How Does Diversification of Producer Services Agglomeration Help Reduce Carbon Emissions Intensity? Evidence from 252 Chinese Cities, 2005–2018

Abstract

Mitigating carbon emissions intensity (CEI) and promoting carbon neutrality at the city level are essential for addressing the challenges of global climate change and advancing sustainable development. This study examines the influence of producer services agglomeration diversification (PSAD) on CEI using an unbalanced panel dataset including 252 Chinese prefectural-level cities from 2005 to 2018 for empirical analyses. We find that improving PSAD in a city can significantly mitigate CEI. Stronger PSAD accelerates a city’s industrial structure transformation from secondary- to tertiary-dominated in addition to boosting green development capabilities, both of which are confirmed to have concrete carbon emissions reduction effects. Furthermore, PSAD only significantly alleviates CEI in non-eastern cities in China, and the benefits of carbon emissions reduction are stronger after 2010. Our policy insights highlight land utilization in shaping the intracity layouts of producer services agglomerations (PSAs) and stress regional-level measures. Recognizing regional differences and integrating PSAs allocation with broader institutional measures can amplify PSAD’s benefits.

1. Introduction

Climate deterioration, greenhouse effects, and global warming have become pivotal components of global environmental discourse in recent years. Carbon dioxide (CO₂) emissions are unavoidable by-products of human economic activities, particularly during industrial production, which is the leading contributor to global climate change [1]. According to the Our World in Data (OWID) database (Global Carbon Budget (2023)—with major processing by Our World in Data, CO₂ emissions: https://ourworldindata.org/co2-emissions(accessed on 12 December 2023)), global CO₂ emissions from fossil fuels was 4.62 billion tons in 1945, and had reached 37.15 billion tons by 2022, representing a massive increase of nearly seven times. Advancing carbon reduction and achieving carbon neutrality have become a common sustainable development goal around the world. The world’s major economies have developed a series of agreements or consensus for addressing global climate change, such as the 1992 United Nations Framework Convention on Climate Change, the Kyoto Protocol in 1997, the Cancun Agreements in 2010, and most recently, the well-known Paris Agreement in 2016. Cities are spatial entities with highly concentrated economic activities and intensive land use, contributing the vast majority of total global CO₂ emissions [2]. For example, Moran et al., (2018) found that the highest emitting 100 urban areas, defined as contiguous population clusters, account for nearly 18% of the global carbon footprint in 2015 [3]. Therefore, effectively reconciling the contradiction between economic growth and environmental pollution at the city level contributes significantly to advancing global sustainability.

Since the reform and opening up, China’s economy and urbanization has developed rapidly, becoming the largest industrialized nation, which has been accompanied by enormous CO₂ emissions [2,4]. According to the OWID database, after 2000, China’s total carbon dioxide emissions rapidly increased, gradually becoming the world’s largest carbon emitter (see Figure 1). Specifically, China’s annual CO₂ emissions in 2006 reached 6.49 billion tons, surpassing those of the United States (US) (6.05 billion tons) and becoming the largest emitter in the world. In 2022, China contributed 30.79% of all global CO₂ emissions, at nearly 2.26 times that of the second-place US (13.61%). In response, China has extended various commitments for CO₂ reduction. In recent years, mitigation efforts have been made in terms of carbon emissions in trading markets, new energy vehicles, green finance, and reforestation programs in China [5]. In addition, China is also cultivating new forms of environmental co-governance with other countries worldwide, inspired by its initiative of the Community with a Shared Future for Mankind [6]. Thus, identifying effective approaches for accelerating carbon reduction and achieving carbon neutrality is crucial for China’s economic transformation and development and can also provide valuable lessons for the rest of the world.

In this study, we argue that leveraging the benefits of producer services agglomeration diversification (PSAD) may be one possible approach for optimizing China’s carbon emission activities. On the one hand, it has been well documented that external producer services help manufacturing firms save costs, improves in-house productivity [7,8], and promotes innovation in manufacturing through knowledge spillover effects [9,10]. According to China’s National Bureau of Statistics (NBS), in 2020, the nation had 290,000 large-scale producer services enterprises, with 26.88 million employees, total assets amounting to 110 trillion yuan, and operating revenue reaching 84 trillion yuan. As with other industrial agglomerations, producer services are unevenly distributed across space [11], contributing considerably to regional differences in manufacturing activities [8]. Since China’s accession to the WTO in 2001, the rapid expansion of its manufacturing industry has significantly contributed to the increase in carbon emissions [12,13]. From 1997 to 2015, the CO₂ emissions from China’s manufacturing sector increased by 2.84 times, constituting over 50% of China’s total CO₂ emissions [14]. On the other hand, the carbon emission intensity (CEI) indicator, typically measured as the CO₂ emissions per unit of GDP, can comprehensively reflect a country’s energy efficiency and economic performance. CEI also serves as a crucial metric of China’s international emissions reduction commitments [14]. For example, China proposed to cut its CEI by 60–65% by 2030 compared to the 2005 level [5]. Accordingly, we ask, does the producer services agglomerations help reduce the CEI at the city level in China?

This study investigates the mitigating effects of PSAD on CEI in China. Specifically, we construct a panel data set including 252 prefectural-level Chinese cities from 2005 to 2018 for empirical analysis. Our CO₂ emissions data are primarily obtained from the Centre for Global Environmental Research (CGER) in Japan and China’s Urban Statistical Yearbooks to construct an indicator representing PSAD. We find higher PSAD to be correlated with significant CEI reduction effects at the city level in China. We also conduct robustness tests, such as changing the carbon emissions data source, recalculating CEI, and readjusting the lag period of core explanatory variables, the results of which validate our main conclusions. We then investigate two possible underlying mechanisms, showing that stronger PSAD in a city significantly promotes industrial transformation from the secondary to the tertiary sector. Diversification also has significant boosting effects for urban environmental governance and green innovation capabilities. Finally, we identify some heterogeneous effects, determining that PSAD only significantly promotes mitigation of CEI in the non-eastern cities in China, and the carbon reduction benefits from PSAD are stronger after 2010. To harness the decarbonization benefits of PSAD, especially in advancing green development capabilities, our policy insights emphasize the critical role of land utilization in shaping intracity PSAs layouts and underscore the importance of measures considering regional perspectives. Furthermore, recognizing regional differences and integrating PSAs allocation with broader institutional measures can also amplify the benefits of PSAD.

Considerable research has emphasized the critical role of industrial structure on carbon emissions activities. Some studies have used indicators of output or employment proportion in macro-level industrial sectors in a region, regarding which the most common conclusion is that industrial structure upgrading, i.e., transforming from a secondary- to a tertiary-dominated economy, will promote carbon emissions efficiency [1,2,15,16,17,18]. Previous studies have also identified the key carbon-emitting sectors based on input–output analysis [19,20]. However, these studies do not consider the spatial agglomeration of industrial activities. In this regard, some studies have analyzed the influence of regional industrial agglomeration on CEI. For example, based on a panel dataset of 187 Chinese cities from 2005 to 2013, Chen et al., (2018) [21] found that stronger industrial agglomeration led to higher total CO₂ emissions while also reducing cities’ CEI. Wu et al., (2021) [22] analyzed the underlying mechanisms, indicating that the industrial agglomeration increases CO₂ emissions through the scale effect but reduces CO₂ emissions through technique and composition effects.

A limited number of studies have examined how regions’ PSAD influences CEI. For instance, examining national-level evidence, Jin et al., (2022) [23] found that domestic producer services development could significantly reduce the carbon intensity of the manufacturing sector with highly integrated producer services. Based on the simultaneous equation model, Zhao et al., (2021) [24] found that specialized PSAs can lower carbon emissions in Chinese provinces by improving economic scale, while strengthening technology effects and optimizing industrial structure alleviate CO₂ emissions. Using a panel dataset of 282 Chinese cities from 2003 to 2017, Xu et al., (2023) [25] found that the influence of co-agglomeration of manufacturing and producer services on carbon productivity is U-shaped, indicating that further enhancing carbon productivity requires stronger industrial co-agglomeration.

Compared with the existing literature, the main contribution of this study is its investigation of how PSAD mitigates CEI, providing credible empirical evidence based on Chinese city-level data. First, we focus on the role of PSAD. Zhao et al., (2021) [24] used a locational entropy index to quantify PSA, primarily capturing an overall specification of all producer services industries in a city. Considering the functional characteristics and diverse types of producer services, we argue that it is more reasonable to examine carbon reduction effects from diverse perspectives. To do so, drawing on Combes, (2000) [26], we introduce a modified version of the inverse Herfindahl index to measure the comprehensive degree of diversity regarding various producer services industries in a city. In addition, we further examine two underlying mechanisms. A majority of the previous literature has used a framework that decomposes the carbon reduction impacts of industrialization or urbanization into three channels of scale, composition, and technology [1,22,24]. We argue that PSA advances carbon reduction by transforming regional industrial structure and promoting green development capabilities. Empirically, we investigate changes in industrial structure from the perspective of new firm entries and analyze city’s green development capabilities by constructing a comprehensive pollution index and using green innovation patent information.

The remainder of this paper is organized as follows. Section 2 details the study’s theoretical foundations. Section 3 presents the empirical framework. Section 4 reports the results of the baseline model and robustness tests. Section 5 examines the results of the mechanism analyses, and Section 6 presents our heterogeneity analyses. Section 7 discusses our results and points out policy implications. Finally, we provide concluding remarks in Section 8.

2. Theoretical Analyses

In this section, we detail the theoretical foundations of how PSAD influences regional carbon emissions. Figure 2 presents the theoretical framework of this study.

Agglomeration economies represent the most important underlying law for the spatial concentration of industrial activities. Theoretically, economic entities in close spatial proximity can harness agglomeration externalities through sharing, matching, and learning, which facilitates increasing returns to scale for the regional economic system [27]. Specialization and diversification, corresponding to Marshall externalities and Jacobs externalities, respectively, are the two commonly used dimensions for analyzing or quantifying agglomeration effects [28]. Specialization measures a ratio, specifically the share of a certain industry’s activities in a city compared to the average level across all cities. Diversification generally reflects the variety of all industries within a city compared to the average level. Producer services generally refer to multiple activities that support producers such as research and development, transportation and communication, financial services, business services, and digital services [23]. Thus, diversification is a more reasonable measure for analyzing the influence and function of producer services from a regional perspective.

We argue that diversified PSAs can generally promote CEI reduction in a city. It is imperative to explain the key mechanisms driving this effect. Generally, the CO₂ emissions in a region can be decomposed into three channels, namely, scale, composition, and technique effects [15,24,29]. Based on the calculation method, the CEI indicator has already incorporated scale variables such as GDP. Thus, the PSAD’s CEI reduction impact is primarily rooted in structural and technique channels.

The first mechanism is accelerating the transformation of the industrial structure. It has been well documented that secondary industries, particularly manufacturing, contribute the most to China’s CO₂ emissions [1,13,30]. Purchasing diversified external producer services can help manufacturing firms save transaction costs, enjoy labor sharing benefits, and mitigate the operation risks, thereby enhancing their internal productivity and external competitiveness [7,8,31,32]. Compared to the incumbents, new manufacturing firms may be obstructed by the more intense competitive environment caused by PSAs in a city. Moreover, firms in the secondary sector are more sensitive to increases in land prices or wages, which often result from highly clustered producer services [33]. As a result, PSAD above a certain level could continue to squeeze out industrial activities in the secondary sector, promoting servitization in the region. In this regard, the industrial structure transition from a secondary- to a tertiary-dominated form, which is related to the composition effect, can mitigate CEI [1,22,24].

The second mechanism is improvement in the region’s green development capabilities. Clusters of diversified industries in a city can cultivate cross-fertilization and spillovers of creative ideas, technologies, and innovations [28], which is called Jacobs externalities. Nearby diversified PSAs can help manufacturing firms promote innovation in manufacturing processes through knowledge spillover effects [7,8,9,10]. In a better innovation environment, polluting enterprises are more likely to promote sustainable and environmentally friendly production through technological improvement [34,35,36]. On the other hand, given that PSAs can enhance manufacturing productivity, previous research also demonstrated that improving productivity can optimize firms’ carbon emissions efficiency and movement toward more sustainable production modes [37,38,39]. Therefore, PSAD can enhance a city’s overall green development capacity, which is related to the technique effects. Many studies further argued that green innovation or eco-friendly development can promote CEI reduction or enhance the carbon emissions efficiency in a region, whether in developed countries [40] or in the developing world [41,42,43].

Based on the above analyses, we propose the following three hypotheses to be empirically tested:

Hypothesis 1.

Stronger PSAD in a city generally mitigates its CEI.

Hypothesis 2.

Improving the diversification degree of producer services agglomeration can accelerate the transformation of a city’s industrial structure, which has been widely proven to mitigate CEI.

Hypothesis 3.

Improving the diversification degree of producer services agglomeration can enhance a city’s green development capabilities, which has been widely proven to be able to mitigate CEI.

3. Empirical Framework

3.1. Econometric Models

To investigate the average effects of PSAD on CEI, we estimate the baseline model as follows:

$$\begin{array}{l}{LnCEI}_{c,t}={\alpha }_0+{\alpha }_1{LnPSAD}_{c,t-1}+{\alpha }_2{LnPSE}_{c,t-1}+{\alpha }_3{LnME}_{c,t-1}+{\alpha }_4{LnGDP}_{c,t-1}+{\alpha }_5{Indusstr}_{c,t-1}\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}+{\alpha }_6{Lnwage}_{c,t-1}+{\alpha }_7{LnPatent}_{c,t-1}+{\theta }_c+{\pi }_t+{\epsilon }_{c,t}\end{array}$$

(1)

where $c$ denotes a prefectural-level city in mainland China comprising a total of 252 units. $t$ denotes a year, from 2005 to 2018. Our dependent variable (${LnCEI}_{c,t}$) represents the natural logarithm of the CEI of city $c$ in year $t$. Our core explanatory variable (${LnPSAD}_{c,t-1}$) is the natural logarithm value of the PSAD indicator. ${\alpha }_1$ is the coefficient of interest, which we expect to be negative.

We introduce a series of controls to mitigate endogeneity problems in terms of omitted variables. ${LnPSE}_{c,t-1}$ and ${LnME}_{c,t-1}$ control for influences of local industrial size. As for potentially disturbing factors concerning overall local development status, ${LnGDP}_{c,t-1}$ and ${Indusstr}_{c,t-1}$ in our model control for the overall scale and composition effects, respectively, while ${Lnwage}_{c,t-1}$ and ${LnPatent}_{c,t-1}$ both represent the overall technique effects. We lag the core explanatory variable and all control variables by one time period to mitigate reverse causality. To further mitigate the omitted variables problem, we introduce city fixed effects ${\theta }_c$ and year fixed effects ${\pi }_t$. ${\alpha }_0$ represents the constant term and ${\epsilon }_{ct}$ is the error term.

We argue that PSAD enhancement reduces CEI by promoting industrial structure transformation and green development capabilities. Since new endogeneity issues are likely to arise, we do not use the mediating effect specification to examine the role of the proposed mechanisms. As we established in Section 2, the mitigating effects of industrial transformation and green development on CEI have been supported by a large body of empirical evidence. Building upon Equation (1), we introduce additional investigations using Equations (2) and (3) to examine these two mechanisms, respectively:

$$\begin{array}{l}{Industrans}_{c,t}={\beta }_0+{\beta }_1{LnPSAD}_{c,t-1}+{\beta }_2{LnPSE}_{c,t-1}+{\beta }_3{LnME}_{c,t-1}+{\beta }_4{LnGDP}_{c,t-1}+{\beta }_5{Indusstr}_{c,t-1}\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}+{\beta }_6{Lnwage}_{c,t-1}+{\beta }_7{LnPatent}_{c,t-1}+{\theta }_c+{\pi }_t+{\epsilon }_{c,t}\end{array}$$

(2)

$$\begin{array}{l}{Greendeve}_{c,t}={\gamma }_0+{\gamma }_1{LnPSAD}_{c,t-1}+{\gamma }_2{LnPSE}_{c,t-1}+{\gamma }_3{LnME}_{c,t-1}+{\gamma }_4{LnGDP}_{c,t-1}+{\gamma }_5{Indusstr}_{c,t-1}\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}+{\gamma }_6{Lnwage}_{c,t-1}+{\gamma }_7{LnPatent}_{c,t-1}+{\theta }_c+{\pi }_t+{\epsilon }_{c,t}\end{array}$$

(3)

where ${Industrans}_{c,t}$ denotes the industrial structure transformation of the city $c$ in year $t$, for which we construct three variables to quantify agglomeration changes of secondary and tertiary industries; ${Greendeve}_{c,t}$ represents the city $c$’s green development capability, which is also quantified using two separate variables. As in Equation (1), our interest is in the coefficient ${LnPSAD}_{c,t-1}$, assuming that stronger PSAD will accelerate the city’s industrial transformation from the secondary to the tertiary sector and promote the city’s green development. All other settings in Equations (2) and (3) are the same as in Equation (1).

Finally, we also examine differential impacts of the PSAD using the following model:

$$\begin{array}{l}{LnCEI}_{c,t}={\omega }_0+({\mathbf{D}}_{\mathbf{c}}^{\mathbf{\prime }}\times {\mathit{LnPSAD}}_{c,t-1})\mathsf{\phi }+{\omega }_1{LnPSE}_{c,t-1}+{\omega }_2{LnME}_{c,t-1}+{\omega }_3{LnGDP}_{c,t-1}+{\omega }_4{Indusstr}_{c,t-1}\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}+{\omega }_5{Lnwage}_{c,t-1}+{\omega }_6{LnPatent}_{c,t-1}+{\theta }_c+{\pi }_t+{\epsilon }_{c,t}\end{array}$$

(4)

where ${\mathbf{D}}^{\prime }$ denotes characteristics in the heterogeneity analysis; for example, ${\mathbf{D}}^{\prime }$ may include dummies indicating whether a city is in eastern China or other regions. Other variables and regression settings are defined in the same way as those in Equation (1).

3.2. Variables and Data

3.2.1. Measuring Carbon Emissions Intensity

The data source for CO₂ emissions is the Open-Data Inventory for Anthropogenic Carbon dioxide (ODIAC) from the CGER in Japan (Open-Data Inventory for Anthropogenic Carbon dioxide, Centre for Global Environmental Research: https://db.cger.nies.go.jp/dataset/ODIAC/ (accessed on 12 December 2023)). Based on nighttime light data and individual power plant emissions profiles primarily representing fossil fuel combustion, cement production, and gas flaring, the ODIAC offers a precise estimation of the spatial distribution of fossil fuel CO₂ emissions globally from 2000 to 2021. The original data are offered in a GeoTIFF version (1 km × 1 km) and we aggregate these data into a panel of total CO₂ emissions (in tons) for each city in China from 2005 to 2018.

CEI refers to CO₂ emissions per unit of some economic variable, typically referring to output or energy consumption. Lower CEI values generally indicate better energy efficiency or green development in a region or industry. Referencing several relevant studies [21,23,24], we calculate the CO₂ emissions per nominal GDP of the city as the CEI indicator, using other measures in the robustness tests. The data for nominal GDP are collected from China’s Urban Statistical Yearbooks for each year from 2005 to 2018.

Figure 3 and Figure 4 present the mean of the log total CO₂ emissions and CEI from 2005 to 2018 for each city in our sample set, respectively, revealing obvious spatial differences in carbon emissions activities. Notably, the spatial patterns exhibited by the two indicators differ. For example, coastal cities in eastern China have relatively higher total emissions but lower intensity. In contrast, the total emissions and CEI of cities in northeast China, Shanxi, Inner Mongolia, Ningxia, and Gansu are higher.

3.2.2. Measuring Producer Services Agglomeration Diversification

Referencing Combes, (2000) [26], we introduce a modified version of the inverse Herfindahl index to measure the comprehensive degree of diversity regarding PSAD in city $c$ in year $t$, which is denoted by ${PSAD}_{c, t}$ as follows:

$${PSAD}_{c, t}=\sum _{i\in I}\frac{{psem}_{i,c,t}}{{cityem}_{c,t}}(\frac{\frac{1}{\sum _{\begin{array}{c}j\in I\\ j\ne i\end{array}}{\left[\frac{{psem}_{j,c,t}}{{cityem}_{c,t}-{psem}_{i,c,t}}\right]}^2}}{\frac{1}{\sum _{\begin{array}{c}j\in I\\ j\ne i\end{array}}{\left[\frac{{psem}_{j,t}}{{cityem}_t-{psem}_{i,t}}\right]}^2}})$$

(5)

where $i$ represents a certain type of producer service industry and $I$ is the set of producer services categories. According to the NBS and previous research [8,10,32], we classify five SIC sectors as producer services industries, including financial services; rental and business services; scientific research and technology services; transportation, storage, and postal services; and transmission, software, and information technology services. Here, ${psemp}_{i,c,t}$ denotes the employment size of producer services industry $i$ in city $c$ in year $t$, and ${psem}_{j,c,t}$ is the employment of a differing producer services industry ($j$) from producer services industry $i$; similarly, ${psem}_{i,t}$ represents the total employment of $i$ in all cities, and ${psemp}_{j,t}$ is the total employment of any producer services industry $j$ different from $i$. Finally, ${cityemp}_{c,t}$ represents the total employment of city $c$, and ${cityemp}_t$ denotes the total employment of all cities. Higher ${PSAD}_{c, t}$ indicates stronger diversification and we take its natural logarithm to construct our core explanatory variable, $Ln{PSAD}_{c, t}$. Data referring to the employment size of producer services industries are obtained from China’s Urban Statistical Yearbooks for each year from 2005 to 2018.

3.2.3. Mechanism Variables

We first want to investigate whether PSAD clusters promote urban servitization, which can advance the industrial composition effect. We use three variables regarding new firm entry rates for secondary and tertiary sectors to quantify cities’ industrial transformation, which is denoted by ${Industrans}_{c,t}$ in Equation (2). The first is the log of new secondary sector firms divided by the city’s total incumbent employment, and the second variable denotes new industrial firms divided by the city’s secondary sector employment. The third simultaneously concerns new entries in the secondary and tertiary sectors, which is the log of new firms per employment in the tertiary sector divided by new firms per employment in the secondary sector. The data for the employment are retrieved from China’s Urban Statistical Yearbooks from 2005 to 2018, and the new firm entries dataset is obtained from the registry database of the State Administration for Industry and Commerce from 2005 to 2018.

Second, diversified producer services clustering, which is denoted by ${Greendeve}_{c,t}$ in Equation (3), may also mitigate carbon intensity by promoting green development. PSAD may enhance the effectiveness of urban environmental governance. Using principal component analysis (PCA), we construct an index to comprehensively measure cities’ pollution emissions. The basic variables include sulfur dioxide, nitrogen oxide, smoke and dust, and wastewater emissions from industrial production processes (The Bartlett’s test for sphericity shows that these selected variables are significantly intercorrelated. The Kaiser–Meyer–Olkin index is 0.773, indicating that conducting the PCA using these variables is appropriate. According to the PCA results, two principal components, denoted by $f_1$ and $f_2$, are selected to calculate the pollution index using the following formula as follows: $pollution index$ =$(0.6820\times f_1+0.1450\times f_2)/0.8270$). We expect a negative relationship between PSAD and cities’ pollution emissions intensity. The pollution data are retrieved from China’s Urban Statistical Yearbooks from 2005 to 2018. We posit that PSAD enhances cities’ green innovation capabilities. We identify green patents based on classification number information using the classification standards issued by the World Intellectual Property Organization (WIPO) (World Intellectual Property Organization IPC Green Inventory: https://www.wipo.int/classifications/ipc/green-inventory/home (accessed on 12 December 2023)). We then calculate the log of total green patent applications to represent the city’s green innovation capability. The source of micro-level patent data is the China National Intellectual Property Administration. We expect a positive relationship between PSAD and cities’ green innovation capability.

3.2.4. Controls

A series of controls are needed to alleviate problems with omitted variables. One set of controls concerns the local industrial size, for which we introduce ${LnPSE}_{c,t-1}$ and ${LnME}_{c,t-1}$ to control for the overall employment size of producer services and manufacturing industries in city $c$, respectively. Another series of controls refers to factors of overall local development status, which have been frequently used in previous studies. Specifically, the impact of local economic development on CEI is decomposed into scale, composition, and technology effects [15,24,29]. Scale effects are controlled by ${LnGDP}_{c,t-1}$ in our models, which is denoted by the log real GDP. Specifically, using the year 2000 as the baseline period, we calculate a city’s real GDP based on GDP deflator indices of the province where the city is located. The variable ${Indusstr}_{c,t-1}$, reflecting the city’s overall industrial structure, controls for composition effects. For technology effects, we introduce the log average wage (denoted by ${Lnwage}_{c,t-1}$) to control for the city’s overall productivity and industrial firms’ production costs. Finally, we use the log total number of patent applications (denoted by ${LnPatent}_{c,t-1}$) to control for the city’s innovation capabilities.

We use the OLS method for estimation and conduct model analysis based on Stata 17.0.

Table 1. Definition and statistical description of the main variables in the baseline model.

Variables	Definitions	Observations	Mean	Std. Dev.
$LnCEI$	Log carbon emissions intensity per nominal GDP	3772	8.520	0.697
$LnPSAD$	Log producer services agglomeration diversification	3772	0.202	0.091
$LnPSE$	Log producer services employment	3772	6.554	0.858
$LnME$	Log manufacturing employment	3772	10.724	1.117
$LnGDP$	Log real GDP	3772	15.677	0.963
$Indusstr$	Ratio of the tertiary industry output to secondary industry output	3772	0.844	0.404
$Lnwage$	Log average wages	3772	10.263	0.575
$LnPatent$	Log total green patent applications	3772	17.896	5.480

presents the definitions and basic statistical information for the main variables in our baseline model.

4. Baseline Results and Robustness Analyses

4.1. Baseline Results

Table 2. Baseline results.

	$\mathbf{DV}: \mathit{l}\mathit{n}\mathit{C}\mathit{E}\mathit{I}$
	(1)	(2)	(3)	(4)
$LnPSAD$	−1.695 ***	−0.302 ***	−0.181 ***	−0.145 ***
	(0.257)	(0.048)	(0.055)	(0.037)
$LnPSE$			−0.043 **	−0.023 *
			(0.017)	(0.012)
$LnME$			−0.081 ***	−0.057 ***
			(0.011)	(0.009)
$LnGDP$				−0.621 ***
				(0.036)
$Indusstr$				0.050 ***
				(0.013)
$Lnwage$				−0.107 ***
				(0.026)
$LnPatent$				−0.006 ***
				(0.001)
City fixed effects	Yes	Yes	Yes	Yes
Year fixed effects	No	Yes	Yes	Yes
Constant	8.861 ***	8.580 ***	9.707 ***	20.201 ***
	(0.052)	(0.010)	(0.118)	(0.538)
Observations	3780	3780	3780	3772
Adj. R-squared	0.743	0.961	0.963	0.977

Notes: Numbers in parentheses are robust standard errors; ***, **, and * denote significance at 1%, 5%, and 10% levels, respectively.

presents our baseline results from estimating Equation (1). Controlling only city fixed effects, column (1) presents the result of a univariate regression of the dependent variable ($lnCEI$) on the core explanatory variable ($LnPSAD$), indicating strong carbon reduction effects from PSAD. Following column (1), columns (2)–(4) sequentially add year fixed effects, local industrial size, and overall local development status controls into the model. The findings reveal that with the addition of control variables and fixed effects terms, the adjusted R-squared for each specification presents a gradual upward trend, increasing from 0.743 to 0.977. Although the absolute value decreases markedly from 1.695 to 0.145, the estimated elasticities of PSAD remain significantly negative throughout all specifications, indicating stable and strong carbon intensity reduction effects. The regression in column (4), corresponding to our baseline specification shown in Equation (1), shows that controlling for other conditions, a 1% increase in PSAD will significantly reduce a city’s CEI by about 0.145%.

4.2. Robustness

To further test the rationality of our baseline results, we conduct robustness tests from two aspects. First, we reconstruct the dependent variable, $lnCEI$. We first use new carbon emissions data provided by the Emissions Database for Global Atmospheric Research (EDGAR) from the European Commission to recalculate the CEI (The website is https://edgar.jrc.ec.europa.eu/emissions_data_and_maps (accessed on 12 December 2023)). As shown in column (1) in

Table 3. Results of robustness tests.

	$\mathbf{DV}: \mathit{l}\mathit{n}\mathit{C}\mathit{E}\mathit{I}$
	(1)	(2)	(3)	(4)	(5)
$LnPSAD$	−0.126 ***	−0.203 ***	−0.078 ***	−0.053 *	−0.143 ***
	(0.039)	(0.054)	(0.029)	(0.028)	(0.047)
$LnPSE$	0.009	−0.026 *	−0.003	−0.008	−0.016
	(0.014)	(0.015)	(0.010)	(0.009)	(0.015)
$LnME$	−0.054 ***	−0.074 ***	−0.047 ***	−0.054 ***	−0.075 ***
	(0.010)	(0.011)	(0.008)	(0.007)	(0.011)
$LnGDP$	−0.715 ***	−0.679 ***	−0.642 ***	−0.828 ***	−0.385 ***
	(0.042)	(0.041)	(0.036)	(0.018)	(0.031)
$Indusstr$	0.112 ***	0.458 ***	0.007	0.029 **	0.061 ***
	(0.017)	(0.021)	(0.012)	(0.012)	(0.014)
$Lnwage$	−0.268 ***	−0.213 ***	−0.049 **	−0.073 ***	−0.111 ***
	(0.032)	(0.033)	(0.023)	(0.020)	(0.029)
$LnPatent$	−0.010 ***	−0.008 ***	0.000	−0.003 ***	−0.009 ***
	(0.002)	(0.002)	(0.001)	(0.001)	(0.001)
City fixed effects	Yes	Yes	Yes	Yes	Yes
Year fixed effects	Yes	Yes	Yes	Yes	Yes
Constant	24.473 ***	22.869 ***	19.942 ***	23.011 ***	16.677 ***
	(0.635)	(0.634)	(0.518)	(0.336)	(0.520)
Observations	3772	3772	3772	4024	3522
Adj. R-squared	0.974	0.967	0.979	0.983	0.972

Notes: Numbers in parentheses are robust standard errors; ***, **, and * denote significance at 1%, 5%, and 10% levels, respectively.

, PSAD still exhibits significant carbon reduction effects based on the EDGAR data. Second, instead of dividing it by the nominal GDP, we replace the denominator with the value-added of the secondary industry and with the real GDP, respectively. The results in columns (2) and (3) in Table 3 both confirm that the carbon reduction effect of PSAD remains robust under different CEI calculation methods. Next, we change the lag periods of all explanatory variables. In column (4), we use all explanatory variables’ present value, and the results reveal an insignificant influence from PSAD. In contrast, lagging the present results by 2 years in column (5) reveals a significant carbon reduction effect with close degrees to our baseline result.

Based on our baseline results in Table 2 and Table 3, we can conclude that stronger PSAD in a city generally mitigates its CEI, validating Hypothesis 1.

5. Mechanism Analyses

Table 4. Results of mechanism analyses.

	$\mathbf{DV}: \mathit{I}\mathit{n}\mathit{d}\mathit{u}\mathit{s}\mathit{t}\mathit{r}\mathit{a}\mathit{n}\mathit{s}$			$\mathbf{DV}: \mathit{G}\mathit{r}\mathit{e}\mathit{e}\mathit{n}\mathit{d}\mathit{e}\mathit{v}\mathit{e}$
	(1)	(2)	(3)	(4)	(5)
$LnPSAD$	−0.470 **	−0.545 ***	4.991 ***	−0.377 **	0.414 **
	(0.196)	(0.208)	(1.710)	(0.158)	(0.202)
$LnPSE$	−0.353 ***	−0.222 ***	−0.570	0.014	0.253 ***
	(0.043)	(0.041)	(0.378)	(0.043)	(0.046)
$LnME$	−0.128 ***	−0.389 ***	2.973 ***	0.067 **	0.011
	(0.024)	(0.026)	(0.219)	(0.034)	(0.031)
$LnGDP$	−0.213 ***	−0.248 ***	1.794 ***	0.161 **	0.217 ***
	(0.044)	(0.046)	(0.455)	(0.072)	(0.077)
$Indusstr$	−0.053 *	0.030	−0.097	0.004	−0.054
	(0.029)	(0.032)	(0.248)	(0.059)	(0.050)
$Lnwage$	0.636 ***	0.529 ***	1.147	0.157 *	−0.033
	(0.066)	(0.068)	(0.698)	(0.089)	(0.093)
$LnPatent$	0.001	−0.005	−0.002	0.003	0.135 ***
	(0.003)	(0.004)	(0.034)	(0.005)	(0.005)
City fixed effects	Yes	Yes	Yes	Yes	Yes
Year fixed effects	Yes	Yes	Yes	Yes	Yes
Constant	4.702 ***	9.228 ***	−60.297 ***	−4.925 ***	−3.159 **
	(0.814)	(0.881)	(8.922)	(1.406)	(1.447)
Observations	3772	3772	3772	3772	3721
Adj. R-squared	0.807	0.832	0.750	0.907	0.949

Notes: Numbers in parentheses are robust standard errors; ***, **, and * denote significance at 1%, 5%, and 10% levels, respectively.

presents the results from estimating Equations (2) and (3). The results in columns (1)–(3) correspond to the mechanisms of the industrial structure transformation. As indicated in columns (1) and (2), PSAD strongly hinders the entry rates for the secondary sector, which includes most industries with the highest carbon dioxide emissions [1,30]. Controlling for other factors, a 1% increase in the PSAD can lead to a significant 0.47% and 0.545% reduction in new secondary sector firms per city total employment and per secondary sector employment, respectively. The results in column (3) provide more direct evidence indicating that improved PSAD will help a city develop from a secondary- to a tertiary-dominant economy. Based on the above results, it can be argued that higher PSAD can have concrete structural transformation effects, which has been proven in a considerable body of literature to have a significant influence on carbon reduction [1,22,24]. Thus, Hypothesis 2 is validated.

The results in columns (4) and (5) correspond to the green development mechanism. Column (4) reveals that controlling for other conditions, a 1% increase in PSAD significantly alleviates cities’ comprehensive pollution degree by 0.377, presenting convincing evidence that producer services clusters promote cities’ environmental governance. As most carbon emissions come from the same production processes as other pollutants, we can assume that a firm’s control over pollution emissions also affects its carbon emissions behavior. Column (5) indicates that a 1% improvement in the degree of PSAD will significantly lead to a 0.414% increase in the city’s total green innovation patent applications. Accordingly, many studies have confirmed that green innovation can contribute considerably to carbon reduction and neutrality in many countries [40,41,42,43]. Thus, Hypothesis 3 is validated.

6. Heterogeneity

We consider heterogeneity across two dimensions. Heterogeneity is estimated using Equation (4), where the variable D’ represents two sets of dummies.

Table 5. Results of heterogeneity analyses.

	$\mathbf{DV}: \mathit{l}\mathit{n}\mathit{C}\mathit{E}\mathit{I}$
	(1)	(2)
$\mathrm{Non}-\mathrm{eastern} \mathrm{regions}\times LnPSAD$	−0.006
	(0.035)
$\mathrm{Eastern} \mathrm{regions}\times LnPSAD$	−0.239 ***
	(0.066)
$\mathrm{Before} 2010\times LnPSAD$		−0.136 ***
		(0.038)
$\mathrm{After} 2010\times LnPSAD$		−0.165 ***
		(0.043)
All controls	Yes	Yes
City fixed effects	Yes	Yes
Province–year fixed effects	Yes	Yes
Constant	20.097 ***	20.192 ***
	(0.538)	(0.539)
Observations	3772	3772
Adj. R-squared	0.977	0.977

Notes: Numbers in parentheses are robust standard errors; *** denotes significance at 1% level.

presents the results of heterogeneous impacts. To save space, we only report the coefficients of the interaction terms.

The first dimension concerns regional disparities. Based on the NBS classification standard, we divide the sample cities into eastern and non-eastern regions according to province locations. According to the results in column (1), improving PSAD only significantly mitigates the CEI in cities within non-eastern China. There may be two reasons for this. First, the effect of PSAD may be marginally decreasing, so increasing PSAD should have significant carbon mitigation effects on inland cities with higher CEI compared with eastern coastal cities in China (as shown in Figure 3). Second, the technology effects of the PSAD may also have rebound effects, meaning that an improved technology level for a region is not only helpful to carbon reduction, but also may cause new energy demand Based on the specification of column (1) in Table 5, we replace the dependent variable $lnCEI$ by the log total CO₂ missions. The results show that higher PSAD leads to significantly more carbon emissions in cities within eastern China. Therefore, eastern regions with stronger technology capabilities are more likely to have rebound effects in terms of carbon emissions, which may be the reason for the insignificant effect of PSAD on eastern cities.

Second, we examine heterogeneous influences across time, specifically investigating the differences in the influence of PSAD before and after 2010. As indicated in column (2), the carbon mitigating effects of PSAD are significant throughout the observation period, but notably stronger after 2010. Around 2010, China surpassed the US, becoming the world’s largest manufacturing country. Meanwhile, China began to encourage industrial structure upgrading through a series of policies and actively explore sustainable development paths. For example, Cao et al., (2014) [44] found that the average CEI decreased considerably during the 12th Five-year Plan period in China, which is attributed to the implementation of multiple policies regulating coal usage. Therefore, during historical periods when the national policy environment is more inclined to promote green and sustainable development, the carbon reduction effect of PSAD may be amplified.

7. Discussion and Policy Implications

Our baseline empirical results indicate that stronger PSAD in a city can generally mitigate its CEI. Contrary to direct governmental interventions, the impacts of PSAD operate through inter-industry connections and market dynamics. The existing literature has suggested many challenges with solely government-led decarbonization efforts, such as conflicting urban planning strategies [45] and efficiency limitations [46]. Given that secondary industries are the primary carbon emitters, a bottom-up market-driven approach could better enhance efficiency. Our findings indicate that using market-driven policies to promote diversified PSAs can effectively reduce the CEI at the city level.

Examining more specific targeted policies, it is apparent that understanding PSAD’s mechanisms is crucial. Our empirical findings highlight that PSAD limits secondary sector growth in cities, driving a shift towards service-based economies, which leads to notable composition effects in carbon reduction. This shift is driven by rising factor costs, notably land, crowding out more cost-sensitive industries. Concurrently, PSAD can also augment pollution control efficiency and foster innovations in green and decarbonized production. We contend that leveraging PSAD’s promoting effects of green development capabilities is paramount. Under the industrial transition mechanism, PSAD displaces or phases out certain high-carbon emitting industrial activities from the city. This may not only elevate the CEI in areas receiving these relocated industries, thereby compromising regional decarbonization efforts, but also exert pressures on the economic system, such as short-term employment strains and disruptions in supply chains. Given this, policymakers should prioritize strengthening green development mechanisms. We propose that targeted policies should focus on the following perspectives:

First, the paramount direction should center on land utilization to refine the industrial spatial layout within cities. Building on earlier discussions, ensuring the efficacy of PSAD necessitates geographical proximity between diversified PSAs and major industrial carbon emitters. Land serves as the basic spatial medium for industrial activities, with urban areas generally exemplifying intensive land use. The extant literature underscores that land utilization significantly impacts carbon emissions in cities [47,48]. Within the institutional context of China, it is typically the responsibility of local urban governments to plan land use and industrial spatial arrangements within their jurisdictions. Hence, policymakers should prioritize strategies related to land supply and allocation when deliberating on specific policy measures to harness the benefits of PSAD. For example, the city government could expand commercial land within current industrial clusters or create new industrial parks, ensuring coordinated land allocation for both service and manufacturing sectors.

Second, policies should also be viewed from a regional perspective. Our findings suggest that PSAD can reduce CEI by displacing some secondary industries within a city. Regionally, industrial CO₂ emissions often shift from large cities to nearby smaller urban areas or towns. Consequently, while large cities reduce CEI through structural adjustments, surrounding areas might see heightened emissions, potentially offsetting regional carbon efficiency gains. Moreover, due to limited and specialized PSAs in smaller cities, leveraging the local PSAD for green development becomes challenging. We propose alleviating this issue through a multifaceted policy approach. For example, bolstering intercity transportation or digital infrastructure can help reduce delivery costs for producer services, allowing smaller cities easier access to diversified PSAs from neighboring large cities. Higher-level government bodies can guide the allocation of producer services resource to smaller cities, potentially through the development of regional industrial parks or offering tax incentives to encourage major service firms to expand into these areas.

Third, policymakers should consider regional heterogeneity in the impacts of PSAD. Our results indicate that PSAD’s carbon reduction effects are mainly significant in non-eastern cities. Due to geographical advantages, China’s coastal areas have larger-scale manufacturing agglomerations and higher PSAD levels. Comparing Figure 3 and Figure 4, while coastal cities emit more CO₂, their CEI remains lower, hinting at great scale effects. We believe this is mainly because the eastern regions likely possess a more carbon-efficient industrial composition and innovation capabilities. While PSAD has an impact, its significance reduces after accounting for composition and technological advancements. Thus, leveraging PSAD’s benefits is essential for eastern cities, achievable through policies like financial subsidies and land allocations to support green service firms and low-carbon industrial chains. For inland cities, strategies could include dedicating service-oriented land within industrial clusters.

Fourth, integrating PSA allocation with other institutional measures is crucial. Our analysis reveals that post-2010, PSAD’s impacts on CEI reduction have strengthened, likely due to China’s shift towards green sustainable development, thereby fostering a more favorable policy environment. Hence, complementary institutional policies can enhance PSAD’s benefits. Regionally, adjacent cities could collaborate on green infrastructure or carbon trading markets. Local city governments could implement a carbon management system for industrial land parcels and promote redevelopment of high-emission sites into low-carbon parks. Regarding firm-level policies, tax incentives, subsidies, and eco-friendly financing alternatives can effectively advance low-carbon production innovations.

8. Conclusions

Exploring effective approaches to reduce carbon emission intensity at the city level through market mechanisms is crucial for promoting sustainable development. This study presents novel evidence exploring an important influencing factor for the marked spatial variations in carbon emissions activities across cities in China. Specifically, based on a panel data set including 252 Chinese prefectural-level cities from 2005 to 2018, we empirically investigate the mitigating effects of the PSAD on CEI. The results show that stronger PSAD has significant mitigating effects on CEI at the city level. As for the mechanisms, improving PSAD promotes the industrial transformation from secondary to tertiary sector and cities’ green development capabilities, which are both confirmed to have concrete influence on carbon reduction. Finally, PSAD only significantly promotes CEI mitigation in non-eastern cities in China, and the carbon reduction benefits from PSAD clusters are stronger after 2010 in China. Our results primarily demonstrate the importance of leveraging existing PSAD as a strategy for guiding carbon emissions behavior and improving carbon emissions efficiency.

To effectively exploit PSAD’s carbon-reducing benefits, especially in promoting green development capabilities, our policy recommendations emphasize the critical role of land utilization in shaping intracity PSA layouts and the need for measures that consider regional perspectives. Furthermore, recognizing regional differences and integrating PSA allocation with broader institutional measures can amplify PSAD’s benefits.