The Impact of Green Innovation on Carbon Emissions: Evidence from the Construction Sector in China

Abstract

Green innovation has been identified as a viable strategy to mitigate the tension between industrial change and pollutant emissions. Relevant research should focus more on the impact of eco-friendly innovations on carbon emissions. Consequently, using a panel data set that covers 30 Chinese provinces from 2005 to 2020, we examine whether there is a carbon-abatement effect of green innovation in the construction sector. According to the findings of a two-way fixed-effects model, green innovation substantially positively influences reduction in dioxide emissions in the Chinese construction industry. The results of a panel threshold model show a significant non-linear relationship between green innovation and carbon emissions when the environmental regulation intensity is used as the threshold variable. Furthermore, we discovered that green innovation cuts carbon emissions considerably through environmental regulation. Finally, the impact of abatement varies significantly between regions, innovation elements, and types of development.

1. Introduction

Global warming has been a major study issue worldwide [1], and carbon emissions have considerably contributed to climate change [2]. According to the Sixth IPCC Assessment Report, global warming is anticipated to approach or surpass 1.5 °C in the next 20 years, based on projections of average temperature increases. Carbon emissions have recovered primarily to pre-Newcastle pneumonia epidemic levels, reversing the temporary drop caused by the pandemic in 2020, according to BP’s World Energy Statistics Yearbook (2022). This growth is primarily driven by emerging economies, with China accounting for 76.9% of the increase. As a result, China’s environmental regulations face the complex and urgent challenge of reducing CO₂ emissions. Achieving China’s goal of a sustainable low-carbon economy is contingent on cutting carbon dioxide emissions [3,4].

In response to local and United Nations climate-discussion criticism, the Chinese government is reducing its carbon dioxide emissions. It has committed to cutting carbon emissions in a variety of bold ways. For instance, within the framework of the Paris Agreement, China pledged to achieve carbon peaking by 2030 and carbon neutrality by 2060. As the world’s largest developing nation, China will swiftly transition from carbon peaking to carbon neutrality and become a world leader in carbon abatement.

It is crucial to determine the binding influence of technological progress on carbon dioxide emissions in the construction sector [5]. As a major pillar sector in the country’s growing urbanization, the construction industry has contributed tremendously to the nation’s economic growth. However, there are issues of excessive energy use and emissions [6]. The building industry accounts for around 40% of worldwide carbon emissions. The Chinese construction industry is the largest in the world. According to the China Building Energy Consumption and Carbon Emissions Study (2022), the total energy consumption of the entire building process in 2020 was 2.27 billion tons, accounting for 45.5% of the country’s total energy consumption; the total carbon emissions of the building process as a whole in 2020 was 5.08 billion tons of carbon dioxide, accounting for 50.9% of the country’s carbon emissions. Using toxic fuels, primary energy consumption, and antiquated technology have increased hazardous emissions and environmental damage [7], resulting in severe emission problems [8]. However, as in all other industries, when new technologies and equipment are generated and put into the construction process, high-value-added, low-energy-consuming products are produced. Energy consumption is more efficient, and construction costs are reduced with lower carbon emissions [9]. New technology-based tools and equipment imply the development of technological innovations [10,11,12]. As a result, green innovation is a viable path for the construction industry to realize energy-saving and low-carbon development [13]. This has a positive impact on economic, social, and ecological benefits.

Scholars have researched whether green innovation can cut carbon emissions extensively, but a consensus has yet to be reached. Most academics believe that green innovation may reduce carbon emissions while improving environmental quality [14,15]. According to empirical research, green innovation is an effective strategy for achieving a “win-win” situation with respect to high-quality economic growth and environmental conservation [16]. The correlation between green innovation and carbon abatement has also been demonstrated [17,18]. For instance, Du et al. (2019) [19] found that a single threshold effect regulated the impact of environmental innovation on the performance of carbon dioxide emissions. Conversely, other researchers contend that green technology cannot reduce emissions [20,21,22]. According to Weina et al. (2016) [23], Italy’s green innovations may have enhanced environmental output but have not appreciably reduced carbon emissions.

Furthermore, the correlation between green technologies and greenhouse-gas emissions must be better studied, especially in developing countries [24,25]. Other studies have found a trend of increasing carbon pollution across economies, particularly in developing countries undergoing rapid economic growth and industrialization [26,27]. The effect of green technology on CO₂ reduction has been found to differ dramatically between economies [28]. For instance, Dauda et al. (2019) [29] found that innovation has reduced CO₂ emissions in the G6 while increasing them in the Middle East, North Africa, and the BRICS. Chen et al. (2020) [30] found that technological innovation in high-income, high-tech, and high-CO₂-emitting countries can significantly reduce emissions in neighboring countries, while R&D intensity in other countries can even increase CO₂ emissions. Many European countries are also facing economic transformation [31]. It should be mentioned that analogous research relies on a worldwide sample and does not include China. Nevertheless, research in this area is necessary for adopting and improving decarbonization measures in the world’s largest rising economy.

The bulk of studies have concentrated on the effect of environmentally friendly innovation on environmental quality. Consequently, further research is necessary to directly demonstrate the carbon-abatement benefits of green technology innovations in the construction industry. Numerous studies have investigated the factors that influence CO₂ emissions in the building sector. However, most of the literature focuses on energy consumption [32], urbanization [33], economic growth [34], environmental regulation [35,36], and carbon rights [37]. Fewer studies consider green innovation as a central explanatory factor. Most domestic and international experts and scholars [6,8] have concluded that technological innovation and industrial carbon emissions are closely linked through different empirical studies. Green technologies are critical to attaining sustainable development goals while having the least detrimental impact on the natural environment [38,39,40,41].

Based on current research, we pose two questions: (1) Can green innovation cut carbon emissions effectively? (2) Does green innovation have a non-linear or linear relationship with dioxide emissions? To address the above topics, using panel data from 30 provinces, the impact of green innovation on CO₂ emissions in the construction sector between 2005 and 2020 is empirically explored. The results of green innovations above and below the threshold are validated using a two-way fixed-effects model followed by a double-threshold model. This study establishes a link between carbon dioxide emissions and ecologically friendly innovation. According to the two-way fixed-effects model findings, green innovation has a substantial positive influence on CO₂ emissions reduction in the Chinese construction industry. The panel threshold-effects model results suggest a double threshold effect of green innovation and carbon emissions when the intensity of environmental regulation is used as the threshold variable. Furthermore, we discovered that green innovation cuts carbon emissions considerably through environmental regulation. Finally, the abatement impact varies significantly between regions, innovation elements, and types of development. Although this article found a significant impact of green technology innovation on the total carbon emissions reduction in the construction industry, we did not reach a significant conclusion regarding the proportion of emissions reduction. This could guide regions in China to realize the carbon-reduction effects of green innovation.

The following are four major contributions: First, using a unified statistical analysis method, we investigate the correlations between carbon dioxide emissions and green innovation, both linearly and non-linearly. The global effect of environmentally friendly innovation on environmental quality, notably carbon emissions, is questionable. In this paper, we study the influence of green innovation on CO₂ emissions using a two-way fixed-effects model followed by a double-threshold model. Environmental regulations are included in the study framework for the role of green technology in reducing carbon dioxide emissions. This may contribute to the advancement of research in linked domains. Second, this essay expands the study of green innovation in carbon emissions based on statistics from the Chinese construction sector. We aim to determine whether green technologies provide the building sector with viable ways to combat climate change. A comprehensive evaluation of the role of green innovation in abating emissions in China’s construction sector facilitates the creation of decarbonization strategies. In addition, it may serve as a guide for other developing nations as they transition to low-carbon economies. We have discovered a probable mechanism. Numerous studies have yet to examine the indirect impact mechanisms of green innovation in reducing CO₂ emissions in the construction industry. Using a mediation-effect model based on environmental regulation, the mechanism of influence of green innovation on dioxide emissions is examined. Finally, we investigate the varying effects of green innovation on CO₂ emissions in different regions, innovation variables, and province types. Due to the potential diversity of green innovation and dioxide emissions in the building sector, only some studies have assessed whether the spatial relationship varies. This facilitates the formulation of green development plans tailored to local circumstances.

The remaining sections of the paper are organized as follows. Section 2 describes the theoretical analysis and research hypotheses. Technical and research data are presented in Section 3. Section 4 shows the estimated benchmark regression findings, the mediation effect, and heterogeneity analysis results. The results of the study are discussed further in Section 5. The Section 6 summarizes the findings and provides policy suggestions.

2. Theoretical Analysis and Basic Assumptions

2.1. Green Innovation and Carbon Emissions

According to modern economic growth theory, technological progress is one of the major factors in economic growth. Innovation is a requirement for technology development which can reduce the harmful effects of economic growth on the environment [42]. The research and application of green innovations can drive technological progress and industrial transformation, thereby achieving the goal of reducing carbon emissions. This has been well documented internationally. For example, the United Nations Framework Convention on Climate Change (UNFCCC) has embraced “green technology transfer” as a key tool in the fight against climate change. Specifically, green innovations in energy, transport, and buildings can significantly curtail CO₂ emissions by improving energy efficiency and promoting new energy sources.

The relationship between green innovation and CO₂ emissions will vary between the initial and subsequent phases. At the initial stage, green innovation has a relatively small effect on CO₂ emissions abatement. This is because it takes time for a new technology or product to become widespread and diffused. In the subsequent stages, however, the effect of CO₂ emissions abatement will gradually increase as green innovation is widely used and its market share increases. The diffusion of green innovations should therefore be a long-term and gradual process. Innovations and technological developments may result in hazardous waste [43] but may also pave the way for eco-friendly technologies with lighter emission rates. Thus, the relationship can be either positive or negative.

There may be a complex non-linear relationship between green innovation and carbon abatement, with the intensity of environmental regulation being a significant threshold variable. Innovation asymmetry theory suggests that the degree of environmental regulation can create innovation differentiation in the marketplace. When environmental regulations are less stringent, firms may adopt more traditional technologies and experiences to meet environmental requirements. Only adopting more advanced and cutting-edge technologies and innovative solutions can give firms a comparative advantage in the competitive marketplace when the environmental regulation intensity increases. This advantage is more difficult to replicate. Therefore, the higher the environmental regulation intensity, the greater the differentiation of the innovation. The following research hypotheses were generated based on the above analysis:

Hypothesis 1. 

Green innovation can significantly reduce carbon emissions in the construction industry.

Hypothesis 2. 

There is a non-linear relationship between green innovation and carbon reduction.

2.2. The Mediating Role of Environmental Regulation

Innovation theory suggests that firms tend to invest more in eco-friendly technological innovations under the influence of environmental regulatory intensity. When environmental policies and regulations are strengthened, the cost to companies of implementing carbon reductions and reducing their environmental impact increases. As a result, companies tend to adopt more optimal and cost-effective methods to reduce carbon emissions through innovation. Through extensive and in-depth research into environmental regulatory oversight, scholars have found that it can significantly affect technological innovation [44]. The higher the environmental regulation intensity, the tighter the environmental constraints on companies. Companies will have to face stricter emissions standards and environmental responsibilities, which will lead to innovations in technology and processes to improve energy efficiency and cut CO₂ emissions.

The “Porter Hypothesis” posits that environmental regulation may force firms to innovate and enhance their competitive advantage in the marketplace [45]. Adequate environmental regulation has a significant impact on green innovation. In a competitive market environment, firms transform elements of environmental regulation into regional factors of green innovation, enhancing the disincentives to waste emissions. According to the Porter Hypothesis, the relationship between environmental regulation, eco-friendly technologies, and environmental pollution has been explained more scientifically and clearly. In addition, studies have revealed an inverted U-shaped relationship between environmental regulation and environmental pollution [46]. Therefore, environmental regulation plays a mediating role when studying the CO₂ abatement effects of technological innovation. Therefore, we propose the following research hypothesis:

Hypothesis 3. 

Green innovation reduces carbon emissions by promoting the intensity of environmental regulation.

3. Materials and Methods

3.1. Model Specification

3.1.1. Carbon-Emissions Measurement

Numerous research studies have investigated the quantification of CO₂ emissions from the building industry. Generally, these studies can be categorized into three groups: the IPCC approach, the input–output analysis technique, and the life-cycle assessment method. First, carbon emissions are evaluated using energy consumption and carbon dioxide emission factors [47,48]. Furthermore, carbon emissions are estimated by assembling input–output tables and building related mathematical models describing the link between initial input, intermediate input, intermediate output, total input and total output, and final output for each economic system sector [49]. Finally, bottom-up emissions of a product or process are evaluated throughout its life cycle by circling carbon emissions, accounting limits, and collecting carbon emissions data [50,51]. In recent years, the amount of academic research using IPCC methods to study the carbon emissions from the building sector has increased significantly. The primary reasons include the IPCC method’s more flexible data selection, straightforward accounting methods, and more reliable computation outcomes. Consequently, the IPCC approach was adopted to estimate carbon emissions in this article.

Based on the CEADS database and regarding Shan et al. (2016) [52], following the IPCC approach, the province-specific dioxide emissions from the energy used by the building sector were calculated.

$$C{E}_{ij}=A{D}_{ij}\times NC{V}_i\times C{C}_i\times O_{ij}$$

(1)

where $C{E}_{ij}$ are the CO₂ emissions from burning fossil fuel $i$ in the sector $j$; $A{D}_{ij}$ represents the fossil-fuel usage for the particular fossil-fuel type and industry; $NC{V}_i$ is the net calorific value generated per physical unit of fossil-fuel burning; $C{C}_i$ (carbon content) is the dioxide emissions per net calorific value produced by fossil fuel $i$; and $O_{ij}$ is the oxidation rate, which is the oxidation ratio during the burning of fossil fuels.

3.1.2. STIRPAT Model

The IPAT model was originally proposed by Ehrlich and Holdren (1971) [53] and extended into the STIRPAT model by Dietz and Rosa (2003) [54]. The model can explain or evaluate the different influences, sources, and magnitudes of environmental stress. The IPAT model can be written as $I=PAT$, where $I$, $P$, $A$, and $T$ denote environmental impact, population, wealth per capita, and technology, respectively.

$$I_i=a{P}_i^b\times A_i^c\times T_i^d\times e_i$$

(2)

where $\mathrm{a}$ is a constant term; $b$, $c$, and $d$ are the elastic coefficients of the three variables $P$, $A$, and $T$, respectively; and $e$ is an error term and represents random variables, indicating uncontrollable or unobservable random variables. The model can be taken logarithmically on both sides and transformed into a linear equation.

We can use the STIRPAT model [55,56] to understand the main factors that influence carbon emissions from buildings.

$$C{E}_{it}=\alpha P_{it}^{{\beta }_1}\times A_{it}^{{\beta }_2}\times T_{it}^{{\beta }_3}\times e^{{\mu }_{it}}$$

(3)

where $\alpha$ and $\beta$ are model parameters and $\mu$ is a random disturbance parameter.

In the STIRPAT model, both the coefficients can be estimated as parameters and the effects can be appropriately decomposed [57].

The factors affecting carbon emissions are not only population increase, economic growth, and technological progress. Explanatory variables, core explanatory variables, and control variables can be added to the model above.

3.1.3. Two-Way Fixed-Effects Model

The following regression equation was developed to test the CO₂ abatement effect of green innovation in the construction sector. To control for time and area effects, a two-way fixed-effects model was chosen in this paper.

$$C{E}_{it}={\beta }_0+{\beta }_{1}Tec{h}_{it}+{\beta }_2{X}_{it}+{\alpha }_i+{\gamma }_t+{\epsilon }_{it}$$

(4)

where $i$ and $t$, respectively, represent the province and year; $C{E}_{it}$ denotes the dioxide emissions from the building sector in the province $i$ in period $t$; $Tec{h}_{it}$ is the number of patents for green inventions in the province $i$ in the period $t$, denoting the level of green innovation; $X_{it}$ is a collection of factors that impact carbon dioxide emissions; ${\alpha }_i$ denotes the province fixed effects, controlling for all province-level factors that do not vary over the region; ${\gamma }_t$ denotes the year fixed effects, controlling for time-level factors that do not vary by time; and ${\epsilon }_{it}$ is the residual term. In the above equation, ${\beta }_1$ is this paper’s estimated coefficient of interest. If negative and significant, ${\beta }_1$ indicates that green innovation may significantly cut CO₂ emissions from the inter-provincial construction sector.

3.1.4. Panel Threshold Model

The economic rationale for the non-linear impact may be that while an area is in its early phases of development, its major development aim is economic expansion. Currently, technical advancement is more likely to enhance production. Although it has a slight damping impact on carbon emissions [58], the outcome is mediocre. When an area achieves a certain degree of economic development and prioritizes green and sustainable development, technological innovation becomes more focused on green technological innovation, as green innovation supports economic development and improves environmental quality and CO₂ abatement intensity. This study employed the Hansen threshold model [59] to investigate whether there is a non-linear link between green innovation and carbon dioxide emissions. As a threshold variable, we used the intensity of environmental regulation. The single-threshold panel model presented in this study is described below.

$$C{E}_{it}={\beta }_0+{\beta }_{1}Tec{h}_{it}\times I\left(Env{r}_{it}\le \theta \right)+{\beta }_{2}Tec{h}_{it}\times I(Env{r}_{it}>\theta )+{\beta }_3{X}_{it}+{\alpha }_i+{\gamma }_t+{\epsilon }_{it}$$

(5)

We extended the single-threshold model to a multi-threshold model, considering the possibility of multiple thresholds.

$$\begin{array}{cc}C{E}_{it}={\beta }_0+\hfill & {\beta }_{1}Tec{h}_{it}\times I\left(Env{r}_{it}\le {\theta }_1\right)+{\beta }_{2}Tec{h}_{it}\times I\left({\theta }_1<Env{r}_{it}\le {\theta }_2\right)+\cdots \hfill \\ & +{\beta }_{n}Tec{h}_{it}\times I\left({\theta }_{n-1}<Env{r}_{it}\le {\theta }_n\right)\hfill \\ & +{\beta }_{n+1}Tec{h}_{it}\times I(Env{r}_{it}>{\theta }_n)+{\beta }_{n+2}{X}_{it}+{\alpha }_i+{\gamma }_t+{\epsilon }_{it}\hfill \end{array}$$

(6)

where $Env{r}_{it}$ represents the severity of environmental regulation and serves as the threshold variable in year $t$ and province $i$; $\theta$ is the specific threshold value, when $Env{r}_{it}\le \theta$, $I\left(·\right)=0$ and when $Env{r}_{it}>\theta$,$I\left(·\right)=1$; and $I\left(·\right)$ is an indicator function whose value relies on the link between the threshold variable ($Env{r}_{it}$) and the threshold value ($\theta$). In model (6), $Tec{h}_{it}$ is the explanatory variable affected by $Env{r}_{it}$ and ${\beta }_1$~${\beta }_{n+1}$ are the coefficients of the effect of $Tec{h}_{it}$ on $C{E}_{it}$ when $Env{r}_{it}$ is in different intervals.

3.2. Variable Definitions

3.2.1. Explained Variable

This study used the construction industry’s carbon emissions (CEs) as the explanatory variable. Due to the non-uniformity of physical energy categories consumed in different provinces, all energy consumed in each province was selected for data completeness, excluding energy sources outside the CEADS emission factor table. Secondary energy sources, such as electricity and heat, were omitted to avoid double-counting electricity and heat emissions. This article focuses on absolute carbon reduction, so continuous total data were used. We also tried to use the logarithmic of carbon reduction for the robustness tests, and although doing so is controversial, the results are robust.

3.2.2. Core Explanatory Variable

Green innovation (Tech) [60] was taken as the fundamental explanatory variable. Three variables can be used to quantify green innovation: patents, R&D expenditure, and total factor productivity. These metrics indicate the innovation’s inputs, outputs, and performance. Patents are a reasonably accurate predictor of environmental innovation [61]. Therefore, the number of patents for green innovation was utilized as a measuring stick. This article focuses on the impact of increasing the number of green patents on carbon reduction, that is, how much carbon emissions can be reduced by adding a green patent.

3.2.3. Threshold Variable and Mechanism Variable

Environmental regulation intensity (Envr/lnEnvr) was used as both a threshold and a mechanism variable. Existing research needs to have defined criteria for quantifying the rigorousness of environmental regulation, which is a policy variable. Some scholars [62] utilize the ratio of regional GDP to regional total energy consumption as a measurement. The bigger the ratio, the more pronounced the environmental regulation’s effect on energy conservation and emissions abatement at a given GDP level. Therefore, we used the ratio of GDP to energy consumption in the construction sector for measurement.

3.2.4. Control Variables

Complex variables influence carbon emissions in the building business. To control the potential impact of other factors and mitigate the endogenous problems caused by omitted variable bias, with reference to the established literature, construction GDP (lnCongdp), the number of employees in the construction industry (lnConjo) [11], the level of urbanization (Town) [63], the degree of openness to the outside world (Open) [64], the R&D intensity (Rede) [65], the industrial structure (lndus) [66], the economic development level (Pgdp) [2], and the efficiency of science and technology innovation (Pate/Scit/Expe/Budge) [67] were selected as control variables.

Table 1. Variable Definitions.

Variable	Name	Variable Measure	Data Sources
	Dependent variable
CE	Carbon emissions from construction [48]	IPCC	A
	Independent variable
Tech	Green innovation [60]	Number of patents for green inventions	B
	Threshold variable
Envr	Environmental regulation intensity [62]	Construction GDP/energy consumption	C and A
	Mediating variable
lnEnvr	Environmental regulation intensity [62]	Ln (construction GDP/energy consumption)	C and A
	Control variables
lnCongdp	Construction GDP	Ln (construction GDP)	C
lnConjo	Number of people employed in the construction industry [11]	Ln (number of employees in construction companies)	C
Town	Urbanization level [63]	Urbanization rate	D
Open	Degree of openness to the outside world [64]	FDI/total GDP	D
Rede	R&D intensity [65]	R&D expenditure of industrial enterprises/total GDP	D
lndus	Industrial structure [66]	Construction GDP/total GDP	D
Pgdp	Economic development level [2]	Ln (per capita GDP)	D
Pate	STI efficiency [67]	General budget expenditure	D
Scit	STI efficiency [67]	Domestic patent applications granted	D
Expe	STI efficiency [67]	Science and technology	D
Budge	STI efficiency [67]	Science and technology expenditure/general public budget expenditure	D

Notes: In the data sources column, “A” indicates the China Energy Statistical Yearbook, the China Environmental Statistical Yearbook, and the CEADS database; “B” indicates the China Research Data Service Platform (CNRDS) and the statistical reports of the State Intellectual Property Office; “C” indicates the China Statistical Yearbook of the Construction Industry; “D” indicates the China Statistical Yearbook.

contains a listing of all factors considered for this study.

Due to the unavailability and scarcity of data, four regions—Tibet, Hong Kong, Macau, and Taiwan—were excluded from this paper. The data from the final panel for thirty Chinese provinces from 2005 to 2020 were selected and interpolated to fill in the missing data for specific provinces. Descriptive statistics for all variables are presented in

Table 2. Descriptive Statistics of Variables.

Variable	Obs.	Mean	Std. Dev.	Min.	Max.
CE	480	802.149	765.214	13.008	4059.920
Tech	480	725.144	1303.389	0	10,010
Envr	480	18.197	43.817	2.209	489.816
lnEnvr	480	2361.909	819.364	792.690	6194.031
lnCongdp	480	6.774	0.924	3.954	8.784
lnConjo	480	13.640	1.117	10.912	15.962
Town	480	0.552	0.140	0.269	0.896
Open	480	0.022	0.018	0.000	0.082
Rede	480	94.037	52.366	5.833	231.865
lndus	480	0.071	0.023	0.019	0.147
Pgdp	480	10.501	0.658	8.528	12.013
Budg	480	3666.702	2926.775	0.201	17,430.790
Pate	480	41,179.920	75,230.880	79	709,725
Scit	480	90.485	137.640	0.48	1168.793
Expe	480	5055.355	33,142.230	2.237	72.019

Notes: “Std. Dev.” denotes the standard deviation.

.

4. Results

4.1. Two-Way Fixed-Effects Regression Analysis

To test the CO₂ abatement effect of green innovation in the building industry, a two-way fixed-effects model was used as the baseline regression model. In addition, in order to see more clearly whether other factors also influenced the two, we divided the regression model into one that did not include control variables and one that did. To strengthen the credibility of the two-way fixed-effects model, we present the regression results with bootstrap standard errors and control for province–year interaction fixed effects in

Table 3. Results of the Two-Way Fixed-Effects Regression Analysis.

	Dependent Variable: CE
	(1)	(2)	(3)	(4)
	FE	FE_r	FE_bootstrap	FE_dt
Tech	0.024	−0.184 **	−0.184 ***	−0.076 *
	(0.083)	(0.073)	(0.049)	(0.046)
lnCongdp		387.472	387.472	292.618 *
		(377.136)	(251.326)	(163.222)
lnConjo		−35.282	−35.282	89.532
		(216.077)	(97.173)	(73.198)
Town		493.041	493.041	1140.732
		(1832.267)	(1151.909)	(1480.447)
Open		−6933.200 **	−6933.200 ***	−2984.050 **
		(3112.535)	(1741.324)	(1318.868)
Rede		−0.790	−0.790	0.832
		(3.663)	(1.838)	(1.165)
lndus		−5426.363	−5426.363 *	333.546
		(4097.149)	(3141.189)	(1868.609)
Pgdp		−78.562	−78.562	184.983
		(462.698)	(288.796)	(181.879)
Pate		−0.000	−0.000	−0.002 **
		(0.001)	(0.001)	(0.001)
Scit		1.123	1.123	−1.318 **
		(1.321)	(0.794)	(0.519)
Expe		0.002	0.002 **	0.000
		(0.002)	(0.001)	(0.001)
Budge		0.134	0.134 ***	−0.024
		(0.123)	(0.046)	(0.046)
Constant	454 ***	−251	−251	−4672 ***
	(88)	(4837)	(2262)	(1739)
Year FE	Yes	Yes	Yes	Yes
Province FE	Yes	Yes	Yes	Yes
R2	0.231	0.354	0.354
N	480	480	480	480

Notes: (1) Robust standard errors are in parentheses; (2) “***”, “**”, and “*” denote statistical significance at the 1%, 5%, and 10% levels, respectively.

. The regression results are shown in Table 3. Column (1) does not contain control variables, column (2) displays results for fixed effects with conventional robust standard errors, column (3) displays regression results with bootstrap standard errors, and column (4) displays regression results controlling for province–year interaction fixed effects. The carbon-reduction effect of green innovation is significant, regardless of the total or proportion of carbon emissions (Appendix B).

The regression coefficients for Tech are 0.024, −0.184, and −0.184, as shown in columns (1) and (3), with columns (2) and (3) passing the 1%-significance-level test. For example, column (1) illustrates that the green-innovation regression coefficient is positive and insignificant before including control variables. Upon addition of the control variables, the estimated coefficient of Tech is −0.184, which is statistically significant at the 1% level. This indicates that additional variables obscure the results of the baseline regression. According to the regression findings in column (2), a rise in green innovation decreases dioxide emissions in the construction sector across all provinces. From the results for the control variables, the coefficient of Open is significantly negative, indicating that the degree of openness to the outside world contributes to the reduction in carbon emissions in the construction industry. Due to economic globalization, trade between countries has become more frequent. The introduction of advanced foreign experience and technology in environmental protection, increased awareness of environmental protection, and increased demand for high quality of life have reduced carbon emissions in the industry.

4.2. Panel Threshold Regression Analysis

According to model (6), whether the threshold-effect test is passed or failed determines whether the panel threshold model can be further manipulated. This threshold effect tests for a non-linear association between the variables. There may be a lagging effect on the impact of Tech on CE. The model was re-estimated using Tech’s first-order lag term for accounting for endogeneity and for validating the non-linear relationship further. Based on Hansen’s hypothesis, the threshold-effect test for Tech and the first-order lag term of Tech were used. Repeating the sample 300 times, the bootstrap technique determined the asymptotic distribution of F-values and the accompanying p-values and critical values. The results of the threshold tests indicated that single and double thresholds passed the hypothesis test at the 1% significance level (

Table 4. Threshold-Effects Tests for the Existence.

Independent Variable	Number of Thresholds	Threshold Value	F-Statistics	p-Value	BS Times	95% Confidence Interval
Tech	Single threshold	4.381	132.11 ***	0.000	300	[4.261, 4.413]
Tech	Double threshold	7.658	75.60 ***	0.000	300	[7.624, 7.671]
Tech	Three-fold threshold	3.499	43.41	0.700	300	[3.461, 4.447]
L.Tech	Single threshold	3.917	103.96 ***	0.000	300	[3.671, 3.921]
L.Tech	Double threshold	7.658	53.10 **	0.020	300	[7.643, 7.671]
L.Tech	Three-fold threshold	3.549	53.47	0.603	300	[3.501, 3.606]

Notes: “***” and “**” denote statistical significance at the 1% and 5% levels, respectively.

). In contrast, the three-threshold effect was insignificant, so the two-threshold model was chosen for analysis.

According to Table 4, the F-statistic is statistically significant at the level of significance of 1% in the double-threshold model. There are two thresholds, with Tech thresholds of 4.381 and 7.658 and L.Tech thresholds of 3.917 and 7.658, respectively. Figure 1 and Figure 2 show the likelihood statistics. As shown, the above threshold values are valid.

Based on the threshold test, we chose environmental regulation as the threshold variable and performed a two-threshold-model regression analysis. In column (1) of

Table 5. Results of the Panel Threshold Regression Analysis.

	Dependent Variable: CE
	(1)	(2)
	Tech	L.Tech
Tech (Envr ≤ threshold1)	1.025 ***
	(0.099)
Tech (threshold1 < Envr ≤ threshold2)	0.194 ***
	(0.056)
Tech (Envr > threshold2)	−0.152 ***
	(0.032)
(Tech)−1 (Envr ≤ threshold1)		1.225 ***
		(0.135)
(Tech)−1 (threshold1 < Envr ≤ threshold2)		0.195 ***
		(0.066)
(Tech)−1 (Envr > threshold2)		−0.195 ***
		(0.039)
Constant	577	1831
	(1658)	(1771)
Control variables	Yes	Yes
Year FE	Yes	Yes
Province FE	Yes	Yes
R2	0.531	0.487
N	480	450

Notes: (1) Robust standard errors are in parentheses; (2) “***” denotes statistical significance at the 1% level; (3) the number of bootstrapping iterations for the likelihood ratio test was 300.

, the results for Tech are displayed. When Envr is below the threshold value of 4.381, the regression coefficient of Tech on CE is 1.025, and the result is significantly positive; when Envr is between the threshold value of 4.381 and 7.658, the coefficient of Tech is 0.194, and the result is significantly positive; when Envr is above the threshold value of 7.658, the coefficient of Tech is −0.152, and the effect is significantly negative. This shows that when environmental regulations are weak, green innovation has a negligible influence on CO₂ abatement in the construction sector. However, as environmental regulation grows more stringent, the coefficient of Tech decreases. Eventually, it becomes negative and plays a larger role in lowering CO₂ emissions. In addition, the same outcomes are realized (see column (2) of Table 5). Therefore, the double-threshold model results suggest a non-linear relationship between green innovation and CO₂ emissions.

4.3. Endogenous Discussion

An inverse causal link between green innovation and construction-related carbon emissions may exist. In other words, higher construction carbon emissions may push the construction industry into green innovation, or there may be measurement errors or omitted variables. The existence of an endogenous problem means that the estimation results of model (4) are biased. As a result, we estimated our model (4) using the generalized two-step method-of-moments (GMM) estimator developed by Arellano and Bond (1991), which allowed us to correct endogeneity using instruments. This included constructing instrumental variables (IVs) using historical data on green innovation.

Green innovation’s first-order and second-order lag terms were used as IVs in the two-step GMM regression. In

Table 6. Instrumental-Variable Regression Results.

Dependent Variable:	Tech	CE	Tech	CE
Instrumental Variable:	L.Tech		L2.Tech
	(1)	(2)	(1)	(2)
	First	Second	First	Second
Tech		−0.187 ***		−0.177 ***
		(0.047)		(0.050)
L.Tech	1.025 ***
	(0.080)
L2.Tech			0.987 ***
			(0.115)
F-test	164.760		74.000
p-value	0.000		0.000
Under-identification test
Kleibergen–Paap rk LM-statistic		35.207 ***		36.411 ***
Weak instrumental variable test
Cragg-Donald Wald F-statistic		1627.022		753.154
Stock–Yogo weak ID test critical values		16.380 (10%)		16.380 (10%)
Control variables	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
Province FE	Yes	Yes	Yes	Yes
R2		0.327		0.290
N	480	450	420	420

Notes: (1) Robust standard errors are in parentheses; (2) “***” denotes statistical significance at the 1% level; (3) the main function of the under-identification test is to check whether the instrumental variables are sufficiently valid to avoid unreliable estimation results due to inadequate instrumental variables. Under-identification of instrumental variables may result in imprecise, statistically insignificant, or biased estimates of regression coefficients; (4) the main function of the weak instrumental variable test is to test whether the instrumental variable is sufficiently strong to avoid unreliable estimation results due to the instrumental variable being underpowered. If the instrumental variable is strong, it may lead to accurate estimates of the regression coefficients, large standard errors, or biased estimation effects.

, “first” and “second” indicate the results of the first- and second-stage regressions, respectively. As expected, each IV mentioned above had a substantial favorable impact on Tech. Furthermore, the IVs were plausible because the F-statistics in Table 6 exceed 10. This demonstrates that the IV selection is logical. The findings indicate that Tech considerably decreases CE, validating the dependability of the baseline estimate results. At the bottom of Table 6, you will see the validation test results for the IVs. The findings imply that concerns over inadequate variable identification and IVs can be ruled out. Therefore, the estimations for GMM2s are reliable.

4.4. Robustness Check

We executed a set of robustness assessments to further validate our findings’ dependability.

Table 7. Robustness-Check Results.

	Dependent Variable:
	F.CE	F2.CE	CE	CE	CE	CE	CE	CE
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
L.Tech				−0.191 **
				(0.078)
Tech	−0.187 ***	−0.171 ***	−0.226 **		−0.182 **	−0.168 **	−0.185 **	−0.186 **
	(0.067)	(0.060)	(0.082)		(0.072)	(0.073)	(0.070)	(0.072)
D_T_1						349.458
						(222.789)
D_T_2							152.979
							(165.617)
D_T_3								519.620 *
								(299.110)
Constant	−1252	−2653	68	−87	−295	68.042	40.484	423.074
	(4644)	(4514)	(4948)	(4935)	(4851)	(4717.596)	(4853.225)	(4444.671)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Province FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
R2	0.301	0.267	0.356	0.324	0.356	0.381	0.357	0.387
N	450	420	480	450	476	480	480	480

Notes: (1) Robust standard errors are in parentheses; (2) “***”, “**”, and “*” denote statistical significance at the 1%, 5%, and 10% levels, respectively. (3) “D_T_1”, “D_T_2”, and “D_T_3” denote the regression results of joining the first batch of low-carbon provincial and municipal pilot policies, carbon-emissions trading policies, and energy-use-rights trading pilot policies, respectively.

displays the results of the regression.

(1)

The time window was extended. This paper extended the time examination window by treating the explained variable CE in a period ahead and two periods ahead. Table 7’s columns (1) and (2) suggest that both estimated coefficients for Tech were notably negative. This suggests that the study findings remained unchanged after extending the time window.

(2)

The control variables were indented. To exclude outliers in the control variables, all control variables were bilaterally shrunk by 1% at the top and bottom. In column (3) of Table 7 are shown the regression findings for the independent-variable tailoring, with a significantly negative regression coefficient for Tech.

(3)

We evaluated the lagging terms of the Tech and the extreme values of the CE. Given the possible non-randomness, outliers, and time specificity of the study sample, all of which may impact the study results, the robustness of the study results was examined using two approaches, lagged terms and outliers, to further enhance the scientific validity of the analysis results. It may take years for green innovation to reach its full potential, so green innovation variables with a one-period lag were selected to examine the possible time-lag effect. To explore possible variability in the CO₂ abatement process in the construction industry, the highest carbon emission value of 1% was eliminated. Table 7 contains the regression findings in columns (4) and (5). The Tech regression coefficient is significantly negative, and there is a temporal lag between Tech and the CE process in the building business. The Tech regression coefficient remains notably negative when eliminating extreme values.

(4)

The impact of other environmental policies was considered. China implemented its first pilot policies for low-carbon provinces and cities in July 2010 and for carbon emissions trading in 2013 and a pilot policy for energy-use-rights trading in 2016. These measures have contributed to decreased energy use in the construction industry and increased green innovation. They may encourage communities to construct an industrial system with a smaller carbon footprint and to promote green and low-carbon lifestyles and consumption patterns. This may skew the appraisal of environmentally friendly innovations toward carbon reduction. To accurately identify the environmental effects of green innovation, interference from other similar policies must be excluded. This article includes interaction variables in the baseline regression for provinces participating in the pilot project between 2005 and 2020. Considering the first batch of low-carbon provincial and municipal pilot policies, carbon-emissions trading policies, and energy-use-rights trading pilot policies in turn, columns (6) to (8) of Table 7 reflect the impact of Tech on CE. The results show that the conclusions remain robust after excluding other policy disturbances.

(5)

The impact of the COVID-19 pandemic was also considered. This paper excluded data for 2019 and 2020 from the COVID-19 period and generated dummy variables for 2019 and 2020 for separate two-way fixed-effects model regressions.

Table A1. Regression Results after Excluding Time and Adding Dummy Variables.

	Dependent Variable: CE
	(1)	(2)
	FE1	FE2
Tech	−0.244 ***	−0.184 **
	(0.076)	(0.073)
D_T		−742.877
		(623.941)
Constant	2844.587	−251.175
	(4243.115)	(4837.315)
Control variables	Yes	Yes
Year FE	Yes	Yes
Province FE	Yes	Yes
R2	0.384	0.354
N	420	480

Notes: (1) Robust standard errors in parentheses; (2) “***” and “**” denote statistical significance at the 1% and 5% levels, respectively; (3) column (1) indicates the exclusion of 2019 and 2020, and column (2) indicates the inclusion of year dummy variables in the two-way fixed-effects model.

in Appendix A reports the results. The results were still significantly negative. This indicates that the results of this paper were not influenced by COVID-19.

4.5. Intermediary Effect

The effect of Tech on the CE of the construction sector was examined. The bootstrap test and the recursive equation can be used to conduct tests for mediating effects. In pursuit of the robustness of results, using the bootstrap test and the recursive equation, the panel data were fitted to validate the mediating function of environmental regulation in its transmission process. The first three columns of

Table 8. Results of Mediation-Effect Analysis.

	Dependent Variable:
	CE	lnEnvr	CE	CE
	(1)	(2)	(3)	(4)
Tech	−0.184 **	0.233 ***	−0.026	−0.026
	(0.073)	(0.048)	(0.057)	(0.031)
lnEnvr			−0.674 ***	−0.674 ***
			(0.242)	(0.036)
ACME				−0.157 ***
				[−0.206, −0.108]
ADE				−0.026
				[−0.078, 0.025]
TOTAL				−0.184 ***
				[−0.258, −0.109]
Percentage mediated (%)				85.669
Constant	−251	−116	−329	193
	(4837)	(4101)	(2947)	(1698)
Control variables	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
Province FE	Yes	Yes	Yes	Yes
R2	0.354	0.474	0.648	0.895
N	480	480	480	480

Notes: (1) Robust standard errors are in parentheses; (2) “***” and “**” denote statistical significance at the 1% and 5% levels, respectively; (3) “ACME”, “ADE”, and “TOTAL” denote average causal mediation effect, average direct effect, and total effect, respectively. The estimation results (standard-deviation effects) and 95% confidence intervals (in square brackets) were obtained based on Imai et al. (2010) [68] and Tingley et al. (2015) [69]; the number of simulations was 1000.

show the stepwise regression method’s results. In contrast, the regression results of the Sobel and bootstrap method are displayed in column (4).

Table 8 shows the test results of the mediating effect of environmental regulations. At the 5% significance level, the technology in column (1) was considered negative, showing that Tech has a major impact on CE. The Tech value in column (2) is very positive, indicating that green innovation is integrally tied to environmental legislation. Including environmental regulation in column (3) resulted in a coefficient of -0.026 for Tech. However, the data for lnEnvr passed the 1% significance test. This indicated that environmental regulation can reduce carbon emissions and act as a mediator in Tech and CE’s influence mechanism. In column 4, the effect of Tech on CE is statistically distinguishable from zero. In addition, the ACME of environmental regulation decreases carbon emissions by 0.157 standard deviations, which is empirically significant (see confidence intervals for 95%) and represents 85.669% of the overall impact, making the overall effect greater than the ADE. Despite this, the direct effect of Tech on CE is statistically insignificant (see 95% confidence ranges). Consistent with earlier regression findings, Tech can only provide carbon reduction benefits via lnEnvr. This facilitates the enhancement of carbon abatement. According to Porter’s premise, green innovation further reduces carbon emissions in the building sector through environmental legislation. The fully mediated role of lnEnvr suggests that the more stringent the environmental regulation intensity, the greater the carbon-reduction effect of green innovation.

4.6. Heterogeneity

4.6.1. Regional Heterogeneity

Depending on economic growth, population size, and other macroeconomic factors, the effect of Tech on CE in the building industry may also be varied. Therefore, it was also necessary to perform a heterogeneity analysis for these differences.

Referring to standard practices in the existing literature, we divided the study sample into three regions, East, Central, and West, denoted as region1, region2, and region3.

Table 9. Regional Heterogeneity Regression Results.

	Dependent Variable: CE
	(1)	(2)	(3)
	Eastern	Central	Western
Tech	−0.121 ***	0.194 *	0.311
	(0.035)	(0.087)	(0.198)
Constant	−3380	−4097	13,175 ***
	(3829)	(4231)	(2124)
Control variables	Yes	Yes	Yes
Year FE	Yes	Yes	Yes
Province FE	Yes	Yes	Yes
R2	0.400	0.878	0.876
N	192	144	144

Notes: (1) Robust standard errors are in parentheses; (2) “***” and “*” denote statistical significance at the 1% and 10% levels, respectively.

presents the results of the regression.

The regression findings of Table 9 suggested that Tech was substantially negative at the 1% level and highly positive at the 10% level in the Eastern and Central areas. In the Eastern area, carbon emissions fell by 12.1 units for every 1 unit rise in the Tech factor. Every 1-unit rise in the Tech factor in the Central region increased CO₂ emissions by 19.4 units. The Eastern region is more economically developed and has a better potential to absorb R&D funding than the Central and Western regions. It also has a huge influx of research expertise, making it easier for technical innovation to play a big role in carbon reduction in the building industry.

4.6.2. Heterogeneity of Innovation Factors

Financial input in science and technology (FIST) and human capital support green innovation and are important factors driving the industry to achieve carbon-emissions reduction. We investigated whether the CO₂ abatement effect of green innovation is characteristically varied between different levels of FIST and human capital. The number of survey and design institutions and personnel in the construction industry were used to represent human capital. The proportion of the public budget allocated to science and technology expenditures shows the financial investment in research and technology. Based on the medians, provinces were divided into high and low phases.

Table 10. Heterogeneity of Innovation Factors and Heterogeneity of Development-Type Regression Results.

	Dependent Variable: CE
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Type1	Type2	Type3	Type4	Type5	Type6	Prov1	Prov2
Tech	−0.139 **	0.584 **	−0.217 *	−0.000	−0.184 *	−0.438	−0.108 **	−0.781 *
	(0.060)	(0.273)	(0.125)	(0.299)	(0.094)	(0.348)	(0.048)	(0.353)
Constant	−7825	2688	−7230	7288	−14,212 *	5213	−363	2718
	(9873)	(5589)	(5633)	(5012)	(7797)	(3391)	(5933)	(6203)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Province FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
R2	0.364	0.510	0.411	0.436	0.446	0.507	0.511	0.563
N	240	240	240	240	240	240	320	160

Notes: (1) Robust standard errors in parentheses; (2) “**” and “*” denote statistical significance at the 5% and 10% levels, respectively; (3) “Type1” denotes high-tech financial inputs, “Type2” denotes low-tech financial inputs, “Type3” denotes high human capital-I, “Type4” denotes low human capital-I, “Type5” denotes high human capital-II, “Type6” denotes low human capital-II, “Prov1” denotes non-resource-based, and “Prov2” denotes resource-based, respectively; (4) “I” and “II” denote the number of survey and design institutions and personnel in the construction industry in human capital, respectively.

’s columns (1) and (2) show the regression findings for financial input to research and technology. Table 10’s columns (3), (4), (5), and (6) detail the human-capital regression findings. The findings indicate that an increase in Tech considerably decreases CE in provinces with high-tech financial inputs and human resources.

4.6.3. Heterogeneity of Development Types

The economic growth of resource-based provinces is primarily based on factor inputs, such as labor and mineral resources, and industries are mostly heavy-chemical industries. There are low levels of technological innovation in these provinces. Specifically, the growth of green innovation needs to be improved in resource-rich regions, which resources contribute to their high carbon emissions. So, then, have green solutions significantly decreased carbon emissions in the building sector in areas with abundant natural resources? The regression results for developmental-type heterogeneity are presented in the last two columns of Table 10. The estimated coefficients for Tech are significantly negative in columns (7) and (8). This means that the increase in eco-friendly technologies may significantly cut CO₂ emissions from the construction industry in resource-poor areas. Still, the impact will be greater in natural-resource-rich provinces.

5. Discussion

Based on this study, we have made some new findings.

(1)

The growth in green innovation has successfully decreased CO₂ emissions in the inter-provincial building sector. CO₂ emissions come largely from fossil-fuel combustion in industries with high carbon emissions. The production benefits of higher levels of technological innovation may compensate for the investment costs of combating emissions [70]. Aware of the global warming challenge caused by “fast growth, high energy usage, and high carbon dioxide emissions”, the Chinese government has rapidly applied technological advances to improve industrial efficiency and curb CO₂ emissions. Through technology development, the application of low-carbon products, expanded green productive capacity, multiple different spillover effects, cost reductions, and specialized labor division, green innovation can reduce the amount of energy consumed in the construction sector and the CO₂ emissions caused by low productivity, high energy consumption, and high input. Therefore, green technical innovation can optimize the industrial structure, eliminate polluting industries, improve energy-consumption efficiency, and meet the objective of attaining energy conservation, which will influence the industry’s carbon emissions ineluctably. In short, the carbon emissions of the building industry can be drastically reduced by increasing the degree of green innovation.

(2)

A certain amount of environmental control facilitates enhancing the potential of green technologies to reduce dioxide emissions. This may have varying effects depending on the extent of environmental control. (1) As the amount of environmental regulation increases, so does the demand for punitive measures, the attractiveness of the incentives, and the motivation to cut carbon emissions. Therefore, more construction firms will reduce some of their production-expansion funds for research and development, while also tending to use new technologies to improve the cleanliness of construction production, thereby achieving CO₂ emissions reduction and meeting governmental environmental regulatory requirements. (2) In places where environmental regulations are lax, construction companies are more willing to invest in expanding production. Since the earnings from such growth may more than balance the costs of environmental regulation, construction companies are less likely to research and implement technical solutions to minimize carbon emissions in the building sector. China’s strong environmental policies have helped to reduce carbon emissions globally. However, environmental management is a long process that should not be achieved by regulatory policies alone, and the link between technological innovation and environmental regulations should be emphasized [71].

(3)

Green innovation exploits environmental regulations to provide additional incentives for carbon emissions in the building sector. With a high degree of environmental regulation, construction businesses think that the penalties for polluting emissions are too severe, causing the costs of emissions to far outweigh the costs of lowering emissions. Consequently, businesses are more ready to invest in research and development or adopt new eco-friendly technologies to increase the cleanliness of building production, thus lowering CO₂ emissions and internalizing external expenses.

(4)

Research reveals that (1) green innovation substantially influences dioxide reduction in the eastern building sector. In Eastern China, the economy is more developed, the awareness of the environment is greater, technology is more sophisticated, and technical research and development facilities are well-established. Establishing stringent environmental legislation and green innovation might amplify the impact of CO₂ reduction in the building sector. For the Central and Western regions, which are primarily focused on the development of heavy industries and have long relied on a high-energy-consumption economic growth model, the effect of technological innovation guidance needs to be strong. (2) Increased levels of green innovation significantly reduce carbon emissions in high-tech financial investment and high-human-capital provinces. Following China’s regulations regarding subsidies for energy conservation and dioxide-abatement projects, the fourteenth 5-year plan has continued to increase expenditures in research, technology, and human capital. Through autonomous research and development of cleaner production technologies and the absorption of energy-saving technologies, the “double-high” companies will create suitable energy-saving technologies and cleaner production models. (3) Resource-based provinces have greater potential to reduce emissions than non-resource-based provinces. In other words, both forms of carbon emissions have been reduced, but the decline in the former is much more pronounced. With economic growth and regional mobility, the degree of eco-friendly technologies in the construction sector in resource-based provinces has progressively grown, as has the expansion of technological applications. Regarding national policy and transitional development, their industries are progressively moving from a resource-intensive, highly polluting, and energy-intensive kind to a technology-intensive one. They are continually overcoming the “resource curse” challenge, and the degree to which eco-friendly technologies are integrated into the growth of these firms is increasing. However, the increasing investment in scientific and technical expertise and funding for R&D in the building industry of these provinces has increased their growth in terms of green innovation. The energy-consumption structure has moved from using fossil fuels to using clean and renewable energy. Due to these characteristics, green innovation may majorly impact carbon reduction.

This article comprehensively reviews the Chinese provincial building industry’s efforts to reduce dioxide emissions. It helps important ministries, governments, and enterprises comprehend technological techniques for lowering dioxide emissions. Based on the viewpoint of emissions-reduction pathways, the findings may serve as a policy reference for the future direction of CO₂ emissions abatement in China’s building sector. This paper’s study framework and technique apply to examination of the construction industry in China and other sectors and nations.

6. Conclusions and Implications

6.1. Conclusions

In the context of carbon peaking and neutrality, green innovation is essential for lowering carbon emissions in the context of carbon peaking and neutrality. This study investigated the linear and non-linear relationships between green innovation and CO₂ emissions in the construction sector using panel data from 2005 to 2020 for 30 Chinese regions. Meanwhile, we investigated the moderating influence that environmental regulation has on green innovation. In addition, we empirically examined the potential heterogeneity of green innovation in terms of CO₂ emissions by reference to geography, innovation factors, and types of development. The following are the primary findings and policy suggestions.

We obtained four results:

(1)

There is a negative association between green innovation and dioxide emissions in China’s construction business. This study found a significant impact on the total amount but did not find a significant impact on the proportion of emissions reduction. Further research is needed.

(2)

The greater the severity of environmental regulation above the second barrier, the greater the abatement effect of green innovation.

(3)

Environmental regulation substantially and beneficially moderates the influence of green innovation on CO₂ emissions.

(4)

Only in the East can technology substantially negatively affect regional-construction-related dioxide emissions. In contrast, the degree of green innovation in high-tech financial investment and high-human-capital regions can cut carbon emissions by a substantial amount. Increased green innovation can significantly cut carbon emissions in resource-poor provinces but has a greater abatement impact in resource-rich provinces.

6.2. Implications

The above conclusions have some policy implications:

(1)

Policy guidance should be considered to strongly support the introduction, research, and development of green technologies.

(2)

We should rationalize environmental regulation policies and improve environmental regulations. Accelerating the implementation of environmental tax policies and establishing a strict technology and industry access mechanism is important.

(3)

We should enhance the technical content of emissions abatement and produce a “win-win” scenario for regional growth and technological advancement in CO₂ emissions abatement in the construction industry.

The limitations of this article may indicate future research topics. First, this study only covered provincial research scales. Selecting smaller scales, such as the city, would have made the results more comprehensive. Carbon emissions at the microscopic level might be a topic of future study. Second, all the energy each province’s building sector utilized was taken for the computation of direct carbon emissions. However, secondary energy carbon emissions, such as electricity consumption and thermal energy, were not considered. Specific indirect consumption and emissions were also excluded. This research only proves that total growth in green innovation can bring about carbon-reduction effects and does not show the impact of proportional changes in green innovation. One possible reason is that the patterns discovered in this article are typical non-linear relationships. Considering regional differences, the impact of a 1% increase in green patent levels may exhibit significant differences. It is necessary to investigate further why this pattern only appears at the total level of green patent data and not proportionally. Future research should expand the scope of carbon-emissions measurement and consider indirect energy-consumption and emissions data.