Industry Clustering, Industry Specialization, and the Utilization Efficiency of Urban Land:Evidence from Prefectural-Level Cities in China

In this paper, a land utilization efficiency evaluation model, which takes environmental loss into consideration, has been structured via taking advantage of the slack-based measure (SBM) model. Meanwhile, based on the panel data from 280 prefecture-level cities in China from 2003 to 2013, the paper thoroughly probed into, and discussed, the effect imposed by industry clustering and specialization on the utilization efficiency of urban land. Research results indicate several conclusions, as follows: (1) Taking environmental loss into account, the land utilization efficiency of prefecture-level cities in China is generally low. During the research period, the average value of the land utilization efficiency of prefecture-level cities in China is only 0.349, with, first, a declining trend, and then a rise. Geographically speaking, the land utilization efficiency presents a “depression in the center” phenomenon which means the land utilization efficiency of prefecture-level cities in the central China are relatively lower than in the east and west. Now, the difference among the urban land utilization efficiency in China significantly reflects the distinctions among Eastern, Western, and Central China. Moreover, the contribution degree of the difference of the land utilization efficiency among cities of central China to the aggregation difference shows an ascending momentum. Additionally, the relation between the population scale and land utilization efficiency in cities manifests as a U shape; (2) theoretically speaking, the relation between industry clustering and urban land utilization efficiency presents an inverted-U shape. However, this kind of relation is not significant in Western and Central China and medium-sized cities. Moreover, most of cities are still relatively far away from the inflection point or the critical value; and (3) the industry professional level has imposed a positive influence on urban land utilization efficiency. However, that influence is not significant in Eastern China and large cities. Consequently, strengthening the industry professional development of Western and Central China and small and medium-sized cities, facilitating diversified development of industries in Eastern China and large cities, and accelerating industrial clustering, all of these measures above will be conducive to improving urban land utilization efficiency in China.


Introduction
The development path of new-style urbanization and industrialization is a crucial supporting point for continuous growth of China's economy. Urban land is the spatial carrier of the economy, society, and environment of cities [1]. However, blind and mindless urban expansion has also generated a series of issues for China in terms of land utilization, such as construction land increase and the threat of the warning limit of arable land, the tension of development land and regional exceeds the optimal urban economic scale. This situation will bring about a crowded effect and increase the consumption of raw materials which will hinder economic growth and exacerbate environmental pollution. Therefore, it is clear that there may exist similar development phases as above for the relation between industry clustering and urban land utilization efficiency.
Since the 1980s, with constant development of the new economic growth theory, urban economic scholars started to focus on the effect of industry clustering structure (namely, industry specialization and industry diversification) on the urban economy. Subsequently, two different points of view have been generated as follows [4][5][6]: On one hand, some scholars believe that local industry specialization will significantly facilitate knowledge spillover and enhance economic growth, and this point of view is considered as MAR spillover ① . On the other hand, others hold that industry diversification will be more conducive to knowledge spillover and economic growth, which is called Jacobs spillover. Therefore, it can be found that the industry clustering structure, via knowledge spillover, imposes influence on urban land utilization efficiency. However, there are few studies probing into which kind of knowledge spillover is more important, the one caused by industry specialization, or the one brought about by industry diversification.
In conclusion, there may be a non-linear relation between industry clustering and land utilization efficiency. Additionally, it has been discovered that the industry clustering structure, via knowledge spillover, imposes influence on urban land utilization efficiency. However, given the context of China's situation, the practical influence should be discussed further.

Measuring Methods of Land Utilization Efficiency
In recent years, the Data Envelopment Analysis (DEA) method has been frequently adopted in ① MAR spillover comes from Marshall (1890), Arrow (1962) and Romer (1986Romer ( , 1990)'s contribution and stresses the externalization of localization with the core idea of specialization spillover. 3 of 17 plenty of research literature in order to discuss and research land utilization efficiency [7][8][9][10].
Through using DEA, the essential characteristics of land utilization have been fully taken into consideration, and land utilization has been considered as an input-output system. Meanwhile, there is no need to presume a certain production function for DEA, and there will be no influence from the dimensional unit; therefore, resulting errors caused by function setting and index weight setting can be avoided [11][12][13].
Currently, the studies employing DEA to evaluate the urban land utilization efficiency possess features as follows: At first, based on the perspective of input and output variables, land, capital, labor, etc., have been regarded as input elements. Meanwhile, GDP, industry production value, and so on have been considered as output variables. Secondly, focusing on research methods, it can be discovered that the BCC and CCR models of the DEA approach have been adopted in most studies [8][9][10][11][12][13][14][15], while the super-efficiency DEA [16] and bootstrap-DEA [17] approaches have been utilized in a few cases. However, during the process of industrialization and urbanization in China, due to the spread of extensive development patterns, tremendously fast economic growth has been achieved with the heavy costs of excessive resource consumption and environmental deterioration.
Consequently, when land utilization efficiency is being evaluated, it is debatable, in terms of output, to only take GDP and industry production value into account, neglecting resources and environmental constraints. Thus, there is room for further improvement.
Scholars have made various attempts in order to make use of the DEA method to measure the economic efficiency considering undesirable output. In the existing literature, there are several methods which have been adopted in order to bring environmental factors into the economic efficiency model analysis; for instance, the method [18] which sets undesirable outputs as the input elements, the hyperbolic method [19], and the method [20] which takes undesirable outputs as output elements. Among those, the method which considers environmental emissions as undesirable outputs and then brings them into the production process has been widely applied [21].
For example, Chung et al. have incorporated polluted emissions into the production process and proposed the analysis model of environmental regulation behavior, which is based on a directional distance function, and this model preferably tackled the efficiency evaluation issue of undesirable output [22]. Nevertheless, this kind of directional distance function is a radial and oriented DEA model when it is applied to evaluate efficiency, which cannot completely take the slackness of the input and output into consideration. At the same time, it is necessary to make input-oriented or output-oriented choices; namely, a certain aspect, either output or input, may be ignored.
Furthermore, there may be errors in the measured efficiency value [23]. In order to deal with the issue above, Tone has proposed a non-radial ad non-oriented slack-based measure (SBM) model. Therefore, in this paper, environmental loss has been included and considered in the evaluation system, and the SBM model has been adopted to measure the land utilization efficiency of each prefecture-level city. In this paper, each city has been regarded as a DMU (decision-making unit). It has been assumed that there are m different kinds of input x y b .
Consequently, the production possibility set measuring the land utilization efficiency can be Afterwards, based on the research from Tone, following SBM model can be structured: Especially, x , y , b are, respectively, the slacks quantity of input, desirable output, and undesirable output. ρ < means that the evaluation unit is inefficient, and there is room for improvement of the output and input.

Index selection and Data Processing
Based on the statistical analysis principle of comparability and feasibility, thoroughly and comprehensively considering data availability, 280 prefecture-level cities from 2003 to 2013 have been selected for the research ② . The data mainly derives from the China City Statistical Yearbook and China Urban Construction Statistical Yearbook over the years, and it also comes from the statistical yearbook of every province, city, and municipality in the related years. Part of the missing data has been complemented by the interpolation method.

Input Factors
The classical production function manifests that the basic input of production includes capital and the labor force. At the same time, the land utilization is the spatial carrier of the urban economy.
Therefore, in this paper, input elements contain the labor force, land, and capital stock. Especially, the labor force is represented as the number of unit employees in cities at the end of the year, and land is represented as the size of urban built-up area. Moreover, capital stock is calculated referring to the method of Zhang [24]. Commodity prices can be replaced by the price index of the province, city, and municipality where the sample city is located: (Investment's deflating price index = 0.45 × FAIPI (Fixed Assets Investment Price Index) + 0.55 × CPI (Consumer Price Index)). In addition, it has been assumed that the growth rate of capital stock and the increasing rate of actual investment is the same in order to calculate the capital stock of the base year (2003 is the base period).

Output Factors
Desirable output: GDP has been selected as the indicator measuring economic output.
Meanwhile, the CPI of the province, city, and municipality where the sample city is located has been utilized to be deflated (adopting the year of 2000 as the base period).
Undesirable output: Frequently-mentioned negative or bad output contains the emission amount of three industrial wastes [25]: sulfur dioxide [26], chemical oxygen demand (COD) [27], and carbon dioxide [28]. Moreover, in order to avoid the negative output elements being excessively singular, in this paper, the emission sum of urban industrial waste water and industrial sulfur dioxide have been taken as the undesirable outputs.

General Characteristics
According to the discussions above, the general situation of the land utilization efficiency of

Spatial Characteristics
From the perspective of the individual differences of cities, the lowest average value of efficiency is only 0.162, which is less than 1/6 of the frontier efficiency value. Figure 1 shows the spatial distribution of the land utilization efficiency of 280 prefecture-level cities in China. Through

Regional Characteristics
In order to research and observe the characteristics of regional difference of Chinese urban land utilization efficiency under the environmental restrictions, the Theil coefficients of the land utilization efficiency has been calculated and, furthermore, the difference among eastern, central, and western areas of China, as well as the difference inside cities of different areas, and their according contribution to the total difference, has been calculated. The results are shown in Table 1. The Theil coefficients across the nation in Table 1  the difference inside certain areas shows that the contribution rate of eastern areas to the total difference is around 35%, for the central areas it is approaching 31%, and for the western areas it is about 29%. However, generally speaking, the contribution rate of eastern areas presents a fluctuating decreasing trend, and central areas act relatively stably, while western areas show an increasing trend.

Analysis of Scaling Difference
According to The State Council's Notice on City Scale Classification Standard Adjustment, which has been issued since 2014, the prefecture-level cities in China have been classified into three categories (based on the total population number of each city at the end of 2013), namely, small-sized cities (with an urban permanent population under 500,000), medium-sized cities (with an urban permanent population between 500,000 and 1,000,000), and large-sized cities (with an urban permanent population over 1,000,000). The distribution of the average land utilization efficiency of cities at different scales from 2004 to 2013 can be seen in Table 2. Based on the data in Table 2 Consequently, promoting the development of small-sized cities will be conducive to improving Chinese urban land utilization efficiency.

Empirical Model and Variables
(1) Empirical Model In order to inspect and research the concrete influence of industry clustering and its according structures on land utilization efficiency, based on the existing research literature and given that there may be a non-linear relation between industry clustering, urban land utilization efficiency, and the interactive relation between industry specialization and industry diversification ③ , the quadratic term of the industry clustering level has been introduced in this paper. Moreover, an empirical model, as follows, has been formulated, taking the industry specialization as the proxy variable of the industry clustering structure. SS is the industry specialization level. In the existing research literature, concerning the measurement of the urban industry specialization, the approaches of Duranton and Puga [29] have frequently been referred to, namely, the professional Gini coefficient. Therefore, in this paper, the index of urban industry specialization has been defined as follows: In particular, The urban infrastructure situation is represented by road area per capita (m 2 ), and the urbanization level is referred to as the urban population density (10,000 people/km 2 ). In this paper, the industrial waste water amount per GDP and the varying rate of sulfur dioxide emission have been used to measure the environmental regulation. In addition, the approaches from Sun and Wang's research [ 30 ] have been adopted to measure the comprehensive indices. Moreover, according to the research of Wang and Liu [31], the linear term, quadratic term and cubic term of environmental regulation have been included in the model.

Empirical Results Analysis of Total Samples
The evaluation approaches of the panel data generally contain fixed effect and random effect regression. Since, within a relatively short period, the cross-section sample of the data selected in this thesis is relatively large, the fixed effect model will be more suitable. Meanwhile, a Hausman test also manifests that the original assumption of random effect has been rejected with a significance level of 1%. In the total samples, via the stepwise regression treatment to Equation (3), and the variables mentioned above having been respectively evaluated, the author expects to probe the relation between industry clustering, industry specialization, and land utilization efficiency. The evaluation results can be found in table 3.  same. At the same time, there are several points concluded from Table 3.
Firstly, the coefficient of the linear term of industry clustering is positive, while the coefficient of the quadratic term is negative. Meanwhile, they are all significant at the 1% level in different models, which indicates that, theoretically speaking, there is a non-linear relation between industry clustering and the urban land utilization efficiency. The fact that the quadratic term is significantly negative indicates that the increase of the industry clustering level is conducive to knowledge spillover and, further, to increasing the land utilization efficiency when it has not reached the critical value. However, when the industry clustering level reaches and surpasses the critical value, the knowledge spillover effect will be weakened and, further, it will lead to the decrease of the land utilization efficiency. However, the coefficient of the quadratic term is relatively small, at 0.00571, Secondly, the regression coefficient of industry specialization is positive with a significance level of 1%. This manifests that, from a general perspective, the increase of industry specialization level is conducive to improving the urban land utilization efficiency. In other words, the industry clustering structure in China currently presents as MAR spillover, namely, the knowledge spillover brought about by the industry specialization level improvement.
Thirdly, the regression coefficient of the infrastructure situation is 0.01, which is significant at the level of 10%. Therefore, generally speaking, the completion and improvement of urban infrastructures is beneficial to increasing the urban land utilization efficiency. In addition, the regression coefficient of the urbanization level is 0.0824, which is significant at the level of 5%, and this shows that active enhancement of new-style and human-centered urbanization will be conducive to improving the urban land utilization efficiency.
Fourthly, the first term coefficient of environmental regulation is -2.327, the according quadratic term coefficient is 2.749, and the cubic term coefficient is -1.217, which are, respectively, significant at the levels of 1%, 1%, and 5%. These results show that the relation between environmental regulation and the land utilization efficiency is in line with an inverted N-shaped relation, and this conclusion is similar with the research results of Wang and Liu [31]. Namely, when the environmental regulation is relatively weak, social and environmental costs will be reduced and, accordingly, innovative motivation will be weakened, which will further decrease the urban land utilization efficiency. However, when the environmental regulation or restriction degree sits in a reasonable range, the social innovation will be promoted and this will further increase the urban land utilization efficiency. However, when the environmental regulation degree exceeds the affordable limit of society, the land utilization efficiency will also be reduced. Consequently, reasonable and suitable degrees of environmental regulation will be conducive to improving the urban land utilization efficiency.

Empirical Results Analysis for Cities from Various Areas at Different Scales
In China, there exist relatively large gaps in terms of the development situation among cities from different areas and at different scales. In order to observe and inspect these two types of differences, the author further exercises evaluations, respectively. The results are shown in Table 4.    Table 4, it can be determined that, although the influencing direction of each variable on the urban land utilization efficiency is almost in line with the total samples', there are still differences in the coefficients and the significance of industry clustering and industry specialization.
At first, there are regional differences in the influence of industry clustering on the urban land utilization efficiency. Specifically, although, theoretically speaking, there should be a significant inverted U-shaped relation in the eastern area, the critical value is 53.709. Although the quadratic term coefficients of the central and western areas are negative, they are statistically non-significant.

of 17
The results above indicate that the enhancement of appropriate industry clustering will obviously facilitate the improvement of urban land utilization efficiency. Moreover, for central and western areas, active advancement of industry clustering is an effective approach to improve the land utilization efficiency.
Secondly, there are regional differences in the influence of industry specialization Fourthly, speaking of the industry specialization, there are relatively large differences among cities at the three scale levels. Namely, the small-sized cities possess the largest coefficients ④ , and they are significant at the level of 1%, which indicates that the improvement of industry specialization of small-sized cities imposes a more significant improving effect on the land utilization efficiency, whereas, for large-sized cities, it is statistically non-significant compared with other city sizes, which manifests that it will be more beneficial to improving the land utilization efficiency of large-sized cities via facilitating diversified industry development.

Robustness Test
In order to test the robustness of the evaluation results, a further measurement of the land utilization efficiency, irrespective of environmental loss, has been exercised. Meanwhile, the total samples have been re-evaluated. Table 5 reports the evaluation results of the robustness test, and the results of Tables 4 and 5 are basically consistent, which means, theoretically speaking, there is a non-linear relation between industry clustering and urban land utilization efficiency. However, the quadratic term coefficients are relatively small and non-significant, which partly indicates that the current diseconomies of industrial agglomeration mainly present as environmental loss. At the same time, industry specialization is conducive to improving urban land utilization efficiency. In ④ Coefficients here refer to the influencing coefficient of different cities' industry specialization on the urban land utility efficiency.
addition, the influencing directions of the infrastructure situation, urbanization level, and environmental regulation on land utilization efficiency are basically consistent, while there are still some differences in terms of the variations in coefficients, and the evaluation of the infrastructure situation is not significant. Notes: ***, **, and * respectively represent the evaluation values being significant at the level of 1%, 5%, and 10%. The values in the round brackets are the t statistical amount under the robust standard errors.

Conclusions
In order to analyze the relation between industry clustering, industry specialization, and the there is no city whose efficiency value has exceeded the critical value. From the perspective of the scales of cities, the industry professional level improvement of small-and medium-sized cities imposes a more significant increasing influence on the urban utilization efficiency. Additionally, complete infrastructure and active enhancement of new-style and human-centered urbanization will be substantially conducive to improving the urban land utilization efficiency. Meanwhile, there is an inverted N-shaped relation between environmental regulation and the land utilization efficiency.
Scientific planning and designing of urban land is an essential approach for improving the land utilization efficiency. Furthermore, the planning of industry land is an important part of the land utilization planning. Based on the research conclusions of this paper, there are several findings and suggestions, as follows: To begin with, it is vital to effectively link land utilization planning and industry planning. For most cities, it is necessary for the land utilization planning to lead the industry professional development and actively facilitate the developmental strategy of "one city, one industry", which embodies the idea of specially developing one unique, advantageous industry in one city. Meanwhile, for eastern areas and large-sized cities, it is necessary to make land utilization planning guide diversified industry development. In conclusion, via the convergent development of industry and cities, led by land utilization planning and through increasing industry clustering intensity, land utilization efficiency of Chinese cities will be further improved.