Investigating operations of industrial parks in Beijing: efficiency at different stages

Abstract Industrial parks enjoy significant importance in many countries and regions. This study presents a multi-stage operational process to evaluate the efficiency of parks at each stage using an empirical study of Beijing. The study finds that only three of 22 parks were efficient overall during 2006–2008 and two of 22 were efficient during 2009–2012. The promotion of business, facilitation of production, and rewards of economic returns are highly correlated stages for efficiency performance. The results suggest that Beijing’s government should expend more effort developing the potential to generate outputs given current land and investment inputs. In addition, it provides a tool to strengthen the organisational capacity development of industrial parks by emphasising their multi-dimensions in inputs and outputs, selecting the right competitors at the right organisational stage, locating sources of efficiency and inefficiency, and understanding progression and balance of internal stages during operation.


Introduction
Industrial parks are pioneering new economic activities and industrial organisations, and thus, are receiving a lot of attention (Curl & Wilson, 2015;Ramos & Fonseca, 2016;Shen & Tsai, 2016). An industrial park is considered a place in which a group of firms is concentrated to realise the potential of economies of scope and to facilitate both tangible and intangible industrial linkages, knowledge exchange, and technology spillovers to boost local economic growth. This high expectation confers privileges to industrial parks in land provision, infrastructure investment, and fiscal budget. According to the database of the International Labour Organization (ILO), the number of industrial parks worldwide increased from 75 in 1975 to 3500 in 2006 (Boyenge, 2007). By 2007, there were more than 130 countries implementing industrial park schemes (Farole & Akinci, 2011) and the development of various industrial parks has become a strategy to support local economic development, KEYWORDS industrial park; network DEa; organisational capacity; operation; Beijing The remainder of the paper is organised as follows. Section 2 offers a multi-stage analysis of industrial park operation by drawing on economic, business, and management literature. Section 3 describes the methodology of S.B.M. network D.E.A. for industrial park management. Section 4 presents the results of efficiencies at the main operational stage of 24 industries in Beijing during [2005][2006][2007][2008][2009][2010][2011][2012]. The implications of these findings are discussed in Section 5, and Section 6 concludes by recapping the findings and limitations, and suggesting directions for future research.

Multiple stages of industrial park operation
As an industrial organisation, the industrial park develops interactions with resources and business environments. From the perspective of business management, it can be divided into the processes of promotion, facilitating production, and generating tax for government. Measuring the efficiencies of these divisions has become critical. Potential can be located effectively in the improvement of overall performance, enabling management in the areas from strategic planning to organisational learning, and eventually improving the competitive advantage of the organisation (Avkiran, 2009;Shen & Tsai, 2016;Yang, Cai, Ottens, & Sliuzas, 2013).
As such, evaluating performance and efficiency of industrial parks is an important issue. Löfsten and Lindelöf (2003) measure the effect of resources, innovation, risk, and strategies on the growth of firms in science parks in Sweden. C. J. Chen and Huang (2004) and Hu et al. (2009) evaluate the performance of high-tech industries in Taiwan's industrial parks. Nosratabadi, Pourdarab, and Abbasian (2011) examine industrial parks' performance in Iran. Rivera, Sheffi, and Knoppen (2016) examine logistics parks in Spain. Gradually, the D.E.A. method has gained popularity in evaluating industrial parks' performance. For example, C.-J. Chen, Wu, and Lin (2006) apply D.E.A. and the Malmquist index to investigate the performance of six industries in a science park in Taiwan. Liu, Tian, Chen, Lu, and Gao (2015) evaluate environmental performance of national eco-industrial parks in China using D.E.A. Izadikhah and Saen (2015) propose a single virtual approach to rank the performance of 17 Iranian industrial parks. Network D.E.A. has also received interest from researchers, such as Khodakarami et al. (2014), who measures the sustainability of industrial parks in Iran by two-stage D.E.A.; and Hu et al. (2010), who analyses science and technology industrial parks in mainland China applying a four-stage D.E.A. approach. Different form Hu et al. (2010) and Khodakarami et al. (2014), who examine both production and environmental efficiency, this research mainly focuses on business management and the economic operation of industrial parks.
From the point of view of business management, it is essential for industrial parks to attract new firms and investment as sources of parks' continuous growth. Marketing and branding have become increasingly important in strategic planning all over the world; these can be regarded as a plurality of efforts to create a corporate image based on firms' and parks' distinctive characteristics, and through this, to attract investments and specialised human resources (Metaxas, 2010;Mudambi, Doyle, & Wong, 1997). This is particularly so for industrial parks in the globalised economy characterised by competition in industrial locations (Bessho & Terai, 2011;Shen & Tsai, 2016;Wilsher, 1994;Yang et al., 2013;Yang, Liang, & Cai, 2014). This constitutes the first stage of industrial park operation, that is, promotion of business.
Successfully managed industrial parks never stop innovating operations after attracting firms. At the second stage, the facilitation of production includes shaping forward and backward linkages, enhancing business support, developing entrepreneurship, and creating opportunities for improving labour skills. Efforts to achieve these features can be observed in the creation of incubators and science parks for start-ups and technological innovation, local institutional reform, strength cooperation between industries and universities, and stimulating the growth of supporting industries (Phan, Siegel, & Wright, 2005;Salvador, Mariotti, & Conicella, 2013). These efforts aim to leverage the production of firms in the park to reap high industrial outputs and revenues.
Economic returns are the last stage of industrial operation. This represents the rewards for the establishment of industrial parks. In addition, because governments quite often invest significantly in infrastructure and public facilities, fiscal revenue is an important prerequisite for the financial sustainability of the industrial park, including the ability to pay its own operating costs (Geng, Zhang, Côté, & Qi, 2008) and to evolve continuously (Peddle, 1993), especially in a decentralised institutional environment.
These stages constitute a relatively complete process of industrial park operation, although there are still limited studies exploiting the stages of promotion and economic returns. These three stages occur coherently and simultaneously, not in exclusion to each other, in the profit-generating process of the park, from attracting firms to harvesting the fruits of establishing the park. In addition, some factors may work at one or all stages; for example, knowledge milieu can both attract firms and facilitate production. Some factors need balancing, subject to the phase and type of industrial parks; for instance, taxation reduction was a main method used in Chinese industrial parks to lure foreign companies, but hampered the generation of fiscal revenue, causing budgetary problems for industrial construction and maintenance, which was abandoned gradually after 1999 (Kynge, 1999).
Therefore, in order to be successful, industrial park operation needs to monitor each stage's efficiency. This is a crucial step to reveal the strengths and weaknesses of operations, activities, and processes (Maleyeff, 2005). Anand and Kodali (2008) claim that efficiency measurement is a continuous analysis of strategies, functions, processes, products or services, and performances with the intention of assessing an organisation's current standards and, thereby, carrying out self-improvement.

Method
By looking at both inputs and outputs, efficiency evaluation has received a lot of attention, particularly using the D.E.A. method. D.E.A. constructs the best performance 'frontier' and reveals the relative shortcomings of inefficient decision-making units (D.M.U.s) (Xu & Yeh, 2014;). It measures efficiency by generating the maximum outputs obtainable from the given inputs consumed or by minimising inputs for generating the given outputs under the current status of technology available . This method shows several merits in practice; for instance, it can evaluate multiple inputs and multiple outputs and thus, can account for multi-dimensionality in management, requires no a priori production function, and distinguishes the best performers for each heterogeneous group rather than against the average of all groups (Kumar Mandal & Madheswaran, 2010;Richard, Devinney, Yip, & Johnson, 2009;Sueyoshi, Goto, & Sugiyama, 2013;Tone, 2001). Despite these merits, standard D.E.A. is denounced as a black box because it provides inadequate information to identify the specific sources of inefficiency (Färe & Grosskopf, 1997). Therefore, network D.E.A. has been developed to open the black box (Avkiran, 2009;Laurens, De Bram, Bart, Filip, & Jeroen, 2013), and has received increasing attention in operational studies (Cook, Liang, & Zhu, 2010;Despotis, Sotiros, & Koronakos, 2015;Park & Sung, 2016). This study attempts to apply a network slack-based network D.E.A. model to investigate the special multi-stage organisational process of industrial parks.
The general idea of applying network D.E.A. is to locate the (in)efficiencies at different stages of the operation of industrial parks. At each stage, the development of industrial parks is featured by multiple inputs (e.g., land and fixed asset investment) and multiple outputs (e.g., to realise economic growth and add fiscal revenues). Moreover, the output in the previous stage could be the input in the next stage, for instance the increase of firms and investment located in the park. Differences in the efficiency of industrial parks at different stages could provide much clearer information on the target of improving an industrial park's operation.
In this study, each industrial park is analogous to a D.M.U., whose efficiency measures multi-dimensional decision problems that can be resolved by D.E.A. Assume a sample that covers n D.M.U.s ( j = 1, 2, … ., n), with m inputs and s outputs on each D.M.U. For the i-th D.M.U., X ij and Y rj are vectors of inputs and outputs. Given that the efficiency of industrial parks can be affected by minimising inputs or maximising outputs, and that inputs and outputs may not change proportionally, an S.B.M. is proposed to solve the non-radial problem in D.E.A. (Tone, 2011). As all industrial parks seek to maximise their outputs under given conditions of inputs, an output-oriented model is selected. In addition, considering the change of scale would affect efficiency, the output-oriented S.B.M. model is chosen under the assumption of variable returns to scale (Tone, 2011). (1) X = (x 1 , x 2 , … , x n ) ∈ R m×n and Y = (y 1 , y 2 , … , y n ) ∈ R s×n , and X > 0 and Y > 0 where S i -, and S r + are slack vectors corresponding to input excesses (input slacks) and output shortfalls (output slacks) in the i-th D.M.U. x i0 , y r0 represent the input and output, respectively, in the frontier unit 0, and ρ * is the efficiency of the D.M.U.
In running the D.E.A. model, a problem arises in that all D.M.U.s are non-efficient or efficient with a score of 1. That means the resulting efficiency score is highly skewed, and efficient D.M.U.s cannot be distinguished. To solve this, a super-efficiency ( * O ) model of D.M.U.s is calculated (Tone, 2011): The network approach considers the existence of several stages in industrial park operations, each of which consumes its own set of inputs and produces its own set of outputs; intermediate products are defined as inputs for some stages and are the outputs for others, as linking activities in between stages. Specifically, if D.M.U.s (j = 1, 2, …, n) can be divided as k stages (k = 1, 2, …., K), L is denoted as the link set from stages k to h by (k, h), m k and r k are the numbers of inputs and outputs, respectively, to stage k, and the production possibility set is defined as where x k j is input resources to DMU j at stage k x k j ∈ R mk + (j = 1, 2, … ., n; k = 1, 2, … ., K), y k j is output of DMU j at stage k y k j ∈ R rk + (j = 1, 2, … ., n; k = 1, 2, … ., K), and z (k,h) the number of items in the link (k, h), and k ∈ R n + is the intensity vector corresponding to stage k (k = 1, 2, … ., K) (Tone & Tsutsui, 2009).
In the analysis, the linking activities are freely determined and maintain continuity between inputs and outputs. The output-oriented efficiency of DMU o is solved by the following linear programme: where * o is the overall output efficiency of DMU o , w k is the relative importance of k division, and ∑ K k=1 w k = 1. The output-oriented efficiency score at each division k is calculated by In output-oriented S.B.M. network D.E.A., the overall efficiency score is the weighted harmonic mean of the divisional scores In order to compare D.M.U.s with efficiency scores of 1, we follow a super-efficiency approach to obtain super-efficiency scores for the overall process and each stage.

Study area and data
Beijing is the capital city of China, with a population of more than 21 million and a land area of about 16,140 km 2 . It is widely reported that industrial parks have become an engine of the Beijing economy (Yang, Sliuzas, Cai, & Ottens, 2012). .
At present, there are 24 industrial parks in Beijing, of which 10 are at national level, including eight sub-parks of the Z.G.C., the Beijing Development Area (B.D.A.), and the Tianzhu Bond Area, and 16 are at municipal level. Correspondingly, there is a huge amount of input in establishing and developing the industrial parks. The land size on which parks have been implemented increased from 9849 ha in 2006 to 13,591 ha in 2012, although this accounts for only 37% of the planned area, implying that 63% of the land remains undeveloped (Beijing Statistical Bureau, 2013). Meanwhile, the accumulated fixed asset investment in the parks, spent mainly on road construction and public facilities, was as high as RMB 82.1 billion in 2012, an increase of nearly four times compared to (Beijing Statistical Bureau, 2006. Therefore, the first empirical question examined in this study is as follows.
Q1: Are Beijing's industrial parks economically efficient so that economic inputs can be used optimally?
As the global financial crisis since 2008 has significantly affected economic development, the second research question this study investigates is as follows.
Q2: Do the relative performances of Beijing's industrial parks change before and after the financial crisis? Furthermore, as discussed in the literature review, industrial park operation involves different stages. The second empirical question in this study is as follows.
Q3: Do Beijing's parks have different efficiencies at different operational stages?
Answering these questions is expected to generate detailed information for improving the internal processes of industrial park operation.
In light of theoretical analysis, the operational processes of an industrial park can be divided into three key stages: the promotion of business, the facilitation of production, and the rewards of economic returns. The key indicators and the three stages are illustrated in Figure 1. The implemented land area is underdeveloped land or land facilitated for industrial park use, and is assumed as a key input in the whole process. Fixed asset investment is the total accumulated fixed investment since the establishment of the park until the year analysed. As the main output of promotion, investment attracted includes domestic investment and F.D.I. in the park. Registered capital refers to the amount of capital registered at commercial and business bureau when firms are established in the park. This is an important indicator of risk and profit share of firms. Revenue is the total sum of money obtained by firms in the park through services, and industrial output is the value obtained by firms engaged in manufacturing activities. Economic returns are reflected by profits made by the  Table 1 presents descriptive statistics of the data-set.

Results
The S.B.M. network D.E.A. approach measured the efficiency of each operational stage of industrial parks in Beijing. As the promotion and production stages are more important than the taxation stage, the weight of each of these stages, w k , is 0.4, 0.4 and 0.2, respectively. Appendix 1 shows the overall network efficiency estimates, and provides a breakdown of each operational score in each year. By considering possible fluctuations and in order to understand long-term effectiveness, the mean of the efficiency score of all periods is calculated by the sum of all scores divided by the number of years observed (Table 2). In addition, the efficiency of industrial parks was compared for the periods 2006-2008 and 2009-2012 to examine differences of their performances before and after the global financial crisis.

Comparing relative performances of industrial parks
The findings show that different parks performed quite differently in terms of efficiencyfrom Z. The performance of industrial parks is substantially different before and after the global financial crisis, which addresses the second empirical question. Table 2 shows that six parks maintain their ranks before and after the financial crisis, nine parks improve their ranks, and seven parks are downgraded. This reflects that industrial parks, as a connection between local and global economies and an engine of local economic growth, are sensitive to this effect of the financial crisis.
Changes of the performances of industrial parks reflect their ability to compete in the market and to some extent their growth momentum in the future, since industries and firms are intensively engaged in technological innovation and upgrading in order to counteract the effect of the financial crisis. The first science park, Z.G.C.-Haidian park, ranks first before and after 2008 because a handful of high-tech firms, such as Lenovo, IBM, and Microsoft, are located in the park. Z.G.C.-Fengtai maintains its third place mainly due to the      increasing importance of its firms engaged in environmental protection technology. BDA, the most important manufacturing base in Beijing (Yang et al., 2013), increases its rank by three places compared to other parks, as it has transformed from traditional manufacturing activities to modern ones. Badaling E.D.Z. increases its rank by 16 places as it quickly develops new energy and environmental protection industries. By comparison, traditional manufacturing industrial parks are degraded significantly, including Linhe E.D.Z., which is mainly for auto-parts production, and Fangshan E.D.Z., which is engaged primarily in the oil industry. The lower ranking of Z.G.C.-Desheng by seven places could be because its finance industries (backup offices) were affected during the financial crisis. The network D.E.A. approach enables a closer analysis of the performance of industrial parks at each operational stage, which allows us to test the third empirical question. Taking  Furthermore, Kendall's tau coefficient was performed to measure correlation among the rank of overall performance and each stage's performance (Table 3). This showed that the rank correlation coefficients have significance levels of no less than 0.7, except for the coefficients between promotion of business and economic returns, which were 0.58 during 2008-2012 and 0.68 during 2006-2008, implying that the efficiency of the previous stage could significantly affect the next stage.
Detailed analysis could help to detect the main sources of efficiency or inefficiency of parks. If the overall efficiency rank is used as a baseline, the difference of each stage rank from the baseline could be depicted as Figure 1, which could roughly be used to understand the main contributor of efficiency or inefficiency in a relative sense. Take Z.G.C.-Desheng as an example. The key stages to improve its performance were the enhancement of economic returns during 2006-2008, and the facilitation of production and the enhancement of economic returns during 2009-2012. Having a better understanding about the key stages of performance would help industrial parks to improve their operations and become more efficient.

Operating parks according to divisional efficiency
The stage analysis greatly enhances industrial park management from strategic planning to organisational capacity development. In particular, stage analysis makes four contributions to industrial park operation, especially parks that are government projects. First, the analysis helps monitor multiple I-O relationships during industrial park development. The D.E.A. method measures efficiency based on multiple inputs and outputs, generating an objective and consistent approach to evaluate park performance, thereby providing information shared by different stakeholders with different interests. As Yip et al. (2009) argue, difficulties in sustaining long-term performance arise not just from the competitive environment but also from subsequent problems in measuring the multi-dimensional characteristics of performance. Depending on data availability, the method can be performed regularly, which helps managers and policymakers to grasp the progress of park operations against their peers or competitors.
Second, by using frontier technology, D.E.A. selects competitors or divides the D.M.U.s as several groups, which share similar inputs or outputs. Therefore, the park can learn from peers that share the same frontier, rather than from the best of the entire group. This enables delivery of information to facilitate the learning process of the organisation. This information is detailed further by examining the operational stages of the industrial parks. Third, divisional efficiency locates the source of efficiency performance of the park, which facilitates more target measures. Accordingly, specific efficiency-enhancing strategies can be fostered for the individual components of the production process (Lewis & Sexton, 2004). For example, even though Z.G.C.-Haidian enjoys the highest efficiency score, it could improve its performance at the production stage, as it ranks next to Z.G.C.-Shijingshan during 2008-2012. Z.G.C.-Shijingshan has more problems with the facilitation of production than with the promotion of business and enhancement of economic returns. This internal strengthening process could eventually contribute to overall performance enhancement.
Last but not the least, the divisional analysis helps us to understand the progression of improving performance of industrial parks, and provides a clue to balance the weight of each stage. In this study, focus was given to attracting firms and production of the park, but this by no means indicates that the last stage is not important. As Figure 2 shows, taxation is the main contribution to inefficiency of the Tianzhu Bond Area during 2006-2008. Owing to the special trade policy in this park, taxation is much lower than expected. Although stimulating production and promotion processes, a preferential tax policy needs to be assessed carefully during the development of the Tianzhu Bond Area, as it could affect budgetary issues and the financial sustainability of the park. However, the key issue for the Tianzhu Bond Area changes to the promotion of business due to high marketing competition after 2008 (Figure 3).

Conclusions
This study presents a multi-stage efficiency analysis of industrial parks in Beijing using S.B.M. network D.E.A. It shows that few parks were efficient overall during the analysis periods, and therefore, the Beijing government should expend more effort developing the potential to generate outputs given current land and investment inputs. Furthermore, the study suggests that the promotion of business, the facilitation of production, and the enhancement of economic returns are successive and highly correlated processes. Inefficiency in one stage may lead to underperformance in the next stage.
This study significantly advances industrial park management from strategic planning to organisational capacity development. Theoretically, by synergizing the literature, the concepts of the promotion of business, the facilitation of production, and the enhancement of economic returns highlight the main stages of industrial park operation, taking academic analysis closer to the reality of management. In addition, this study applies D.E.A. to a sophisticated level. Standard D.E.A. deals with one-stage production processes, in which the operation to a large extent is a black box. On the other hand, this study employs S.B.M. network D.E.A. successfully to reveal that the internal operational structure, by referring to the multi-stage processes and emphasising the flow of the intermediate measures among the stages, plays a key role in the efficiency assessment.
In practice, this study contributes to a performance assessment of industrial parks based on the multi-dimensions of inputs and outputs, selecting the right competitors at the right organisational stage. More importantly, it can help managers and policymakers to locate the stage, and identify the sources of efficiency and inefficiency of industrial park development. Network analysis improves the understanding of the progression and balancing of internal stages during operations and therefore, contributes to improving internal management with clearer evidence to strengthen the performance of different offices responsible for marketing and promotion, production of firms, and reward returns to the park. Given the fast growth of industrial parks, especially in developing regions, this study should have wide applicability.
This study has some limitations. The D.E.A. method measures relative not absolute efficiency. All measures should be in a relative sense, and therefore, it is difficult to compare periodical change of efficiency scores. A trade-off is to compare the ranking orders, which is, however, subject to change of the backdrop. For example, if the overall performances of Beijing industrial parks were to decrease, one park with one order improved would not imply that this park improved its performance. Second, although this network approach to some extent makes us prone to discovering the black box, this is highly dependent on our understanding of the internal operational structure and the analysis is subject to data availability. A more customised structure needs to be proposed according to the particular organisation and economic activity of the park; for example, technological promotion and commercialisation could be the main point of research of a science park. Furthermore, the multi-stage analysis is constrained by data availability: the analysis requires more detailed and internal flows of data in an organisation, which are not easy to obtain. For instance, at our last stage, we confine economic returns to direct outcomes to the park and use tax as an indicator, yet this would be more meaningful if wage data were available. Nevertheless, this study proves that multi-stage analysis and network D.E.A. can be powerful tools for management to investigate internal and coherent operational processes. Balancing and weighing the internal stages would be an interesting topic for future research.

Disclosure statement
No potential conflict of interest was reported by the authors.
Funding year.