Does the Digital Economy Promote Green Land Use Efficiency?

Abstract

Land is a critical factor of production that contributes significantly to economic growth. However, conventional land use pattern in China has resulted in serious environmental pollution. Now enhancing green land use efficiency (GLUE) has emerged as an effective strategy for improving environmental quality. The development of the digital economy (DE), characterized by low cost and high efficiency, has demonstrated considerable potential in reducing environmental pollutants and enhancing resource allocation. This study employs an extensive analytical framework to analyze the impact of DE development on GLUE across 267 cities in China from 2011 to 2019. The results show that DE exerts a significant effect on improving GLUE, which remains valid after the execution of endogeneity and robustness tests. The research on mechanisms indicates that this promotional effect is primarily achieved through the innovation in green technology and the optimization of industrial structure. Extended empirical tests indicate there is a nonlinear trend, wherein the positive effect increasingly intensifies after green industry innovation and industrial structure optimization exceeds threshold values. There is also a significant short term spillover effect of DE on GLUE, supplemented by long term effects. These findings substantially improve our comprehension of the connection of DE and land use, while providing practical policy recommendations for promoting environmentally sustainable development and land green utilization.

1. Introduction

Increasingly serious ecological and environmental problems have received wide attention. Words such as “green”, “low-carbon” and “environmental protection” frequently appear on various occasions. The 17 Sustainable Development Goals (SDGs) proposed by the United Nations list environmental protection as a priority [1]. Accordingly, the pursuit of green and sustainable development has become a consensus among all countries [2,3]. As the carrier for economic and social advancement [4] and the scarce factor in the production process, land is the key resource for economic development [5] and human well-being [6,7]. China’s traditional land use mode of “seeking development by land acquisition” has resulted in a miracle of economic growth [8,9]. Still, it has resulted in the inefficient use of resources and environmental degradation [10,11], which is contrary to the concept of sustainability [12,13]. As China’s economic development has entered a new normal, the Chinese government has placed greater emphasis on environmental preservation, prioritizing green development [14]. Therefore, the traditional land use pattern, characterized by high costs, high pollution, and low efficiency [15], poses a challenge to supporting the green economy [5]; thus, it needs to be urgently transformed into a new pattern. The purpose of green land use is “high efficiency, high yield, less energy use, reduced pollutants, and diminished carbon emissions” [16]. Although the government has enacted a series of land protection and regulation policies aimed at improving land use efficiency [17], China’s rapid urbanization and industrialization have significantly impacted sustainable land use [16], and the overall level of the GLUE is rather low in China [7]. In this context, a new driving force is urgently needed to enhance the GLUE.

In recent years, with the rapid advancement of big data and internet technologies, DE has become a prominent research focus [18,19]. Digital economy has permeated all aspects of socioeconomic activities and daily life [20], establishing itself as a crucial driver for both socioeconomic advancement [21,22] and technological innovation [23]. DE, characterized by low cost and high efficiency, makes a positive contribution to low-carbon environmental protection [24,25,26] and enhances the optimization of resource consumption and allocation [18,27]. Previous research has researched the positive impact of DE on general resource allocation [28] or specific resources, including R&D resources [29] and the resources allocation within the service industry [30]. Nevertheless, inadequate attention has been paid to its impact on land allocation. In China, in the context of eco-friendly and sustainable advancement, can DE in China also promote GLUE in China? If the answer is yes, then how does DE influence GLUE? These questions need to be empirically tested. Clarifying the aforementioned questions holds considerable theoretical significance in enhancing GLUE and advancing green development.

The vigorous development of DE has attracted wide attention. The existing literature about it can be divided into two categories: the first one deals with measuring and calculating it. Most scholars evaluate DE by constructing comprehensive indicators from different dimensions based on their research focus. The dimensions of the assessment index system mainly include three types: the first type focuses on information improvement, internet development, E-commerce development, and digital finance [28]; the second type emphasizes digital infrastructure, digital industrialization, the digital innovation ecosystem, and financial inclusion [31]; the third type focuses on digital industrialization and industrial digitization [32]. Another category is to research the impact brought by it. Most of the existing literature focuses on its economic benefits. Existing research results show that DE can enhance economic growth [33], foster high-quality economic development [34,35], and improve total factor productivity [36,37]. Certain literature additionally emphasizes the impact of DE on technological innovation [25], industrial sector upgrading [38], as well as the optimization of resource allocation [18,27]. In comparison, only a few studies have explored its impact on ecological benefits, which primarily focus on environmental quality [24,39] and carbon emissions [25,26]. Existing literature has shown that DE can both directly and indirectly enhance environmental quality.

With the prevalence of green development, GLUE has also gained attention among researchers and scholars. The existing academic research on DE can be broadly classified into two categories. The first category of literature focuses on how to measure and evaluate it. Since a single index is difficult to comprehensively reflect the status of GLUE, most existing literature adopts a comprehensive index valuation system [40]. The comprehensive indicators primarily include both desirable and undesirable inputs and outputs within the evaluation system. The measurement methods of the extensive index evaluation system, such as SFA [41,42,43] and DEA [40,44,45], are widely used [46,47], as they can incorporate both desired and undesirable outputs into the measurement model. After the GLUE was calculated, the literature continued to analyze its dynamic change characteristics. The research results indicate that GLUE presents dynamic characteristics of continuous rise overall, as well as positive spatial auto-correlation and spatial agglomeration characteristics [48,49]. The second category analyses the determinants influencing GLUE. The research results indicate that economic development [50,51], industrial structure optimization [52,53], urbanization [54,55], and city size [49] can all have significant impact on land use efficiency.

Prior research has examined DE and land use efficiency, offering valuable good references for this study. Some recent studies find that DE could exert a promoting effect on the green use efficiency of urban land [56,57,58]. However, there are some aspects that should be improved, such as the potential nonlinear impact and the spatial effect, which both should be explored. The innovative and marginal contributions of this study are as follows: (1) In addition to the linear impact of DE on GLUE, this study also systematically verifies the nonlinear impact and spatial effect of DE on GLUE, which can deepen the research question from multiple perspectives and make it more specific. (2) This study conducts the heterogeneity test based on the city type, development stage of DE, and marketization level of the city. The above detailed research will be more conducive to putting forward practical and corresponding green development policy recommendations.

2. Theoretical Analysis and Research Hypotheses

2.1. Influencing Mechanism of DE on GLUE

With the support of the Internet and digital innovation, DE can improve LUGE directly. First, it serves as a platform to facilitate data sharing and information exchange within the market, utilizing the Internet information technology as a bridge and intermediary. In addition to significantly lowering production cost and transaction cost for businesses, this helps mitigate the risk of information asymmetry among market participants [59,60]. This can enable the enterprise to have sufficient capital for ecological conservation investments and green management and further improve GLUE. Second, DE has enhanced the production quality and efficiency of enterprises and optimized their management and production processes [61]. Based on available data, enterprises can utilize cloud computing and intelligent analysis to simulate production processes and adjust the proportion of various production factors. In this way, economic, social, and ecological benefits can be guaranteed, energy consumption can be reduced, resource allocation can be optimized, and GLUE can be improved. Third, the development of DE has overcome geographical and time limitations, thereby increasing the degree of marketization and enhancing trans-regional resource flow efficiency. This can further intensify market competition and guide the capital flow into environmentally friendly industries, thereby forcing high-polluting enterprises to increase their GLUE. Fourth, with the aid of big data and cloud computing, environmental protection authorities can continuously monitor ecological quality [62,63]. DE also provides online platforms for the public to participate in ecological supervision [25]. Open and transparent environmental information significantly enhances the efficiency of ecological supervision. Therefore, the government can formulate corresponding environmental management policies [64] to reduce environmental pollution and improve GLUE. Therefore, the first research hypothesis is put forward as follows:

H1. 

DE could improve GLUE directly.

DE can also promote GLUE by enhancing green technology innovation. Existing research shows that DE can significantly promote green technology innovation [65,66]. In terms of innovation resources and innovation factors, DE’s advantages of cross-space transmission and its deep integration with various fields can accelerate the communication and exchange of green innovation knowledge and innovation factors among different regions. This can improve various innovation resource efficiencies [25] and shorten the innovation process, which contributes to releasing the innovation value of these innovation resources. Additionally, green technology innovation is a time-consuming and risky process [67]. In terms of innovative subject, enterprises require financial support to implement green technology innovation [68,69]. Emerging digital technology has generated various online fundraising platforms. This broadens finance support availability for enterprises [70] and eases the financial pressure faced by enterprises [71,72]. Thus, it can mitigate the risk of green technology innovation and encourage them to participate in eco-friendly technology innovation. Green technology innovation is crucial to solve environmental pollution. It can diminish pollution emissions and mitigate the environmental impact of energy consumption in the production process [39]. In this way, the misallocation of resources caused by negative ecological externalities can be reduced [66]. Therefore, the second research hypothesis is put forward as follows:

H2. 

DE could promote GLUE through green technology innovation.

DE can also promote GLUE through optimizing industrial structure. The development of DE can stimulate the formation of new industries patterns [73], such as low-carbon emission industries, low-energy consumption industries, and low-pollution industries [74]. This can contribute to the transformation of the production structure of enterprises, further promoting GLUE. First, DE can be incorporated into the manufacturing process as a special production factor. Unlike traditional factors of production, which inevitably lead to diminishing marginal returns, DE can overcome this inherent deficiency by its advantage of higher efficiency, lower cost, replicability, and massive acquisition [75], and its marginal cost can gradually decrease and may even tend to zero. This not only solves the limitation of diminishing the marginal returns of traditional production factors, but it can also bring a scale economy effect to the enterprise, change the current production and management mode, and contribute to industrial structure optimization and upgrading. Second, as a new form of economy, DE can enhance industrial intelligence through industrial digitization and digital industrialization [75], and further promote the transformation of industrial structures [76]. Through digital transformation, enterprises can eliminate outdated production and management modes, giving rise to a new business model. Third, DE also enables traditional industries to achieve advancements and transformations [77,78]. Industrial structure is considered a resource converter [79] that can optimize the combination of production factors, like land, labor, and capital. Digital transformation in industry can change the traditional mode of resource utilization, improving the efficiency of land resource utilization [80]. Simultaneously, industrial structure optimization and upgrading is a dynamic process that achieves high-output and low-energy consumption [80,81]. Therefore, it can encourage high-pollution and high-energy-consumption enterprises to switch their production pattern to a greener production pattern, thus reducing environmental pollution and improving GLUE. Therefore, the third research hypothesis is put forward as follows:

H3. 

DE could promote GLUE through industrial structure optimization.

The theoretical analytical framework about the influencing mechanism of DE on GLUE is presented in Figure 1.

2.2. Nonlinear Effect of DE on GLUE

DE’s development process is gradual [25]. Consequently, the effect of DE on GLUE might not be linear. Firstly, during the early stage of DE development, the network scale is small, and the internet technology is not yet mature enough. Thus, the role of DE on GLUE is limited. As network infrastructure and communication technologies develop, DE develops rapidly and becomes more mature. This process can break the barrier and significantly decrease the marginal cost of information acquisition, thereby the promotion effect of DE on GLUE also shows a marginally increasing trend. Secondly, based on the Internet, DE has the characteristics of network effects of information technology [82]. The value of DE keeps increasing as the number of internet users grows. Therefore, its impact on GLUE has the trend of marginal increase. Thirdly, a nonlinear relationship exists between DE and carbon emissions or pollution emissions [83,84]. These emissions are undesired outputs of land use. Thus, it can be inferred that DE might affect GLUE in a nonlinear way. Fourthly, existing research has indicated that the relationship between LUGE and the upgrading and optimization of industrial structures is not linear [52]. This implies that DE, which promotes GLUE through optimizing industrial structures, may also have a nonlinear relationship with GLUE. Therefore, the fourth research hypothesis is put forward as follows:

H4. 

DE exerts a marginally increasing nonlinear influence on the enhancement of GLUE.

2.3. Spatial Spillover Effect of DE on GLUE

DE exerts spatial spillover effect on economic growth, resource allocation [85], and technological innovation [86]. Then it may also have similar effect on GLUE. Firstly, its advantages of information transmission and data sharing can break the boundaries of traditional economic activities [79], promote information exchange and factor flow efficiency between neighboring regions, and optimize resource allocation efficiency. Moreover, there exists competition and imitation among local governments [53]. The prosperity and advancement of DE in adjacent regions will compel the local government to imitate and catch up and subsequently improve the local GLUE. Secondly, air pollution, water pollution, and carbon emissions have the characteristics of spatial diffusion. DE can significantly reduce these emissions [26,87,88,89] and mitigate the spread of emissions to other regions. In addition, the expansion of DE in the nearby areas can reduce local pollution emissions and improve local environmental quality, which will further promote GLUE by spatial spillover effect. Thirdly, as previous analysis shows, DE can improve GLUE by promoting green technology innovation, which has spatial spillover effect [90]. The Internet serves as a platform for exchanging innovative ideas and factors, effectively facilitating local innovation activities. Technological innovation in adjacent areas can affect local innovation activities through its spatial spillover effect, driving local green technology innovation and further improving its GLUE. Therefore, the fifth research hypothesis is put forward as follows:

H5. 

DE exerts a spatial spillover impact on GLUE.

3. Research Methods and Data

3.1. Research Methods

3.1.1. Empirical Model

To test the effect of DE on GLUE, this study constructs the following panel regression model:

$${\mathrm{GLUE}}_{\mathrm{i},\mathrm{t}}={\alpha }_0+{\alpha }_1{\mathrm{DE}}_{\mathrm{i},\mathrm{t}}+{\alpha }_2{\mathrm{X}}_{\mathrm{i},\mathrm{t}}+{\mu }_{\mathrm{i}}+{\delta }_{\mathrm{t}}+{\epsilon }_{\mathrm{i},\mathrm{t}}$$

(1)

In Formula (1), i is city, t is year, $\mathrm{G}\mathrm{L}\mathrm{U}\mathrm{E}$ is green land use efficiency; $\mathrm{D}\mathrm{E}$ is the development level of digital economy; X denotes a set of control variables; ${\mu }_{\mathrm{i}}$ and ${\delta }_{\mathrm{t}}$ are city and time fixed effect, respectively; ${\epsilon }_{\mathrm{i},\mathrm{t}}$ denotes the error term.

This study adopts the following model to confirm the influencing mechanism of DE on GLUE:

$${\mathrm{M}}_{\mathrm{i},\mathrm{t}}={\beta }_0+{\beta }_1{\mathrm{DE}}_{\mathrm{i},\mathrm{t}}+{\beta }_2{\mathrm{X}}_{\mathrm{i},\mathrm{t}}+{\mu }_{\mathrm{i}}+{\delta }_{\mathrm{t}}+{\epsilon }_{\mathrm{i},\mathrm{t}}$$

(2)

$${\mathrm{GLUE}}_{i,t}={\gamma }_0+{\gamma }_1{\mathrm{DE}}_{i,t}+{\gamma }_2{\mathrm{M}}_{i,t}+{\gamma }_3{X}_{i,t}+{\mu }_i+{\delta }_t+{\epsilon }_{i,t}$$

(3)

According to the formula above, M denotes the mechanism variables, and the meanings of the remaining variables are consistent with Equation (1).

3.1.2. Other Research Methods

We also employ other research methods [17], such as analysis, comparison, and induction. We employ the analysis method to analyze the empirical results and determine whether they are consistent with the research hypotheses. We use the comparison method to analyze the research contents of the existing literature and discuss the findings of this study with the existing literature. We summarize the study’s research findings, draw conclusions from the research, and offer recommendations using the inductive analysis method.

3.2. Description of Variables

3.2.1. Digital Economy (DE)

The prevailing academic consensus regarding the exact definition and measurement of DE is ambiguous. Bukht and Heeks (2017) proposed a tripartite conceptual framework, categorizing DE into three layers [91]: the fundamental core layer, the intermediate narrow layer, and the extensive broad layer. The Chinese Academy of Communications further elaborates this conceptualization through a four-dimensional paradigm, examining digital industrialization, industrial digital transformation, digital governance frameworks, and data valuation mechanisms. Nevertheless, the pervasive integration of digital technologies across multiple industrial value chains presents significant challenges for quantitative assessment. Consequently, the existing literature has adopted various measurement approaches, with some scholars focusing exclusively on internet penetration and digital financial services [26,37,92], while others utilizing information technology infrastructure indicators as proxies [93,94].

By referring to existing research [95], this study evaluates DE using the multi-index construction method, with a focus on Internet development. Due to the availability of data, we construct an evaluation index system from five aspects, see Table 1. The reason for taking the Digital Finance Inclusive Index as an evaluation indicator is that this indicator is a key dimension for measuring the social value, technological penetration and economic efficiency of DE. If ignored, it may result in the evaluation index system placing excessive emphasis on technical aspects, thereby failing to adequately capture the inclusive nature of DE. Refer to Table 1 for specific indicators. The weight of each indicator is measured by principal component analysis in this study.

3.2.2. Green Land Use Efficiency (GLUE)

The GLUE is assessed using the SBM model which includes undesired output in this study. Taking into comprehensive consideration the features of green land use and previous research [16], the input indicators of GLUE are chosen from four important factors: capital, land, labor and energy. The perpetual inventory method is employed to calculate the capital stock, which is based on the actual total investment in fixed assets over the years. The calculation method of initial capital stock is referred to in reference [96]. The following are desirable outcomes: ecological, social, and economic. Undesirable outputs including industrial sulfur dioxide (SO2), wastewater, and soot emissions are represented by the Environmental Pollution Composite Index, calculated by the entropy approaches. Taking inflation into account, we take 2005 as the base year for constant price treatment. In addition, we perform a 1% tail reduction on the GLUE. For a comprehensive description of the indicators, please refer to Table 1.

Table 1. Measurement index system of DE and GLUE.

First-Level Indicator		Second-Level Indicator	Third-Level Indicator	Unit	Weight
DE		Internet penetration rate	Internet users per 100 people	/	0.0865
		Internet-related employees	The percentage of employees in the computer and software industry	%	0.1896
		Internet-related output	Total number of telecommunications services available per capita	/	0.0166
		Mobile Internet Users	Mobile phone users per 100 individuals	/	0.0358
		Digital finance development	China Digital Financial Inclusion Index	/	0.6716
GLUE	Input	Capital	Investment in urban fixed assets	Million RMB	/
		Land	Urban construction land area	km2	/
		Labor	Employees in the secondary and tertiary industries	Ten thousand people	/
		Energy	Coal, oil, natural gas, and electricity are converted to standard coal	Ten thousand tons	/
	Desirable output	Economic output	Value added by the secondary and tertiary sectors	Million RMB	/
		Social output	General government budget revenue	Million RMB	/
		Ecological output	Park and green area	m2	/
	Undesirable output	Environmental pollution	Index of environmental pollution	/	/

Table 1. Measurement index system of DE and GLUE.

First-Level Indicator		Second-Level Indicator	Third-Level Indicator	Unit	Weight
DE		Internet penetration rate	Internet users per 100 people	/	0.0865
		Internet-related employees	The percentage of employees in the computer and software industry	%	0.1896
		Internet-related output	Total number of telecommunications services available per capita	/	0.0166
		Mobile Internet Users	Mobile phone users per 100 individuals	/	0.0358
		Digital finance development	China Digital Financial Inclusion Index	/	0.6716
GLUE	Input	Capital	Investment in urban fixed assets	Million RMB	/
		Land	Urban construction land area	km2	/
		Labor	Employees in the secondary and tertiary industries	Ten thousand people	/
		Energy	Coal, oil, natural gas, and electricity are converted to standard coal	Ten thousand tons	/
	Desirable output	Economic output	Value added by the secondary and tertiary sectors	Million RMB	/
		Social output	General government budget revenue	Million RMB	/
		Ecological output	Park and green area	m2	/
	Undesirable output	Environmental pollution	Index of environmental pollution	/	/

3.2.3. Other Variables

Other variables are as follows:

(1)

Green technology innovation (Pat): The quantity of patent applications could represent well the quantity of technological innovations of an enterprise [97,98]. Since exterior design patents are not considered green patents, this study focuses exclusively on the quantity of the utility model and green invention patents. Thus, the logarithm of the total number of the above two types of patent applications represents green technical advancement.

(2)

Industrial structure optimization (Str): This study represents it by dividing the added value of the secondary industry by that of the tertiary industry [56].

(3)

Control variables (CVs): Referring to the existing literature [20,99], this study selects the following control variables: The economic development (PGDP) is denoted by the logarithm of GDP per capita. The ratio of educational expenditure to government financial expenditure denotes the level of education (Edu). Opening (Open) is denoted by the proportion of actual foreign capital utilized relative to GDP. The logarithm of population density represents urban population size (Pop). The ratio of the deposit and loan balance of financial institutions to GDP denotes financial development (Fin). The logarithm of road mileage data in each city represents the level of infrastructure (Road).

3.3. Data Sources

Empirical research was conducted on 267 Chinese cities from 2011 to 2019, excluding cities with unavailable data. The source of green patent data is the National Intellectual Property Database. We match them with the WIPO International Patent Classification list. The China Digital Financial Inclusion Index comes from the Digital Finance Research Center of Peking University. The primary sources of original data for the remaining variables are the China Urban Statistical Yearbook, the China Energy Statistical Yearbook, the Wind Database, and the CSMAR Database. We use the interpolation method to supplement some missing data.

Table 2. The descriptive statistics of variables.

Variable	N	Mean	SD.	Min	Max
GLUE	2383	0.667	0.092	0.392	1.131
DE	2383	4.025	0.936	1.196	8.198
Pat	2383	5.238	1.641	0.693	10.507
Str	2383	0.960	0.499	0.176	5.154
PGDP	2383	10.727	0.584	8.773	15.675
Edu	2383	0.178	0.040	0.036	0.356
Open	2383	0.017	0.018	0.000	0.199
Pop	2383	5.791	0.891	1.629	7.882
Fin	2383	2.372	1.131	0.588	21.301
Road	2383	9.288	0.661	6.291	12.068

provides the descriptive statistics of the variables.

In Table 2, the values of DE range from 1.2 to 8.2, and the explained variable, GLUE, ranges from 0.4 to 1.13. The other control variables also exhibit a narrow range between their maximum and minimum values. This indicates that the data used in this study has a limited degree of variability during the study period.

In addition, we use the variance inflation factor test to test whether there is multi-collinearity among the variables, and the results showed that the variance inflation factor of each variable is far less than 5, indicating that the constructed regression model has no obvious multi-collinearity problem.

4. Empirical Results

4.1. Benchmark and Mechanism Results

Table 3. Benchmark regression results.

Variable	(1)	(2)
DE	0.0437 *** (0.00539)	0.0198 *** (0.00478)
CVs	N	Y
City fixed	Y	Y
Time fixed	Y	Y
N	2383	2383
adj. R-sq	0.065	0.255

Note: *** denotes statistical significance at the 1%; and the values in parentheses are robust standard errors clustered at the city level.

shows the results of the influence of DE on GLUE based on Equation (1). Column (1) and Column (2) display the results without adding and with adding control variables, respectively. It is clear that whether control variables are added or not, DE’s influence coefficient remains positive and at 1% significance level. Thus, the first hypothesis has been validated.

The influence mechanism of DE on GLUE is verified by Equations (2) and (3), and the results are shown in

Table 4. Mechanism regression results of DE on LUGE.

Variable	(1)	(2)	(3)	(4)
	Pat	GLUE	Str	GLUE
DE	0.362 ***	0.0113 **	0.109 ***	0.0148 ***
	(0.0609)	(0.00440)	(0.0235)	(0.00461)
Pat		0.0236 ***
		(0.00247)
Str				0.0464 ***
				(0.00838)
CVs	Y	Y	Y	Y
City fixed	Y	Y	Y	Y
Time fixed	Y	Y	Y	Y
N	2383	2383	2383	2383
adj. R-sq	0.543	0.307	0.228	0.273

Note: *** and ** denote statistical significance at the 1% and 5% level, respectively; and the values in parentheses are robust standard errors clustered at the city level.

. Column (1) shows the results of DE on green technology innovation. The coefficient of DE is 0.362, which is significant at the 1% level. Column (2) shows the regression results of both DE and green technology innovation as explanatory variables on GLUE. The result shows that their coefficients are positive and significant at the 5% and 1% level, respectively. The coefficient of DE in column (2) is smaller than that in the benchmark regression results. The above result indicates that green technology innovation, as a mediating variable, can enhance GLUE through DE, and it plays a partial mediating role, accounting for 43.14%. This result confirms Hypothesis 2.

Likewise, when industrial structure optimization serves as a mechanism variable, the results are presented in columns (3) and (4). The influence coefficient of DE on industrial structure optimization is 0.109, which is significant at 1% significance level. It shows that DE can enhance the optimization of industrial structure. In column (4), the industrial structure optimization coefficient on GLUE is positive, and at 1% significance level. At this time, the coefficient of DE on GLUE remains significantly positive but decreases to 0.0148. This indicates that DE’s promotion effect on GLUE is weakened. The industrial structure optimization, as a mechanism variable, can promote GLUE by DE. And it plays a partial role, accounting for 25.54%. This result confirms Hypothesis 3.

4.2. Robust Test

4.2.1. Endogeneity Test

The empirical results of this study verify that DE can improve GLUE. The improvement of GLUE can also stimulate production efficiency and drive economic development, which may, in turn, influence the development of DE. Drawing upon prior research [100], the instrumental variables in this study are the number of post offices per million people and the quantity of fixed telephones per 100 people in the year 1984 in each city. The development of the Internet will be perpetually influenced by the historical advancement of communication technologies, such as post and telecommunications. Thus, it satisfies the relevance requirement of instrumental variables. Meanwhile, the historical postal and telecommunication numbers do not directly influence the current GLUE. And the year 1984 marked the turning point for the reform of China’s telecommunication industry. These satisfy the exogeneity of the instrumental variables. The sample for this empirical test is panel data, while the instrumental variables selected are cross-sectional data. As a result, panel instrumental variables are constructed by multiplying the two cross-sectional instrumental variables from 1984 by a time-changing variable. According to the setting method of Nunn and Qian (2014) [101], this study employs the number of Internet users in a city in the last year to multiply its historical data in 1984 as instrumental variables. Subsequently, a 2SLS regression is conducted (see the results in

Table 5. Robust test results.

	Instrument Variable		GMM Model	Change GLUE	Exclude Municipalities and Sub-Provincial Cities
Variable	IV1	IV2
	(1)	(2)	(3)	(4)	(5)
L. GLUE			0.800 ***
			(0.107)
DE	0.0855 ***	0.0830 ***	0.0887 ***	0.0214 ***	0.0189 ***
	(0.00652)	(0.0058)	(0.0331)	(0.00487)	(0.00482)
AR (1)			−4.28
			[0.000]
AR (2)			0.41
			[0.685]
Sargan test			22.27
			[0.175]
Kleibergen–Paap RK LM statistics	53.578	95.894
	[0.000]	[0.000]
Kleibergen–Paap rk Wald F statistics	97.073	246.789
	{16.38}	{16.38}
CVs	Y	Y	Y	Y	Y
City fixed	Y	Y	Y	Y	Y
Time fixed	Y	Y	Y	Y	Y
N	2383	2383	1841	2383	2212
adj. R-sq	0.373	0.379	/	0.264	0.269

Note: *** denotes statistical significance at the 1% level; and the values in parentheses are robust standard errors clustered at the city level. The values in brackets are p value, the values in brace are the critical values at the 10% significance level based on the Stock–Yogo weak identification test.

).

Column (1) and column (2) in Table 5 display the estimation results with instrumental variables added, respectively. We can see the influence coefficient of DE is not only positive at 1% significance level, but also larger than the influence coefficients without instrumental variables. This implies that GLUE can still be significantly promoted by DE, even after accounting for endogeneity. The Kleibergen–Paap rk LM test results and the Kleibergen–Paap rk Wald F test results show that the model successfully passes the insufficient identification test of instrumental variables and the weak instrumental variable identification test. The results of the above two tests also illustrate the rationality of selecting instrumental variables.

4.2.2. Change the Empirical Model

This study adds GLUE’s lag one period into the model to avoid the influence of different empirical models on the results and consider the temporary continuity of land use. The dynamic panel model is as follows:

$${\mathrm{GLUE}}_{\mathrm{i},\mathrm{t}}={\alpha }_0+{{\alpha }_3\mathrm{LGUE}}_{\mathrm{i},\mathrm{t}-1}+{\alpha }_1{\mathrm{DE}}_{\mathrm{i},\mathrm{t}}+{\alpha }_2{\mathrm{X}}_{\mathrm{i},\mathrm{t}}+{\mu }_{\mathrm{i}}+{\delta }_{\mathrm{t}}+{\epsilon }_{\mathrm{i},\mathrm{t}}$$

(4)

The meanings of the variables are the same as in Equation (1).

In this study, the differential GMM model is used, and the regression results are detailed in column (3) in Table 5, which suggests DE still has a significant promotion effect on GLUE. Moreover, the choice of the dynamic panel model is also justified by the coefficient of lagged GLUE, which shows that GLUE in the last year can enhance GLUE in the current year. According to the differential GMM model test results, the residual term shows no evidence of second-order autocorrelation, as reflected by the p-value of AR (2) of 0.685. Additionally, the p value is 0.175. It is associated with the Sargan test, indicating that the selected instrumental variables are valid. The above results indicate the validity of benchmark regression results.

4.2.3. Replace the Explained Variable

To mitigate the influence of different calculation methods on regression results, we use the DEA model with non-expected output calculated by the DEA-Solver Pro5.0 software to measure GLUE. The results are reported in column (4) in Table 5. We can see the result remains robust.

4.2.4. Exclude Municipalities and Sub-Provincial Cities

A city with superior administrative levels may exhibit a greater propensity to obtain preferred policies. Consequently, this study excludes municipalities and sub-provincial cities with higher administrative levels from the sample data and subsequently conducts regression on the remaining samples. The results are presented in column (5) in Table 5, which suggest that the results are still robust.

4.3. Heterogeneity Analysis

Due to China’s extensive land coverage and geographical differences among cities, factor endowments, policies, and regulations are different among cities [102]. This study conducts heterogeneity analysis on the research samples from different perspectives;

Table 6. Results of heterogeneity regression.

Variable	(1)	(2)	(3)	(4)	(5)	(6)
	Central	Peripheral	2011–2013	2014–2019	Low_mkt	High_mkt
DE	0.0272 **	0.0193 ***	0.012 ***	0.0172 ***	0.0215 ***	0.015 **
	(0.0126)	(0.00355)	(0.00433)	(0.00640)	(0.00561)	(0.00741)
CVs	Y	Y	Y	Y	Y	Y
City fixed	Y	Y	Y	Y	Y	Y
Time fixed	Y	Y	Y	Y	Y	Y
N	278	2105	793	1590	1105	1278
adj. R-sq	0.063	0.173	0.097	0.182	0.307	0.221

Note: *** and ** denote statistical significance at the 1% and 5% level, respectively; and the values in parentheses are robust standard errors clustered at the city level.

shows the details of the results.

Because the industrialization level [100], factor endowment, and resource utilization efficiency [50] of different types of cities are different, this study divides the sample into two groups: central cities and peripheral cities. Municipalities, provincial capitals, and sub-provincial cities belong to central cities, and other cities belong to peripheral cities. Columns (1) and (2) in Table 6 present the regression results. Although the coefficient of DE in central cities and peripheral cities is different, they are all significantly positive. This result indicates that DE has a significant promotion effect on cities of different levels.

The year 2013 was a turning point for the rapid expansion of digital finance and was bound to drive the rapid growth of DE. Taking 2013 as the time dividing point, the sample is classified into two groups: the early development stage and the mature development stage. Columns (3) and (4) in Table 6 show the results of temporal heterogeneity. We can see that the coefficient of DE is 0.017 at the mature stage, which is larger than 0.012 at the early stage. This result implies that the promoting effect of DE on GLUE is constantly strengthening along with the development of DE.

As the marketization can impact the circulation efficiency of elements, the research sample is divided into high and low marketization cities by the median of the marketization index. The marketization index is from the Report of China’s Marketization Index at the provincial level. Columns (5) and (6) in Table 6 show the regression results. The coefficients in both groups are significant and positive, which indicate that DE’s promotion effect on GLUE is significant in different degrees of marketized cities.

5. Extended Analysis

5.1. Threshold Effect Analysis

Threshold model is a type of econometric model used to analyze the structural changes in variable relationships within different intervals which are divided by threshold values. It assumes that when the threshold variable crosses a specific threshold, the relationship among the model variables will change significantly. The following panel threshold regression model is set up to confirm Hypothesis 4, which states that DE has a nonlinear impact on GLUE:

$$\begin{gathered} {\mathrm{GLUE}}_{\mathrm{i},\mathrm{t}}={\phi }_0+{\phi }_1{\mathrm{DE}}_{\mathrm{i},\mathrm{t}}\times \mathrm{I}\left({\mathrm{Thv}}_{\mathrm{i},\mathrm{t}}<\theta \right)+{\phi }_2{\mathrm{DE}}_{\mathrm{i},\mathrm{t}}\times \mathrm{I}\left({\mathrm{Thv}}_{\mathrm{i},\mathrm{t}}\ge \theta \right) \\ +{\phi }_3{\mathrm{X}}_{\mathrm{i},\mathrm{t}}+{\mu }_{\mathrm{i}}+{\delta }_{\mathrm{t}}+{\epsilon }_{\mathrm{i},\mathrm{t}}\# \end{gathered}$$

(5)

In Equation (5), ${\mathrm{Thv}}_{\mathrm{i},\mathrm{t}}$ represents the threshold variable of city i in year t. Other variables are the same in Equation (1). Equation (5) illustrates the scenario of a singular threshold value, while the situation with multiple threshold values can be extrapolated based on the sample size. Based on the above analysis, this study identifies DE, green technology innovation, and industrial structure optimization as threshold variables for nonlinear effect analysis.

First of all, it is essential to determine the existence of threshold variables, as well as the number and the values of the variables. This study performs the threshold existence test as per Hansen (1999) [103]. The test results are reported in

Table 7. Results of the threshold existence test.

Threshold Variable	Threshold Number	Threshold Value	F Statistic	p
DE	Single	6.60	40.97	0.000
	Double	5.95	10.31	0.210
Pat	Single	5.38	50.97	0.000
	Double	8.49	48.05	0.000
	Triple	4.41	11.64	0.707
Str	Single	1.29	48.27	0.000
	Double	0.70	33.15	0.000
	Triple	0.36	18.53	0.590

. Based on the F statistic value and p value, DE successfully passes the single threshold test but fails to meet the criteria for the double threshold test. It indicates that there is a single threshold value for DE, which is 6.60. Similarly, green technological innovation only successfully passes the double-threshold test at the 1% significance level, which indicates there is a dual threshold for green technology innovation. Consistent with technological innovation, the industrial structure optimization also has two threshold values.

Then, Equation (5) is used to test the nonlinear effects of DE on GLUE. Column (1) in

Table 8. Results of panel threshold regression.

		Threshold Variables
Variable		(1)	(2)	(3)
		DE	Pat	Str
Threshold value	q1	6.60	5.38	0.70
	q2		8.49	1.29
DE * I (Thv < q1)		0.0186 *** (0.00344)	0.0165 *** (0.00341)	0.0125 *** (0.00352)
DE * I (q1 ≤ Thv < q2)		0.0273 *** (0.00409)	0.0219 *** (0.00344)	0.0170 *** (0.00341)
DE * I (Thv ≥ q2)			0.0277 *** (0.00355)	0.0223 *** (0.00341)
CVs		Y	Y	Y
City fixed effect		Y	Y	Y
Time fixed effect		Y	Y	Y
N		2356	2356	2356
adj. R-sq		0.163	0.184	0.183

Note: *** denotes statistical significance at the 1% level; and the values in parentheses are robust standard errors.

are the results of DE as a threshold variable. When DE passes the threshold value (6.60), the regression coefficient of DE (0.0273) is greater than that (0.0186) when not passing the threshold value. Both coefficients have passed significance level test. This implies that with the continuous development of DE, it has an increasing nonlinear promoting impact on GLUE. Column (2) in Table 8 are the results of green technology innovation as a threshold variable. When the value of green technological innovation is less than the first threshold value (5.38), between the first and second threshold values, or greater than the second threshold value (8.49), the influence coefficient of DE on GLUE keeps increasing, which are 0.0165, 0.0219, and 0.0277, respectively, and all significant at the 1% level. This implies that the impact of DE on GLUE shows a slightly increasing nonlinear effect as green technology innovation advances. Column (3) in Table 8 are the results of industrial structure optimization as a threshold variable. As the value of industrial structure optimization continues to increase, the regression coefficient of DE on GLUE also constantly increases. Above results show that the promoting impact of DE on GLUE exhibits an increasing nonlinear relationship with the strengthening of industrial structure optimization. Above all, the results verify Hypothesis 4.

5.2. Spatial Spillover Effect Analysis

This study employs the spatial model to test the potential spatial spillover effect. Given that land use is an ongoing process, the lag phase of GLUE is added into the model, which is specified as follows:

$$\begin{gathered} \begin{array}{c} {\mathrm{GLUE}}_{\mathrm{i},\mathrm{t}}={\vartheta }_0+{\rho }_1{\mathrm{W}\ast \mathrm{GLUE}}_{\mathrm{i},\mathrm{t}}+{\vartheta }_1{\mathrm{GLUE}}_{\mathrm{i},\mathrm{t}-1}+{\rho }_2\mathrm{W}\ast {\mathrm{DE}}_{\mathrm{i},\mathrm{t}}+{\vartheta }_2{\mathrm{DE}}_{\mathrm{i},\mathrm{t}}\end{array} \\ \begin{array}{c}+{\rho }_3\mathrm{W}\ast {\mathrm{X}}_{\mathrm{i},\mathrm{t}}+{\vartheta }_3{\mathrm{X}}_{\mathrm{i},\mathrm{t}}+{\mu }_{\mathrm{i}}+{\delta }_{\mathrm{t}}+{\epsilon }_{\mathrm{i},\mathrm{t}}\end{array} \end{gathered}$$

(6)

In Equation (6), ${\rho }_1$ represents the spatial autoregressive coefficient of GLUE, and W represents the spatial weight matrix. ${\rho }_2$ and ${\rho }_3$ represent the spatial spillover coefficients of DE and control variables, respectively. The remaining variables are the same in Equation (1).

This study examines the spatial relationship between DE and GLUE by calculating Moran’s I, utilizing a spatial geographic adjacency matrix in which W equals 1 if two cities share a boundary and 0 otherwise. In

Table 9. The Moran’s I of DE and GLUE.

Year	DE		GLUE
	Moran’s I	Z	Moran’s I	Z
2011	0.071 ***	13.103	0.039 ***	7.523
2012	0.059 ***	11.005	0.026 ***	5.222
2013	0.061 ***	11.286	0.029 ***	5.904
2014	0.049 ***	9.261	0.026 ***	5.271
2015	0.049 ***	9.303	0.02 ***	4.308
2016	0.05 ***	9.373	0.011 ***	2.592
2017	0.052 ***	9.778	0.022 ***	4.525
2018	0.052 ***	9.689	0.016 ***	3.434
2019	0.049 ***	9.31	0.03 ***	5.998

Note: *** denotes statistical significance at the 1% level.

, the results show that the Moran’s I of DE and GLUE both were significantly positive during 2011–2019, which indicate both DE and GLUE exhibit spatial autocorrelation phenomena.

A specific spatial econometric model should be chosen next. Based on the results of the LM test, LR test, and Hausman test, this study finally decides to use the fixed-effect dynamic spatial Dubin model. And because DE is a key economic indicator, the geographical spatial matrix may not accurately reflect the economic development gap among various cities. To conduct spatial econometric analysis, this study utilizes an inverse economic distance spatial weight matrix which is measured by the reciprocal of the absolute value of the difference in per capita GDP between the two cities during the study period.

Table 10. Results of the spatial Dubin model.

Variable	Main Spatial Effect	Short Term Spatial Effect			Long Term Spatial Effect
		Direct	Indirect	Total	Direct	Indirect	Total
	(1)	(2)	(3)	(4)	(5)	(6)	(7)
L.GLUE	0.711 ***
	(0.0216)
DE	0.00703 **	0.00717 **	0.0208 *	0.0279 **	0.0316 **	0.315	0.346
	(0.00308)	(0.00298)	(0.0108)	(0.0116)	(0.0124)	(0.442)	(0.448)
W * DE	0.0151 *
	(0.00853)
Spatial rho	0.208 ***
	(0.0300)
Variance sigma2_e	0.00135 ***
	(0.0000379)
CVs	Yes	Yes	Yes	Yes	Yes	Yes	Yes
City fixed effect	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	2016	2016	2016	2016	2016	2016	2016
R2	0.647	0.647	0.647	0.647	0.647	0.647	0.647

Note: ***, ** and * denote statistical significance at the 1%, 5%, and 10% level, respectively; and the values in parentheses are robust standard errors.

shows the results of the dynamic spatial Dubin model.

Column (1) in Table 10 shows the spatial regression coefficient of DE on GLUE is significantly positive when the spatial weight matrix is considered. This implies that DE has a spatial spillover effect on GLUE.

However, simply relying on the spatial interaction term coefficient of DE is not enough to accurately assess the spatial spillover effect of it [104]. Therefore, it requires additional verification through both direct and indirect effects. The direct, indirect, and total effects of DE on LUGE in short term effects are shown in Table 10 in columns (2) to (4), respectively. We can see that in the short term, the indirect spatial coefficients of DE on GLUE are significantly positive. Thus, it can be inferred that DE exhibits a spatial spillover effect on GLUE in the short term. The results above verify Hypothesis 5.

From columns (5) to (7) in Table 10, we can see the spatial spillover effect differs over the long term. The direct spatial coefficient of DE is 0.0316 and passes 5% of the significance level test, while the indirect spatial coefficient of DE fails to pass the significance level test. Furthermore, the total spatial coefficient is not significant. The long term spatial test results indicate in the long term, DE can significantly promote local GLUE, but its spatial spillover effect is not obvious.

6. Discussion

The low-carbon and environmentally friendly advantages of DE significantly improve resource allocation and utilization efficiency. Concurrently, green land use helps achieve a win–win goal for environmental protection and economic growth, which is essential for achieving green development. This study theoretically and empirically analyzes the promoting effect, nonlinear effect, and spatial spillover effect of DE on GLUE. The research results provide a basis for DE to enhance GLUE and broaden scholarly investigation within the digital economy field.

This study empirically demonstrates that the development of DE can enhance GLUE, with green technology innovation and industrial structure optimization as effect channels. Existing research has shown that urban GLUE is positively impacted by DE [58], and our research conclusion aligns with these findings. Unlike utilizing national big data comprehensive pilot zones as a quasi-natural experiment [57], this study measures DE by establishing a comprehensive evaluation index system; therefore, the concept is more precise and accurate. This study utilizes cities as the research sample, in contrast to the previous literature taking provinces as research samples [57], resulting in a larger sample size and more reliable conclusions.

DE has different impacts on the improvement of GLUE during different development stages. Research findings indicate that the role of DE in enhancing GLUE is constantly strengthening along with the development of DE. A possible reason for this is that DE requires large-scale investment but gains relatively low returns in the early stage of development. In the mature stage, the marginal cost of DE keeps decreasing while the marginal benefit keeps increasing, thus its promoting effect on LUGE is constantly strengthening.

DE has a spatial spillover effect on GLUE. Relevant studies have also confirmed that it has a spatial spillover effect on GTFP [105] and carbon emission reduction [20,26]. Nonetheless, some differences exist between the results of this study and existing research, which is that DE has significant spatial spillover effect on GLUE in the short term, but not in the long term. The growth of network and broadband infrastructure plays a prominent role in DE development. Therefore, the development of local DE will, in the short term, drive the construction of networks and other infrastructure in the nearby region, which will further promote DE development in that region. When the infrastructure construction is completed, the driving effect will gradually become weakened, so the spillover effect of DE on GLUE in nearby regions is not obvious in the long term. Based on technological catch-up [106] and government competition theories [53], if one region is ahead of its neighbors in digital technology and the Internet, its neighbors will find ways to imitate. In the short term, this behavior can stimulate the development of DE in nearby regions, to further improve the GLUE. However, in the long run, either the nearby regions have caught up and developed their digital economy at a similar level, or the gap continues to widen, leaving little incentive to stimulate and catch up in nearby regions. The above two conditions will lead to DE’s insignificant spatial spillover effect on GLUE.

The research on the impact of DE on GLUE is conducted in the context of China, and the empirical findings are derived from Chinese data; thus, the conclusions should be applied to other countries with caution. Simultaneously, China’s DE is developing rapidly, significantly outpacing that of other countries. The unique land use management system in China markedly differs from other governance models. Consequently, limited extant literature analyzes the impact of DE on land use efficiency by considering countries beyond China as an example.

7. Conclusions and Recommendations

7.1. Research Conclusion

The main conclusions of this study are as follows: (1) DE can significantly enhance GLUE. This conclusion remains robust after considering endogeneity problems and conducting a serious of robustness tests. An examination of the influence mechanism reveals that DE enhances GLUE by driving green technological innovation and industrial structure optimization. (2) The promoting effect of DE on GLUE is constantly strengthening along with DE development. (3) As the threshold for DE, green technology innovation, and industrial structures optimization rise, DE has a nonlinear promoting effect on GLUE. (4) DE has a spatial spillover effect on GLUE, and this effect is only significant in the short term.

7.2. Policy Implications

Firstly, DE makes a significant contribution to low-carbon environmental protection and sustainable land utilization. As it develops, its beneficial impact on GLUE shows marginally increasing characteristics, and its promoting effect on green and sustainable development becomes increasingly evident. Consequently, governments should actively support the development of the Internet and its related industries to facilitate the improvement of DE. They should also take advantage of the role of big data in land use supervision [5], combined with market and government control, to effectively improve the land use efficiency [107].

Secondly, governments should formulate and implement relevant policies in accordance with the actual conditions of their localities [5]. Considering the local disparities in the effects of DE on GLUE, local governments should progressively adopt digital economy development policies that align with local resource advantages. The construction of Internet infrastructure can be further strengthened and a new industrial development model integrated with local industry and digital technology can be formed. As a result, DE can fully play its driving role in promoting sustainable and green development.

Thirdly, since green technology innovation is crucial for sustainable land use, governments should incentivize enterprises to engage in green technology innovation activity. Enterprises can adjust their production structure and mode by adopting green technology, and further deal with the negative externalities of environmental pollution resulting from production progress. To be specific, the governments can utilize a combination of various environmental regulatory tools, such as pollution charges and environmental subsidies [71], to fully stimulate enterprises’ innovation enthusiasm. Such actions can accelerate the formation of high-tech development zones, then improve the land use efficiency [108]. Polluting industries and enterprises should be guided to improve resource utilization efficiency through industrial structure optimization and transformation, to shift from traditional production to a green and clean production mode.

Fourthly, the government should adopt rational land space allocation to relieve the pressure caused by the irrational use of land on the environment [109,110]. Cities with high levels of digital economy and GLUE are located near one another. These cities should continue to leverage their agglomeration advantages to develop towards higher goals of sustainable land use. For those cities with low GLUE, it is necessary to learn from the experience of green land use in neighboring areas, and to eliminate backward industries and strengthen the regulation of high-pollution land usage, which can guide the industry towards green development and ultimately improve GLUE.

7.3. Limitations and Future Research Directions

This study conducted an analysis on the promoting effect of DE on GLUE. However, this study may still have some limitations. First, when constructing the evaluation indicator system for DE and GLUE, there may be some missing indicators or multi-collinearity among indicators. Future research should improve the evaluation indicator system by integrating a literature review with field investigations and expert consultations. Second, the land in this study covers all types of land within the jurisdiction and is not classified according to different types of land utilization. As different land use types accommodate various industrial activities, DE may have different effects on different types of land utilization, such as industrial land [21], agricultural land [111], and construction land [61]. Future research should prioritize examining the impact of DE on GLUE of different industries, with findings facilitating comparative analysis of its heterogeneous impact on GLUE across various industries.