Threshold Effect of Manufacturing Agglomeration on Eco-Efficiency in the Yellow River Basin of China

Abstract

Research on the impact of industrial development on the ecological environment of the Yellow River Basin plays a significant role in accelerating the high-quality development of that key region of China. Since the impact of industrial agglomeration on eco-efficiency is very complex, this study constructs a panel threshold model of the impact of manufacturing agglomeration on eco-efficiency and analyzes the heterogeneity of different industries. The results led to the following conclusions: The optimal range for the industrial agglomeration level is 0.37 to 0.40. When the industrial agglomeration level is in that optimal range, the manufacturing agglomeration has a significant positive effect on eco-efficiency, and the eco-efficiency level increases by 2.87% for every 1% increase in the agglomeration level. The agglomeration of high-energy-consuming manufacturing has obvious negative externalities for eco-efficiency; however, this negative effect weakens after the threshold value is reached. However, the impact of the agglomeration of low- and medium-energy-consuming manufacturing industries on eco-efficiency is manifested as a significant positive effect, though when the agglomeration degree is low, the effect is not significant.

1. Introduction

The Yellow River Basin plays an important role in China’s ecological strategy. The rapid development of the Basin led to a series of ecological problems, such as resource shortages and ecological destruction, and these issues have greatly hindered its development. Improvement of the ecological environment and control of pollution emissions have become intrinsic requirements for its high-quality development. At this key moment in promoting the ecological protection and high-quality development of the Basin, it is of great practical significance to analyze the relationship between industrial development and the level of eco-efficiency in the Basin in order to promote the transformation and upgrading of industries and realize the coordinated development of ecology and economy. What are the levels of eco-efficiency and industrial agglomeration in the provinces of the Basin? What are the laws and characteristics of industrial agglomeration and eco-efficiency in time and space? As an important engine of economic development, what effect does industrial agglomeration have on eco-efficiency? The discussion of these questions can aid in effectively judging the ecological impact of industrial development. Analyzing the main factors that affect eco-efficiency in the Yellow River Basin plays a very important role in formulating the path of industrial transformation and upgrading, as well as in achieving coordinated development of the eco-economy.

At present, there are still some controversies about the environmental impact of industrial agglomerations. At the national level, industrial agglomeration can effectively improve environmental conditions. From the urban perspective, the agglomeration of industries, concentration of population, and expansion of urban scale will lead to an increase in production and total consumption (Cleveland et al., 1984) [1]. Frank et al. (2001) concluded that the scale of industrial agglomeration was significantly correlated with the air quality of a city based on an empirical analysis of environmental pollution data for 200 urban agglomeration areas in the European Union [2]. Verhoef et al. (2002) conducted an analysis of the relationship between industry and environmental pollution using a spatial balance model. When considering industrial layout as a factor in environmental pollution in industrial clusters, it was concluded that industrial clusters contribute to environmental pollution [3]. The negative externalities generated by industrial agglomeration lead to excessive consumption of resources, increased emissions of various pollutants, and reduced environmental carrying capacity, ultimately causing increased environmental pollution and increasing the difficulty of environmental management (Ehrenfeld J, 2003) [4].

On the one hand, industrial agglomeration can increase environmental pollution. The continuous accumulation of industrial agglomeration has resulted in a series of ecological environmental degradation problems, such as deterioration of the water environment, air pollution, soil pollution, and a reduction in biodiversity. The supporting capacity of the environment for industries has gradually decreased, and the negative environmental externalities of industrial agglomeration have increased (Wang et al., 2008; Shapiro et al., 2018) [5,6]. Subsequently, scholars have conducted more accurate empirical studies on the relationship between industrial agglomeration and the environment. Qin et al. (2014) conducted an empirical study and found that the continuous increase in economic activity concentration causes carbon emission intensity to present an obvious “inverted U-shaped” curve feature, where it initially rises but then decreases, and that industrial structure, technological progress, and the economies of scale effect are the main factors producing this influence [7]. Wang (2018) showed in his study that there exists a U-shaped relationship between environmental efficiency and industrial agglomeration. In this relationship, the former gradually deteriorates with the beginning of the latter; however, it then improves as the latter proceeds. The author provided evidence to validate this U-shaped pattern using different estimation strategies [8]. Miao et al. (2019) showed that an increase in the level of manufacturing agglomeration and productive services agglomeration would aggravate the degree of environmental pollution; however, the collaborative agglomeration of industries would reduce environmental pollution [9]. Andersson et al. (2011) started from the perspective of examining the externalities of industrial agglomeration, and they found that the crowding effect and even the scale effect created by industrial agglomeration would aggravate the degree of environmental pollution [10]. Chen et al. (2020) concluded that there exists an “inverted U-shaped” relationship between the agglomeration of industry and the emission of industrial wastes by studying the data of Chinese cities [11]. The study of Du et al. (2021) used a sample of 2158 coal-fired generating units of different types in more than eight hundred coal-fired power plants, and they ultimately came to the conclusion that technological emission reductions lead to negative synergistic effects on CO₂ emissions from both electric power consumption and the emissions from chemical reactions at the end of emission control pipeline measures [12].

On the other hand, industrial agglomeration can curb environmental pollution. Specifically, industrial agglomeration indirectly suppresses environmental pollution through certain action mechanisms and intermediate variables, such as technological innovation. Lucas (1988) showed that urbanization could enhance energy efficiency and reduce the pollution of the environment through the spillover effects of green technology agglomeration [13]. Zeng et al. (2009) believed that industrial agglomeration could encourage enterprises to adopt cleaner and more environmentally friendly production technologies through technological innovation and the spillover effect, thereby reducing the degree of environmental pollution caused by industrial development and improving environmental quality [14]. However, He (2006) found that, while in the short term, industrial agglomeration is beneficial for reducing pollution in the environment, in the long term there is no significant causal relationship between the two [15]. Deng et al. (2021) confirmed that environmental pollution in China had obvious features of positive spatial autocorrelation, which gradually weakened along with distance [16]. Yuan et al. (2020) conducted an analysis on data from nearly 290 cities in China using a spatial panel Durbin model and a mediation effect model. They ultimately discovered a significant “U-shaped” relationship between manufacturing agglomeration and green eco-efficiency [17]. Xie et al. (2021) also confirmed that the current industrial clustering in China could promote green development [18]. Technological innovation plays a crucial intermediary role between air pollution and industrial agglomeration. Most studies suggest that the latter can promote the adoption of more environmentally friendly production technologies by many enterprises within industrial clusters through the spillover effects of technological innovation, thereby reducing pollution and improving the quality of the environment. Zheng et al. (2018) suggested that the concentrated use of pollution control equipment and the sharing of energy-saving and emission-reducing technologies within manufacturing industry clusters were key factors leading to the positive externalities of industrial agglomeration on air pollution [19]. Song et al. (2023) argued that industrial agglomeration and air pollution do not have a linear relationship but that a moderate level of agglomeration is beneficial for emissions reduction to some extent [20]. Although most cities are still between the first and second inflection points, the positive externalities of industrial agglomeration are in play.

Reviewing the existing research, it is clear that there are many studies on industrial agglomeration and eco-efficiency, and a relatively mature research system has been formed. At the same time, the environmental effects of industrial agglomeration are also a research focus. The results of the existing research provided the basis for this study; however, in several aspects, it needed to be deepened and expanded. Regional eco-efficiency studies mostly focus on an overall analysis of impact factors. Based on the multiple characteristics of industrial agglomeration externalities, their impact on eco-efficiency is clearly more complex. Therefore, this study focuses on the impact of industrial agglomeration on eco-efficiency in the Yellow River Basin.

The contributions of this paper include the following aspects: Firstly, this study establishes an undesirable output superefficiency slacks-based measure model to evaluate the ecological efficiency levels of various provinces in the Yellow River Basin. According to the 2030 target for peak carbon and the 2060 carbon neutrality concept, emissions of carbon dioxide are also included in the undesired output index. Secondly, this study establishes a panel threshold model to analyze the impact of manufacturing agglomeration in the Yellow River Basin on the ecological environment. Thirdly, this study examines the heterogeneity of the impact of the manufacturing industry on eco-efficiency in the Yellow River Basin. This study focuses on analyzing the influence of industrial agglomeration on eco-efficiency from a more specific research perspective.

The following is the arrangement of the remaining parts of this study. The impact mechanisms of industrial agglomeration on eco-efficiency are analyzed in Section 2. The panel threshold model of the impact of manufacturing agglomeration on eco-efficiency is presented in Section 3. The empirical analysis is explained in Section 4. The final section presents the main conclusions of the paper.

2. Impact Mechanisms of Industrial Agglomeration on Eco-Efficiency

2.1. “Positive Externality”: The Positive Effect Mechanism of Industrial Agglomeration on Eco-Efficiency

The positive effect of industrial agglomeration on eco-efficiency is mainly based on the competitive advantage theory and the theory of external-scale economies. There are two types of external-scale economies: one based on the industry level and the other on the city level. The former is due to the expansion of the size of a whole industry, and the latter is due to the expansion of the size of the city where an enterprise is located. Industrial agglomeration indirectly inhibits environmental pollution through a number of action mechanisms or intermediate variables, such as technological innovation.

Based on the above-mentioned theoretical mechanism, assuming that industrial agglomeration has a positive effect on eco-efficiency in the Yellow River Basin, the reason for this effect may be the external scale economies brought by factor agglomeration. The agglomeration of human capital and basic factors in the Yellow River Basin is conducive to high-quality development. Labor agglomeration increases the labor supply rate and reduces the employment costs of enterprises in agglomeration areas. The faster the industrial agglomeration, the faster the development of knowledge and technology. The agglomeration of basic factors such as talent, capital, knowledge, and technology effectively increases the level of regional eco-efficiency. The development of industrial agglomeration also promotes innovation in governance concepts, and enterprises pay more attention to improving their own governance effectiveness, which further promotes the eco-efficiency level. The mechanism by which industrial agglomeration positively influences eco-efficiency is illustrated in Figure 1.

2.2. ”Negative Externality”: The Negative Effect Mechanism of Industrial Agglomeration on Eco-Efficiency

The negative effect of industrial agglomeration on eco-efficiency is mainly based on the crowding theory. When the scale of agglomeration exceeds a certain threshold, there will be a crowding effect, and the agglomeration will produce the economic phenomenon of “internal diseconomy”. The “crowding effect” means that when the degree of agglomeration is too high, the agglomeration will have a restrictive effect on the development of a region or industry. The mechanism by which industrial agglomeration negatively impacts eco-efficiency is illustrated in Figure 2.

Based on the theoretical mechanism analysis above, this study hypothesizes that industrial agglomeration in the Yellow River Basin may have a negative impact on eco-efficiency. The main reason for the “crowding effect” may be the overconcentration of resource-based industries. The Yellow River Basin is rich in energy and mineral resources, and it is an important energy supply region in China. However, a single-type industrial layout with a heavy chemical industry as the main industry aggravates the burden on the regional ecological environment and increases the risk of ecological system collapse to a certain extent. The problems of an inexpedient industrial layout and excessive growth in the scale of industry have severely limited the improvement of the ecological environment of the region. Ren et al. (2021) also concluded that industrial development and the ecological environment had not reached a state of coordinated development in the Yellow River Basin [21].

3. Equations and Mathematical Expressions

3.1. Construction of Threshold Effect Model of Manufacturing Industry on Eco-Efficiency

In view of the complexity of the influence of industrial agglomeration on eco-efficiency in the Yellow River Basin, this study makes the following assumption based on the empirical theory of the “Environmental Kuznets Curve”: this impact may take the form of a quadratic function or a piecewise function. There exists a threshold between the level of industrial agglomeration and the level of eco-efficiency. It is the existence of this threshold that makes the mechanism of the effect of agglomeration on eco-efficiency appear as both a “positive externality” and a “negative externality”.

Industrial agglomeration and economic growth can have complex effects on eco-efficiency. On the one hand, industrial agglomerations may have positive externalities on the latter due to scale effects, competitive advantages, knowledge spillovers, and the like. On the other hand, it may also have negative externalities due to crowding effects, pollution concentration, and other factors. Therefore, this study utilized a panel threshold model to explore the impact of industrial agglomeration on eco-efficiency in the Yellow River Basin. Additionally, the Hansen threshold regression model was also employed to test for the presence of a non-linear relationship between the above two types of effects.

The panel threshold model of industrial agglomeration on eco-efficiency was set as follows:

$$\begin{array}{lll}e{e}_{it}=& =& {\beta }_0+{\beta }_{1}id{d}_{it}I(id{d}_{it}\le \gamma )+{\beta }_{2}id{d}_{it}I(id{d}_{it}>\gamma )+{\beta }_3\ln gd{p}_{it}+{\beta }_4\ln fd{i}_{it}\\ & & +{\beta }_{5}i{s}_{it}+{\beta }_6\ln u_{it}+{\beta }_{7}c{r}_{it}+{\beta }_8\ln por{t}_{it}+{\epsilon }_{it}\end{array}$$

(1)

Here $i$ is the region, and $t$ here is the time. $I$ is a characteristic function that identifies a function as 1 when the values in the set of functions are in this range and 0 when they are not, by using a threshold value to determine the threshold value in the form of $\gamma$ value assignment. In addition, $ee$ is the eco-efficiency level of undesirable output; $idd$ is the agglomeration level of manufacturing industry in each province; $\ln gdp$ is the level of economic growth; $\ln fdi$ is the foreign direct investment level (FDI); $is$ is the industrial structure; $\ln u$ is human capital input; $cr$ is the level of urbanization; $\ln port$ is the import and export rate; and $\epsilon$ is the error term.

At the same time, referring to Qin (2016) [22], this study examined whether there is a critical effect of the economic growth level on eco-efficiency. The economic growth level was treated as a threshold variable. And the threshold model for the impact of economic growth on eco-efficiency $ee$ with undesirable output was as follows:

$$\begin{array}{lll}e{e}_{it}& =& {\beta }_0+{\beta }_1\ln gd{p}_{it}I(\ln gd{p}_{it}\le \gamma )+{\beta }_2\ln gd{p}_{it}I(\ln gd{p}_{it}>\gamma )\\ & & +{\beta }_3\ln fd{i}_{it}+{\beta }_4\ln port+{\beta }_{5}i{s}_{it}+{\beta }_6\ln u+{\beta }_{7}c{r}_{it}+{\epsilon }_{it}\end{array}$$

(2)

Referring to the discussion on the relationship between FDI and eco-efficiency in Gong (2018) [23], the level of FDI was set as a threshold variable and $fd{i}_{it}$ was the foreign direct investment level in the region $i$ in the year. The panel threshold model of the effect of manufacturing agglomeration on the eco-efficiency $ee$ of undesirable outputs was as follows:

$$\begin{array}{lll}e{e}_{it}& =& {\beta }_0+{\beta }_{1}id{d}_{it}I(\ln fd{i}_{it}\le \gamma )+{\beta }_{2}id{d}_{it}I(\ln fd{i}_{it}>\gamma )+{\beta }_3\ln gd{p}_{it}+{\beta }_{4}i{s}_{it}\\ & & +{\beta }_5\ln port+{\beta }_6\ln u+{\beta }_{7}c{r}_{it}+{\epsilon }_{it}\end{array}$$

(3)

3.2. Data Description

3.2.1. Explained Variable

The eco-efficiency level was selected as the explained variable. Referring to Wang (2023) [24], the measurement method for the eco-efficiency of each province was as follows:

$$\begin{gathered} \rho =\min \frac{1+\frac{1}{m}{\displaystyle \sum _{i=1}^m\frac{s_i^x}{x_{i0}}}}{1-\frac{1}{s_1+s_2}({\displaystyle \sum _{k=1}^{s_1}\frac{s_k^y}{y_{k0}}}+{\displaystyle \sum _{l=1}^{s_2}\frac{s_l^z}{z_{l0}}})} \\ s.t.\{\begin{array}{l}{x}_{i0}\ge {\displaystyle \sum _{j=1,\ne 0}^n{\lambda }_j}{x}_j-s_i^x,\forall i;\\ y_{k0}\ge {\displaystyle \sum _{j=1,\ne 0}^n{\lambda }_j}{y}_j-s_k^y,\forall k;\\ z_{l0}\ge {\displaystyle \sum _{j=1,\ne 0}^n{\lambda }_j}{z}_j-s_l^z,\forall l;\\ 1-\frac{1}{s_1+s_2}({\displaystyle \sum _{k=1}^{s_1}\frac{s_k^y}{y_{k0}}+{\displaystyle \sum _{l=1}^{s_2}\frac{s_l^z}{z_{l0}}}})>0\\ s_i^x\ge 0,s_k^y\ge 0,s_l^z\ge 0,{\lambda }_j\ge 0\end{array} \end{gathered}$$

(4)

where $s_i^x\in R^m$ is the excess of inputs; $s_l^z\in R^{s_2}$ is the excess of undesirable outputs; $s_k^y\in R^{s_1}$ represents the lack of desirable output; $m$ is the number of variables of input; $s_1$ is the desirable output; $s_2$ is the undesirable output; and $\rho$ represents the efficiency value of the decision-making unit. Here, $s_i^x,s_k^y,s_l^z$ there are no slack variables in the traditional sense. This study took capital investment, labor input, energy consumption, and technological input as the input indicators. This study divided the output indicators into desired outputs and undesired outputs. The former included indicators of economic, innovation, and environmental benefits, while the latter included wastewater emissions, sulfur dioxide emissions, and carbon dioxide emissions.

Using model (4), the eco-efficiency level of the provinces in the Yellow River Basin could be measured, including the undesirable output scenario. The result is shown in Figure 3. The overall eco-efficiency level of every province showed a fluctuating upward trend. First of all, Qinghai, Sichuan, and the Ningxia Autonomous Region had higher eco-efficiency values, and the next provinces that had relatively high values were Henan and Shandong. Additionally, Shaanxi’s ecological efficiency level was rapidly increasing, with a value exceeding 0.7 in 2020. Furthermore, Gansu’s eco-efficiency in 2020 increased by 12.6% compared to 2019, reaching the highest value in recent years.

3.2.2. Core Explanatory Variable

The manufacturing agglomeration level was selected as the core explanatory variable. The measurement methods for industrial agglomeration mainly included the EG index, location entropy index, industry concentration degree, spatial Gini coefficient, and regional development index. Location entropy represents a special economic ratio that can be utilized to measure the degree of specialization of an industry sector and its spatial distribution in a particular region. If the entropy of an industry exceeds 1, this indicates that it has a higher degree of aggregation in this region. Therefore, the level of manufacturing industry agglomeration was measured with the location entropy method. The level of manufacturing agglomeration was measured using the 31 manufacturing industries used by the National Bureau of Statistics of China as the standard. The number of manufacturing employees in the nine provinces of the Basin and the total amount of regional employment were calculated as the base values. The method for using location entropy to measure the agglomeration level of a manufacturing industry is as follows:

$$Idd=\frac{p_{it}/s{p}_{it}}{P_t/S{P}_t}$$

(5)

where $Idd$ is the level of agglomeration of an industry in a region $i$ in the year $t$$p_{it}$ represents the number of persons employed in the industry in that region; $s{p}_{it}$ represents the total number of persons employed in that region; $P_t$ and $S{P}_t$ represents the number of persons employed in that particular industry nationally and the total number of persons employed nationally in the year $t$, respectively. The higher the value $Idd$, the higher the concentration of the manufacturing industry in that region, and vice versa.

The agglomeration level of the manufacturing industry can be obtained by using Formula (5), as shown in Figure 4. The concentration of manufacturing industry in Shandong Province and Henan Province has been at a high level. The degree of industrial agglomeration in Gansu Province was the lowest, and the overall trend was at first weakening and then increasing. In general, the differences in the manufacturing agglomeration levels among the nine provinces mentioned gradually decreased, and the trend of collaborative development among provinces was obvious year by year.

3.2.3. Controlled Variables

(1)

Human capital: According to the practice of most scholars, the published average years of schooling for each province were used to represent their levels of human capital.

(2)

Industrial structure: Referring to the research conclusions of Gu (2020), the share of secondary sector output in total output was used to represent the province’s industrial structure [25].

(3)

Degree of import and export dependence: Imports and exports have a direct impact on the overall development process, thus affecting the level of domestic eco-efficiency. This study used each province’s total imports and exports as a proportion of its GDP to measure its degree of dependence on imports and exports.

(4)

Economic growth level: The layout of industries affects the level of economic growth, which in turn directly influences the eco-efficiency of a region. Based on the combined empirical research experience of domestic and foreign scholars, the logarithm of GDP was chosen to represent the level of economic development of the region. To exclude the effects of factors such as inflation on the results, this paper used GDP in 2005 as the base period for deflating.

(5)

Urbanization rate: Luo et al. (2013) concluded that the urbanization level and regional eco-efficiency are significantly correlated in China [26]. Therefore, the urbanization rate was used as a control variable in the research model, and the urbanization rate was treated logarithmically.

(6)

FDI level: According to the study conducted by Yang et al. (2015) on the impact of FDI on eco-efficiency, the level of FDI in each province in this study was reflected by the logarithm of the total amount of foreign investment [27].

The above data were obtained from the China Statistical Yearbook, the China Environmental Statistical Yearbook, the China Compendium of Statistics, and the Statistical Bulletin on Human Resources and Social Security Development.

3.3. Test of Correlation Variables

In this study, unit root tests were performed on the correlation variables. Using panel data, LCC, and IPS inspection methods, the results showed that all variables were significant at the 5% level, suggesting that no unit root existed. Before the empirical analysis of panel data, the Hausman test was carried out on the panel data to determine whether to choose a fixed panel model or a random panel model. The results are shown in

Table 1. Hausmann test results.

	CHI-SQ.DF	CHI-SQ. STATISTIC	PROB
Hausmann test	7	33.30	0.0000

. The Chi-square value of the Hausmann test was significantly greater than 0, rejecting the null hypothesis at the 1% significance level; therefore, the fixed panel model was chosen.

4. Empirical Analysis

4.1. Existence Test for Threshold of Industry Agglomeration Level

Before conducting empirical analysis using the threshold model in this study, it was necessary to examine whether thresholds exist, as well as the existence of threshold effects and the number of thresholds. The relevant code used the threshold model developed by Wang (2015) to test the threshold effect on the impact of industrial agglomeration on eco-efficiency [28]. The distribution was obtained by sampling 500 times using the bootstrap method of “autonomous sampling”. Whether the threshold effect of manufacturing agglomeration on eco-efficiency existed was determined by using the F-statistic and the p-value obtained from the test. The number of thresholds where thresholds exist was estimated. Industrial agglomeration was taken as the independent variable and also as the threshold variable. The results of the industrial agglomeration threshold model test are shown in

Table 2. Results of the identification of manufacturing agglomeration thresholds.

Identification Models	F-Statistic	p-Value	Bootstrap Sampling Numbers	Critical Value
				10%	5%	1%
Single Threshold	36.11 ***	0.002	500	14.65	17.29	23.96
Double Threshold	33.38 ***	0.0001	500	12.90	17.73	25.16
Three Threshold	22.94	0.176	500	36.42	49.82	97.42

Note: *** represents a significance level of 1%.

.

The double threshold for manufacturing agglomeration is shown in

Table 3. Estimated double threshold for manufacturing agglomeration.

Independent Variable	Threshold Variable	Threshold	Estimated Value	95% Confidence Interval
$idd$	$idd$	${\gamma }_1$	0.3667	[0.3545, 0.3674]
		${\gamma }_2$	0.3978	[0.3949, 0.4013]

. The first threshold for manufacturing agglomeration was 0.3667, which corresponded to a 95% confidence interval of [0.3545, 0.3674], and the second threshold was 0.3978, which corresponded to a 95% confidence interval of [0.3949, 0.4013]. After analyzing the likelihood ratio function of a manufacturing agglomeration double threshold, this study selected a double-threshold model when analyzing the role of manufacturing industry agglomeration on eco-efficiency.

To further consider whether there is a threshold effect of economic growth level on eco-efficiency in the absence of environmental constraints, the threshold value of model (2) was identified. The results are shown in

Table 4. Results of eco-efficiency threshold identification.

Identification Models	F-Statistic	p-Value	Bootstrap Sampling Numbers	Critical Values
				10%	5%	1%
Single Threshold	28.82 *	0.058	500	22.53	28.17	36.19
Double Thresholds	29.14 **	0.020	500	17.78	20.10	30.24
Three Thresholds	19.07	0.565	500	44.85	52.14	71.82

Note: * represents a significance level of 10%, ** represents a significance level of 5%.

.

According to the identification results for the eco-efficiency threshold, in the single-threshold test, the F-statistic was 28.82 and the p-value was 0.058, which was significant at the 10% level, rejecting the original hypothesis and thus demonstrating a single-threshold effect of economic growth on eco-efficiency without undesirable output. And in the double-threshold test, the F-statistic was 29.14 and the p-value was 0.020, which was significant at the 5% level, indicating that economic growth has a double-threshold effect on eco-efficiency. In the three-threshold test, the F-statistic was 19.07 and the p-value was 0.565; therefore, the original hypothesis could not be rejected, indicating that economic growth does not have a three-threshold effect on eco-efficiency without undesirable outputs. The results for the thresholds are shown in

Table 5. Estimated eco-efficiency double threshold.

Independent Variable	Threshold Variable	Threshold	Estimated Value	95% Confidence Interval
$\ln gdp$	$\ln gdp$	${\gamma }_1$	7.0098	[6.9250, 7.0853]
		${\gamma }_2$	7.9864	[7.9280, 8.0089]

. The first threshold value for economic growth was 7.0098, which corresponded to a 95% confidence interval of [6.9250, 7.0853], and the second threshold value was 7.9864, which corresponded to a 95% confidence interval of [7.9280, 8.0089]. After analyzing the likelihood ratio function of an ecological efficiency double threshold, a double-threshold model was chosen to analyze the role of economic growth on eco-efficiency without undesirable output.

Considering the possible mutual relationship of influence between the level of FDI and the degree of manufacturing agglomeration, the bootstrap method of “autonomous sampling” was used to sample 500 times to determine whether manufacturing agglomeration under the influence of FDI has a threshold effect on eco-efficiency. Manufacturing agglomeration was used as the independent variable and the level of FDI as the threshold variable in model (3). The test results for the threshold model are shown in

Table 6. FDI level threshold identification results.

Identification Models	F-Statistic	p-Value	Bootstrap Sampling Numbers	Critical Value
				10%	5%	1%
Single Threshold	50.75 ***	0.004	500	21.91	26.79	43.67
Double Thresholds	8.51	0.334	500	16.87	21.42	33.21
Three Thresholds	17.09	0.078	500	14.02	21.44	135.27

Note: *** represents a significance level of 1%.

.

The identification results of the FDI threshold level indicated that the single threshold exists and is significant at the 1% level. Therefore, the original hypothesis was rejected, demonstrating that there is a single-threshold effect of FDI level on eco-efficiency. The F-statistic in the double threshold test was 8.51, and the p-value was 0.334, showing that there is no double-threshold effect on the level of FDI. In the three-threshold test, the F-statistic was 17.09 and the p-value was 0.07; however, since the double threshold was invalid, the results of the three-threshold test were also invalid, indicating that there is only a single-threshold effect of FDI level on eco-efficiency.

As can be seen from

Table 7. Estimated value of single threshold of FDI level.

Independent Variable	Threshold Variable	Threshold	Estimated Value
$idd$	$fdir$	$\gamma$	0.0005

, the threshold value of the level of FDI was 0.0005; that is, the proportion of total FDI in the gross product was 0.05%. Therefore, the single-threshold model was chosen to analyze the impact of manufacturing agglomeration on eco-efficiency under the regulation of FDI.

4.2. Threshold Effect of Manufacturing Agglomeration on Eco-Efficiency

According to the threshold effect test above, regression analysis of the double-threshold model was conducted for models (1) and (2), and regression analysis of the single-threshold model was conducted for model (3). The results are shown in

Table 8. Regression results of threshold models.

	Model (1)		Model (2)		Model (3)
$idd<0.37$	0.0010	$\ln gdp<7.99$	−2.6449 *	$idd$	−8.7257 ***
	(0.00)		(−1.75)	$\left(idd<0.0005\right)$	(−6.26)
$0.37<idd<0.40$	2.8713 *	$7.99\le \ln gdp\le 9.05$	−2.5228 *	$idd$	−0.5785
	(1.80)		(−1.69)	$\left(fdir>0.0005\right)$	(−0.71)
$idd>0.40$	−1.5263 *	$\ln gdp\ge 9.05$	−2.9342 *
	(−1.71)		(−1.96)
$\ln u$	5.1340	$\ln u$	5.6169 *	$\ln u$	5.4346 *
	(1.60)		(1.89)		(1.86)
$\ln port$	1.0242 ***	$\ln port$	0.4805	$\ln port$	0.2017
	(3.06)		(1.49)		(0.59)
$cr$	1.1885	$cr$	1.4372	$cr$	1.5932
	(1.12)		(1.48)		(1.63)
$is$	−0.7371	$is$	−0.2304	$is$	−0.4318
	(−0.25)		(−0.08)		(−0.16)
$\ln gdp$	−2.6494	$\ln gdp$		$\ln gdp$	−3.0518 **
	(2.28)				(−2.04)
$\ln fdi$	0.0555	$\ln fdi$	0.0238	$\ln fdi$
	(1.22)		(0.55)
N	135	N	135	N	135
R2	0.2292	R2	0.3106	R2	0.3331
F	3.8648	F	6.6446	F	8.4901

Note: * represents a significance level of 10%; ** represents a significance level of 5%; *** represents a significance level of 1%.

.

As shown in Table 8, this study took manufacturing industry agglomeration as the independent and threshold variable. When the level of this indicator was below 0.3667, the coefficient of influence of the agglomeration indicator on eco-efficiency was small. This indicated that, when the level of manufacturing industry agglomeration is low, its impact on eco-efficiency is small and the effect is insignificant. When the level was between 0.37 and 0.40, the effects of the agglomeration measure on eco-efficiency were significantly positive, and this is the optimal agglomeration interval to improve the eco-efficiency level. When the manufacturing industry concentration level increases by 1%, the level of ecological efficiency can be increased by 2.8713%. There is an optimal interval for the improvement of eco-efficiency by industrial agglomeration; however, the optimal interval in the Yellow River Basin is smaller. When the manufacturing agglomeration level exceeded 0.40, the impact of the agglomeration measure on eco-efficiency was negative. For every 1% decrease in the manufacturing agglomeration level, the eco-efficiency level decreased by 1.5263%. The influence of the manufacturing agglomeration level on eco-efficiency in the Basin changes from positive to negative after it crosses the threshold value, indicating that under the current industrial development mode, the negative externality of manufacturing agglomeration on eco-efficiency in the Basin is greater than the positive externality when it exceeds the threshold value, thus showing the significant negative effects.

According to the regression results of model (2) in Table 8, the following conclusions can be obtained when economic growth is taken as the independent variable and threshold variable and eco-efficiency is taken as the dependent variable. When the economic growth level is below 7.99, that is, the GDP is below CNY 295.1 billion, economic growth has a significant negative effect on eco-efficiency; for every 1% of economic growth, eco-efficiency decreases by 2.64%. When the economic growth level is between 7.99 and 9.05, that is, the GDP is between CNY 295.1 and 851.9 billion, economic growth has a significant negative effect on eco-efficiency; however, the effect is weakened. For every 1% increase in the economic growth level, the eco-efficiency level decreases by 2.25%. When the economic growth level exceeds 7.99, that is, the GDP exceeds CNY 851.9 billion, economic growth has a significant positive impact on eco-efficiency. For every 1% increase in the economic growth level, the eco-efficiency level decreases by 2.93%. In general, according to the current development model, the economic growth of the Basin is not conducive to the improvement of the eco-efficiency level, and the economic growth level has the smallest negative effect on eco-efficiency.

According to the regression results of model (3) in Table 8, the following conclusions can be drawn when taking manufacturing agglomeration as the independent variable and the foreign direct investment level as the threshold variable. When the proportion of foreign investment is less than 0.05% of the total output value, manufacturing agglomeration has a significant negative impact on eco-efficiency. According to the above theoretical mechanism analysis, it shows that when the foreign direct investment level is low, the knowledge spillover and competitive advantage generated by industrial agglomeration are less than the centralized environmental pollution or crowding effect caused by agglomeration, making the negative externalities greater than the positive externalities, which is shown as a significant negative impact. When the proportion of foreign investment is greater than 0.05% of the total output value, the impact of the agglomeration level of the manufacturing industry on eco-efficiency is no longer significant, indicating that the impact of agglomeration on eco-efficiency has faded or that the positive and negative externalities offset each other, thus presenting an insignificant result. These empirical results are consistent with the conclusion of Xu et al. (2012) [29] that foreign investment can improve environmental pollution to a certain extent. For the Yellow River Basin, foreign investment does have an impact on the effect of industrial agglomeration on eco-efficiency; however, its role in improving environmental pollution is limited; therefore, after crossing the threshold, industrial agglomeration has no obvious positive effect, which reflects the scientific nature of China’s policy of actively introducing foreign capital. However, when the foreign investment level is low, the agglomeration of the manufacturing industry is seriously detrimental to the improvement of eco-efficiency.

The robustness of the regression results was tested as follows to further demonstrate their reliability.

(1) Replacement of the indicators for measuring industrial agglomeration: The measurement index of industrial agglomeration was replaced. This study replaced the measurement of industrial agglomeration based on the number of employees with the measurement of industrial agglomeration based on the value of output. The original model was used to conduct regression analysis on the replacement data, again referring to

Table 9. Robustness results.

Variable	Model (1)	Variable	Model (2)
$idd3<0.37$	0.5946	$idd$	1.2814 ***
	(0.29)		(6.37)
$0.37\le idd3\le 0.40$	3.3923 **	$id{d}^2$	−0.4578 ***
	(2.19)		(−4.07)
$idd3>0.40$	−1.0640
	(−1.30)
$\ln u$	5.6803 *	$\ln u$	−2.1586 ***
	(1.76)		(−8.27)
$\ln port$	0.9664 ***	$\ln port$	−0.0317
	(2.90)		(−1.05)
$cr$	5.1519	$cr$	−0.5966 *
	(0.50)		(−1.89)
$is$	−1.4235	$is$	0.4736 **
	(−0.49)		(2.19)
$\ln gdp$	−1.8108 *	$\ln gdp$	−0.0180
	(−1.89)		(−1.35)
$\ln fdi$	0.0434	$\ln fdi$	−0.0045
	(0.96)		(−0.72)

Note: * represents a significance level of 10%; ** represents a significance level of 5%; *** represents a significance level of 1%.

. The test results demonstrated that the role of the core explanatory variables was consistent, and the role coefficient had not changed significantly; only some of the other control variables had small changes in significance. The positive properties, the negative properties, and the size of the regression coefficients had not changed significantly; therefore, the regression results of this study were robust.

(2) Adding the quadratic term of the industrial agglomeration measure into the benchmark model for regression analysis: According to model (2) in Table 9, the impact of the quadratic term of industrial agglomeration on eco-efficiency was significant, proving the nonlinear relationship between the two, further verifying the threshold effect, and indicating the robustness of the results.

4.3. Model Estimation Heterogeneity of the Threshold Effect of Manufacturing on Eco-Efficiency

To further analyze the impact of manufacturing industry clustering on eco-efficiency, this study classified the manufacturing industry according to energy consumption. Using the classification criteria in the National Economic and Social Development Statistics Report, this paper classified five types of manufacturing industries as high-energy-consuming industries, namely, the petroleum processing and coking industry and nuclear fuel processing industry, the chemical raw materials and chemical products manufacturing industry, the ferrous metal smelting and rolling processing industry, the non-ferrous metal smelting and rolling processing industry, and the non-metallic mineral products industry.

In addition to the above five categories of high-energy-consumption manufacturing, other manufacturing industries were divided into medium- and low-energy manufacturing. The agglomeration measures for high-energy-consuming manufacturing industries as well as medium- and low-energy-consuming manufacturing industries are shown in

Table 10. Identification results of eco-efficiency threshold of undesirable output.

Industry Category	Identification Model	F-Statistic	p-Value	Sample	Critical Value
					10%	5%	1%
High energy consumption industry	Single threshold	21.90 **	0.034	500	16.32	20.12	41.36
	Double threshold	17.57 **	0.048	500	14.05	17.51	27.03
	Triple threshold	5.70	0.744	500	45.81	52.39	66.00
Medium and low energy consumption industry	Single threshold	33.40 **	0.028	500	21.28	27.17	39.13
	Double threshold	14.37	0.268	500	22.29	30.49	36.00
	Triple threshold	6.71	0.894	500	97.29	121.55	166.64

Note: ** represents a significance level of 5%.

.

4.3.1. Identification of the Threshold Effect of the High-Energy-Consuming Industry as Well as the Medium- and Low-Energy-Consuming Manufacturing Industry on Eco-Efficiency

Before using the threshold model for empirical analysis, the existence of the threshold values for the agglomeration of high-energy-consuming manufacturing industries and for the agglomeration of medium- and low-energy-consuming manufacturing industries was tested. It was shown that the threshold effects included the impact of the eco-efficiency of undesirable output in Table 10.

A single threshold for the agglomeration level of the high-energy-consuming manufacturing industry is shown in Table 10. It was significant at the 5% level, rejecting the original hypothesis and proving that there is a single-threshold effect of high-energy-consuming manufacturing industry agglomeration on eco-efficiency. In the double-threshold test, the F statistic was 17.57 and the p value was 0.048, which was significant at the 5% level, indicating that there is a double-threshold effect of high-energy-consuming manufacturing agglomerations on eco-efficiency. In the three-threshold test, the F statistic was 5.70, and the p value was 0.744. The original hypothesis cannot be rejected, indicating that the agglomeration of high-energy-consuming manufacturing industries does not have a triple-threshold effect on eco-efficiency. There was a single threshold for the agglomeration level of the medium- and low-energy-consuming manufacturing industry, and it was significant at the 5% significance level; however, the F statistic in the double-threshold test was 14.37, which proved that there is no double-threshold effect; that is, there is only a single-threshold effect for the agglomeration level of the medium- and low-energy-consuming manufacturing industry. Therefore, this study selected the double-threshold model and the single-threshold model to analyze the impact of the agglomeration of high-energy-consuming and medium- and low-energy-consuming manufacturing industries on eco-efficiency, respectively.

The double-threshold values of high-energy-consuming manufacturing agglomerations are shown in

Table 11. Estimation of manufacturing agglomeration threshold.

Industry Category	Independent Variable	Threshold Variable	Threshold	Estimates	95% Confidence Interval
High-energy- consumption industry	$idd1$	$idd1$	${\gamma }_1$	0.5190	[0.4984, 0.5280]
			${\gamma }_2$	0.5495	[0.4491, 0.5958]
Medium- and low- energy-consumption industry	$idd2$	$idd2$	${\gamma }_1$	0.1647	[0.1605, 0.1671]

. The first threshold value of high-energy-consuming manufacturing agglomeration was 0.5190, and its corresponding 95% confidence interval was [0.4984, 0.5280]; the second threshold value was 0.5495, and its corresponding 95% confidence interval was [0.4491, 0.5958]. The single threshold of low- and medium-energy-consuming manufacturing industry agglomeration was 0.1647. Therefore, this paper chose the single-threshold model to study the impact of low- and medium-energy-consuming manufacturing industry agglomerations on eco-efficiency.

4.3.2. Regression Result Analysis of Threshold Model

Based on the results of the threshold effect test discussed above, this study selected both the double-threshold model and the single-threshold model to analyze the impact of the agglomeration of high-energy-consuming and medium- and low-energy-consuming manufacturing industries on eco-efficiency, respectively. The results of the analysis are shown in

Table 12. Regression results of heterogeneity analysis.

High-Energy-Consuming Manufacturing Industry		Medium-and Low-Energy Consumption Manufacturing Industry
$idd1<0.52$	−4.8110 ***	$idd2<0.17$	−0.6428
	(−3.42)		(−0.92)
$0.52\le idd1\le 0.55$	−1.7689 ***	$idd2\ge 0.17$	9.8657 ***
	(−2.73)		(3.42)
$idd1\ge 0.55$	−1.1305 ***
	(−2.69)
$\ln u$	6.4738 **	$\ln u$	1.5354
	(2.00)		(0.46)
$\ln port$	1.0359 ***	$\ln port$	0.9963 ***
	(3.09)		(2.98)
$cr$	4.8491	$cr$	1.3112
	(0.47)		(1.19)
$is$	−1.9544	$is$	0.5816
	(−0.66)		(0.18)
$\ln gdp$	−1.8392	$\ln gdp$	−2.3503
	(−1.14)		(−1.40)
$\ln fdi$	0.0881 *	$\ln fdi$	0.0520
	(1.88)		(1.15)

Note: * represents a significance level of 10%; ** represents a significance level of 5%; *** represents a significance level of 1%.

.

The impacts of various energy-consuming manufacturing agglomerations on eco-efficiency were very different. The agglomeration of high-energy-consuming manufacturing in the Basin had an obvious negative externality effect on eco-efficiency, while the agglomeration of medium- and low-energy-consuming manufacturing had a positive externality effect on eco-efficiency. From the double-threshold regression results of the agglomeration of high-energy-consuming manufacturing industries, it could be concluded that when the agglomeration level of these industries in the Basin is below 0.52, the impact on eco-efficiency will be significantly negative, indicating that the development mode of high-energy-consuming manufacturing industries in the initial stage is relatively extensive, the production concept is relatively backward, and the production mode that only seeks to maximize production capacity has a large negative impact on eco-efficiency; therefore, the negative externalities of the initial agglomeration effect on environmental pollution and ecological disruption are greater than the positive externalities of the economic benefits.When the level of agglomeration exceeds 0.52, its negative impact on eco-efficiency will be weakened. The cluster development of the high-energy-consuming manufacturing industry can effectively reduce its negative externalities and reduce the adverse impact on ecology in the process of industrial development. When the agglomeration level rises to above 0.55, its negative impact on eco-efficiency will be further weakened, and the development mode of industrial clustering and specialization will reduce the negative externalities of high-energy-consuming manufacturing. Since the development of energy-intensive industries mainly depends on resource endowment, the impact of other control variables, such as urbanization level, is not significant. Zhao et al. (2021) also verified that its resource endowment inhibits the green development of the region [30]. The exploitation of resources intensifies the pressure on the ecological environment. The pollution control level of the manufacturing industry with high energy consumption is poor, and the advantage of resources restricts the sustainable development of the region.

From the regression results for a single threshold of the agglomeration of medium- and low-energy-consuming manufacturing industry in the Yellow River Basin, it could be concluded that, when the concentration level of medium- and low-energy-consuming manufacturing industry is lower than 0.17, the impact of agglomeration on eco-efficiency is not significant, which is different from the theoretical analysis. The positive and negative effects may offset each other, which may lead to an insignificant result. When the agglomeration level of the medium- and low-energy-consuming manufacturing industry exceeds 0.17, the impact of agglomeration on eco-efficiency is significantly positive. The cluster development of the medium- and low-energy-consuming manufacturing industry will produce significant positive externalities. Technology spillovers and knowledge spillovers brought by industrial agglomeration are more prominent in electronic information, high-end equipment manufacturing, and other industries, while labor force sharing and transportation cost reduction brought by the agglomeration effect enable medium- and low-energy-consuming manufacturing industries to gain more competitive advantages. The specialized industrial parks generated by the industrial agglomeration also further accelerate the gathering of talent, technology, and capital. The centralized treatment of pollutants and centralized learning about environmental protection also accelerate the efficiency and levels of pollution treatment. With the government’s policy support, it is conducive to an overall improvement in eco-efficiency; therefore, it shows significant positive externalities.

5. Conclusions and Recommendations

5.1. Conclusions

The target group of the research was all the provinces of the Yellow River Basin. This study provided a detailed analysis of the impact of manufacturing industry agglomeration in all the provinces in the Basin on eco-efficiency from an empirical perspective. This study examined the agglomeration of manufacturing industries in the provinces within the Basin. This study was based on the theoretical mechanism of the impact of industrial agglomeration on eco-efficiency. By constructing a panel threshold model, it explored the impact of the former on the latter in the Basin. What is more, this study also analyzed the heterogeneity of different industries. The conclusions are as follows:

(1) When the manufacturing agglomeration level is low, its impact on eco-efficiency is small. When the manufacturing agglomeration level is in the optimal agglomeration interval, the agglomeration measure has a positive effect on eco-efficiency; when the manufacturing agglomeration level increases by 1%, the eco-efficiency level can increase by 2.8713%. When the manufacturing agglomeration level exceeds the optimal interval, the impact of the agglomeration measure on eco-efficiency is negative; for every 1% reduction in the manufacturing agglomeration level, the eco-efficiency level will decrease by 1.5263%. After surpassing the critical threshold, the impact of manufacturing industry agglomeration on the eco-efficiency of the Basin turns into a negative effect. This is due to the fact that, under the current industrial development mode, the negative externalities of manufacturing industry agglomeration on eco-efficiency in the Basin exceed the positive externalities, thus exhibiting a significant negative impact. When the proportion of foreign investment is less than 0.05% of the total output value, manufacturing agglomeration has a significant negative impact on eco-efficiency, which indicates that when the foreign direct investment level is low, the knowledge spillover and competitive advantages generated by industrial agglomeration are less than the crowding effects and centralized environmental pollution caused by agglomeration, making the negative externalities greater than the positive externalities and resulting in a significant negative impact. When the proportion of foreign investment is greater than 0.05% of the total output value, this impact fades away, making it no longer significant.

(2) In the industry heterogeneity analysis of the threshold effect of manufacturing agglomeration on eco-efficiency, the impact varies widely for different energy consumption types of manufacturing. On the one hand, the agglomeration of high-energy-consuming manufacturing industries has an obvious negative impact on eco-efficiency. On the other hand, the agglomeration of medium- and low-energy-consuming manufacturing industries has a positive impact. From the double-threshold regression results of the agglomeration of high-energy-consuming manufacturing industries, it can be concluded that, when the agglomeration level of these industries in the Basin is below 0.52, the impact on eco-efficiency is significantly negative. When the agglomeration level exceeds 0.52, its negative impact on eco-efficiency will be weakened, and the cluster development of the high-energy-consuming manufacturing industry can effectively reduce its negative externalities. When the level of agglomeration rises to above 0.55, its negative impact on eco-efficiency will be further weakened, and the development mode of industrial clustering and specialization will reduce the negative externalities of high-energy-consuming manufacturing. From the single-threshold regression analysis of the agglomeration of low- and medium-energy-consuming manufacturing industries, it can be concluded that the impact of low- and medium-energy-consuming manufacturing industries on eco-efficiency is positive but not significant at a low level of agglomeration. When the agglomeration level of the medium- and low-energy-consuming manufacturing industry exceeds 0.17, the impact of agglomeration on eco-efficiency shows a significant positive effect. The cluster development of the medium- and low-energy-consuming manufacturing industry will have a significant positive externality effect, which is conducive to the overall improvement of eco-efficiency.

5.2. Policy Recommendations

The construction of a modern industrial system and the promotion of coordinated development in the Yellow River Basin need to be accelerated. From a basin-wide perspective, it is essential for the government to accelerate the transformation and upgrading of traditional industries. For instance, traditional industries such as chemicals and steel should focus on deep integration with new information technologies. Additionally, in order to promote the efficient and intensive development of the supply chain, local governments should also focus on creating highly centralized and closely coordinated industrial clusters.

Local governments should promote the classification of functional zones and strengthen the ecological security pattern. Ecological protection and restoration should be emphasized in areas that perform ecological functions such as water conservation and being a green barrier. For Zhengzhou, Jinan, and other urban areas, the degree of aggregation should be strengthened, labor allocation should be optimized, and the regional population carrying capacity should be improved. The main functional zones should not only adapt to local conditions but should also promote the integrated protection of ecological space, giving full play to the synergistic governance effect and strengthening the overall ecological security pattern of the whole Basin.

The level of human resources in the urban agglomerations of the Yellow River Basin currently leaves much to be desired. In order to achieve coordinated development of industrial innovation in various enterprises, the government needs to focus on strengthening the leading position of enterprises in innovation and encourage them to establish technological innovation alliances in relevant industries. Local governments should facilitate the adoption of various types of innovation resources and elements in enterprises and should promote their innovation decision-making, research and development investment, and achievement transformation to create regional development advantages.

Local governments need to strengthen regionally coordinated development and improve the level of openness. It is important to focus on integrating population, industry, and market factors and to strengthen the complementary advantages and coordinated development between central cities and surrounding cities. Additionally, in the diffusion of technological innovation and the promotion of regional economic transformation, the government should fully leverage the active role of central cities such as Zhengzhou and Xi’an. The government should also accelerate construction in each city and improve the level of openness, fully utilizing the advantages of foreign investment and actively aligning it with the national strategy for high-quality development of the Yellow River Basin.

5.3. Deficiencies and Prospects

This paper analyzes the impact of manufacturing agglomeration on ecological efficiency among provinces in the Yellow River Basin; however, this could not be carried out for all industries due to the lack of data. In the future, the index of the service industry and other industries could be added to make the research conclusions more comprehensive and objective and to put forward more targeted industrial layout suggestions for the Yellow River Basin.