How Does Land Finance Influence Vegetation Dynamics in China?

Abstract

Land-based financing plays an essential role in urbanization in the developing world, and it is widely recognized to have profound environmental effects. However, there have been relatively few research endeavors on the impact of land finance on vegetation dynamics. This study applies fixed effects models and an instrumental variable approach to examine the impact of land finance on vegetation status and mechanisms of influence, using data for 286 Chinese cities between 2011 and 2022. The nonlinear relationship between land finance and vegetation conditions at different levels of economic development is investigated by estimating panel threshold models. The findings show that land finance exerts an inhibiting impact on vegetation conditions. The restraining effect of land finance on vegetation status tends to be more pronounced in western China or in secondary industry-led cities. The analysis of mechanisms of influence indicates that land finance negatively affects vegetation conditions by speeding up urban expansion, suppressing innovation, reducing land use efficiency, and distorting the fiscal expenditure structure. The analysis of the threshold effect suggests that land finance exerts a stronger curbing effect on vegetation status as the economic development level rises. The findings have significant policy implications for deepening reform of the fiscal system and promoting vegetation protection and restoration.

1. Introduction

Vegetation has a series of ecological functions, such as preventing soil erosion, acting as carbon sinks, protecting biodiversity, and purifying the air [1,2]. However, deforestation and vegetation destruction have posed severe threats to global sustainable development in recent decades. As reported by the United Nations, global forest coverage declined from 31.9% in 2000 to 31.2% in 2020 [3]. As the world’s largest developing country, China has endeavored to promote vegetation protection and restoration and exhibits the world’s largest area of planted forests (0.088 billion hectares) and grasslands (2.637 billion hectares) in 2024. However, rapid industrialization and urbanization puts forward serious challenges to vegetation conservation in China.

Land-based financing exists widely (although it takes different forms) in the developing world where the demand for infrastructure is substantial and the government funding is insufficient [4,5]. In Hyderabad, India, betterment charges are levied to finance on-site and off-site amenities, and an impact fee is imposed to fund city capital improvements [4]. In Bengaluru, India, cess on new layouts/developments and cess on additional floor area ratios are levied to finance the mass rapid transit system [4]. In Ho Chi Minh City, Vietnam, the Land-for-Infrastructure mechanism, which is devised under the Build-Transfer model of public-private partnership, is applied to fund road construction [5]. China’s land finance system refers to a local public finance system in which the revenue from the conveyance of Land Use Rights (LURs) constitutes a substantial portion of local government revenue [6]. Even during a real estate market downturn, the land sales revenue reached CNY 5540.7 billion in 2023, with the proportion of land sales revenue to budget revenue being 0.495:1. Land finance is of vital importance in China’s economic life and urbanization and is widely recognized to have profound environmental effects through influencing land use patterns and the behaviors of governments, enterprises, and individuals [7].

Since a substantial proportion of land development occurs in peri-urban areas, it is evident that land finance materially drives rural-urban land conversion and results in a decrease in vegetation coverage [8]. However, in addition to speeding up urban expansion, researchers tend to neglect that land finance can impact vegetation dynamics by influencing the innovative behaviors of governments, enterprises and individuals, and resource utilization efficiency. Technological innovation contributes significantly to improving resource utilization efficiency and preventing pollution [9,10], thus playing a vital role in encouraging vegetation preservation. Under the land finance system, local governments are highly motivated to promote short-term economic growth and tend to bias fiscal spending toward infrastructure development rather than innovation investment [11,12]. In addition, high housing prices induced by the land finance system hinder high-skilled immigration and crowd out enterprises’ innovation investment, hence having an inhibiting impact on innovation activities [13,14].

In this research, we explore the nexus between land finance and vegetation dynamics by using panel data from 286 Chinese cities between 2011 and 2022. We use the normalized difference vegetation index (NDVI) to measure the level of vegetation cover and apply fixed effects models and an instrumental variable approach to examine the impact of land finance on vegetation changes and mechanisms of influence. We also use panel threshold models to explore the nonlinear impact of land finance on vegetation dynamics. The potential contributions of this research are as follows. First, this research adds to an emerging stream of literature on the institutional factors affecting vegetation dynamics. Previous research primarily concentrates on the effects of environmental policies on vegetation changes [1,15], and little attention has been devoted to the effects of the local public finance system on vegetation dynamics. Since land-based financing exists widely in the developing world [4,5], China’s experience is also relevant for policymakers in other developing countries. Second, this study is conducive to a more complete and in-depth understanding of the mechanisms through which land finance influences vegetation dynamics. As analyzed earlier, it is manifest that land finance can bring about urban expansion and lead to a reduction in vegetation coverage. However, researchers tend to neglect that land finance can have a negative effect on resource utilization efficiency and vegetation conditions by suppressing innovation. We delve into how land finance influences the innovative behaviors of governments, enterprises, and individuals, and investigate the innovation-inhibiting mechanism through which land finance adversely affects vegetation status, thereby enriching the theoretical system of the determinants of vegetation dynamics and related empirical evidence. Third, this research explores the nonlinear impact of land finance on vegetation changes, thereby providing a more scientific basis for promoting vegetation protection and restoration.

2. Literature Review

2.1. Determinants of Vegetation Dynamics

Vegetation dynamics is influenced primarily by climatic and environmental factors, socio-economic factors, and institutional factors. Climatic and environmental factors affecting vegetation dynamics include precipitation, temperature, soil moisture status, solar radiation, carbon emissions, extreme climatic events, environmental pollution, etc. [16]. Spatial patterns of temperature and precipitation play an important role in determining vegetation types (e.g., forests, grasslands, deserts), and their inter-annual fluctuations affect ecosystem productivity. The effect of soil moisture status on vegetation exhibits significant variations across regions [17]. The effects of elevated carbon dioxide concentrations on vegetation include both positive (e.g., enhancing photosynthesis and increasing water use efficiency) and negative effects (e.g., inducing nutrient dilution and inducing unbalanced interspecies competition) [18]. Extreme climatic events such as droughts, heat waves and floods can destabilize ecosystems, alter species composition, and reduce the carbon sink function of vegetation [19]. Pollutants causing damage to forests and crops mainly include sulfur dioxide, nitrogen oxides, ozone, etc., and Asia has been identified as the region where severe damage to vegetation brought about by air pollution is most likely to occur [20].

Socio-economic factors influencing vegetation changes include urbanization, population growth, economic growth, etc. The direct effect of urbanization on vegetation tends to be negative and refers to the conversion of vegetated land surfaces to impervious ones. The indirect effects of urbanization on vegetation include both positive (e.g., air, water, and soil pollution and habitat fragmentation) and negative effects (e.g., urban green projects and longer growing seasons in urban environments) [21]. In poor areas that are highly dependent on natural resources, overgrazing or fuelwood collection causes vegetation degradation [22]. Brandt et al. (2017) find that rapid population growth gives rise to deforestation of the humid areas in sub-Saharan Africa due to rising demand for arable land and timber [23].

Institutional factors affecting vegetation dynamics include land use and ecological conservation policies, nature reserve systems, international agreements, and transnational cooperation, etc. With respect to land use and ecological conservation policies, the implementation of the Returning Farmland to Forest Program brings about an increase in forest cover in China, but the lack of variety in tree species under the program can pose a threat to biodiversity [24]. The enforcement of the Brazilian Forest Code contributes significantly to the decrease in deforestation in the Brazilian Amazon [25]. The establishment of nature reserves plays a crucial part in preventing vegetation loss within their boundaries [1,26]. With respect to international agreements and transnational cooperation, under the Reducing Emissions from Deforestation and Forest Degradation initiative, Norway and Brazil signed an agreement on funding for decreasing emissions from deforestation of USD 1 billion [27]. According to the “Aichi Targets” set by the parties to the Convention on Biological Diversity, governments committed to preserving more than 17% of terrestrial environments globally through protected area systems or other protection measures [28].

2.2. Unfavorable Effects of Land Finance in Economic, Environmental and Social Aspects

As the costs and risks of land finance system become increasingly evident, many studies have investigated the unfavorable effects of land finance in economic, environmental, and social aspects. It is noteworthy that the economic, environmental, and social effects of land finance tend to interact with each other. In terms of economic effects, existing studies have demonstrated that land finance gives rise to land resource mismatch [7,29], has an unfavorable effect on productivity [30], pushes up housing prices and triggers a housing affordability problem [31], increases local debt risks [32], distorts the structure of local governments’ fiscal expenditures [11,33], and exacerbates macroeconomic fluctuations [7]. Zhang, Geltner, and de Neufville (2018) point out that land supply constraints under the land finance system amplify the price effects of other market features such as speculative demand [31]. Hu and Qian (2017) find that local governments exhibiting stronger dependence on land finance devote a lower proportion of land to affordable housing construction [34]. Guo, Liu, and Zhao (2015) find that local government dependence on highly procyclical land sales revenue to finance public spending can considerably magnify business cycle fluctuations [35].

With respect to environmental effects, existing studies have suggested that land finance increases carbon emissions and exacerbates pollution. Wang, Wu, and Hao (2020) point out that land finance increases local governments’ fiscal capacity and promotes fiscal decentralization, thus enhancing the ability of local governments to determine the direction of their fiscal expenditures [36]. For the purpose of fostering economic growth, local governments tend to bias public spending towards infrastructure construction, relatively ignore environmental protection expenditures, and relax environmental regulations. Applying dynamic spatial Durbin models, Yang et al. (2023) find that growth in local governments’ land sale revenue leads to an increase in PM_2.5 concentrations in China [37]. Compared with land conveyance through tender, auction, and listing, land conveyance through negotiation exerts a greater impact on PM_2.5 concentrations. Based on evidence from Chinese cities, Cao et al. (2023) indicate that land finance not only positively influences local haze pollution, but also exerts an enhancing impact on haze pollution of neighboring cities [38]. The positive effect of land finance on haze is greater in cities exhibiting higher levels of local government competition.

In terms of social effects, previous studies have shown that land finance harms farmers’ welfare and induces social instability [39,40], widens urban-rural income inequality [7], produces rent-seeking opportunities, and gives rise to corruption [41]. Ong (2014) argues that landless farmers who experience land expropriation can undergo substantial welfare loss if they obtain insufficient compensation and are not given access to the full range of social welfare [40]. Wang et al. (2019) indicate that inadequate compensation for land expropriation substantially increases rural-urban income disparity [42].

According to the above analysis, previous research on institutional factors influencing vegetation dynamics primarily concentrates on the effect of environmental policies, and there is little exploration of the effect of the local public finance system on vegetation changes. Studies concerning the environmental effects of land finance mainly concentrate on carbon emissions and environmental pollution, and limited research endeavors are dedicated to examining the effect of land finance on vegetation dynamics and the mechanisms of influence. Thus, this research contributes to the enrichment of both the theoretical system of the determinants of vegetation dynamics and the environmental impacts of land finance.

3. Theoretical Analysis and Hypotheses

Under the land finance system, prefectural-level and county-level local governments first expropriate collectively owned rural land or repurchase the allocated LURs from existing urban land users, conduct preliminary land development (mainly including site clearance and infrastructure construction), and then convey LURs of developable urban land via tender, auction, listing, or negotiation [43]. Revenue from the conveyance of the LURs constitutes a significant proportion of local governments’ revenue. Government land sales directly induce conversion from non-construction land to built-up land and give rise to a reduction in the area of vegetation cover [44]. Activities such as housing and infrastructure development, which are closely linked to government land sales, consume primary products such as timber and crops and result in vegetation depletion. High housing prices induced by the land finance system impede the geographical concentration of innovative labor and crowds out enterprises’ innovation investment [13,14], thereby inhibiting innovation activities and adversely affecting resource utilization efficiency. Land finance can adversely influence land use efficiency by causing disorderly urban expansion, hindering urban renewal, and giving rise to insufficient industrial land use. In addition, land finance can distort the structure of local governments’ fiscal expenditures [11,33]. More specifically, land finance can make local governments bias fiscal expenditures in favor of infrastructure construction, and relatively neglect innovation investment [12]. Thus, land finance can exert a restraining impact on vegetation status through accelerating urban expansion, curbing innovation, decreasing land use efficiency, and distorting the fiscal expenditure structure. It is noteworthy that land finance is viewed as closely associated with the factor-driven development mode characterized by an incremental expansion-oriented approach to land development and relatively low resource utilization efficiency. As the economic development level rises, the curbing impact of land finance on innovation becomes more prominent, leading to stronger adverse environmental effects. Thus, land finance can exert a nonlinear effect on vegetation conditions under different stages of economic development. Figure 1 illustrates the mechanisms through which land finance impacts vegetation conditions.

3.1. Land Finance and Urban Expansion

Under the current fiscal system and performance assessment system for local government officials in China, local governments exhibit strong incentives to acquire more land sales revenue and promote economic growth by stimulating urban expansion. In terms of land use, the essence of urban expansion entails the process of converting non-construction land, such as arable land, forest land, garden land, and grassland, to built-up land, such as industrial, residential, and commercial land. Rural-urban land conversion brings about a decrease in the area covered by vegetation and causes damage to the ecological environment. The higher the degree of local governments’ reliance on land finance, the stronger the motivation for local governments to generate land sales revenue and drive economic growth through urban expansion. On the one hand, after the 1994 tax-sharing reform in China, the proportion of local fiscal revenue has decreased substantially, but the expenditure responsibilities of local governments (infrastructure development, provision of public services, and social security, etc.) have not declined accordingly and have even increased. Facing a mismatch between fiscal capacity and spending responsibilities, local governments, while striving to obtain various fiscal transfers from the central government, have exploited new sources of revenue, such as land conveyance revenue [32,45]. While the compensation for demolished housing and resettlement and transaction costs for urban land redevelopment are relatively high, the land acquisition costs and transaction costs entailed in rural land expropriation and greenfield development are comparatively low. Thus, local governments tend to favor greenfield development rather than brownfield redevelopment and are keen on fostering the development of new urban areas, such as economic development zones and industrial parks at various administrative levels [46]. It is noteworthy that local governments tend to adopt different strategies for the conveyance of LURs for different types of land, i.e., strategies of ‘conveying LURs of industrial land at low prices and LURs of commercial and residential land at high prices’ [47]. Manufactured products are often sold in national or even global markets, and manufacturing enterprises are cost-sensitive and highly mobile. Hence, confronted with fierce regional competition, local governments need to attract manufacturing investment by providing low-price industrial land. Unlike manufacturing industries, many producer and consumer service industries are characterized by synchronized production and consumption [48], and service enterprises mainly provide non-tradable services to local consumers. Thus, the commercial and residential land market tends to exhibit characteristics of a localized seller’s market, and local governments are able to raise the prices of commercial and residential land by intentionally controlling land supply [49]. On the other hand, under the performance assessment system for local government officials, which employs GDP growth rates as an essential evaluation indicator, local governments have strong incentives to foster local economic growth [50]. Since infrastructure development can not only markedly boost investment growth in the short term but also facilitate economic growth in the medium to long term by increasing the marginal productivity of factors of production and attracting capital inflows, local governments have a high willingness to raise funds for infrastructure development through the conveyance of LURs [7]. The improvement of productive infrastructure and living amenities will produce a significant capitalization effect, raising land prices and increasing land sales revenue. Thus, a cyclical pattern of ‘generating wealth with land sales and raising land prices with wealth’ can be formed, reinforcing local governments’ dependence upon land finance. In addition, growth in land sales revenue helps to relieve financial constraints faced by local governments and enhance their ability to provide tax subsidies to new and existing enterprises [51], thereby increasing attractiveness to investors and promoting local economic growth.

3.2. Land Finance and Innovation

Innovation plays an essential part in fostering green transition and is vital to facilitating the improvement of the ecological environment and vegetation protection [52]. Innovations in production technologies help to enhance resource utilization efficiency and drive resource recycling. Digital innovations such as artificial intelligence, big data, cloud computing, and the industrial Internet can optimize resource allocation and production processes in areas such as industrial and agricultural production and building construction, enhancing the levels of intelligence, personalization, and greening at various stages of production, trading, consumption, and recycling. Digital innovations can also heighten the government’s capacity to monitor and manage the ecological environment in real-time and with precision. Hence, both innovations in production technologies and digital innovations are conducive to reducing overconsumption of primary products, such as timber and crops, and thus have a favorable impact on the preservation and restoration of vegetation. Technological innovations in new energy, energy-saving, and pollution prevention help to decrease the harmful effects of environmental pollution on plant growth. However, high housing prices brought about by the land finance system hamper high-skilled immigration and crowd out enterprises’ innovation investment, thereby having a disadvantageous effect on innovation activities.

From the perspective of innovative labor, positive agglomeration externalities such as knowledge spillovers and technological diffusion generated by the geographical concentration of high-skilled workers are a crucial source of the enhancement of urban innovative capacities. Under the land finance system, high housing prices brought about by local governments’ strategies of ‘conveying the LURs of commercial and residential land at high prices’ through intentionally controlling land supply augment housing costs and reduce the relative utility of labor, which in turn hinders high-skilled immigration and the geographical concentration of innovative labor [14]. From the perspective of enterprises, land finance pushes up housing prices and profit margins in the real estate industry and attracts excessive flows of capital into the real estate sector, thus crowding out enterprises’ innovation investment [53]. The issuance of special bonds by local governments using land sales revenue as sources of debt-servicing funds and the use of LURs as collateral for loans have taken up limited credit resources, making it more difficult for enterprises to finance innovation activities [12]. In addition, at the current stage, there is still widespread government intervention in the conveyance of LURs (especially in land conveyance through listing) [29]. Government intervention in land markets can distort market competition mechanisms and enterprises’ incentives, inducing enterprises to invest valuable resources in rent-seeking activities and reducing enterprises’ innovation efforts.

3.3. Land Finance and Land Use Efficiency

Land finance can exert unfavorable effects on land use efficiency by inducing disorderly urban expansion, hampering urban renewal, and bringing about insufficient industrial land use. Lower land use efficiency generally results in greater damage to the ecological environment. First, in order to obtain more land sales revenue, local governments are eager to promote the development of new urban areas, often resulting in excessive and disorderly urban expansion. While the land area of economic development zones and industrial parks at various administrative levels increases considerably in the short run, upstream and downstream industrial chain support, the skilled labor pool, and public services, including education, health care, etc., tend to be neglected, leading to the emergence of a large amount of idle or underutilized land and relatively low land use efficiency in new urban areas. Second, under the land finance system, local governments prefer greenfield development to brownfield redevelopment, hindering the improvement of the utilization efficiency of existing urban land. From the perspective of cost recovery and investment cycle, by expropriating rural land, converting it to construction land, and conveying LURs, local governments can obtain a considerable amount of one-time land conveyance revenue, and the process is relatively short (usually completed in 1–2 years). By comparison, old district renovation and the reuse of underutilized industrial land generally entails demolition and resettlement, negotiation of property rights, pollutant treatment, etc., and thus its investment cycle can be as long as 5–10 years. From the perspective of the degree of difficulty of land development, local governments mainly negotiate with rural collectives in rural land expropriation, and the transaction cost is comparatively low. By contrast, urban land redevelopment (e.g., redevelopment of urban villages and aging industrial zones) generally involves multiple holders of property rights (e.g., rural collectives, enterprises, and public institutions) and the transaction cost can be noticeably high. Third, low-price industrial land supplied by local governments weakens the incentives for enterprises to utilize land intensively and undertake technological upgrading, giving rise to relatively low industrial land use efficiency.

3.4. Land Finance and Fiscal Expenditure Structure

Government expenditures on science and technology promote technology development and application in areas such as vegetation monitoring and early warning techniques, ecological restoration techniques, and data-driven techniques for precise vegetation management, thereby playing a crucial role in facilitating vegetation protection and restoration. Local governments’ reliance on land finance distorts the local fiscal expenditure structure and weakens governments’ support for science and technology. More specifically, for the purpose of acquiring more land sales revenue and facilitating short-term economic growth, local governments tend to bias fiscal expenditures toward infrastructure development that can significantly increase land prices in the short term and produce obvious investment-pulling and investment-attracting effects [36], while relatively ignoring investment in science and technology, which exhibits a comparatively long investment cycle [12]. According to the above analysis, the following research hypotheses are proposed.

Hypothesis 1:

Land finance exerts a negative effect on vegetation status.

Hypothesis 2:

Land finance exerts a depressing effect on vegetation status through accelerating urban expansion, curbing innovation, reducing land use efficiency, and distorting the fiscal expenditure structure.

3.5. Nonlinear Effect at Different Levels of Economic Development

Land finance is rooted in an incremental expansion-oriented approach to land development, which conflicts with the characteristics of advanced stages of economic development, such as efficient and intensive utilization of resources and sustainable improvement of the ecological environment, to some extent. Thus, it is reasonable to expect that the effect of land finance on vegetation dynamics can vary across different stages of economic development, exhibiting non-linear characteristics. First, at advanced stages of economic development, the mode of economic growth shifts from factor-driven to efficiency- and innovation-driven more rapidly, and the role of innovation in enhancing resource utilization efficiency and reducing ecological and environmental damage becomes more prominent. However, under the land finance system, local governments tend to favor projects with a short investment cycle, such as infrastructure construction, which significantly raises land prices in the short run and generates an obvious investment-pulling effect, rather than projects that are highly risky with a relatively long cost recovery cycle, such as innovation activities. In addition, high housing prices induced by the land finance system hamper the geographical concentration of innovative labor, significantly weakening the momentum of the green transition at advanced stages of economic development. Second, land prices, especially residential and commercial land prices, tend to increase significantly as the economic development level rises, thus enhancing incentives for local governments to drive rural-urban land conversion and urban expansion, and bringing about more severe damage to the ecological environment. Third, many cities already possess relatively well-developed infrastructure at advanced stages of economic development, and the marginal return on large-scale infrastructure investment supported by the land finance system can decline substantially. Excessive infrastructure investment not only fails to notably increase productivity but also causes excessive consumption of resources and generates unfavorable environmental effects. Based on the above analysis, the following research hypothesis is proposed.

Hypothesis 3:

As the economic development level rises, land finance exerts a higher adverse impact on vegetation status.

4. Methodology and Data

4.1. Model Specification

Based on the theoretical analysis presented in the previous section and the econometric analysis framework proposed by existing studies [15,54], a baseline model is developed to investigate the effect of land finance on vegetation dynamics:

$$Ve{g}_{it}={\beta }_0+{\beta }_{1}L{f}_{it}+{\beta }_2{X}_{it}+u_i+v_t+{\epsilon }_{it}$$

(1)

where Veg represents the level of vegetation cover, Lf represents the extent of local governments’ reliance on land finance, and X denotes a vector of control variables. Subscripts i and t represent city and year, respectively. ${\epsilon }_{it}$ is the error term. All variables are logarithmically transformed to reduce potential errors caused by heteroscedasticity and skewness.

As analyzed in Section 3, land finance can have a dampening impact on vegetation conditions through speeding up urban expansion, suppressing innovation, reducing land use efficiency, and distorting the fiscal expenditure structure. Referring to existing studies [13,36], the following models are developed to investigate the mechanisms through which land finance influences vegetation dynamics:

$$M_{it}={\alpha }_0+{\alpha }_{1}L{f}_{it}+{\alpha }_2{X}_{it}+u_i+v_t+{\epsilon }_{it}$$

(2)

$$Ve{g}_{it}={\gamma }_0+{\gamma }_{1}L{f}_{it}+{\gamma }_2{M}_{it}+{\gamma }_3{X}_{it}+u_i+v_t+{\epsilon }_{it}$$

(3)

where M represents the mechanism variables. Model (2) is employed to examine the effects of land finance on mechanism variables, and model (3) investigates the effects of mechanism variables on vegetation changes.

The panel threshold model proposed by Hansen (1999) is employed to investigate the nonlinear association between land finance and vegetation status [55]:

$$Ve{g}_{it}={\beta }_0+{\beta }_{1}L{f}_{it}I(q_{it}\le \gamma )+{\beta }_{2}L{f}_{it}I(q_{it}>\gamma )+{\beta }_3{X}_{it}+u_i+v_t+{\epsilon }_{it}$$

(4)

where q denotes the threshold variable, $\gamma$ denotes the threshold value to be estimated, and $I(•)$ represents the indicator function.

4.2. Methods for Robustness Tests

To verify the reliability of empirical findings, the following robustness tests are carried out by replacing variables and samples. First, following the approach proposed by Chen and Liang (2023), we sum up the average growing-season NDVI of all pixels within the administrative area of a city to obtain the total NDVI and employ it as the explained variable [15]. Second, net ecosystem productivity (NEP), which is defined as net primary productivity of vegetation minus carbon emissions from soil microbial respiration and is a measure of carbon sequestration capacity [56], is used as the explained variable. The NEP data are sourced from the Loess Plateau SubCenter, National Earth System Science Data Center, National Science & Technology Infrastructure of China (http://loess.geodata.cn). Third, considering the differences in administrative levels between the municipalities directly under the central government and prefecture-level cities, the former are removed from the sample. Fourth, to eliminate the estimation errors induced by extreme values, all variables are trimmed at the 1st and 99th percentiles.

Since the climatic, environmental, and economic factors affecting vegetation changes are spatially linked, there tends to be a strong spatial correlation in the vegetation conditions of neighboring areas. First, neighboring cities are usually located in the same climate zone and may have similar climatic factors, such as precipitation and temperature. Second, the natural elements that shape the ecological substrate, such as soil, topography, and water resources, are spatially continuous. Third, interregional socio-economic interaction, such as labor movement, trade, and regional integration policies, causes the homogenization of human interference that plays a crucial role in affecting vegetation conditions. Thus, the following spatial autoregressive (SAR) model is developed to take into account the spatial effects:

$$Ve{g}_{it}={\beta }_0+\rho {\displaystyle \sum _{j=1}^n{W}_{ij}Ve{g}_{it}+}{\beta }_{1}L{f}_{it}+{\beta }_2{X}_{it}+u_i+v_t+{\delta }_{it}$$

(5)

where W denotes the n × n spatial weight matrix, and $W_{ij}$ represents the spatial weight between cities i and j. Since spatial linkages between cities are affected by both geographical distance and gaps in economic development, the economic geography weight matrix is employed in this study. Elements in the spatial weight matrix are specified as follows:

$$W_{ij}=1/d_{ij}\times 1/\left|pcGD{P}_i-pcGD{P}_j\right|$$

(6)

where $d_{ij}$ refers to the geographical distance between cities i and j. $pcGD{P}_i$ and $pcGD{P}_j$ refer to the average per capita GDP between 2011 and 2022 for cities i and j, respectively.

4.3. Variable Measurement

4.3.1. The Explained Variable

Average growing-season NDVI is employed to measure the level of vegetation cover. In accordance with existing literature, the growing season in China is from April to October [16,57]. NDVI is calculated based on plants’ visible-light absorption and near-infrared reflection characteristics and is an extensively used indicator of the level of vegetation cover [21,54]. More specifically, this research employs PKU GIMMS NDVI, which has a spatial resolution of 1/12° [58]. The red band (visible band) and near-infrared band are used in the NDVI calculation. PKU GIMMS NDVI has a longer time series than other NDVI datasets and effectively removes apparent orbital drift and sensor degradation impacts.

4.3.2. The Core Explanatory Variable

From a broad perspective, local governments’ land-related revenue includes both land sales revenue and land-related tax revenue. Since land sales revenue constitutes a predominant proportion of land-related revenue and the data for land sales revenue are readily available, researchers usually use land sales revenue to measure land finance [13,30]. This research employs the proportion of land sales revenue to budget revenue to measure the extent of local governments’ reliance on land finance (Lf). Land sales revenue is calculated using land transaction data collected from the China Land Market Website (https://www.landchina.com/).

4.3.3. The Mechanism Variables and Threshold Variable

In accordance with the theoretical analysis presented in Section 3, the following mechanism variables are selected. The first mechanism variable is the extent of urban expansion (Urexp), which is measured by the proportion of built-up area to the total area of the city [59]. The second mechanism variable is the level of innovation output (Innov), which is measured by the quantity of patents granted per capita. The third mechanism variable is land use efficiency (Lue), which is measured by the value added of secondary and tertiary industries per unit land area [60]. The fourth mechanism variable is the share of government expenditure on science and technology (St). Economic development level (Ed), measured by per capita GDP, is employed as the threshold variable.

4.3.4. Control Variables

Following existing studies [15,17,54], precipitation (Rain), annual average temperature (Tem), industrial SO₂ emissions (SO), economic development level (Ed), population density (Pop), intensity of government expenditure (Gov), and financial development level (Fin) are introduced as control variables. Precipitation and annual average temperature are measured based on the Global Surface Summary of the Day (GSOD) dataset provided by the National Oceanic and Atmospheric Administration (NOAA) of the United States. More specifically, the gridded data on precipitation and temperature with a resolution of 0.1° are generated through spatial interpolation and converted into raster data, which is then aggregated and averaged to obtain the annual data for each city. We use per capita GDP to measure Ed, the ratio of budget expenditure to GDP to measure Gov, and the proportion of loan balance to GDP to measure Fin.

4.4. Study Areas and Data Sources

This research uses panel data from 286 Chinese cities (whole city-region) between 2011 and 2022. Socioeconomic data are mainly collected from the China City Statistical Yearbook series and the China Urban Construction Yearbook series. Descriptive statistics of the variables are presented in

Table 1. Descriptive statistics of the variables.

Variable	Variable	Mean	S. D.	Min	Max
Veg	Average growing-season NDVI	0.623	0.126	0.157	0.820
Lf	Land sales revenue/local public budget revenue	0.682	0.488	0.009	6.081
Urexp	Built-up area/the total area of the city	0.020	0.043	0.000	0.486
Innov	Quantity of patents granted per capita	11.190	16.172	0.063	157.891
Lue	Land use efficiency (ten thousand yuan/km2)	56,572.07	95,432.21	2243	139,782
St	Share of public expenditure on science and technology	0.0173	0.0178	0.0005	0.2068
Rain	Precipitation (mm)	1155.611	562.511	54	3234
Tem	Annual average temperature (°C)	14.843	5.153	0.000	25.877
SO	Industrial SO2 emissions (t)	31,732.900	42,933.070	0.470	531,340
Ed	Per capita GDP (ten thousand yuan)	5.832	3.546	0.697	28.219
Pop	Population/the total area of the city(ten thousand persons/km2)	0.049	0.065	0.001	0.885
Gov	Local public budget expenditure/GDP	0.204	0.105	0.043	1.129
Fin	Loan balance of financial institutions/GDP	1.075	0.632	0.132	6.707

.

5. Results and Analysis

5.1. Measurement Results of Land Finance and Vegetation Status

The extent of local governments’ dependence on land finance (Lf) and vegetation status (Veg) in China between 2011 and 2022 is demonstrated in Figure 2 and Figure 3. The national average of extent of local governments’ dependence on land finance fluctuated during the sample period, declining from 0.770 in 2011 to 0.429 in 2016, increasing between 2016 and 2020, and then falling to 0.619 in 2022. Cities with a high extent of local governments’ dependence on land finance are mainly located in coastal areas and along the Yangtze River. The national average of growing-season NDVI shows a continuous upward trend, rising from 0.603 in 2011 to 0.626 in 2016 and further increasing to 0.629 in 2022. Growing-season NDVI is notably higher in South China than in North China. Figure 4 shows the kernel density estimation for Lf and Veg. While the regional disparities in Lf decrease during the sample period, the regional disparities in Veg stays relatively stable.

5.2. Baseline Estimation Results

Since the Hausman test rejects the random effects assumption (Chisq = 439.320, p value = 0.000), the fixed effects models are employed, and the estimation results are reported in

Table 2. The results of baseline regression.

Variables	(1)	(2)	(3)	(4)
Lf	−0.003 ***	−0.005 ***	−0.003 ***	−0.004 ***
	(0.001)	(0.001)	(0.001)	(0.001)
Rain			0.033 ***	0.032 ***
			(0.002)	(0.002)
Tem			−0.001	−0.001
			(0.003)	(0.003)
SO			−0.005 ***	0.000
			(0.001)	(0.001)
Ed			0.041 ***	−0.007
			(0.004)	(0.008)
Pop			0.016 **	−0.001
			(0.008)	(0.008)
Gov			0.021 ***	0.022 ***
			(0.004)	(0.004)
Fin			−0.014 ***	−0.021 ***
			(0.003)	(0.003)
Constant	−0.503 ***	−0.539 ***	−0.654 ***	−0.714 ***
	(0.001)	(0.002)	(0.034)	(0.034)
City fixed effects	Yes	Yes	Yes	Yes
Time fixed effects	No	Yes	No	Yes
Number of observations	3432	3432	3432	3432

Note: *** and ** denote significance at the 1% and 5% levels, respectively. Standard errors are presented in parentheses.

. As shown in Columns (1)–(2) of Table 2, when city and time fixed effects are considered but no control variables are included, the coefficient on Lf is significantly negative, preliminarily suggesting that land finance adversely influences the level of vegetation cover. The results in columns (3)–(4) of Table 2 demonstrate that the magnitude and significance of the coefficients on Lf do not change substantially after the inclusion of control variables, indicating that estimation result is relatively robust and that land finance has an unfavorable effect on vegetation status. This can be due to the fact that land finance can speed up urban expansion and reduce the area covered by vegetation, and discourage innovation activities that is tightly linked to resource utilization efficiency and the level of pollution prevention.

5.3. Robustness Tests

5.3.1. Replacing Variables and Samples

To verify the reliability of the empirical findings, robustness tests are carried out by replacing variables and samples. The estimation results in

Table 3. The results of robustness tests.

Variables	Total NDVI as Explained Variable	NEP as Explained Variable	Removing Municipalities	Trimming Variables
	(1)	(2)	(3)	(4)
Lf	−0.004 ***	−0.005 ***	−0.004 ***	−0.003 ***
	(0.001)	(0.001)	(0.001)	(0.001)
Constant	4.253 ***	−0.501 ***	−0.712 ***	−0.518 ***
	(0.035)	(0.007)	(0.034)	(0.048)
Control variables	Yes	Yes	Yes	Yes
City fixed effects	Yes	Yes	Yes	Yes
Time fixed effects	Yes	Yes	Yes	Yes
Number of observations	3432	3432	3384	3432

Note: *** denote significance at the 1% level. Standard errors are presented in parentheses.

suggest that the coefficient of Lf is still negative and passes the 1% significance level test, suggesting that the empirical results are reliable.

5.3.2. Spatial Econometric Analysis

Regression results of the spatial panel data model (

Table 4. The estimation results of the spatial panel data model.

Variables	Regression Coefficient	Direct Effect	Indirect Effect	Total Effect
Lf	−0.004 ***	−0.004 ***	−0.001 ***	−0.005 ***
	(0.001)	(0.001)	(0.000)	(0.001)
$\rho$	0.227 ***
	(0.027)
Control variables	Yes
City fixed effects	Yes
Time fixed effects	Yes
Number of observations	3432

Note: *** denote significance at the 1% level. Standard errors are presented in parentheses.

) demonstrate that spatial autoregression coefficient is significantly positive, suggesting that there is a significantly positive spatial correlation in vegetation conditions. After accounting for spatial effects, the coefficient of Lf remains significantly negative, suggesting a dampening impact of land finance on vegetation conditions.

5.4. Endogeneity Issues

There can be bidirectional causality between land finance and vegetation status, and the model can be susceptible to omitted variable bias; thus, instrumental variable (IV) regression is used to address potential endogeneity issues. The interaction term between the proportion of land area classified as unsuitable for development (Sland, land area with slopes greater than 15 degrees), and the one-period lag of Lf is used as an instrumental variable. The share of land area with slopes greater than 15 degrees is calculated based on the data from ASTER Global Digital Evaluation Model V003. The instrumental variable is constructed based on the following considerations. First, in accordance with planning regulations issued by China’s planning and building authorities, areas with slopes greater than 15 degrees are classified as unsuitable for development. The proportion of land area classified as unsuitable for development is a desirable indicator of land scarcity and is closely related to the potential supply of developable land and the potential land sales revenue. Second, slopes are natural-geographical features with good exogenous properties. Third, the interaction term between the share of land area classified as unsuitable for development and the one-period lag of Lf varies across cities and over time and is suitable for employment in a panel data model. The instrumental variable is constructed based on the following considerations. First, in accordance with planning regulations issued by China’s planning and building authorities, areas with slopes greater than 15 degrees are classified as unsuitable for development. The proportion of land area classified as unsuitable for development is a desirable indicator of land scarcity and is closely related to potential supply of developable land and the potential land sales revenue. The higher the proportion of land area classified as unsuitable for development, the lower the potential supply of developable land tends to be, and thus, the higher the land price tends to be. Second, slopes are natural-geographical features with good exogenous properties. Third, the interaction term between the share of land area classified as unsuitable for development and the one-period lag of Lf varies across cities and over time and is suitable to be employed in panel data model. The results of IV regression are presented in

Table 5. The results of IV regression (2SLS).

Variables	The First-Stage Regression	The Second-Stage Regression
	(1)	(2)
	Lf	Veg
Sland × L.Lf	−0.134 ***
	(0.014)
Lf		−0.013 ***
		(0.005)
Control variables	Yes	Yes
City fixed effects	Yes	Yes
Time fixed effects	Yes	Yes
Kleibergen–Paaprk LM	84.847
Kleibergen–Paaprk Wald F	88.567
Number of observations	3146	3146

Note: *** denote significance at the 1% level. Standard errors are presented in parentheses.

. The Kleibergen–Paaprk LM test and Wald F test reject the null hypothesis of under-identification and weak instruments, respectively; hence, verifying the validity of instrumental variables. The results of the second-stage regression suggest that Lf negatively influences Veg after dealing with endogeneity, thus further validating the findings of the baseline regression.

5.5. Heterogeneity Analysis

Considering that climatic conditions and ecological vulnerability vary significantly across regions, the regression results for different regions are reported in

Table 6. The regression results of heterogeneity analysis.

Variables	(1)	(2)	(3)	(4)	(5)	(6)
	Eastern China	Central China	Western China	Northeast China	Secondary Industry-led Cities	Tertiary Industry-led Cities
Lf	−0.003 **	−0.004 *	−0.005 **	−0.003	−0.005 ***	−0.001
	(0.002)	(0.002)	(0.002)	(0.002)	(0.001)	(0.002)
Constant	−0.052	0.287 **	−0.950 ***	−0.620 ***	−0.605 ***	−1.009 ***
	(0.123)	(0.137)	(0.122)	(0.114)	(0.049)	(0.070)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes
City fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Time fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Number of observations	1032	960	1032	408	1887	1545

Note: ***, **, * denote significance at the 1%, 5% and 10% levels, respectively. Standard errors are presented in parentheses. The same below.

. Land finance exerts a significant negative impact on vegetation status in eastern, central, and western China, and the magnitude of the effect is highest in western China. Since the ecological environment is relatively fragile in western China (especially in northwest China and the Tibetan Plateau), land development can have a more pronounced destructive effect on vegetation in that region. Since the industrial structure is tightly linked to resource utilization mode and efficiency, sample cities are divided into the group of secondary industry-led cities and that of tertiary industry-led cities. The curbing effect of land finance on vegetation conditions is more marked in secondary industry-led cities. Resource utilization tends to be more extensive, and economic growth tends to be more resource-dependent in secondary industry-led cities. Thus, land development can have a more noticeable disadvantageous impact on the ecological environment in those cities. Considering that the impact of land finance on vegetation dynamics can be linked to land development patterns, land finance is disaggregated by sources of land revenue and is divided into land finance based on industrial land and that based on residential and commercial land. Land finance based on industrial land has a greater adverse impact on vegetation status. Compared with residential and commercial land supply, industrial land supply tends to rely more heavily on greenfield development; thus, it can give rise to more severe damage to the ecological environment.

5.6. Mechanism Analysis

As analyzed in Section 3 and Section 4, land finance can have an inhibiting impact on vegetation conditions through accelerating urban expansion, curbing innovation, reducing land use efficiency, and distorting fiscal expenditure structure, and the extent of urban expansion (Urexp), the level of innovation output (Innov), land use efficiency (Lue), and the share of government expenditure on science and technology (St) are selected as mechanism variables to investigate the mechanisms through which land finance influences vegetation status. The regression results of mechanism analysis are presented in

Table 7. The regression results of mechanism analysis.

Variables	Urban Expansion		Innovation		Land Use Efficiency		Science and TechnologyExpense
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Urexp	Veg	Innov	Veg	Lue	Veg	St	Veg
Lf	0.008 *	−0.004 ***	−0.031 **	−0.004 ***	−0.009 ***	−0.003 ***	−0.056 ***	−0.004 ***
	(0.005)	(0.001)	(0.015)	(0.001)	(0.003)	(0.001)	(0.017)	(0.001)
Urexp		−0.013 ***
		(0.004)
Innov				0.007 ***
				(0.001)
Lue						0.062 ***
						(0.007)
St								0.001 ***
								(0.000)
Constant	−2.768 ***	−0.751 ***	0.905 **	−0.721 ***	11.109 ***	−0.029	−0.292	−0.714 ***
	(0.138)	(0.036)	(0.442)	(0.034)	(0.083)	(0.087)	(0.519)	(0.034)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
City fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Time fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Number of observations	3432	3432	3432	3432	3432	3432	3432	3432

Note: ***, **, * denote significance at the 1%, 5% and 10% levels, respectively. Standard errors are presented in parentheses. The same below.

. Land finance significantly contributes to urban expansion, which, in turn, significantly decreases the level of vegetation cover, thus indicating that accelerating urban expansion is one of the channels of impact through which land finance influences vegetation status. Land finance significantly dampens the growth of innovation output, which exerts a significant favorable impact on vegetation conditions, thereby suggesting that inhibiting innovation is another channel of impact through which land finance affects vegetation conditions. Land finance negatively affects land use efficiency, which has an advantageous effect on vegetation status, hence indicating that decreasing land use efficiency is a channel of impact through which land finance influences vegetation status. Land finance exerts a restraining effect on the share of government expenditure on science and technology, which positively influences vegetation conditions, thus suggesting that distorting fiscal expenditure structure is a channel of impact through which land finance affects vegetation conditions.

5.7. Analysis of Nonlinear Effects

The theoretical analysis presented earlier suggests that as the economic development level rises, land finance can exert a greater negative impact on vegetation status. Thus, the economic development level (Ed) is employed to be the threshold variable, and the nonlinear effect of land finance on vegetation conditions is investigated by estimating the panel threshold model. The first step in estimating a panel threshold model is to test for the existence of a threshold effect and determine the quantity of thresholds; the results are shown in

Table 8. The results of threshold effect tests.

Threshold Effect Test	F Statistics	p Value
Single threshold	47.040	0.000 ***
Double threshold	11.030	0.280

Note: *** denote significance at the 1% level. Standard errors are presented in parentheses.

. Since the test for a single threshold is statistically significant and that for a double threshold is not, the regression analysis is conducted based on the single threshold effect. Threshold estimates and confidence interval construction are illustrated in the plot of likelihood ratio (LR) function (Figure 5). The LR statistic hits the zero axis at the point estimate of the threshold value. The 95% confidence intervals of the threshold value are determined by the values of $\gamma$ (threshold parameter) for which the LR statistic lies below the dotted line.

The regression results of the panel threshold models are presented in

Table 9. The regression results of panel threshold models.

Variables	(1)	(2)
$LfdI(q\le \gamma )$	−0.004 ***	−0.003 ***
	(0.001)	(0.001)
$LfdI(q>\gamma )$	−0.029 ***	−0.026 ***
	(0.004)	(0.004)
Constant	−0.539 ***	−0.717 ***
	(0.002)	(0.034)
Control variables	No	Yes
Threshold value	14.530	14.530
City fixed effects	Yes	Yes
Time fixed effects	Yes	Yes
Number of observations	3432	3432

Note: *** denote significance at the 1% level. Standard errors are presented in parentheses.

. The threshold value is 14.530, and land finance exerts a significant adverse effect on vegetation status whether per capita GDP lies below or above the threshold value. The absolute value of the coefficient of land finance is significantly higher when per capita GDP exceeds the threshold value, suggesting that land finance exerts a stronger curbing impact on vegetation conditions when a high level of economic development is reached. In 2022, per capita GDP has exceeded the threshold value in 17 cities, thus more attention should be devoted to the stronger unfavorable environmental impact of land finance at advanced stages of economic development.

6. Conclusions and Policy Implications

In addition to promoting infrastructure development and accelerating urbanization, land finance has profound environmental effects. This research examines the effects of land finance on vegetation dynamics and mechanisms of influence, based on data from 286 Chinese cities between 2011 and 2022. The main research findings can be summarized as follows: (1) Land finance has an inhibiting effect on vegetation conditions. The results are robust to replace the explained variable and samples, spatial econometric analysis, and dealing with endogeneity. (2) Heterogeneity analysis suggests that land finance significantly reduces the level of vegetation cover in eastern, central, and western China, and the magnitude of the effect is highest in western China. The restraining effect of land finance on vegetation status is more pronounced in secondary industry-led cities. (3) The analysis of mechanisms of impact indicates that land finance exerts a curbing impact on vegetation conditions through speeding up urban expansion and suppressing innovation. (4) The analysis of the threshold effect suggests that there is a nonlinear association between land finance and vegetation status. Land finance exerts a stronger dampening impact on vegetation conditions when a higher economic development level is reached. Based on the above findings, the following policy recommendations can be proposed to deepen the reform of the fiscal system, facilitate the transition to a green economy, and encourage the preservation and restoration of vegetation.

First, the extent of the match between local governments’ fiscal capacity and expenditure responsibilities should be enhanced, so as to relieve local governments’ fiscal pressure. In terms of the division of fiscal powers, the sources of local government revenue should be expanded by promoting reforms of the consumption tax and value-added tax, optimizing the division of shared taxes between central and local governments, and augmenting the scope of local government special bonds. In the pilot reform of the consumption tax in Hainan province, the collection of the consumption tax is shifted to the retail stage, with the tax revenue allocated to local governments. In 2024, the Chinese central government decentralized the approval of special bonds to 11 provinces, municipalities, and regions, allowing those areas to review and issue special bonds on their own. Those policy pilots provide practical references for efforts to enhance the fiscal capacity of local governments. In terms of the division of fiscal responsibilities, the proportion of central government fiscal expenditures should be appropriately increased, and fiscal responsibilities entrusted to local governments should be reduced. In addition, social capital should be encouraged and guided to participate in the construction and operation of infrastructure and public utilities, thereby relieving local governments’ expenditure pressure.

Second, the performance assessment system for local government officials should be improved continuously, hence promoting high-quality development and the transition to a green economy. Economic growth targets should be formulated in accordance with the objective law, avoiding over-exploitation of land and damage to the ecological environment as a result of over-pursuing the rate of economic growth. In addition, the assessment of target responsibility for environmental protection should be reinforced, thus stimulating the motivation of local governments to strengthen environmental governance and coordinate economic development and environmental protection.

Third, the scope of the use of land sales revenue should be adjusted and optimized to facilitate ecological civilization construction. The urban-rural distribution pattern of land sales revenue should be further adjusted, and the proportion of land sales revenue used for the construction of high-standard farmland, comprehensive rural land consolidation, protection of arable land, and permanent basic farmland, and ecological protection and restoration should be increased, thus promoting environmental protection and vegetation restoration.

Fourth, greater policy and planning efforts should be made to foster ecological protection and vegetation restoration in ecologically fragile regions. The ecological compensation system should be improved by promoting innovations in market-based compensation mechanisms such as insurance of carbon sink bonds and the conversion of certificated ecological management experience into technology shares, through enriching sources of funding such as earmarking lottery proceeds for ecological compensation, and by providing a scientific basis for ecological compensation based on digital technology. In addition, land use planning for ecologically fragile regions should be formulated and implemented in a more targeted manner. Local governments should endeavor to formulate ecological restoration planning for revegetation, soil and water conservation, wetland protection, etc., optimize the spatial distribution of industries, and promote the development of green industries such as ecotourism, organic agriculture, clean energy, etc.

Admittedly, there remain limitations in this research. Due to data availability, this research mainly investigates how land sales revenue influences vegetation dynamics. As more data become available, future studies can explore how land-related tax revenue and land-related debt impact vegetation conditions. This will be conducive to a more comprehensive and thorough understanding of the environmental impacts of land-based financing. In addition, this research does not distinguish between different types of land development related to land conveyance. Future research can further examine whether greenfield development and brownfield redevelopment exert different effects on vegetation status. The Chinese central government has enacted policies to give priority to agricultural and rural development and rural revitalization in terms of the use of land sales revenue. Efforts should be made to investigate the impacts of those policies on vegetation changes. This will contribute to deepening the reform of the fiscal system and facilitating vegetation protection. Since local governments in China impose direct control over urban land supply, land finance in China is unique in the sense that land sales revenue constitutes a dominant portion of land-related revenue. Future studies can undertake international comparisons and explore whether different forms of land-based financing exert differentiated effects on the ecological environment.