on stochastic frontier analysis. Journal of Shijiazhuang University of Economics, 35(1), 20–24.

58.

Xie

Lijuan

Yang

Wenpeng

Zhou

Hang

. (2009). Study of regional innovative environmental evaluation and disparities Based on DEA. Journal of Xi’an Polytechnic University, (6), 124–128.

59.

Zabala-Iturriagagoitia

Jon M

Voigt

Peter

Gutiérrez-Gracia

Antonio

Jiménez-Sáez

Fernando

. (2007). Regional innovation systems: How to assess performance. Regional Studies, 41(5), 661–672.

60.

Zhang Manyin

Wen Shihui

Dahai

Han

. (2011). Efficiency evaluation of Regional innovation system based on Government-industry-university institute collaboration. Science & Technology Progress and Policy, 28(11), 130–133.

61.

Zhu

Joe

. (2003). Quantitative models for performance evaluation and benchmarking: data envelopment analysis with spreadsheets and DEA excel solver (Vol. 51). Norwell: Springer.