How Does Industrial Agglomeration Affect Exports? Evidence from Chinese Province-Industry Panel Data

Abstract

Industrial agglomeration is currently an important mode of industrial organisation in China’s regional economic development. Many provinces rely on their favourable resource and location advantages to formulate corresponding industrial agglomeration policies. Industrial agglomeration policies produce agglomeration effects, which enable enterprises in the agglomeration area to gain external advantages such as labour, resources and technology, effectively improving productivity and promoting sustainable local economic development. Based on the inter-provincial industry panel data from 2008 to 2017, the static benchmark regression model and dynamic GMM model are constructed to study the relationship between industrial agglomeration, labour productivity and exports, and to explore their interaction mechanism and practical effects. The study shows that, firstly, industrial agglomeration in China significantly contributes to industrial exports. However, in the long run, industrial agglomeration has a positive and then negative effect on industrial exports, with an inverted U-shaped curve. Secondly, the relationship between industrial agglomeration and industrial exports is negatively influenced by labour productivity. Finally, based on the perspective of regional heterogeneity, the impact of industrial agglomeration on exports is not significant in industrial base regions, but shows a significant promotion relationship in non-industrial base regions. Moreover, the impact of industrial agglomeration on industrial exports is not significantly different between coastal port regions and non-coastal port regions. Based on the research results, this paper puts forward targeted policy recommendations such as changing the competition model, increasing the introduction of talent and adjusting the industrial structure.

1. Introduction

China’s economy has now entered a new phase of high-quality development from high-speed growth, and the Chinese government is building a new development pattern with a large domestic cycle as the mainstay and dual international and domestic cycles promoting each other. Industry, as the pillar industry of China’s national economy, is an important vehicle for achieving a more optimised economic structure, a significant increase in innovation capacity and an advanced industrial base. Therefore, efforts to develop high value-added industries and manufacturing industries and the formation of distinctive industrial clusters are inevitable requirements for building a new development pattern at this stage.

In recent years, China’s export-oriented economy has suffered a severe setback due to the impact of COVID-19, with foreign trade enterprises facing the problem of market capture, the insufficient resumption of production by export enterprises and the accelerated outflow of foreign direct investment [1]. For a long time, China has suffered from the problems of the weak competitiveness of foreign trade enterprises [2], products being easily substituted and markets being easily captured by foreign enterprises [3], and once the foreign trade market is lost, it is difficult to recover. Therefore, maintaining the stability of the industrial supply chain and achieving stable growth in foreign trade is crucial to achieving sustainable economic development [4].

Industrial agglomeration (also known as “industrial clusters” or “enterprise agglomeration”) is a process whereby the same industry is highly concentrated in a particular geographical area and the elements of industrial capital are constantly converging in a spatial context. There are three main reasons why industrial agglomeration promotes the trade exports of enterprises. Firstly, industrial agglomeration increases the productivity of enterprises through the formation of scale effects, and enterprises with high productivity have a greater advantage in terms of trade exports [5]. Secondly, industrial agglomeration can reduce production costs through three mechanisms: the industrial linkage effect, the labour agglomeration effect and the knowledge spillover effect, and thus have more competitive advantages in trade exports [6]. Thirdly, a high regional industrial agglomeration may form a monopoly within the industry, which may easily contribute to an absolute advantage and voice in trade exports [7]. Existing research has contributed to the development of theories related to industrial agglomeration and productivity on exports, but there are still shortcomings.

Therefore, based on inter-provincial industrial panel data from 2008 to 2017, this paper constructs a theoretical model to analyse the impact of industrial agglomeration and labour productivity on exports, and discusses the relationship between industrial agglomeration, labour productivity and their interaction terms with exports, taking into account China’s specific national conditions and the characteristics of industrial development at the current stage. And furthermore, based on regional heterogeneity, the impact of industrial agglomeration on exports is studied from two perspectives: a coastal port region and an industrial base region, respectively.

This paper is organised as follows: Section 2 is the literature review and research hypothesis, Section 3 is the research design, Section 4 is the empirical analysis, Section 5 is the discussion and Section 6 is the conclusions.

2. Literature Review

In recent years, scholars at home and abroad have undertaken a lot of work on industrial agglomeration, productivity and exports, generally focusing on two dimensions: the impact of industrial agglomeration on exports or the impact of productivity on exports. On this basis, this paper will conduct an in-depth theoretical analysis and propose research hypotheses.

2.1. Industrial Agglomeration and Exports

There is a mutual reinforcement between industrial agglomeration and exports. In recent years, numerous studies have explored the various factors affecting trade exports [8] and the effects of trade on human activities [9,10,11]. Since China’s reform and opening up, foreign trade has developed rapidly, greatly contributing to its economic and social development. Scholars have summarised the factors influencing foreign trade through their research and forecast the trade development trends of the countries in the study sample [12]. Among them, the industrial policies of trading countries have become an important research direction. This paper will discuss the influence mechanism between industrial agglomeration and exports from the perspective of industrial agglomeration. Industrial agglomeration is an important feature of modern industrial development [13,14]. In the context of global industrial shifts, export-oriented industries have become a way for local governments to achieve faster economic growth, and governments are rapidly developing local export capacity by setting up development zones and introducing special and advanced industries. With higher uncertainty in export markets than at home, firms need access to information on business processes, customer preferences and the legal environment, and the spatial clustering of large numbers of firms offers the possibility of sharing export information and common costs [15]. Malmberg et al. pointed out that export enterprises are more likely to gain benefits from knowledge and information spillover through spatial agglomeration [16]. It has been found that industrial agglomeration not only increases the probability of firms entering export markets by increasing their own productivity through the associated spillover effects [17], but also significantly increases firms’ own propensity to export and the competitiveness of their export products through competitive pressures within the cluster [18,19,20]. Fujita and He found that the growth of exports has reinforced China’s industrial agglomeration, especially in coastal areas [21]. In the era of economic globalisation, it is not only necessary to focus on the role of trade facilitation in promoting global economic and trade growth, but also to establish institutional arrangements within a country to match trade facilitation, otherwise it will inevitably affect a country’s industrial growth. At the same time, as the degree of industrial agglomeration in China is also increasing, with an increasing number of industries with a high degree of agglomeration and a decreasing number of industries with a low degree of agglomeration, and as trade facilitation itself has the characteristic attributes of transaction costs and institutional arrangements, the degree of industrial agglomeration will also affect the extent to which trade facilitation can contribute to the growth of regional enterprises. Trade facilitation reforms can improve the export performance of developing country industrial clusters [22,23,24,25,26]. However, the excessive agglomeration of homogenous exporters leads to an imagined “de-localisation effect” of exporting firms, and the implementation of industrial policies is likely to exacerbate the disincentive effect of agglomeration on trade modal upgrading. That is, excessive industrial agglomeration does not significantly facilitate the transition of firms from processing trade to general trade, which is higher up the chain [27,28].

2.2. Productivity and Exports

Industrial agglomeration can effectively reduce the transaction costs of foreign trade and accelerate commodity import and exports and the cross-border flow of factors, thus promoting the improvement of enterprise productivity and ultimately promoting economic growth [29]. Exports help enterprises import and export more kinds of foreign products, intensify the competition in the domestic market, promote the survival of the fittest in the domestic market and thus accelerate the improvement of enterprises’ own productivity [30]. In addition, exports also accelerate the entry of foreign intermediate inputs with high quality and high technology content, thus replacing domestic inputs, generating higher economic output through the technology spillover effect, promoting the growth of the output value of related industries and further promoting the improvement of enterprise productivity [31].

The results of several studies have found that firms that enter export markets generally have higher productivity [32] and that once a firm enters a foreign market, it is able to learn the advanced production technology and management experience of foreign firms, which will help to further increase the productivity level of the firm [33]. This can also cause the productivity gap between exporting firms and their domestic counterparts to widen further [34,35], while higher productivity has a positive contribution to a firm’s likelihood of being a good exporter [36], and the greater the likelihood of exporting to another market [37]. Other scholars have found through their research that firms must pay fixed entry costs to access foreign markets and that firms must have sufficient mobility to export rather than productivity [38]. Exporting firms are generally more productive than non-exporting firms before exporting begins, and exporting raises productivity growth, but this positive effect soon disappears [39]. Due to local protection and spillover effects, long-term development can lead to lower productivity in some exporting firms than in non-exporting firms, especially in foreign affiliates in developing countries, which are very common [40,41].

The existing research has contributed to the development of theories related to industrial agglomeration and productivity on exports, but there are still shortcomings. Firstly, most studies have looked at exports solely from the perspective of industrial agglomeration or productivity, with few linking the three to consider their interactive effects on each other [42,43]. Secondly, most studies have sunk to the micro level of enterprises, and there are few studies at the macro-regional level [44]. Finally, most studies focus on an empirical analysis of the positive or negative impacts of industrial agglomeration or productivity on trade, but few scholars have explored the long-term impact trends in depth.

The next section of this paper discusses the relationship between industrial agglomeration, labour productivity and their interaction terms with exports by constructing a theoretical model for analysing the impact of industrial agglomeration and labour productivity on exports, and also formulates the research hypotheses.

3. Methodology

The traditional models of international trade theory, the Ricardian model and the Heckscher–Ohlin model, demonstrate the comparative advantages based on production costs from the perspective of labour productivity and factor endowment, respectively. However, the traditional theoretical models of international trade do not consider the role and function of enterprises in international trade. The early 1980s saw the emergence of a “new international trade theory” represented by the Krugman model. The Krugman model considers the role and effect of enterprises, with economies of scale and imperfect competition directly affecting the choice of behaviors of enterprises, but this model assumes that enterprises are “homogenous”, that is, there is no difference between enterprises. The “Heterogeneous firm model” was developed by Melitz, driven by Krugman’s “New international trade theory” and empirical micro-trade research [5]. By introducing the heterogeneity of firms in terms of productivity, Melitz further portrayed the superiority and inferiority effects of trade, thus formally creating the “New new international trade theory”. Since then, the relationship between productivity and firm behaviour has become an important topic in international trade research. Based on the enterprise trade model in Melitz’s new international trade theory [5], this paper expands the theoretical model and puts forward a research hypothesis.

3.1. Industrial Agglomeration and Exports

In this paper, according to the research model of Helpman [45], the enterprise export expected profit objective function is obtained:

$$\mathrm{E}\left(\pi \left(x\right)\right)={\phi }^m\left[p\left(x\right)q\left(x\right)-\frac{\tau q\left(x\right)}{x}\right]+\left(1-{\phi }^m\right)\left[p\left(x\right)q\left(x\right)-s\left(\theta \right)\frac{\tau q\left(x\right)}{x}\right]-f$$

(1)

In Equation (1): $p\left(x\right)$ is x commodity price and q($x$) is x commodity quantity. $\phi$ is the probability that the workforce will perform its own work in pursuit of the desired wage and m denotes the type of skill attribute of the workforce. $\tau (\tau >0)$ means the transportation cost of the enterprise, and f means that the enterprise must invest a certain sunk cost to enter the export market. θ indicates the degree of industrial agglomeration of economic activities in a region. The higher the degree of industrial agglomeration, the more likely it is that firms will gain information from spillover effects and reduce the adjustment costs of exporters $s\left(\theta \right)$. At the same time, when industrial agglomeration increases to a certain level, the rate at which the adjustment cost $s\left(\theta \right)$ of exporters decreases as θ increases will slow down. $s\left(\theta \right)$ is the concave function, so set: $s^{\prime }\left(\theta \right)<0, s^{″}\left(\theta \right)>0$.

Taking the first order derivative of Equation (1) yields the expected number of exports by the firm as follows:

$$\mathrm{E}\left(q\left(x\right)\right)=\left[{\phi }^m+\left(1-{\phi }^m\right)s{\left(\theta \right)}^{-\sigma }\right]a\left(x\right)$$

(2)

where $a\left(x\right)=L^{*}[{\frac{\sigma \tau }{\left(\sigma -1\right)x{p}^{*}}]}^{-\sigma }$, $\left(\sigma >1\right), L^{*}$ and $p^{*}$ denotes the quantity of foreign labour supply and the total price index, respectively.

Taking the first order partial derivative of θ in Equation (2) yields:

$$\frac{\partial \mathrm{E}\left(q\left(x\right)\right)}{\partial \theta }=\frac{\partial \left[{\phi }^m+\left(1-{\phi }^m\right)s{\left(\theta \right)}^{-\sigma }\right]a\left(x\right)}{\partial \theta }=\left(-\sigma \right)a\left(x\right)\left(1-{\phi }^m\right)s{\left(\theta \right)}^{-\sigma -1}{s}^{\text{'}}\left(\theta \right)$$

(3)

according to the assumptions. Due to $\sigma >1,a\left(x\right)>0,s\left(\theta \right)>0,s^{\prime }\left(\theta \right)<0,0<\phi <1$, we can obtain

$$\frac{\partial \mathrm{E}\left(q\left(x\right)\right)}{\partial \theta }>0$$

(4)

From Equation (4), it can be deduced that industrial agglomeration promotes exports to a certain extent.

Based on the existing theories and literature, the reasons can be summarized in the following aspects: Firstly, the agglomeration effect not only makes it easier to use increasing returns to scale to reduce corporate cost bonuses [45], but it can also offset the negative effects of labour cost pressure through the labour pool effect [46]. Secondly, the concentration of enterprises in the region can not only promote the formation of economies of scope in the industry to a certain extent, and increase the total output, but also increase the productivity of some enterprises in the cluster through the spillover effect of the flagship enterprises in the cluster. Thirdly, appropriate industrial agglomeration will also be conducive to the formation of a good competitive environment, forcing enterprises to develop themselves, increase enterprise output and promote exports.

Further, taking the second order partial derivative of θ in Equation (2) yields:

$$\begin{array}{l}\frac{{\partial }^2\mathrm{E}\left(q\left(x\right)\right)}{\partial {\theta }^2}=\frac{{\partial }^2\left[{\phi }^m+\left(1-{\phi }^m\right)s{\left(\theta \right)}^{-\sigma }\right]a\left(x\right)}{\partial {\theta }^2}\\ =\left(-\sigma \right)a\left(x\right)\left(1-{\phi }^m\right)\left[-\left(\sigma +1\right)s{\left(\theta \right)}^{-\sigma -2}{s}^{\prime }\left(\theta \right)+s{\left(\theta \right)}^{-\sigma -1}{s}^{″}\left(\theta \right)\right]\end{array}$$

(5)

according to the assumptions. Due to $\sigma >1,a\left(x\right)>0,s\left(\theta \right)>0,s^{\prime }\left(\theta \right)<0,0<\phi <1,s^{″}\left(\theta \right)>0$, we can obtain

$$\frac{{\partial }^2\mathrm{E}\left(q\left(x\right)\right)}{\partial {\theta }^2}>0$$

(6)

From Equation (6), it can be deduced that as agglomeration continues to rise to a certain level, the promotion effect of industrial agglomeration on exports becomes smaller and smaller until it turns into a suppressing effect.

Synthesizing existing theory and literature, there are several reasons why excessive industrial agglomeration inhibits exports. First: the increased cost of monopolies. Excessive agglomeration tends to create monopolies in industries, and economies of scale and natural monopolies are the main problems that cause market failure [47]. Regional local monopolies also lead to different degrees of market fragmentation in each region, which in turn increases the cost of entering foreign markets [48]. Second: the increase in comprehensive costs. While industrial agglomeration brings about regional economic prosperity, it also brings about a series of problems such as increased taxation, rising prices and higher per capita income, which in turn leads to an increase in various costs and expenses such as prices and labour, resulting in an increase in the comprehensive costs of enterprises [49], which, to a certain extent, restricts exports. Third: the increase in environmental protection costs. On the one hand, with the increase in industrial agglomeration, many enterprises produce a large number of pollutants in a short period of time and accumulate them, leading to various environmental problems, and green production without affecting the environment then requires enterprises to pay additional environmental protection costs [50]. On the other hand, during the period when industries were not clustered and the level of economic development was low, the state did not pay enough attention to environmental protection, and once the regional economy developed and environmental awareness rose, the cost of environmental protection skyrocketed, which in turn increased the production costs of enterprises and, to a certain extent, weakened their exports [51]. Fourth: the decline in international competitiveness. The part of the industry that has low-cost advantage development is large but not strong, and because of policy dependence, innovation becoming weaker and the comparative cost advantage decreasing, industrial upgrading has become difficult, resulting in a decline in international competitiveness [52] and a decline in exports. Even if further mergers and integration may occur in China to expand the production scale, it cannot recover the decline in exports. Therefore, we propose the following hypothesis:

Hypothesis 1.

Industrial agglomeration has a boosting and then inhibiting effect on exports, with an inverted U-shaped curve.

3.2. Productivity and Exports

Based on previous research, we then analysed the relationship between productivity and exports in three dimensions in depth and formulated research hypotheses.

3.2.1. Optimal Consumption and Expenditure Decision

According to the heterogeneous enterprise trade model in Melitz’s trade theory [5], consumer preference is determined by the utility function of product ω, $U={\left[{{\displaystyle \int }}_{\omega \in \Omega }q{(\omega )}^{\rho }d\omega \right]}^{\frac{1}{\rho }}$, $\Omega$ is the number of products and ρ is the alternatives between products, $0<\rho <1$.

The optimal product consumption and expenditure decision model is as follows:

$$q\left(\omega \right)=Q{\left[\frac{p\left(\omega \right)}{p}\right]}^{-\sigma },r\left(\omega \right)=R{\left[\frac{p\left(\omega \right)}{p}\right]}^{1-\sigma }$$

(7)

where $q\left(\omega \right)$ is the output function of the enterprise, $r\left(\omega \right)$ is the profit function of the enterprise and R is the income of the enterprise, $R=PQ={{\displaystyle \int }}_{\omega \in \Omega }r\left(\omega \right)d\left(\omega \right)$.

3.2.2. Enterprise Productivity

Supposing that the enterprise will choose to produce different products $\omega$, there is only $l$ labour factor input in production, and the input is a function of output q: $l=f+q/\varphi$, the enterprise has fixed costs $f>0$, without difference, but different enterprises have different productivity levels. Therefore, when the firm has a residual demand function with elasticity of σ, the firm will choose to maximize the profit plus $\frac{\sigma }{\left(\sigma -1\right)=\frac{1}{\rho }}$.

Therefore, the optimal pricing is

$$p\left(\phi \right)=\frac{\omega }{\rho \phi }$$

(8)

where, $\omega$ is the wage rate, and the profit of the enterprise is

$$\pi \left(\mathsf{\Phi }\right)=r\left(\mathsf{\Phi }\right)-l\left(\mathsf{\Phi }\right)=\frac{r\left(\mathsf{\Phi }\right)}{\sigma }-f$$

(9)

The output ratio of the enterprise is reflected in the productivity ratio:

$$\frac{q\left({\mathsf{\Phi }}_1\right)}{q\left({\mathsf{\Phi }}_2\right)}={\left(\frac{{\mathsf{\Phi }}_1}{{\mathsf{\Phi }}_2}\right)}^{\sigma },\frac{r\left({\mathsf{\Phi }}_1\right)}{r\left({\mathsf{\Phi }}_2\right)}={\left(\frac{{\mathsf{\Phi }}_1}{{\mathsf{\Phi }}_2}\right)}^{\sigma -1}$$

(10)

Therefore, the higher the productivity of an enterprise, the larger the scale and the higher the profit.

3.2.3. Entry and Exit of Enterprises

When enterprises find that an industry has profits, they will enter one after another and fully compete until the industry profit is zero. Assuming that the enterprise inputs unit cost ${\mathrm{f}}_{\mathrm{e}}$, the production efficiency is $\mathsf{\phi }$ ($\mathsf{\phi }$ is a random variable, obeying the distribution function ${\mathrm{G}}_{\mathrm{c}}\left(\mathsf{\phi }\right)$, and the density function is ${\mathrm{g}}_{\mathrm{i}}\left(\mathsf{\phi }\right)$). The possibility that an enterprise will be forced to withdraw from the market due to a negative impact in the production process is δ, the enterprise’s profit per period is π and the expected profit in the whole life cycle is

$${{\displaystyle \sum _{t=0}^{\infty }\left(1-\delta \right)}}^t\pi \left(\mathsf{\Phi }\right)=\frac{\pi \left(\mathsf{\Phi }\right)}{\delta }$$

(11)

Therefore, when an enterprise just enters the market, the conditions for deciding whether to withdraw from the market or continue production and operation meet the following equation.

$$v\left(\mathsf{\Phi }\right)=max\left\{0,\frac{\pi \left(\mathsf{\Phi }\right)}{\delta }\right\}$$

(12)

The minimum productivity level for continuous production and operation is ${\mathsf{\varphi }}^{*}=\inf \left\{\mathsf{\varphi }:\nu \left(\mathsf{\varphi }\right)>0\right\}$; ${\mathsf{\varphi }}^{*}$ is the boundary level when the industry is zero profit.

Through the derivation of the above formula, it is not difficult to see that, theoretically, enterprises can export only when their productivity exceeds the critical value, and the higher the productivity, the larger the export scale, and the enterprises with low productivity only serve the domestic market [19]. Therefore, we can make the following assumptions:

Hypothesis 2.

Labour productivity has a catalytic effect on exports.

3.3. Interaction of Industrial Agglomeration, Labour Productivity and Exports

Industrial agglomeration can help enterprises share specialized inputs for export production and transportation facilities for import and export and reduce export costs, and thus promote enterprises to improve productivity. Secondly, industrial agglomeration attracts a specialized labour force for export enterprises, and reduces the cost of labour search and training for export by individual enterprises, which is conducive to improving the productivity of enterprises. In addition, within the scope of industrial agglomeration, face-to-face communication and human capital flow are conducive to the dissemination of specialized knowledge and management methods needed to enter the international market, which greatly promotes the improvement of enterprise productivity. Potter and Watts found that technology-related industries also have specialized labour markets, specialized supplier markets and knowledge spillovers [53].

In regions with higher productivity, the economy is relatively prosperous and exports have reached a considerable scale. Under the influence of comprehensive and complex conditions, such as higher price levels and stronger environmental awareness, industrial agglomeration, as one of the many factors influencing exports, has a relatively limited and sometimes insignificant impact [39]. Conversely, in regions with lower productivity, the potential for export trade is greater, and lower labour costs and less awareness of environmental protection have led to lower production costs and thus relatively greater scope for industrial agglomeration to enhance exports [54]. At this point, the impact of industrial agglomeration on exports is relatively more pronounced [55]. Therefore, we propose the following hypothesis:

Hypothesis 3.

The relationship between industrial agglomeration and exports is negatively influenced by labour productivity.

4. Research Design

4.1. Model Design

According to the new trade theory and existing research results, this paper includes the control variables affecting the export performance of enterprises, and constructs the production function as follows:

$$\ln Expor{t}_{it}=\alpha +\beta Ag{g}_{it}+\gamma L{p}_{it}+\delta Ag{g}_{it}\ast L{p}_{it}+{\displaystyle \sum {\rho }_{it}}Contro{l}_{it}+{\mu }_{it}$$

(13)

where i represents the region, t represents the year, $Export$ represents the volume of industrial exports, $Agg$ represents the degree of industrial agglomeration and $Lp$ represents the industrial labour productivity. ${\mu }_{it}$ represents the random disturbance term and $Control$ represents the control variable.

$$Contro{l}_{it}=\ln Outpu{t}_{it}+\ln N_{it}+\ln Flagshi{p}_{it}+\ln F{e}_{it}+\ln Es{u}_{it}$$

(14)

where $Output$ represents the total output of the industry, $N$ represents the number of enterprises in the industry, $\mathit{Flagship}$ represents the number of flagship enterprises in the region, $Fe$ represents the expenditure in the financial budget and $Esu$ represents expenditures on scientific matters in the region.

4.2. Variable Data and Descriptive Statistics

4.2.1. Explained Variable

Exports refers to the sale of goods produced or processed in one’s own country to overseas markets. It is measured by the total annual industrial export delivery value in the region.

4.2.2. Explanatory Variables

Industrial agglomeration. Based on the practice of Ellison [56] and the idea of location entropy, this paper uses the number of employees in regional industries to obtain the regional specialization index. The specific indicators and descriptions are as follows:

$$L{Q}_{ij}=\frac{\raisebox{1ex}{$q_{ij}$}\!\left/ \!\raisebox{-1ex}{${\displaystyle \sum _{i=1}^m{q}_{ij}}$}\right.}{\raisebox{1ex}{${\displaystyle \sum _{j=1}^n{q}_{ij}}$}\!\left/ \!\raisebox{-1ex}{${\displaystyle \sum _{i=1}^m{\displaystyle \sum _{j=1}^n{q}_{ij}}}$}\right.}$$

(15)

In the above equation. $L{Q}_{ij}$ is the location entropy of industry i in region j, which is an index reflecting the degree of specialization of industry i in this region, $q_{ij}$ is the number of employees in industry i in region j, $q_j$ is the total number of employees in area j and $q_i$ is the number of employees in i industry nationwide.

Labour productivity is the ratio of the labour output created by workers in a certain period of time to the amount of labour consumed in relation to it. In this paper, we use total labour productivity, which refers to the average number of products produced per unit of time by each employee, calculated according to the value of the product. Currently, China’s total labour productivity is calculated by dividing the value added of an enterprise’s industry by the average number of all the employees in the same period. This paper therefore draws on Lu’s approach [57] and uses the total industrial output value divided by the number of industrial employees in the region to express labour productivity.

4.2.3. Control Variables

According to the above theoretical analysis and the existing research results, this paper intends to add the following control variables. The total industrial output (output) is the measurement of the total annual output value in the region. The number of enterprises (N) is the total number of annual industrial enterprises in the selected region. The number of flagship (Flagship) enterprises is the number of national top 500 enterprises in the region. Financial expenditure (Fe) is selected as the total expenditure within the annual local budget. Expenditure on scientific undertakings (Esu) is chosen as the total amount of local financial expenditure on scientific undertakings. Descriptive statistics for the variables are shown in

Table 1. Descriptive statistics of variables.

Variable	Obs	Mean	Std. Dev.	Min	Max
Exports (100 million dollar)	280	81.70	115.99	1.37	516.41
Industrial agglomeration (percentage)	280	0.93	0.54	0.17	2.90
Labour productivity (10,000 yuan/person)	280	12.22	24.01	0. 30	200.77
Industrial output (10,000 yuan)	280	796.27	555.96	37.06	2533.77
Number of enterprises	280	13,738.27	14,645.01	337.00	65,495.00
Flagship enterprises	280	17.31	20.24	1.00	101.00
Financial expenditure (ten thousand yuan)	280	789,390.00	406,689.90	28,747.58	2,017,400.00
Expenditure for scientific (ten thousand yuan)	280	136,016.70	63,694.56	5371.76	257,834.60

.

Table 2. Correlation coefficient matrix.

Variable	EX	IA	LP	IO	NE	FEN	FEX	ES
Exports (EX)	1.00
Industrial Agglomeration (IA)	0.38	1.00
Labour productivity (LP)	0.51	−0.42	1.00
Industrial Output (IO)	−0.18	0.59 *	0.39	1.00
Number of Enterprises (NE)	−0.01	−0.27 *	0.44	−0.41 *	1.00
Flagship Enterprises (FEN)	0.09	0.31 *	−0.18	0.43 *	−0.38	1.00
Financial Expenditure (FEX)	0.20 *	0.36 *	0.52	0.49 *	−0.28 *	0.60	1.00
Expenditure for Scientific (ES)	0.24 *	0.51 *	0.43	0.42 *	−0.39 *	0.18 *	0.55 *	1.00

Note: * p < 0.05, the absolute value of the correlation coefficient between variables is between [0.01 and 0.60]. At the same time, the variance inflation factor (VIF) is calculated. The average value of the variance inflation factor of each variable is 3.12, and the maximum value is 4.65 (<10). It can be considered that the degree of multicollinearity among variables is relatively acceptable.

shows the correlation test, and the results show that there is no multicollinearity between the variables.

4.3. Data Source

The data used in the empirical analysis of this paper were obtained from the China Industrial Statistical Yearbook, the China Urban Statistical Yearbook and the China Top 500 Enterprises in a Decade of Style from 2008 to 2017. Data from 2008 to 2017 were chosen for this study for two main reasons. The first is because some of the relevant data after 2018 have not yet been published by the Chinese government. The second is that the outbreak of a trade war between China and the US in 2018 and the fact that the US has long been China’s top export destination has led to huge fluctuations in some of the data involving export trade, causing greater interference in the analysis. For the above two reasons, this paper selects data from 2008 to 2017 for empirical analysis. All nominal variables involved in the model are adjusted using the GNP index for the base period of 1978 to transform the nominal variables into real variables. Missing data in the yearbook are also interpolated. The explanatory and control variables are logarithmised in the empirical evidence below in order to standardise the scale of the data.

5. Empirical Analysis

5.1. Benchmark Regression Model Estimates Results

This section empirically tests the relationship between industrial agglomeration, labour productivity and exports. The results are shown in

Table 3. Benchmark regression results.

Variable	Model I	Model II	Model III	Model IV
	lnExport	lnExport	lnExport	lnExport
Agg		0.3910 ***	1.9090 ***	0.5417 ***
		(3.24)	(5.34)	(4.50)
Lp		0.4430 ***	0.4621 ***	0.4580 ***
		(6.69)	(6.70)	(6.34)
Agg*Agg			−0.4172 ***
			(−5.61)
Agg*Lp				−0.2212 ***
				(−2.76)
lnOutput	0.9644 ***	0.7092 ***	0.5433 ***	0.6944 ***
	(5.54)	(4.54)	(3.98)	(5.24)
lnN	−0.0923	−0.1525 **	−0.1815 **	−0.1560 **
	(−0.81)	(−2.16)	(−2.45)	(−2.16)
lnFlagship	0.0808 *	0.0679 **	0.0754 **	0.0783 **
	(1.78)	(2.02)	(2.37)	(2.17)
lnFe	−0.6520 ***	−0.4322 ***	−0.3738 ***	−0.3940 ***
	(−3.37)	(−3.50)	(−3.00)	(−3.41)
lnEsu	0.4682 **	0.3205 ***	0.2576 **	0.2862 **
	(2.51)	(3.24)	(2.11)	(2.55)
Fix Region	Yes	Yes	Yes	Yes
Fix Time	Yes	Yes	Yes	Yes
Wald chi2	536,198.27	1,038,762.42	642,027.42	389,714.29
Prob > chi2	0.00	0.00	0.00	0.00

Note: *, **, *** mean significance at the levels of 10%, 5% and 1%, respectively, and the number in parentheses is the “t” value of the estimated coefficient.

.

Model II tests the impact of industrial agglomeration and labour productivity on exports. Model II shows that the industrial agglomeration and labour productivity coefficients are significantly positive, indicating that China’s industrial agglomeration and labour productivity have significantly promoted industrial exports at this stage, which verifies the hypothesis H2. Model III adds the square term of industrial agglomeration on the basis of model II; the square term coefficient is significantly negative and the agglomeration coefficient is significantly positive, indicating that the impact of industrial agglomeration on exports is first promoted and then suppressed and the curve is in an inverted U shape, which verifies hypothesis H1.

Model IV is a test of the interaction between industrial agglomeration, labour productivity and exports. The results of model IV show that the coefficients of industrial agglomeration and labour productivity are significantly positive, but the coefficients of their interaction terms are significantly negative, indicating the negative moderator effect of labour productivity on industrial agglomeration and exports, that is, compared with regions with high labour productivity, the positive impact of industrial agglomeration on exports in regions with low labour productivity is more obvious. The results support hypothesis H3.

Model I was designed to test the effect of the control variables on the dependent variable. The control variable in benchmark model I shows that the coefficients of industrial output, the number of flagship enterprises and the science expenditure are significantly positive, indicating that industrial output, the number of flagship enterprises and science expenditure have a significant positive impact on exports, which is also in line with China’s national conditions. Firstly, industrial output is one of the key factors affecting industrial exports. Only on the premise that the output is enough to meet domestic demand, will enterprises further develop exports. Secondly, the research shows that flagship enterprises will have a certain “spillover effect” on enterprises in the cluster, actively promoting enterprise innovation and regional economic development, which will then affect the industry exports [58]. Finally, a prosperous economic environment and stable investment in scientific research can improve the productivity of enterprises, and then play a positive role in exports.

The number of industrial enterprises and the financial budget expenditure are negatively related to exports. The reason for this may be the unreasonable distribution of enterprises and the trade distribution of enterprises in China at this stage, the output structure of enterprises cannot meet the needs of domestic and foreign markets, and the labour cost increases year by year, which has a negative impact on exports. The budget expenditure comes from tax. The increase in tax will increase the burden of enterprises, which will affect the productivity of enterprises, and then affect the exports.

5.2. Feasible Generalised Least Squares Regression Estimates Results

In order to verify the robustness of the benchmark regression results, this paper uses the feasible generalised least squares (FGLS) method to test the panel data of 28 provinces in China from 2008 to 2017, and the empirical results are presented in

Table 4. Feasible generalised least squares regression results.

Variable	Model I	Model II	Model III	Model IV
	lnExport	lnExport	lnExport	lnExport
Agg		0.6231 ***	2.5302 ***	0.7557 ***
		(4.12)	(6.69)	(4.63)
Lp		0.4221 ***	0.4134 ***	0.4332 ***
		(5.62)	(5.79)	(5.80)
Agg*Agg			−0.5269 ***
			(−5.45)
Agg*Lp				−0.2163 **
				(−2.07)
lnOutput	0.9037 ***	0.6917 ***	0.3903 ***	0.6655 ***
	(7.02)	(5.23)	(2.84)	(5.05)
lnN	0.0942	−0.1249	−0.2278 **	−0.1271
	(0.95)	(−1.24)	(−2.33)	(−1.27)
lnFlagship	0.1138 **	0.1184 **	0.1162 **	0.1350 **
	(1.97)	(2.26)	(2.34)	(2.57)
lnFe	−0.8925 ***	−0.6460 ***	−0.4746 **	−0.5855 ***
	(−4.24)	(−3.33)	(−2.53)	(−3.01)
lnEsu	0.6121 ***	0.4779 ***	0.3266 *	0.4217 **
	(3.15)	(2.69)	(1.91)	(2.36)
Fix Region	Yes	Yes	Yes	Yes
Fix Time	Yes	Yes	Yes	Yes
Wald chi2	9950.47	12,188.34	13,510.97	12,379.36
Prob > chi2	0.00	0.00	0.00	0.00

Note: *, **, *** mean significant at the levels of 10%, 5% and 1%, respectively, and the number in parentheses is the “t” value of the estimated coefficient.

below.

The robustness tests show that the results of all the above tests are basically consistent with the results of the benchmark regression, indicating that the results are robust. This paper repeatedly intercepted five consecutive years of panel data during 2008–2017 and tested the benchmark regression model using short panel fixed effects, and the results were also largely consistent, fully indicating that the benchmark regression results are more robust.

5.3. Dynamic Panel Regression Model Estimation Results

Static panels are more likely to ignore the influence of factors such as social environment and economic policies, which play an extremely important role in socio-economic development. In this paper, considering the inevitable endogeneity problem of static panels, we propose to introduce the first-order lagged term of exports to construct a dynamic panel model (differential GMM) to correct some of the biases in the results of the static panel model, and the specific results are shown in

Table 5. Dynamic panel estimation results (differential GMM).

Variable	Model I	Model II	Model III	Model IV
	lnExport	lnExport	lnExport	lnExport
lnExportt−1	0.3482 **	0.4048 **	0.2841 ***	0.2825 ***
	(2.12)	(2.15)	(2.76)	(2.85)
Agg		0.2883 *	0.4390	0.9031
		(1.79)	(1.21)	(1.16)
Lp		2.0512 **	4.2140 *	7.9412 **
		(2.21)	(1.81)	(2.46)
Agg*Agg			−0.0030
			(−0.97)
Agg*Lp				−0.0114 *
				(−1.85)
lnOutput	0.4834 ***	0.3532 ***	0.3994 **	0.4148 **
	(5.19)	(4.02)	(2.31)	(2.51)
lnN	0.0167	−0.0783	−0.1519	−0.1898 *
	(0.15)	(−0.43)	(−1.43)	(−1.77)
lnFlagship	0.0700 **	0.0685 ***	0.0615	0.0588
	(2.46)	(2.66)	(1.54)	(1.48)
lnFe	−0.4115 **	−0.3356 *	−0.3967 ***	−0.3579 **
	(−2.14)	(1.71)	(−2.66)	(−2.39)
lnEsu	0.2330	0.1774	0.2617 *	0.2512 *
	(0.94)	(1.22)	(1.87)	(1.81)
AR(1)	0.3037	0.2903	0.2873	0.2307
AR(2)	0.3021	0.3932	0.4026	0.4898
Sargan	26.64	19.82	20.17	18.02
p-value	0.0518	0.1790	0.1656	0.2617

Note: *, **, *** mean significant at the levels of 10%, 5% and 1%, respectively, and the number in parentheses is the “t” value of the estimated coefficient.

.

The AR (1) and AR (2) in Model I–Model IV are both greater than 0.05 (rejecting the original assumption that the model disturbance terms are correlated), indicating that the first-order and second-order model disturbance terms do not have autocorrelation. The sargan test p values for judging the over-identification of instrumental variables are all greater than 0.05, indicating that the instrumental variables selected by the model are all valid. At a significance level of 5%, the first-phase lag term of exports is significant, and the coefficient is greater than 0, indicating that exports are meaningful, and the lag term of exports has a significant promoting effect on the current period.

In general, the statistical results of the dynamic panel data are basically consistent with the static panel, which fully demonstrates the robustness of the model setting, and also demonstrates the previous hypothesis from many aspects.

5.4. Regional Heterogeneity Model Estimation Results

Industrial agglomeration was first proposed by Mashall [59]. The theory holds that the impact of industrial agglomeration on the industry is due to its convenient transportation and spillover effect in the industry cluster. Based on this theory, this paper intends to further test the impact of regional heterogeneity on industrial agglomeration from the perspective of transportation convenience (whether the region contains coastal port cities) and cluster spillover (whether it contains industrial bases).

5.4.1. Heterogeneity Estimation Results Based on Coastal Port Areas

Coastal port cities have certain advantages for exports in terms of both export costs and export convenience due to their convenient locations. Therefore, this paper intends to divide China’s 28 provinces into coastal port areas (Tianjin, Hebei, Liaoning, Shanghai, Jiangsu, Zhejiang, Fujian, Shandong, Guangdong, Guangxi and Hainan) and non-coastal port areas (Beijing, Shanxi, Inner Mongolia, Jilin, Heilongjiang, Anhui, Jiangxi, Henan, Hubei, Hunan, Chongqing, Sichuan, Guizhou, Yunnan, Shaanxi, Gansu and Xinjiang). The regional dummy variables (set to 1 for coastal ports and 0 for non-coastal ports) are set to further examine whether regional differences have any effect on industrial agglomeration and exports.

The results in

Table 6. Regional heterogeneity regression estimation results.

Variable	Model I	Model II	Model III
	Coastal Port Areas	Non-Coastal Port Areas	Overall
	lnExport	lnExport	lnExport
Agg	0.1609 **	1.3793 ***	0.4549 ***
	(2.17)	(7.73)	(3.39)
Lp	0.3412 ***	0.4860 ***	0.4391 ***
	(7.21)	(12.72)	(6.69)
lnOutput	0.5663 ***	0.7045 ***	0.7200 ***
	(4.93)	(7.31)	(5.77)
lnN	−0.0590	−0.2469 ***	−0.1522 **
	(−1.00)	(−4.32)	(−2.14)
lnFlagship	0.0608	0.1251 **	0.0716 **
	(1.63)	(2.13)	(2.12)
lnFe	0.0890	−0.7866 ***	−0.4434 ***
	(0.85)	(−5.18)	(−3.54)
lnEsu	0.0475	0.5931 ***	0.3267 ***
	(0.34)	(4.37)	(2.76)
Fix Region	Yes	Yes	Yes
Fix Time	Yes	Yes	Yes
Wald chi2	26,302.20	119,279.01	433,930.93
Prob > chi2	0.00	0.00	0.00

Note: *, **, *** mean significant at the levels of 10%, 5% and 1%, respectively, and the number in parentheses is the “t” value of the estimated coefficient.

show that industrial agglomeration and labour productivity significantly contribute to exports, whether one is in a coastal port region or a non-coastal port region. The interaction term between industrial agglomeration and the regional dummy variable is negative, but not significant.

5.4.2. Heterogeneity Estimation Results Based on Industrial Base Areas

To assess regional differences, we estimated whether there is a differentiated impact of industrial agglomeration on exports from the perspective of industrial bases. In this paper, the provinces and cities involved in the four major industrial bases (Beijing, Tianjin, Hebei, Liaoning, Shanghai, Jiangsu, Zhejiang, Anhui and Guangdong) in Liaoning, Central and Southern China, Beijing, Tianjin and Tang, Yangtze River Delta and the Pearl River Delta. The other regions (Shanxi, Inner Mongolia, Jilin, Heilongjiang, Fujian, Jiangxi, Shandong, Henan, Hunan, Hubei, Guangxi, Hainan, Chongqing, Sichuan, Guizhou, Yunnan, Shaanxi, Gansu and Xinjiang) are non-industrial base areas to further test whether regional heterogeneity has any impact on industrial agglomeration and exports.

The results in

Table 7. Regression estimates for analyzed regions.

Variable	Model I	Model II	Model III
	Industrial Base Areas	Non-Industrial Base Areas	Overall
	lnExport	lnExport	lnExport
Agg	0.0082	1.1275 ***	0.8211 ***
	(0.49)	(3.12)	(2.99)
Lp	0.2284 ***	0.3916 ***	0.3308 ***
	(4.08)	(3.46)	(4.11)
lnOutput	0.6133 ***	0.5226 ***	0.6062 ***
	(5.81)	(3.97)	(4.68)
lnN	0.0692	−0.2286 ***	−0.1842 *
	(0.90)	(−3.61)	(−1.90)
lnFlagship	0.0719 *	0.0609 **	0.0801 **
	(1.92)	(2.00)	(2.29)
lnFe	−0.1412	−0.5107 ***	−0.4018 ***
	(−1.39)	(−3.08)	(−3.17)
lnEsu	0.0712	0.3812 ***	0.2286 **
	(0.80)	(2.98)	(2.39)
Wald chi2	71,231.32	683,651.40	451,292.08
Prob > chi2	0.00	0.00	0.00

Note: *, **, *** mean significant at the levels of 10%, 5% and 1%, respectively, and the number in parentheses is the “t” value of the estimated coefficient.

show that labour productivity significantly contributes to exports in both industrial base and non-industrial base regions, but industrial agglomeration only presents a significant contribution in non-industrial base regions, and not within industrial base clusters.

6. Discussions

This article focuses on the interrelationships between industrial agglomeration, port productivity and exports, and conducts an in-depth analysis of their impact mechanism testing and heterogeneity testing. Based on the research findings of this article, we will discuss five aspects.

Firstly, there is a positive correlation between industrial agglomeration and exports, indicating that industrial agglomeration has a certain promoting effect on exports. On the one hand, the economies of scale brought about by industrial agglomeration can reduce transportation costs and raw material procurement costs within the region, thereby giving regional enterprises corresponding price advantages in export competition. On the other hand, industrial agglomeration can drive the overall improvement of technological level of enterprises within the cluster through the spillover effect of the technological level, thereby improving the competitiveness of export products [60]. Therefore, China should gradually change its previous model of competing with individual enterprises for the entire global value chain, shift to competing with clusters, improve the competitiveness of export products from the inside out, comprehensively enhance the international perspective of industrial and manufacturing labour and form a development pattern of mutual promotion between the international and domestic markets as soon as possible.

Secondly, the impact curve of industrial agglomeration on industrial enterprises’ exports shows an inverted U-shape, indicating that industrial agglomeration has a promoting effect on exports, but the long-term excessive agglomeration effect will have a negative impact on exports. The main reason for this is that, in the early stages of industrial agglomeration, there are significant technological differences among enterprises within the cluster. Through technology spillovers, the overall technological level of enterprises can be improved, which is conducive to export competition. As the technological gap between enterprises within the cluster gradually narrows, the spillover effects of industrial agglomeration also decrease. In addition, there is a large amount of vicious competition among homogeneous enterprises, which to some extent hinders the exports of enterprises [40]. Therefore, while guiding enterprise agglomeration, the government should focus on grasping the agglomeration situation, adjust policies in a timely manner and avoid vicious competition and the diffusion of unnecessary agglomeration.

Thirdly, labour productivity promotes trade exports, while the relationship between industrial agglomeration and trade exports is negatively affected by labour productivity. On the one hand, the improvement of labour productivity can reduce the production costs of enterprises, thereby promoting their export competitiveness [34]. On the other hand, industrial agglomeration not only affects exports through technology spillovers, but also the talent flow brought by agglomeration is an important factor affecting the competitiveness of enterprises. In areas with low labour productivity, labour is relatively scarce, and there are significant differences in the quantity and quality of labour between enterprises. Industrial agglomeration attracts a large amount of foreign labour, which can effectively supplement the local labour shortage and low efficiency, which is beneficial for enterprise production and thus improving enterprise competitiveness. In areas with high labour productivity, most enterprises have a sufficient labour force and relatively fixed personnel. Many labour productivity bottlenecks are reached, and the foreign labour attracted by industrial agglomeration does not provide substantial assistance to local enterprises. On the contrary, it can also cause vicious competition among personnel and reduce efficiency. Therefore, enterprises should combine their own situations and work together to arrange personnel introduction, comprehensively improve labour productivity, promote the interactive effect of talent and industry agglomeration and fully solve the problem of regional human resource allocation.

Fourthly, whether located in coastal or non-coastal port areas, industrial agglomeration and labour productivity have a significant promoting effect on exports. This is inconsistent with the conclusion of Mashall’s theory. The main reason for this is that China’s transportation and logistics industry has developed to become quite mature in recent years, achieving efficient and low-cost operations, to a certain extent. At the same time, with the implementation of national supply side reform policies and the rapid development of the Internet of Things, the efficiency of domestic product resource allocation has greatly improved, resulting in the cost and time cost of inland product transportation to coastal cities being almost negligible compared with the cost of the product itself and export costs. Therefore, whether it is coastal ports or inland regions, industrial agglomeration has a significant impact on exports [61]. This has also, to some extent, reduced the overall cost of exports and further expanded China’s competitive advantage.

Finally, the impact of industrial agglomeration on exports is significant in non-industrial regions, but not in industrial regions. The main reason for this is the spillover effect of China’s industrial bases radiating outward within a certain range. However, in terms of themselves, due to their high degree of agglomeration and the highly competitive pressure within the cluster, the potential and impact of industrial agglomeration within the region are relatively limited. In areas outside the base, due to their low level of agglomeration and the impact of base spillover effects, the impact of industrial agglomeration is more quickly and effectively exerted, thereby driving industry development and increasing the competitiveness of exports. Therefore, the government should further accelerate the adjustment of industrial structure and industrial layout, accelerate technological updates, and promote the transformation and upgrading of leading industries to high-tech. We should also focus on completing the transfer of key industries and prepare for the formation of a new pattern of high-quality development.

7. Conclusions

Based on China’s 2008–2017 28 inter-provincial industrial panel data, this paper constructs static and dynamic models to study the relationship between industrial agglomeration, labour productivity and exports, and to explore the influence mechanism and heterogeneity among them. The results show that, firstly, industrial agglomeration is positively related to industrial exports in China at the present stage, and the quadratic curve is an inverted U shape, indicating that industrial agglomeration can promote exports in China to a certain extent at the present stage, and the long-term development will show a tendency of first promoting and then inhibiting it. Secondly, labour productivity is positively correlated with industrial exports, and the relationship between industrial agglomeration and exports is negatively influenced by labour productivity, i.e., in regions with low labour productivity, the development of industrial exports is significantly more influenced by industrial agglomeration than in regions with high labour productivity. Finally, based on the perspective of regional heterogeneity, industrial agglomeration does not have a significant impact on industrial exports in industrial base regions, while it has a significant relationship of promoting industrial exports in non-industrial base regions. Industrial agglomeration in coastal port areas and non-coastal port areas significantly contributes to industrial exports, and transport heterogeneity does not affect the effects of industrial agglomeration. This paper expects the findings of the study to provide insights for the government to formulate relevant policies and systems, and also to provide practical references for the sustainable development of enterprises.

Due to the unavailability of data on the relevant variables for some regions, this study analysed data from 2008 to 2017, taking into account the availability of data. The authors’ subsequent study will further extend the sample period through variable measurement and other methods. In addition, this study will consider the impact of multilateral trade on industrial agglomeration and the differences in the effects of industrial agglomeration policies on exports in different countries.