Drivers of Carbon Emissions in China’s Construction Industry: A Perspective from Interregional Carbon Transfer

Abstract

The construction sector is pivotal to China’s economy, with interregional trade driving carbon emission transfers. This study uses 2007–2017 panel data from 30 provinces to analyze provincial carbon transfer patterns, combining input-output analysis and complex network theory. Key findings reveal: (1) Construction emissions remained high nationwide, showing upward trends in most provinces. Major development zones dominated carbon outflows, while demand-driven and balanced development zones received significant inflows; (2) Emission outflow intensity, population size, urbanization rate, economic scale, and mechanical equipment usage positively correlated with emissions, whereas construction workers’ per capita income showed mitigating effects. Other factors demonstrated negligible impacts. These insights enable differentiated mitigation strategies: For outflow-intensive provinces in developed regions, recommendations focus on green supply chain management and advanced emission-control technologies. Inflow-concentrated areas should prioritize low-carbon procurement standards and circular construction practices. Nationwide implementation of carbon accounting mechanisms and interprovincial compensation frameworks is proposed to address transfer inequities. The integrated methodology provides a novel perspective for quantifying emission responsibilities across value chains, supporting China’s dual carbon goals through spatially optimized governance approaches.

1. Introduction

In recent years, China’s carbon emissions have increased dramatically along with its rapid economic development. In 2007, China’s carbon emissions became the world’s highest, surpassing those of the United States, Europe, and other developed economies [1]. To cope with the increasingly serious environmental problems, President Xi Jinping proposed the “dual carbon” goal of “carbon peak” by 2030 and “carbon neutrality” by 2060 at the 75th session of the UN General Assembly [2,3]. The “dual-carbon” goal urgently needs a solution to China’s rapid economic growth. The United Nations Intergovernmental Panel on Climate Change (IPCC) further suggests that the world must achieve zero net CO2 emissions, i.e., become carbon neutral, by 2050 to meet these goals [4]. China urgently needs to address the harmonious unity of rapid economic growth and green low-carbon development.2023 At the end of the year, the China Building Energy Efficiency Association released the “Research Report on Carbon Emissions from China’s Buildings and Urban Infrastructure”, which pointed out that the total carbon emissions from the whole process of the national construction industry amounted to 5.01 billion tons of CO₂, accounting for 47.1% of the country’s energy-related carbon emissions. In addition, in recent years, China has been emphasizing a new development pattern in which the domestic macrocycle is the main body and the domestic and international double cycles promote each other, and more and more attention has been paid to strengthening domestic demand and increasing domestic trade. From the data of the National Bureau of Commerce on the total amount of domestic trade, the total amount of domestic trade in 2017 reached 10 trillion yuan, which is second only to China’s manufacturing industry, and thus domestic trade has become the main way of China’s economic development, tax increase and the main way to increase employment. Therefore, it is very crucial to clarify the carbon transfer caused by domestic interregional trade in the construction industry and its driving factors to control the total amount of carbon emissions in China and to realize China’s energy conservation and emission reduction and the “double carbon” goal, which is of great significance to the realization of the low-carbon development of the construction industry and the “double carbon” goal. It is essential to realize the construction industry’s low-carbon development and the goal of “double carbon”. In carbon emission measurement and carbon emission research, most of the existing studies use the carbon emission coefficient method provided by IPCC to measure carbon emissions, which is simple and easy to calculate and can more intuitively measure the CO₂ emitted by the consumption of each energy type; however, this method is more used for the measurement of carbon emissions of a single region, and it is necessary to combine with other techniques to measure interregional carbon emissions and carbon transfers. The input-output method can more accurately calculate the amount of interregional carbon transfer in the construction industry. Multiregional carbon transfer in the construction industry has spatial linkage. Its carbon emissions can be transferred to other regions through the industrial structure, and the sectors and regions will no longer be isolated individuals [5].

In summary, China is still in a period of rapid development of industrialization and urbanization, and the construction industry, as the most critical leading industry in China, generates massive carbon emissions. As the most important leading industry, the construction industry generates enormous carbon emissions, and it is typical and representative in the world to study the transfer of its carbon emissions and clarify the drivers of carbon emissions [6].

Moreover, “carbon peaking” is a key turning point in the process of achieving “carbon neutrality”, and the decoupling of economic growth and carbon dioxide emissions is its premise and foundation [7]. Therefore, developing carbon reduction and decarbonization in the construction industry and construction sector is the key to China’s fulfilling its commitments [8]. It is essential to analyze the dynamic characteristics and drivers of carbon emissions in China’s construction industry, clarify the trend of its spatial and temporal dimensions and the distribution pattern of carbon emissions in each province, and put forward the key points of carbon emission-reduction policy design to achieve the goals of “carbon peaking” and “carbon neutrality” of the localities and the industry, which is of crucial theoretical significance and practical value [9,10].

1.1. Research Status of Construction Carbon Emissions

With the intensification of global climate change, many scholars have focused their research on carbon emissions and reduction. As a significant source of carbon emissions, the construction industry has become a key area of investigation. Yuan et al. [11] emphasized that clarifying the historical characteristics and influencing factors of construction carbon emissions is crucial for formulating targeted emission-reduction policies. Zhou et al. [12] argued that the substantial carbon emissions from the construction sector hinder emission-reduction efforts and constrain the development of China’s low-carbon construction economy. Wu et al. [13] contended that achieving net-zero greenhouse gas emissions in the construction industry is vital for sustainable development policies. Their regression analysis revealed that previous studies overlooked the digital economy’s mitigating effects on construction emissions, demonstrating its significant potential to reduce urban carbon emissions in China. Zhao et al. [14] integrated carbon emission factors with ternary diagrams from materials science to conduct environmental assessments of resource allocation, developmental outcomes, efficiency, and sustainability. Wang et al. [15] identified the reduction of construction waste and carbon emissions as essential for achieving China’s “dual carbon” goals (carbon peaking and carbon neutrality), proposing regional-level efficiency evaluations as critical benchmarks for sustainable development. Hu et al. [16] characterized the construction sector as a key target for intervention due to its high energy consumption, substantial CO₂ emissions, and low resource efficiency. They advocated using the Malmquist index to tailor policy recommendations to regional attributes and systemic frameworks. Given the industry’s status as a significant global energy consumer and GHG emitter, Liu et al. [17] emphasized the urgency of forecasting construction emission trends. Their study introduced a grey prediction model optimized by particle swarm optimization and metabolic algorithms for accurate national and provincial-level carbon emission forecasting. Luo et al. [18] developed a hybrid model combining GM(1,1) and Mind Evolution Algorithm (MEA)-optimized BP neural networks to predict construction emissions in Jiangsu Province over a four-year horizon. Li et al. [19] investigated recent emission trends in Fujian Province’s construction sector, employing the Generalized Divisa Index Model (GDIM) to identify key drivers. Through entropy-weighted TOPSIS analysis, they quantified the impacts of economic, energy, social, and technological factors on emissions while evaluating time-dependent decarbonization capacities.

There are three primary academic methods to measure carbon emissions: IPCC carbon emission factor method, life-cycle assessment, and actual measurement method. Still, because the actual measurement method is too large and highly complicated and does not apply to the nationwide measurement of carbon emissions in the construction industry, this part omits the reference study of its literature.

The carbon emission factor method multiplies the consumption of various types of energy with the carbon emission factor coefficients corresponding to that energy type so that the CO₂ emissions generated by the consumption of that type of energy by that factor can be derived. This method is simple to calculate, easy to understand, and provides a more intuitive measure of CO₂ emissions from the consumption of each energy type. David et al. [20] combined the IPCC carbon emission factor method, based on the characteristics of the Brazilian coal industry, to construct a carbon emission measurement method specifically for the Brazilian coal industry. De Carvalho Jr et al. [21] explored the degree of correlation between CO₂ emission factors and carbon losses from heavy equipment used to transport ore based on the IPCC Carbon Emission Factor (CEF) methodology. Sperow [22] noted that better to reflect changes in soil organic carbon (SOC), the IPCC methodology is often used to estimate how land use and management changes affect default reference SOC stocks. Hagenbo [23] argued that the IPCC Bio Carbon Tier 1 methodology estimates biochar sequestration based on fixed emission factors. In contrast, the Tier 2 approach allows countries to use local emission factors and climate data to calculate biochar’s contribution to soil carbon sequestration, which is essential for biochar research. Peter et al. [24] constructed a new set of interregional carbon footprint measurement methods based on the IPCC carbon emission factor method according to the characteristics of agriculture.

The Life-Cycle Assessment (LCA) approach is a methodology for assessing the overall environmental impact of a product system throughout its life cycle. Boguski et al. [25] noted life-cycle assessment methods, such as the National Geographic magazine, may be used to evaluate a product’s carbon footprint. Foteinis et al. [26] used the statement cycle method to measure the key factors affecting carbon emissions from the marine industry. Di Ruocco et al. [27] used LCA to assess the extent to which building materials are linked to carbon emissions associated with the production and construction of buildings. Gustavsson et al. [28] analyzed the life-cycle primary energy use and CO₂ emissions of an apartment building. The results showed that building operations accounted for the largest share of the whole life-cycle energy use, which increased with the building’s life span. Kneifel [29] evaluated the energy-saving potential and the cost-effectiveness of carbon emission reduction of a new building over its life cycle utilizing an integrated design approach.

1.2. Research Status on Carbon Transfer

Academics mainly adopt life-cycle assessment and input-output measurement methods for carbon transfer measurement at present.

Measuring carbon transfers using full life-cycle assessment models, most studies have examined the full life-cycle carbon emissions of a particular product or industry, and few studies have been able to quantify the carbon transfer between different regions of an economy due to the complexity of data acquisition. Through LCA modeling, Mueller-wenk and Brandao [30] accurately assessed and measured the carbon transfer between vegetation, soil, and air during land transfer. Dombrowski et al. [31] assessed the carbon emissions of the construction industry during the construction phase of structural works and the carbon transfers to other phases based on LCA modeling. Xu et al. [32] developed a BIM-integrated LCA solution based on a five-tier framework (materials, components, assemblies, floor plans, and buildings) for automated assessment of the carbon content of prefabricated buildings and carbon transfer structures.

The input-output method analyzes the interdependence between different sectors in an economic system. In recent years, with the global concern for climate change, the input-output method has been widely applied to carbon emissions to track direct and indirect carbon emissions from economic activities, assess inter-sectoral carbon flows, and provide a scientific basis for formulating emission-reduction policies. Liu et al. [33] proposed a multiregional input-output analysis and a coupled coordination degree model of the world’s energy centers, including 42 industrial sectors, which was used to analyze further the transfer and interaction of carbon emissions among the world’s energy centers. Xu et al. [34] measured the inflow of carbon emissions from imported goods in Tianjin through a multiregional input-output approach. Long et al. [35] used single-region and multiregion input-output tables to compare differences in carbon footprint accounting in Tokyo, Japan. The results show that single-region input-output tables largely ignore emissions from coal, which can bias the measurement of carbon emissions.

1.3. Research Status on Factors Influencing Construction Carbon Emissions

The key to controlling carbon emissions lies in identifying their primary drivers. Therefore, investigating the factors affecting carbon emissions in the construction sector enables governments to formulate effective policies. Existing literature indicates that studies on these influencing factors predominantly focus on national or provincial levels.

Yin et al. [11] employed the Slope model, Moran’s I index, and standard deviation ellipse to reveal the spatial-temporal characteristics of carbon emissions from the construction industry, and then applied the geographical detector model to identify the main driving factors of carbon emissions. Wang et al. [36] adopted the theory of planned behavior (TPB) based on the three impact factors of behavioral attitude (BA), subjective norms (SNs), and perceived behavioral control (PBC), introduced the two potential impact factors of moral obligation (MO) and government supervision (GS), and also uses structural equation modeling (SEM) to test the impact factors in construction enterprises’ personnel’s carbon emission-reduction intention. Wang et al. [37] utilized quantile regression in 2004–2016 panel data to identify key drivers of CO₂ emissions. Wang et al. [38] studied the spatial-temporal differentiation characteristics, dynamic trend evolution, and driving factors of high-quality development of China’s construction industry from 2006 to 2021 by using the SE-SBM model of unexpected output, GML index analysis, and grey correlation model. Du et al. [39] explored the key factors affecting carbon emissions and the influencing relationships from the perspective of the supply chain, applying structural equation modeling, a hypothesis model composed of ‘social and governmental factors’, ‘market factors’, ‘technical factors’, and ‘supply chain coordination factors’ is evaluated. Han et al. [40] analyzed the influencing factors of CE in the CI based on structural equation modeling. Lai et al. [41] highlighted the critical role of the construction sector in achieving national carbon reduction targets, investigating the relationships and trends among construction carbon emissions, energy consumption, GDP growth rate, and carbon emission intensity in China. Their findings revealed that construction carbon emissions are predominantly influenced by construction GDP growth, with energy consumption emerging as the principal driving force behind carbon emission increases.

1.4. Review of Research

In summary, the issue of carbon emission transfer in the construction industry has garnered widespread attention from government agencies, international organizations, scholars, and other stakeholders globally. Current academic research provides substantial theoretical foundations and policy recommendations. However, most studies on the driving factors of construction-related emissions focus on individual regions’ intrinsic elements and characteristics (or economies). Given the increasing volume of interregional trade within China, greater emphasis should be placed on cross-regional carbon transfer. Adopting an interregional perspective to examine carbon emissions and their driving factors enables the formulation of more comprehensive carbon reduction policies.

The life-cycle approach can provide a relatively accurate assessment of carbon emissions when measuring a particular building or product. Still, its application is limited due to its high data requirements and measurement difficulties. On the other hand, the core strengths of the IPCC Carbon Emission Factor Approach are its standardization, simplicity, and international applicability. It can quickly estimate carbon emissions through uniform emission factors (such as default factors for energy consumption and agricultural activities), with easy access to data and a standardized calculation process, applicable to large-scale accounting in multiple industries and regions, and has been widely adopted by the global climate policy and carbon market to provide basic support for emission-reduction strategies. However, the method also has shortcomings, such as insufficient coverage of complex systems (e.g., the whole life cycle of each industry). However, because of the above research and considering the broad application of the carbon emission factor method and the characteristics of the construction industry, this paper will also adopt the IPCC carbon emission factor method to measure the carbon emissions of China’s construction industry. On this basis, material consumption will also be included in the scope of measurement. This paper will also incorporate the material consumption of the construction industry into the in-depth study to make up for the shortcomings of the carbon emission factor method for the assessment of the carbon emissions of the building materials production stage and synthesize the two methods as much as possible in the evaluation of the carbon emissions of the building materials production stage. This paper will also include the material consumption of the construction industry for an in-depth study to make up for the insufficiency of the carbon emission factor method in the assessment of carbon emission in the production phase of construction materials to assess the carbon emission of the construction industry by combining the two methods as far as possible. In addition, this paper will also include the material consumption of the construction industry for an in-depth study to make up for the inadequacy of the carbon emission factor method in the assessment of carbon emission at the stage of production of construction materials, to combine the advantages of the two methods as far as possible to measure the carbon emission of the construction industry.

In the measurement of carbon emission transfers, The life-cycle assessment method takes the whole life cycle of a product as the research object, through which the environmental impact of an industrial sector from the input of production materials to the output of the product can be evaluated, such as the carbon footprint of the construction industry, agriculture, real estate industry, etc. [28]. The life-cycle assessment method is more suitable for tracking the whole production of a specific industrial sector or product from a micro perspective [29]. The single-region input-output approach is usually used to calculate the carbon footprint of the supply side and the consumption side of the overall production process or the local carbon intensity. The multiregional input-output approach allows for linking different regions and studying their economic linkages because of the global and fluid nature of trade, carbon emissions, and transfers. Therefore, using a multiregional input-output approach is more appropriate for studying an economy’s carbon transfers. Both methods have their scope of application and advantages.

Building on existing research and aligning with the practical context of China’s construction industry, this study first calculates carbon emissions using the IPCC coefficient method combined with construction material data. It quantifies interprovincial carbon transfer through interregional input-output tables, followed by a preliminary spatial analysis. Subsequently, a complex network-based model is constructed to analyze China’s construction sector’s carbon emission transfer network, revealing its structural characteristics. A panel regression model is then employed to investigate the driving forces and impact levels of network metrics and other control variables on construction-related carbon emissions. Finally, targeted improvement measures and policy recommendations are proposed.

2. Materials and Methods

2.1. InterRegional Carbon Emission Transfer Model for the Construction Industry

2.1.1. Carbon Emission Measurement Model of the Construction Industry

Considering the characteristics of the construction industry, a carbon emission accounting model based on the whole process of the construction industry is constructed. The construction industry usually consists of two stages: the production of building materials and the operation of building construction. The building materials production stage is closely related to other upstream and downstream material enterprises. It represents the carbon dioxide released during the production and transportation of building materials, usually as indirect carbon emissions. On the other hand, the carbon emission in the operation stage of building construction is generated by the primary energy directly consumed by the construction industry, so it is expressed as direct carbon emission. Based on the above characteristics, this paper adds the construction material consumption factor based on the IPCC carbon emission coefficient method [42]; the carbon emission coefficients of each energy source are shown in

Table 1. Carbon Emission Factors for Different Energy Types.

Energy	Coal	Crude Oil	Petroleum	Coke	Petrol	Diesel	Diesel Oil	Fuel Oil
Carbon emission factor	2.69 Kg CO2/Kg	2.76 Kg CO2/L	2.09 Kg CO2/m3	3.14 Kg CO2/Kg	3.14 Kg CO2/L	2.56 Kg CO2/L	2.73 Kg CO2/L	3.14 Kg CO2/L

, and the carbon emission coefficients as well as the recycling coefficients of each construction material are shown in

Table 2. Carbon Emissions of Primary Building Materials, Recovery Factors.

Building Material	Clinker	Nylon	Steels	Aluminum	Lumber
Carbon emission factors	0.815 kg/kg	0.9655 kg/kg	1.789 kg/kg	2.6 kg/kg	−842.8 kg/m3
Recovery factor	—	0.7	0.8	0.85	0.2

.

The carbon emissions from China’s construction sector across 30 provinces (2007–2017) were calculated using the IPCC coefficient method. The computational model for construction carbon emissions is expressed as:

$$C=c_1+c_2=\sum _{j=1}^5{M}_j\times q_j\times \left(1-{\alpha }_j\right)+\sum _{i=1}^8{K}_i\times F_i$$

(1)

where $c_1$ denotes the carbon emissions of materials consumed in the construction industry, i.e., indirect carbon emissions from the construction sector (at the stage of production of construction materials) and $c_2$ denotes the carbon emissions of energy substances, i.e., direct carbon emissions from the construction sector (operational phase of building construction). $M_j$ denotes the total consumption of the $j$th type of construction industry, $q_j$ denotes the carbon dioxide emission coefficient of the $j$th type of construction material, and ${\alpha }_j$ denotes the recovery coefficient of the $j$th type of construction material. $K_i$ denotes the total consumption of the $i$th type of energy source, and $F_i$ is the carbon dioxide emission coefficient of the $i$th type of energy source. The calculation formula is as follows:

$$F_i={NCV}_i\times {CEF}_i\times {COF}_i\times {\displaystyle \frac{44}{12}}$$

(2)

where $NCV$ denotes the net calorific value, $CEF$ represents the carbon content per unit heat, $COF$ indicates the carbon oxidation rate during heat combustion, and the constants 44 and 12 correspond to the molecular weights of carbon dioxide (CO₂) and elemental carbon (C), respectively.

2.1.2. Interregional Construction Carbon Emission Transfer Measurement Model

Carbon emission transfer refers to how carbon-intensive industries or activities relocate from one nation/region to another. This phenomenon has intensified alongside the acceleration of international trade, industrial specialization, and globalization. Current methodologies for measuring interregional carbon emission transfers primarily fall into life-cycle assessment (LCA) and input-output analysis [43]. Given this study’s focus on interregional carbon transfers and the superior capability of input-output analysis in quantifying sectoral linkages across regions, we employ multiregional input-output (MRIO) analysis to measure construction carbon transfers among 30 Chinese provinces.

Within the input-output theoretical framework, the fundamental relationship between total output and final demand is expressed as:

$$X={\left(I-A\right)}^{-1}Y$$

(3)

where $X$ represents the total output column vector in the input-output data, $Y$ represents the end-use column vector in the input-output data, $A$ represents the matrix of direct consumption coefficients (i.e., the amount of product produced by one sector that is directly consumed by another industry), and $I$ represents the matrix of units. ${(I-A)}^{-1}$ denotes the Leontief inverse matrix, which has the economic meaning of representing the additional output of one sector that is necessary to satisfy the final demand per unit in the other industry.

(1)

Calculating Construction Carbon Emission Transfer Coefficients

Building upon the Leontief inverse matrix, the formula for determining the total carbon emission coefficient of the construction sector is derived as follows:

$$\epsilon =\theta \times {\left(I-A\right)}^{-1}$$

(4)

where:

$${\theta }^r={\displaystyle \frac{L^r}{X^r}}$$

(5)

where ${\theta }^r$ denotes the direct carbon emission factor of the construction sector in province $r$; $L^r$ denotes the total carbon emissions from the construction sector in province $r$; and $X^r$ denotes the total output of the construction sector in province $r$.

Given that this study focuses exclusively on interprovincial carbon emission transfers within the construction industry across China, we nullify carbon emission coefficients for all non-construction sectors in the transfer coefficient matrix (i.e., set values outside row 24 to zero).

(2)

Calculate the total amount of carbon transfer in the construction industry.

After obtaining the carbon transfer coefficient between provinces, we can further calculate the carbon transfer between provinces; the primary method is to multiply the complete carbon emission coefficient of the construction industry with the end-use portion of each province to other provinces in the interregional input-output table, that is, the interregional carbon emission transfer of the construction industry:

$$O^{rs}=\left[\begin{array}{ccccccc}0& 0& \dots & \dots & \dots & \dots & 0\\ \dots & \ddots & \dots & \dots & \dots & \dots & \dots \\ \dots & \dots & \ddots & \dots & \dots & \dots & \dots \\ {\epsilon }_{\mathrm{24,1}}^r& {\epsilon }_{\mathrm{24,2}}^r& {\epsilon }_{\mathrm{24,3}}^r& \ddots & \dots & \dots & {\epsilon }_{24,n}^r\\ 0& 0& \dots & \dots & \ddots & \dots & 0\\ \dots & \dots & \dots & \dots & \dots & \ddots & \dots \\ 0& 0& \dots & \dots & \dots & \dots & 0\end{array}\right]\left[\begin{array}{c}{y}_1^{rs}\\ y_2^{rs}\\ y_3^{rs}\\ \dots \\ \dots \\ \dots \\ y_n^{rs}\end{array}\right]$$

(6)

where $O^{rs}$ denotes the carbon emissions from the construction industry transferred from province $s$ to province $r$; and $Y^{rs}$ denotes the portion of services and products of the industries in province $r$ that are ultimately used in province $s$ in the multiregional input-output table.

2.2. Carbon Emission Transfer Network Model of the National Construction Industry

Based on all the carbon emission transfer relationships and carbon transfer quantities between provinces in China calculated above, a carbon emission transfer network model for China’s construction industry can be constructed, i.e., taking the provinces where the carbon emission association occurs as the nodes of the network, taking the construction carbon emission relationships existing between provinces as the edges (i.e., connecting the two provincial nodes if there is a carbon transfer occurring between the two provinces), and using the construction carbon emission existing between provinces as the weights of the network. The total amount of carbon emissions from the construction industry between provinces is the weight of the network, and a weighted directed complex network is constructed to obtain the carbon emission transfer network of China’s construction industry.

2.3. Model of Carbon Emission Drivers in the Construction Industry

Since the roles and positions of provinces in China’s carbon transfer network in the construction industry are not static, this paper considers the use of panel regression modeling based on interregional input-output data for the last five years (the latest year of data is 2017) to study the drivers of carbon emissions in the construction industry.

2.3.1. Construction of Panel Regression Model

$$Y_{it}=\alpha +{\beta }_1{X}_{it}+{\beta }_2{control}_{it}+{\epsilon }_{it}$$

(7)

where $Y_{it}$ is the dependent variable, $X_{it}$ is the independent variable, ${control}_{it}$ is the control variable, and ${\epsilon }_{it},$ it is the random perturbation term; see below for the selection of variables.

2.3.2. The Selection of Dependent and Independent Variables

Since this model investigates the driving factors of carbon emissions in the construction industry, the previously calculated Construction Industry Carbon Emissions (CDCE) are selected as the dependent variable. To explore how the positions and characteristics of provinces within the carbon transfer network influence carbon emissions, the following independent variables are incorporated: Construction Carbon Emission Inflow Intensity (INE, measured by the total carbon transfer inflow at network nodes), Construction Carbon Emission Outflow Intensity (OUTE, measured by the total carbon transfer outflow at network nodes), Construction Carbon Emission Mediation Capacity (BC, measured by the betweenness centrality of network nodes), and Degree of Influence (EC, measured by the eigenvector centrality of network nodes). The calculation formulas for these variables are as follows:

• Construction Carbon Emission Inflow and Outflow Intensity (INE, OUTE):

$${INE}_i=\sum _{j\in N_i}{w}_{ji}\left(j\ne i\right)$$

(8)

$${OUTE}_i=\sum _{j\in N_i}{w}_{ij}\left(j\ne i\right)$$

(9)

In the formula, ${INE}_i$ denotes the carbon inflow intensity of the construction industry in province $i$, ${OUTE}_i$ denotes the carbon outflow intensity of the construction industry in province $i$, $w_{ji}$ denotes the total amount of carbon emissions transferred into province $i$ from province $j$, and $w_{ij}$ denotes the total amount of carbon emissions transferred out from province $i$ to province $j$.

• Mediation Capacity (BC):

Since the betweenness centrality of network nodes primarily measures the capacity of each node to mediate transmissions, this study employs betweenness centrality to represent the mediation capacity (carbon transfer mediation capacity) of each province within the network.

$${BC}_i=\sum _{s\ne i\ne t}{\displaystyle \frac{n_{st}^i}{g_{st}}}$$

(10)

where ${BC}_i$ is the median centrality of province $i$, $n_{st}^i$ is the number of shortest paths from node $s$ to node $t$ and through node $i$, and $g^{st}$ is the number of shortest paths from node s to node $t$.

• Degree of Influence (EC):

Since the eigenvector centrality of network nodes primarily measures whether a node is connected to other significant nodes within the network, this model employs eigenvector centrality to represent the degree of influence of each province within the network.

$${EC}_i=x_i=c\sum _{j=1}^N{a}_{ij}{x}_j$$

(11)

where ${EC}_i$ is the eigenvector centrality of province $i$, $x_i$ is the importance value of node $i$, $c$ is the constant of proportionality, and $a_{ij}$ is each element in the adjacency matrix of the complex network.

2.3.3. The Selection of Control Variables

To mitigate potential bias from common factors affecting provincial construction carbon emissions and reduce regression model errors, this study selects control variables—population size (SP, measured by year-end permanent population), urbanization level (UL, urban-to-total population ratio), construction economic level (CCL, per capita construction GDP), per capita net income in construction (CE, average net income of construction employees), and mechanical equipment intensity (MES, construction machinery power per unit built area)—based on data availability, literature review, and sector-specific operational realities.

2.4. Data Sources and Process

Since the number of sections in the input-output table varies yearly and there is missing data for provinces in various years, provinces with missing data in certain years (Tibet, Hong Kong, Macao, and Taiwan) were excluded from this paper before data aggregation. As a result, to facilitate data analysis and organization, this study defines its research scope as 30 Chinese provinces based on data availability, objectivity, and alignment with national and provincial statistical realities. Province codes are listed in

Table 3. Abbreviations of Provinces.

Abbreviations of Province	Province	Abbreviations of Provinces	Province
BJ	Beijing	QH	Qinghai
TJ	Tianjin	NX	Ningxia
HE	Hebei	NM	Inner Mongolia
SX	Shanxi	LN	Liaoning
SH	Shanghai	JL	Jilin
JS	Jiangsu	HL	Heilongjiang
ZJ	Zhejiang	AH	Anhui
FJ	Fujian	SD	Shandong
JX	Jiangxi	SC	Sichuan
HA	Henan	SN	Shaanxi
HB	Hubei	GS	Gansu
GX	Guangxi	XJ	Xinjiang
HI	Hainan	GD	Guangdong
CQ	Chongqing	YN	Yunnan
GZ	Guizhou	HN	Hunan

. The original input-output tables were aggregated into 27 sectors to streamline construction carbon emission calculations, with the aggregation process omitted for brevity. The consolidated process codes, as well as the consolidated input-output tables, can be found in the Supplementary Documentation. Given that the latest interregional input-output tables are available up to 2017, this study utilizes five-year interval interregional input-output tables spanning 2007–2017. Energy data were sourced from the China Energy Statistical Yearbook, provincial control variables from the China Statistical Yearbook, construction material consumption data from the China Construction Industry Statistical Yearbook, and interregional input-output tables for 2007 [44] and 2010 [45] from Liu Weidong’s compilations, while 2012, 2015, and 2017 tables were sourced from CEADs [46]. Since the input-output tables come from different sources and different years, and their sectoral classifications are compiled according to the China National Economic Industry Classification (CNEIC), there are still sectors that are different from each other, so to unify the basis of calculation, we compare and merge the sectors of the input-output tables of the five years to form the interregional input-output table with five sectors and provinces that are the same, and the classification of sectors before and after the merger is shown in the Supplementary Material.

3. Measurement of Interregional Construction Carbon Emission Transfers

3.1. Analysis of Interregional Total Construction Carbon Emissions

To better analyze the construction industry’s carbon emissions across geographical regions in China’s 30 provinces during the study period. However, at present, there is no official categorization of the construction industry at the provincial level according to its characteristics, similar to the agricultural “main production area, main marketing area, and production and marketing balance area”. However, based on the construction industry’s output value, enterprise distribution, market demand, and other data, we can try to generalize and analyze it from the perspective of the regional economy, this study categorizes the 30 provinces into four zones based on practical considerations: Major Development Zones (MDZ), Demand-Driven Zones (DDZ), Balanced Development Zones (BDZ), and Specialized Supplementary Zones (SSZ). These classifications integrate provincial construction output value, urbanization levels, resource endowments, geographical conditions, and national policy orientations. The specific provincial compositions of each zone and their respective construction carbon emission volumes by year are presented in Figure 1.

As shown in Figure 1, construction industry carbon emissions in most Chinese provinces in 2017 exceeded those in 2007, with three distinct patterns observed. First, emissions exhibited a continuous upward trend from 2007 to 2017 in provinces such as Zhejiang, Guangxi, Fujian, Hunan, Anhui, and Chongqing. Second, emissions initially rose and declined in provinces including Heilongjiang, Hebei, Beijing, Shandong, Hainan, Jiangsu, and Shanghai. Third, provinces such as Heilongjiang, Hebei, Beijing, Shandong, Hainan, Jiangsu, Shanghai, Xinjiang, Gansu, Ningxia, Yunnan, Guizhou, Sichuan, Guangdong, Henan, Jilin, Liaoning, Shanxi, and Tianjin experienced a single-year decline amid otherwise rising emissions.

Analysis of the Four Functional Zones reveals that provinces with the highest construction carbon emissions are concentrated in the Major Development Zones (MDZ), where rapid economic growth, large populations, and escalating construction demands—driven by accelerated urbanization and surging real estate investments—directly fuel emission increases. In contrast, lower emissions are clustered in the Specialized Supplementary Zones (SSZ) and Demand-Driven Zones (DDZ). SSZ provinces, characterized by remote geography, weaker economies, smaller populations, and harsh environmental conditions (e.g., high altitude, low temperatures), face significant barriers to construction development. Furthermore, the lack of industrial clusters and supporting infrastructure leads construction-related firms to prioritize more developed central and eastern regions, further reducing activity and emissions in SSZ. DDZ provinces fall into two categories: The Demand-Driven Zones (DDZ) provinces can be broadly categorized into two groups: highly economically developed municipalities and provinces with minimal economic scale and population. In the case of financially advanced megacities, the construction industry has reached regional saturation, and these cities are not inherently dominant in construction-related activities, resulting in relatively low-carbon emissions from the sector. For provinces with sparse populations, the limited demographic size inherently restricts industrial expansion, leading to extremely low demand for construction and, consequently, minimal construction-related carbon emissions.

3.2. Analysis of Interregional Construction Carbon Emission Transfer

Based on the previously calculated interprovincial carbon transfer volumes, this study aggregates and analyzes carbon emissions and transfers across the four major functional zones, visualized using Origin 2025 software in Figure 2 and Figure 3. In these figures, the size of the circles represents the magnitude of carbon outflow and inflow, with larger circles indicating greater carbon transfers. Since there are almost all carbon transfer edges between the 30 provinces in China, close to about 900 edges many of the edges carry a limited amount of carbon transfers and are of little help in analyzing the interregional carbon transfer correlations. This paper, in order to ensure the clarity of the interregional carbon transfer diagram, adopts the common method in complex networks: the two-eight principle for filtering, and only shows the top 20% of the former carbon transfer amount. The carbon transfer relationship is ensured to maintain the clarity of the interregional carbon transfer relationship, which will not be applied to the complex network modeling, only for visualization.

Analysis of Figure 2 reveals that construction carbon outflow in China is predominantly concentrated in the Major Development Zones (MDZ), including Zhejiang, Jiangsu, Shandong, Hubei, Henan, Guangdong, and Sichuan. According to data from China’s National Bureau of Statistics (2023), eastern provinces accounted for 51.7% of the national construction industry output value, while central provinces contributed 24.3%. MDZ provinces encompass economically advanced, densely populated regions (e.g., Zhejiang, Jiangsu, Shandong, Guangdong) and regional hub provinces (e.g., Hubei, Henan, Sichuan). High construction output values correlate with rapid industry growth, driving significant carbon emissions and inevitable carbon outflow or leakage to other provinces. In contrast, the Specialized Supplementary Zones (SSZ) exhibit minimal carbon outflow due to underdeveloped economies, limited infrastructure, and geographical isolation. Most SSZ provinces are resource-exporting regions with low construction output and minimal interprovincial trade in construction-related activities.

Analysis of Figure 3 reveals that, with the exceptions of Guangdong and Henan in 2017, construction carbon inflows in China are primarily concentrated in the three zones outside the Major Development Zones (MDZ), with significant inflows observed in provinces such as Hebei, Beijing, Tianjin, Yunnan, and Shanghai. Most of these provinces belong to the Demand-Driven Zones (DDZ). For instance, Hebei, tasked with supplying construction demands for the Beijing-Tianjin-Hebei region while urgently requiring industrial upgrading, relies heavily on importing high-end, environmentally friendly building products, leading to substantial carbon inflow. Similarly, Beijing, Shanghai, and Tianjin—highly developed regions with robust construction demands (e.g., large public buildings, commercial complexes)—strictly regulate local energy-intensive industries, necessitating importing construction materials (steel, cement, glass, etc.) from high-carbon provinces. Yunnan, experiencing rapid growth in tourism and real estate in recent years, serves as an ecological barrier in southwestern China with limited local building material production capacity. Consequently, it depends on importing materials and energy fuels from neighboring provinces, further increasing carbon inflows. Additionally, the construction industry’s reliance on external resources, equipment, and products from other regions significantly contributes to rising carbon inflows.

When examining carbon transfer pathways collectively, most transfers occur between geographically adjacent regions or within the four functional zones. This pattern stems from the fact that interprovincial trade in construction and related industries favors economic collaboration among neighboring provinces, and carbon transfer pathways predominantly align with regional economic boundaries, reflecting strong geographic proximity effects.

4. Results

4.1. Analysis of China’s Construction Carbon Emission Transfer Network

4.1.1. Construction of the Complex Network for China’s Construction Carbon Emission Transfers

Following the established framework, a weighted directed complex network is constructed to represent China’s construction carbon emission transfer network. In this network, provinces with carbon emission linkages serve as nodes and interprovincial construction carbon transfer relationships form the edges (i.e., edges connect provinces if carbon transfers occur between them). The total volume of construction carbon emissions between provinces defines the edge weights.

To enhance clarity and highlight core network features, edges with cumulative edge weight probabilities below 98% are filtered out, as these redundant or near-zero connections—though inconsequential to the overall network—risk obscuring critical transfer relationships. From 2007 to 2017, the cumulative probabilities of retained edges were 68.01%, 69.99%, 82.29%, 81.06%, and 72.34%, respectively, indicating that the remaining 15% to 35% of edges accounted for merely 3% of total carbon transfers. The interregional construction carbon emission transfer network is illustrated in Figure 4.

Figure 4a–e depict China’s construction of carbon emission transfer networks for 2007, 2010, 2012, 2015, and 2017. Each circle represents a province, with its size corresponding to the node degree—larger circles indicate higher node degrees. Directed edges illustrate the direction of carbon emission transfers. At the same time, their thickness reflects the volume of embodied carbon emissions transferred from one province to another, with thicker edges denoting larger transfer quantities.

4.1.2. Analysis of Construction Carbon Emission Transfer Network Metrics

• Overall Network Characteristics

Using the statistical and computational tools of Gephi-0.10.1 software, this study measured key metrics—including nodes, edges, and related parameters—for the five-year construction of the carbon emission transfer complex network constructed earlier. The results are summarized in

Table 4. Construction Carbon Emission Transfer Network Overall Characteristics.

Year	2007	2010	2012	2015	2017
Node	30	30	30	30	30
Edge	399	414	283	382	382
Density	0.459	0.436	0.325	0.439	0.441
Average path length	1.505	1.417	1.684	1.423	1.444

.

Calculations based on the complex network reveal the overall characteristics of China’s construction carbon emission transfer network. The network density exhibits an initial decline followed by a subsequent rise, with values between 0.3 and 0.5. A higher number of edges indicates a denser network, reflecting greater carbon transfer activity and stronger network integrity. During the study period, the average path length of the network remained between 1.4 and 1.7, with a maximum variation of approximately 0.3 across years. A smaller average path length signifies tighter connectivity between nodes.

• Individual Network Characteristics

• Strength

In the complex network, inflow and outflow strength correspond to each node’s weighted in-degree and out-degree, respectively. As these metrics—provincial inflow and outflow intensity—have already been analyzed in Section 3.2, this section omits further individual network characteristic evaluations.

2.

Mediation Capacity

Each province’s mediation capacity (betweenness centrality) was calculated using Gephi software and visualized with ArcGIS, as shown in Figure 5.

As shown in Figure 5, provinces with higher mediation capacity (betweenness centrality) are predominantly concentrated in the Major Development Zones (MDZ). This is primarily because betweenness centrality reflects the importance of a node (or edge) as a “mediator” or “bridge” within the entire network. MDZ provinces, such as Henan, Jiangsu, Zhejiang, and Shanghai, are core hubs for construction industry development and maintain strong interconnections with other provinces. Henan, located in central China, links eastern coastal regions with the west and central areas, serving as a critical transit hub for energy, raw materials, and energy-intensive products, exhibiting high mediation capacity. As an economic powerhouse, the Jiangsu-Zhejiang-Shanghai (Yangtze River Delta) region is a manufacturing and export base and a key gateway for energy and commodity flows, further enhancing its mediation capacity. Notably, Hebei—though not part of MDZ—demonstrates high mediation capacity due to its role in the Beijing-Tianjin-Hebei economic zone. Hebei acts as a transit center for construction products flowing into and out of Beijing and Tianjin, fulfilling substantial mediation demands driven by the construction needs of these megacities. This intermediary role in regional supply chains elevates its betweenness centrality within the network.

3.

Degree of Influence

Each province’s degree of influence (eigenvector centrality) was calculated using Gephi software and visualized with ArcGIS, as shown in Figure 6. Eigenvector centrality emphasizes the significance of a node’s surrounding connections: a province is deemed influential if connected to numerous provinces holding essential positions within the network.

Figure 6 reveals that Henan and Guangdong in the Major Development Zones (MDZ) exhibit the highest influence levels, closely tied to their strategic geographical locations and large population scales. In the Demand-Driven Zones (DDZ), Beijing and Shanghai stand out as high-influence representatives, driven by their metropolitan scale and dynamic population mobility. In contrast, provinces within the Specialized Supplementary Zones (SSZ) and Balanced Development Zones (BDZ) generally display uniform influence levels, with no exceptionally dominant provinces. This pattern underscores how geographical positioning, demographic factors, and regional economic roles collectively shape the eigenvector centrality of provinces within China’s construction carbon emission network. This is primarily attributed to their enduring carbon linkages with high-carbon provinces within the Major Development Zones (MDZ), which sustain a higher and comparatively stable degree of influence.

4.2. Analysis of Drivers of Construction Carbon Emissions

4.2.1. Benchmark Model Selection

The data used in this paper are panel data. Therefore, the F-test and Hausman test are required to determine the final model. The results of the tests are shown in

Table 5. F-test and Hausman-test Results.

Dependent Variable	F-Test		Hausman-Test
CDCE	F-statistic	p-value	${\chi }^2$	p-value
	5.61	Prob > F = 0.0000	17.77	Prob > F = 0.0068

.

It can be seen from Table 5. Firstly, an F-test is conducted to test which one of the mixed effects model and fixed effect model is more suitable for the selected sample data; according to the test result, it can be seen that the p-value is 0.000, which is significant at 1% level, indicating that the fixed effect model is more suitable for the sample data. Then, the Hausman test is conducted to test whether the random effect model is better than the fixed effect model. According to the test results, the p-value is 0.0068, which is lower than 0.01, so the original hypothesis is rejected at a 1% level, and the fixed effect model is more suitable for the sample data selected in this paper.

4.2.2. Multicollinearity Test

To prevent the existence of serious multicollinearity problems between the variables that will have an impact on the regression results of this paper, this paper carries out the Variance Inflation Factor (VIF) test of each variable; the test results are shown in

Table 6. Results of the Multicollinearity Test.

Variable	VIF	1/VIF
CCL	3.350	0.299
UL	2.720	0.367
CE	2.500	0.401
EC	1.440	0.692
BC	1.350	0.742
OUTE	1.310	0.765
MES	1.310	0.766
SP	1.280	0.778
INE	1.030	0.967
Mean VIF	1.810

, the VIF value of each variable is located in the 3.5 or less, much less than 10, which proves that this paper carries out the hypothesis test used in the model does not exist between the variables of the serious problem of multicollinearity.

4.2.3. Heteroscedasticity Test

As can be seen from the test results in

Table 7. Heteroscedasticity Test Result (BP).

BP Test Statistic	p-Value
43.35	0.0000

, the statistical value of the BP test is 43.35, corresponding to a p-value of 0.0000 < 0.05, so the original hypothesis is rejected, indicating the existence of heteroskedasticity to eliminate the effect of heteroskedasticity, this paper adds a robust standard error to the regression process to circumvent the problem of heteroskedasticity.

4.2.4. Panel Regression Results

Given that the dataset consists of large-N, small-T, and short-panel data, this study omits correlation and stationarity tests [47,48]. The panel regression model was constructed earlier. In the analysis, construction industry carbon emissions are treated as the dependent variable, with explanatory variables including inflow intensity, outflow intensity, mediation capacity, and degree of influence. Control variables encompass population size, urbanization rate, construction economic output, per capita net income in construction, and machinery/equipment intensity) are implemented using Stata 17.0. To mitigate multicollinearity, Descriptive statistics for each variable are shown in

Table 8. Descriptive statistics of the variables.

Variables Type	Variables Name	Variable	Unit	Mean	Standard Deviation	Maximum	Minimum
Dependent Variable	Construction Carbon Emissions	CDCE	t	7.38 × 107	8.81 × 107	4.46 × 108	1.39 × 106
Independent Variables	Carbon Emission Inflow Intensity	INE	t	16,843.352	37,498.199	2.15 × 105	39.394
	Carbon Emission Outflow Intensity	OUTE	t	19,847.692	95,051.726	8.81 × 105	0.000
	Mediation Capacity	BC	-	10.693	21.956	115.985	0.000
	Degree of Influence	EC	-	0.625	0.238	1.000	0.074
Control Variables	Population Size	SP	capita	4505.100	2752.086	12,141.000	552.000
	Urbanization Rate	UL	%	0.544	0.136	0.916	0.282
	Construction Economic Output	CCL	CNY/capita	1.015	0.898	4.438	0.096
	Per Capita Net Income in Construction	CE		37,814.173	16,392.943	99,718.000	13,102.000
	Machinery/Equipment Intensity	MES	Kw/m2	0.091	0.04	0.546	0.018

. The empirical process follows the panel regression methodology outlined by Jiang et al. [49], where each variable in the model is estimated separately. The baseline panel regression results are presented in

Table 9. Panel regression results.

Variable	Model(1)	Model(2)	Model(3)	Model(4)	Model(5)
	CDCE	CDCE	CDCE	CDCE	CDCE
SP	42,783.230 *** (3.089)	41,510.669 *** (3.007)	36,567.397 *** (3.327)	42,358.354 *** (2.933)	42,746.289 *** (3.022)
UL	3.93 × 108 ** (2.378)	3.76 × 108 ** (2.282)	3.69 × 108 ** (2.749)	3.91 × 108 ** (2.270)	3.93 × 108 ** (2.340)
CCL	6.59 × 107 *** (3.815)	6.55 × 107 *** (3.688)	4.98 × 107 *** (3.376)	6.58 × 107 *** (3.750)	6.60 × 107 *** (3.884)
CE	−1504.405 ** (−2.102)	−1487.036 * (−2.040)	−1076.856 * (−1.929)	−1483.795 * (−1.936)	−1504.980 ** (−2.110)
MES	5.77 × 108 ** (2.323)	5.42 × 108 ** (2.093)	4.27 × 108 * (1.966)	5.81 × 108 ** (2.337)	5.79 × 108 ** (2.346)
INE		87.307 (1.214)
OUTE			246.655 *** (7.391)
BC				32,452.000 (0.227)
EC					1.45 × 106 (0.061)
_cons	−3.47 × 108 *** (−2.965)	−3.34 × 108 *** (−2.866)	−3.09 × 108 *** (−3.330)	−3.45 × 108 *** (−2.810)	−3.48 × 108 *** (−3.070)
R2	0.498	0.501	0.639	0.498	0.498
N	150	150	150	150	150

Notes: The table presents panel regression results for the individual fixed-effects model; robust standard errors are applied to mitigate heteroscedasticity during estimation. Values in parentheses represent t-statistics, with *** p < 0.01, ** p < 0.05, and * p < 0.1. Variable definitions are provided in the preceding sections.

.

Analysis of Model (1) reveals that all five control variables in the model passed significance tests in the panel regression for construction carbon emissions. Population size, urbanization rate, construction economic output, and machinery/equipment intensity positively correlate with construction carbon emissions. This suggests that larger populations and accelerated urbanization drive demand for infrastructure such as housing, schools, hospitals, and transportation facilities, while increased mechanization fosters modernization and rapid construction growth. These processes inherently rely on carbon-intensive building materials, significantly elevating sectoral emissions. Furthermore, higher economic output for construction stimulates the expansion of high-carbon industries (e.g., mining, chemicals, transportation) across regions, amplifying emissions throughout the supply chain, which fuels further construction growth. Conversely, per capita net income in construction demonstrates a notable inhibiting effect on emissions. Higher income levels reflect a shift toward capital-intensive production, where green and advanced environmental technologies—requiring skilled professionals for development and implementation—become prioritized. The adoption of such technologies effectively curbs carbon emissions.

For Models (2)–(5), the explanatory variables—construction carbon inflow intensity, mediation capacity, and degree of influence—show a positive but statistically insignificant relationship with total construction emissions, indicating no substantial impact. However, construction carbon outflow intensity significantly promotes emissions. This implies that provinces with high outflow intensity (e.g., Province A to Province B) reflect substantial direct or indirect consumption of products and services from other provinces, driving increased construction output in the originating province. Expanded output correlates with a larger industry scale and higher emissions. Additionally, the analysis highlights unavoidable carbon leakage during interprovincial trade, particularly from central/eastern regions to western regions, as domestic trade intensifies.

4.2.5. Robustness Test

This study adopts two approaches based on existing methodologies to verify the robustness of the construction carbon emission panel regression model and account for data characteristics. First, following prior research practices [50], the 2010 data sample is excluded from Robustness Test I, as the 2008 US subprime mortgage crisis triggered global financial turbulence and a subsequent economic recession, severely impacting China’s construction industry [51]. The results are presented in

Table 10. Robustness Test I.

Variable	Model (1) CDCE	Model (2) CDCE	Model (3) CDCE	Model (4) CDCE	Model (5) CDCE
SP	39,511.201 *** (3.269)	38,098.861 *** (3.139)	33,650.401 *** (3.321)	40,060.276 *** (3.127)	39,690.650 *** (3.165)
UL	4.73 × 108 *** (2.959)	4.62 × 108 *** (2.886)	4.15 × 108 *** (3.235)	4.77 × 108 *** (2.828)	4.75 × 108 *** (2.881)
CCL	6.47 × 107 *** (4.001)	6.40 × 107 *** (3.835)	5.08 × 107 *** (3.640)	6.50 × 107 *** (3.847)	6.44 × 107 *** (4.019)
CE	−1827.381 ** (−2.424)	−1800.270 ** (−2.354)	−1308.195 ** (−2.304)	−1859.672 ** (−2.239)	−1826.209 ** (−2.421)
MES	4.06 × 108 (1.160)	4.23 × 108 (1.269)	3.51 × 108 (1.106)	3.93 × 108 (1.089)	4.07 × 108 (1.144)
INE		93.975 (1.257)
OUTE			241.448 *** (7.429)
BC				−3.72 × 104 (−0.215)
EC					−4.97 × 106 (−0.205)
_cons	−3.59 × 108 *** (−3.305)	−3.49 × 108 *** (−3.206)	−3.12 × 108 *** (−3.597)	−3.62 × 108 *** (−3.149)	−3.57 × 108 *** (−3.378)
R2	0.486	0.489	0.622	0.486	0.486
N	120	120	120	120	120

Notes: The table presents panel regression results for the individual fixed-effects model, with the regression methodology aligning with that in Table 9. Values in parentheses denote t-statistics, where *** p < 0.01, ** p < 0.05.

. Second, drawing on the method of Liu et al. [52], Robustness Test II integrates all core and control variables into a unified regression to assess result consistency, with outcomes also summarized in

Table 11. Robustness Test II.

Variable	Model(1)
	CDCE
SP	33,571.057 *** (3.055)
UL	3.40 × 108 ** (2.455)
CCL	4.89 × 107 *** (3.258)
CE	−990.928 * (−1.705)
MES	4.07 × 108 * (1.777)
INE	100.186 (1.521)
OUTE	250.229 *** (7.296)
BC	98,119.868 (0.817)
EC	6.33 × 106 (0.285)
_cons	−2.89 × 108 *** (−3.151)
R2	0.645
N	150

Notes: The table presents panel regression results for the individual fixed-effects model, with the regression methodology aligning with that in Table 9. Values in parentheses denote t-statistics, where *** p < 0.01, ** p < 0.05, and * p < 0.1.

.

Table 10 displays the panel regression results after excluding post-crisis data (2010). Model (1) regresses all control variables, while Models (2)–(5) sequentially incorporate the four explanatory variables. The results align closely with those in Table 9.

After removing 2010 data, construction carbon inflow intensity, mediation capacity, and degree of influence remain statistically insignificant in explaining provincial construction carbon emissions, with only outflow intensity retaining a significant positive effect. Among control variables, machinery/equipment intensity no longer exhibits a statistically significant impact, while other variables maintain consistent directional and significance trends. This reinforces the baseline findings and underscores the robustness of the outflow intensity’s role in driving emissions.

Table 11 presents regression results incorporating all control and explanatory variables, which align with the findings in Table 9. Model (1) indicates that control variables—population size, urbanization rate, construction economic output, and machinery/equipment intensity—positively drive construction carbon emissions. In contrast, per capita net income in construction exerts an inhibiting effect. Among the explanatory variables, construction carbon outflow intensity significantly increases emissions. In comparison, inflow intensity, mediation capacity, and degree of influence do not statistically affect total construction carbon emissions considerably.

4.2.6. Endogeneity Test

The text performs endogeneity tests on the regression model based on the robustness test to address the possible omission of variables in the econometric model that have an impact on carbon emissions from the construction industry, as well as the fact that changes in carbon emissions from the construction industry may also affect other explanatory variables. In this paper, systematic GMM estimation is used for regression estimation. The explanatory variable (L.CDCE) with one period lag is selected as the GMM instrumental variable for regression estimation, and the results are shown in

Table 12. Endogeneity Test Results.

Variables	(1)
	GMM
L.CDCE	−0.206
	(0.177)
INE	−68.240
	(268.431)
OUTE	217.470 ***
	(58.461)
BC	874,647.945
	(663,335.705)
EC	−5.599 × 107
	(51,537,095.262)
SP	25,445.321
	(46,455.130)
UL	2.347 × 108
	(6.652 × 108)
CCL	71,846,623.447 **
	(30,406,021.534)
CE	−241.031
	(3632.386)
MES	9.972 × 108
	(4.658 × 109)
N	120
AR (1)	0.0545
AR (2)	0.183
Hansen	0.190

Notes: Values in parentheses denote t-statistics, where *** p < 0.01, ** p < 0.05.

. The intensity of carbon emission outflow still significantly affects the increase of carbon emissions from the construction industry. The p-value of AR (1) is less than 0.1. The p-value of AR (2) is greater than 0.1, which can be regarded as the first-order autocorrelation of the perturbation term in the systematic GMM estimation. The autocorrelation of the second-order and higher autocorrelation does not exist. Autocorrelation and the p-value of the Hansen test are greater than 0.1, which indicates that the instrumental variables are valid, thus indicating that the endogeneity problem of the model constructed in this paper has been solved better.

5. Discussion

This study integrates China’s construction carbon emission transfer network with panel regression models to explore how provinces’ roles within the network influence total construction carbon emissions. The panel regression results indicate that only construction carbon emission outflow intensity significantly promotes total emissions. Compared with findings from Jiang et al. [49], who analyzed global trade carbon drivers through a complex network lens, China’s domestic supply chain exhibits distinct characteristics. Within China, local supply and cross-provincial production interactions primarily drive provincial carbon emissions. In contrast, global supply chains introduce additional factors, such as a nation’s influence in the network and its frequency of acting as an intermediary (“middleman”) in carbon transfers, which further shape emission dynamics.

This study has certain limitations. Due to the inherent complexity of compiling interregional input-output tables in China, which demands substantial time and labor resources, coupled with the lack of annual updates for these tables, the temporal resolution of the analysis is constrained. In addition, the inflow intensity, medium transmission capacity (Betweenness centrality), and the degree of influence (eigenvector centrality) in the network characteristics do not have a significant effect on the carbon emissions of the construction industry, which suggests that the development of carbon emission-reduction measures may not need to be designed from the perspectives of the above characteristics in the network structure. Instead, more attention should be paid to the outflow intensity of carbon emissions. In terms of variable selection, the adoption of Building Information Modeling (BIM), the proportion of recyclable building materials, and the existence of local carbon tax policies are extremely important factors in driving carbon emissions in the construction industry, but they are not taken into account in this paper due to the lack of uniformity of data across provinces and the lack of data that can comprehensively measure the 30 provinces in China. In addition, the regression model demonstrates a positive correlation between urbanization and carbon emissions, but it does not take into account the effects of green technologies in highly urbanized areas (over 60%).

Due to differences in economic structures, political systems, and energy mix across nations and regions, this study has inherent limitations in its applicability to carbon reduction research in developed countries like those in Europe and North America, with restricted generalizability. While building operational emissions in developed economies predominantly concentrate on gas-based heating systems, developing countries face pressure from surging embodied carbon emissions amid rapid urbanization, where the carbon intensity of building materials generally exceeds that of European circular renovation models. Although the unique characteristics of China’s case provide valuable references for construction industry transitions in developing nations, its global extension requires detailed studies tailored to their specific construction industry characteristics and institutional frameworks.

Future research on carbon emission reduction should prioritize an in-depth investigation of China’s industrial profile and carbon emission supply-side structure to develop practical, context-sensitive mitigation strategies and policies tailored to China’s unique socio-economic and developmental realities. And for the demand side of the carbon emission-reduction reform can be after the supply side reform. In addition, further research and analysis on the application of green building technology, carbon tax policy, and the application of green emission-reduction technology in major cities should be carried out in future research, so as to further enrich the research on carbon emission in the construction industry.

6. Conclusions and Recommendations

6.1. Conclusions

China’s construction industry carbon emissions continue to rise in most provinces, particularly in the Major Development Zones (MDZ). However, a decline was observed in 2017 for many provinces, likely linked to nationwide efforts to promote green building practices and eco-friendly materials. High-emission provinces are concentrated in MDZ and Balanced Development Zones (BDZ). Demand-Driven Zones (DDZ) exhibit lower emissions due to their limited economic scale and stringent environmental regulations in megacities.

Interregional carbon transfer analysis reveals that carbon outflows predominantly originate from MDZ, a highly developed construction industry. At the same time, inflows are concentrated in BDZ and DDZ, reflecting their economic profiles and regulatory environments. Carbon transfers frequently occur between geographically adjacent provinces and within functional zones, driven by regional economic integration.

The carbon emission outflow intensity of the construction industry, population scale, urbanization process, economic level of the construction sector, and machinery equipment intensity significantly increase its carbon emissions. These factors collectively reflect the environmental pressures arising from resource-intensive characteristics and scale effects during industrial development, with urbanization as the core carrier of modernization exacerbating the carbon footprint by driving the expansion of construction demand and energy consumption. Conversely, per capita net income in the construction industry demonstrates a notable mitigating effect on carbon emissions, indicating that when industry development surpasses specific thresholds, economic advancement fosters transformative drivers, including enhanced investment capacity in clean technologies, awakening environmental awareness, and implementation of green building standards, thereby forming a virtuous cycle of “development promoting governance”. This dual effect highlights the complexity of low-carbon transition in construction, necessitating both technological innovation and management optimization to reduce carbon intensity per unit output while coordinating ecological carrying capacity during urbanization. Ultimately, this dual approach aims to achieve decoupled development between economic growth and environmental burden.

6.2. Recommendations

Based on the research findings and China’s current construction industry landscape, this study proposes the following recommendations:

Accelerate the Development of Green Building Practices. Sustainably promote the green building sector by supporting and scaling up green building standards. Integrate low-carbon principles across the entire building life-cycle—design, material selection, construction, maintenance, and usage—to reduce energy consumption and environmental impact. Enhance recycling of construction waste, wastewater, and low-carbon materials while upgrading construction equipment and techniques. Align emission-reduction efforts with national “dual carbon” goals (carbon peaking and carbon neutrality) through systemic decarbonization strategies.

Refine InterRegional Carbon Responsibility Allocation Mechanisms. Establish a transparent carbon accountability framework to address interprovincial carbon transfers. Provinces with substantial carbon outflows, particularly those in Major Development Zones (MDZ), should proactively assume carbon responsibility for cross-regional trade. Given their economic strength, large populations, and advanced industries, MDZ provinces (e.g., Zhejiang, Jiangsu) transfer significant emissions to Specialized Supplementary Zones (SSZ) and Demand-Driven Zones (DDZ) to meet local demand. To offset these transfers, MDZ provinces should provide technical assistance to recipient regions, such as green building technologies and high-tech industrial solutions.

Leverage Per Capita Income Growth to Drive Green Innovation. Prioritize prefabricated construction, near-zero-energy building systems, and BIM (Building Information Modeling) technology to optimize workflows. Replace energy-intensive machinery with innovative equipment, promote reusable low-carbon materials, and implement digital platforms for complete life-cycle carbon monitoring. The analysis confirms that higher per capita net income in construction suppresses emissions and incentivizes capital-intensive, technology-driven practices. A “technology premium + carbon efficiency subsidy” mechanism can channel income growth into green technology adoption, counterbalancing carbon inertia from urbanization.

These measures aim to harmonize economic growth with emission reduction, fostering a sustainable transition for China’s construction industry.