Research on Land Use Transition in China from the Perspective of Household Livelihood Capital

Abstract

A symbiotic mechanism exists between household livelihood capital and land use transition (LUT). However, previous studies have seldom examined, from a macro perspective, the characteristics of household livelihood capital and its impact on LUT, lacking an in-depth analysis of the spatial spillover effects and regional heterogeneity. Household livelihood capital influences LUT by shaping livelihood strategies. This article incorporates psychological capital into the Sustainable Livelihood Framework (SLF) and utilizes data from the China Family Panel Studies (CFPS) database and land use data spanning 24 provincial-level administrative units in China from 2010 to 2022; this study employs the Spatial Durbin Model to empirically analyze the impact of household livelihood capital on LUT and its regional variability, along with the spatiotemporal evolution patterns of household livelihood capital and land use change. The results demonstrate the following: (1) From 2010 to 2022, household livelihood capital increased, with higher levels of psychological, natural, and human capital, while social, financial, and physical capital were lower. The eastern region exhibited higher livelihood capital levels than the central and western regions, but the gap narrowed annually. (2) Between 2010 and 2022, LUT primarily involved the transition of cropland to construction land, cropland to forest and grassland, and forest and grassland to cropland. The intensity of LUT increased over time and showed a spatial gradient from west to east. (3) LUT positively affected the LUT of adjacent provinces. Various forms of capital and livelihood strategies had different effects on LUT within a province and across neighboring provinces. (4) The heterogeneity analysis reveals that the eastern region is significantly affected by the competitive effect and the spillover effect from neighboring provinces, while the central and western regions are constrained by water resources and inhibited by pure agricultural farmers. The article reveals the mechanism through which household livelihood capital drives LUT and the spatial spillover patterns, providing scientific evidence for regionally differentiated land management and ecological compensation policies, which is of great importance for the sustainable use of regional land resources.

1. Introduction

Since the Earth entered the Anthropocene Era [1], global environmental issues such as soil erosion and desertification have coexisted with socio-economic phenomena including large-scale population migration and rapid urbanization, making the “human–land relationship” increasingly complex [2]. Large-scale human activities have significantly altered the surface morphology and land use patterns. Land use transition (LUT), a new research avenue within land use and land cover change (LUCC), has emerged as a focal area in land system science, offering theoretical and practical insights for managing the complex “human–land relationship” [3]. Within this regional system, livelihood, as the most fundamental human behavior, influences LUT by modifying land use patterns, structures, and efficiencies [2,4]. It provides a distinctive perspective and innovative tools for comprehending the “human–land relationship”. Accurately recognizing the “human–land relationship” between human livelihoods and LUT has become a central research focus in sustainability science [5,6].

Research on LUT began in the 1980s with studies on forest transition [7]. LUT was initially defined by Grainger (1995) as the change in land use types with economic and social development [8]. As land use forms expanded to include both explicit and implicit connotations, LUT is defined as the process of changing regional land use structures and functional forms driven by economic and social changes and innovations [9,10,11,12]. LUT is influenced by natural conditions (e.g., topography, precipitation, and slope [13,14]) and human factors (e.g., economic development [15,16], population migration [17], and national policies [18,19,20]), and is closely related to urbanization [21], rural revitalization [12,22], and rural development transformation [23]. However, macro-level studies may overlook specific information such as processes and human behaviors [24], and policy and environmental changes ultimately take effect through the behaviors of microeconomic agents.

Therefore, it is essential to start from micro-behavioral decision-makers such as farmers to examine the human–land relationship [4]. At the farmer level, land use changes (e.g., land cover [25,26], cropland use [27,28,29], and homestead use [30]) are mainly driven by factors such as livelihood pressure, population growth, non-agricultural employment, and labor transfer [31]. Livelihood capital is a key entry point for understanding farmers‘ land use behavior. Livelihood is defined as the capabilities, assets, and activities required to sustain a living [32]. The Sustainable Livelihood Framework (SLF) proposed by DFID is the most widely applied approach. The SLF suggests that, under the influence of external environments, policies, organizations, and institutions, farmers form different income structures and production activities based on their endowments of human, natural, physical, financial, and social capital, which, in turn, leads to the choice of different livelihood strategies [33], including four types: pure agricultural, agricultural part-time, non-agricultural part-time, and non-agricultural [34,35,36]. In addition, recent studies have incorporated psychological capital into the SLF [19,37,38,39,40,41,42,43].

Research has found a symbiotic mechanism between LUT and household livelihoods, with the two interacting with each other [2,44,45]. On the one hand, land is the capital on which farmers rely for survival and the most basic production factor for rural economic development, and household livelihoods are directly affected by LUT. Studies have confirmed that LUT can alter household livelihood capital structures and increase total household assets [46], promote diversified livelihood transitions by adjusting human–land interactions and factor substitution in agricultural development [47], and effectively increase farmers’ income with significant regional heterogeneity [48]. On the other hand, farmers integrate the roles of producers, consumers, and decision-makers [49], and their livelihood capital (especially physical, human, and natural capital) is a key variable influencing their production decisions (essentially land use behavior) [50], directly causing LUCC and generating LUT [44,51,52] with dual effects. From the positive perspective, livelihood diversification enhances farmers’ resilience, risk resistance, and livelihood sustainability, thereby reducing land cover activities such as deforestation and promoting vegetation and forest restoration [4,53,54]. From the negative perspective, with socio-economic development, farmers engaging in non-agricultural production for economic returns has become an inevitable trend, which affects the land use types and intensification levels [27], leading to land abandonment, landscape fragmentation, and extensive use [55,56,57,58], and even triggering ecological degradation such as soil erosion and forest loss [59,60,61,62,63]. In addition, the off-farm transfer of rural labor drives the transformation of agricultural industrial structure, changing the type, input structure, and cropping structure of cropland use [28,64], and causing local reductions in rural settlements [65].

In summary, the existing macro-level research on LUT tends to overlook essential information such as specific processes and human behaviors, while micro-level research is limited in scope and timespan and lacks an in-depth explanation of the pathways through which livelihood capital influences LUT. To address these gaps, this study introduces the human–land relationship regional system theory and the SLF. The theory of the human–land relationship regional system highlights that LUT arises from the interaction of natural and human factors, with household livelihood serving as the primary link connecting micro-level behaviors and LUT. When internal livelihoods and external environments (such as the Grain for Green Project, and ecological protection policies) undergo significant changes, livelihood capital forms the basis for households to adjust livelihood strategies to adapt to new human–land relationships [6,41]. By determining these strategies, it directly influences households’ land use behaviors (such as land use types, cropland area, woodland area, etc.) [25], resulting in regional LUT. As changes in livelihood strategies occur, LUT aims to improve livelihoods and enhance sustainability, leading to alterations in the structure and level of household livelihood capital. Moreover, LUT further affects the external environment, thus reshaping household livelihood capital structures. This paper integrates the theory of the human–land relationship with the SLF framework, systematically unveiling the driving mechanisms of LUT from the perspective of household livelihood capital (Figure 1).

Since the reform and opening up, China has undergone profound socio-economic transformations. From 2010 to 2022, this period marked rapid economic development, with a 5.45% population growth and an urbanization rate rising from 50% to 65.22%. Rapid urbanization, the Grain for Green Project, and other ecological protection initiatives led to the continuous expansion of construction land and fluctuating changes in cropland, woodland, and grassland [66,67,68]. Concurrently, these 12 years represented a crucial phase for the Chinese government’s targeted poverty alleviation efforts and comprehensive poverty eradication, with the rural revitalization strategy marking a shift towards holistic rural development. Driven by urbanization, rapid economic growth, and a series of pro-agriculture policies, the livelihood levels of China’s rural population have continuously improved [40,69]. Based on the intrinsic relationship between household livelihood capital and LUT, this study employs the above theoretical framework, using China Family Panel Studies (CFPS) data from 2010 to 2022 to characterize regional household livelihood capital levels, and constructs an evaluation index system from six dimensions: human capital, natural capital, physical capital, financial capital, social capital, and psychological capital. The study aims to explore the driving factors of household livelihood capital in LUT and address the following research questions: (1) What are the spatiotemporal evolution patterns of household livelihood capital and LUT in China during the study period? (2) How does household livelihood capital directly or indirectly influence LUT? (3) Is there regional heterogeneity in the impact of household livelihood capital on LUT?

2. Materials and Methods

2.1. Study Area

China comprises 34 provincial-level administrative units, each demonstrating significant spatial heterogeneity in terms of economic development stages, industrial structure characteristics, and resource endowments. Due to geographical peculiarities and limitations in data availability, Beijing, Inner Mongolia, Tibet, Xinjiang, Taiwan, Hong Kong, and Macau were excluded from this study. The provinces were categorized into eastern and central and western regions based on urban geographical location and economic development level. The eastern region comprises Tianjin, Hebei, Liaoning, Shanghai, Jiangsu, Zhejiang, Fujian, Shandong, and Guangdong. The central–western region comprises Shanxi, Jilin, Heilongjiang, Anhui, Jiangxi, Henan, Hubei, Hunan, Chongqing, Sichuan, Yunnan, Guizhou, Shaanxi, Gansu, and Guangxi [38] (Figure 2).

2.2. Data Sources

The data utilized in this paper comprises three types: socio-economic data, GIS data, and household survey data. The GIS data includes national administrative boundary data and land use data. Socio-economic data is sourced from the China Statistical Yearbook. Vector data for administrative boundaries is obtained from the National Earth System Science Data Sharing Platform (https://www.geodata.cn). Land use data is derived from the 1990–2023 China 30 m annual land cover data by the research team led by J. Yang and X. Huang from Wuhan University, which includes cropland, forest, shrub, grassland, water, snow/ice, barren, impervious surface, and wetland [70]. For the research purposes of this paper, the data is resampled to a precision of 1 km, ensuring consistent rows, columns, and pixel numbers across provinces annually. Household survey data is sourced from China Family Panel Studies (CFPS, https://cfpsdata.pku.edu.cn/#/home), which conducts surveys biennially. The survey encompasses comprehensive information on Chinese household economic activities, family relationships, migration, health, and more. The CFPS employs a probability-proportional-to-size (PPS) sampling method based on implicit stratification, with multistage, multilevel sampling. Computer-assisted personal interviewing is used, covering 25 provincial-level administrative units in China, excluding Hong Kong, Macau, Taiwan, Xinjiang, Tibet, Qinghai, Inner Mongolia, Ningxia, and Hainan. These areas account for about 95% of China’s total population, excluding Hong Kong, Macau, and Taiwan, making CFPS a nationally representative sample. This paper uses data from seven rounds of national surveys conducted between 2010 and 2022. To ensure the quality of research data, effective screening was performed, retaining only rural households. On this basis, data that included refusals, blanks, unknowns, or respondents under 20 were all removed, resulting in valid data from 37,865 rural households across 24 provincial-level administrative units. The data was matched in Stata 18.0 software using family ID with the adult and household databases, using the average household livelihood capital of the sample as the representative level of household livelihood capital for that province, constructing panel data.

2.3. Variables

Based on the DFID sustainable livelihood framework’s definition of household livelihood capital, and referencing existing literature on the methods of measuring livelihood capital indicators [34,37,38,40,71], this study incorporates the dimension of psychological capital to construct a household livelihood capital indicator system (

Table 1. Descriptions of variables in the analysis.

Variable		Unit	Description
Natural capital	N1 Land asset	104 yuan	<0.1=1; ≥0.1~1=2; ≥1~3=3; ≥3~5=4; ≥5=5
	N2 Drinking water source	-	Cellar and rainwater = 1; river, lake, pond, and spring water = 2; well and other water = 3; tap water = 4; bottled and purified water = 5
	N3 Domestic fuel	-	Others = 0; firewood, biogas, and coal = 1; liquefied gas, natural gas, and electricity = 2
Human capital	H1 Labor force share	-	The proportion of the labor force within the family population
	H2 Education status of family members	-	University bachelor’s degree and above = 5; college diploma = 4; high school, vocational school, and technical school = 3; junior high school = 2; primary school = 1; illiterate = 0
	H3 Health status of family members	104 yuan	<0.1=5; ≥0.1~0.3=4; ≥0.3~0.5=3; ≥0.5~1=2; ≥1=1
Physical capital	M1 Agricultural machinery value	104 yuan	<0.1=1; ≥0.1~1=2; ≥1~3=3; ≥3~5=4; ≥5=5
	M2 Durable goods value	104 yuan	<1=1; ≥1~3=2; ≥3~5=3; ≥5~10=4; ≥10=5
	M3 House value	104 yuan	<10=1; ≥10~30=2; ≥30~50=3; ≥50~100=4; ≥100=5
Financial capital	F1 Total cash and deposits	104 yuan	<1=1; ≥1~3=2; ≥3~5=3; ≥5~10=4; ≥10=5
	F2 Debt status	-	Whether household in debt. Yes = 0, No = 1
Social capital	S1 Household social status	-	Very poor = 0; low = 1; relatively low = 2; average = 3; relatively high = 4; very high = 5
	S2 Transport and communications expenditure	104 yuan	<0.1=1; ≥0.1~0.3=2; ≥0.3~0.5=3; ≥0.5~1=4; ≥1=5
Psychological capital	P1 Life satisfaction	-	Very poor = 0; low = 1; relatively low = 2; average = 3; relatively high = 4; very high = 5
	P2 Confidence status	-	Very poor = 0; low = 1; relatively low = 2; average = 3; relatively high = 4; very high = 5
Livelihood strategy	LS1 Pure agricultural type	%	The proportion of pure agricultural famers within the sample of the province
	LS2 I-part-time agricultural type	%	The proportion of I-part-time agricultural farmers within the sample of the province
	LS3 II-part-time agricultural type	%	The proportion of II-part-time agricultural famers within the sample of the province
Control variable	C1 GDP	104 yuan/ per capita	Gross National Product per capita
	C2 TS	-	Ratio of tertiary industry output to secondary industry output
	C3 UR	-	Ratio of family laborers to family population size

). Natural capital encompasses the land, water, and biological resources individuals can access to sustain their livelihoods, measured by land assets (N₁) which were calculated from agricultural income according to the CFPS database, drinking water sources (N₂), and domestic fuels (N₃) of the household. Human capital refers to the quantity and quality of family labor, measured by the labor force share of the household (H₁), and the average education status (H₂) and health status (H₃) of family members. Physical capital pertains to basic production materials and infrastructure sustaining livelihoods, measured by the value of agricultural machinery (M₁), durable goods (M₂), and house (M₃) owned by the household. Financial capital involves the financial resources available to farmers for achieving livelihood goals, measured by the total cash and deposits (F₁) owned by the household and whether the household is in debt (F₂). Social capital signifies farmers’ ability to access resources through social networks or broader social structures, measured by the average social status level (S₁) and expenditure of transport and communications (S₂) of the household. Psychological capital embodies the psychological resources enhancing farmers’ growth and performance, measured by the life satisfaction (P₁) and the average confidence status (P₂) of the household.

As previously mentioned, household livelihood capital, being an intrinsic determinant of changes in livelihood strategy [41], influences land use behaviors by shaping these strategies. According to the standards of the National Bureau of Statistics and the Chinese Academy of Social Sciences (

Table 2. The classification standard of Peasant Household Concurrent Occupation.

Livelihood Strategy	Livelihood Method	Income Status
Pure agricultural type	Agriculture-dominated	Agricultural income share ≥ 90%
I-part-time agricultural type	Predominantly agriculture with some non-agriculture	50% > Non-agricultural income share ≥ 10%
II-part-time agricultural type	Predominantly non-agriculture with some agriculture	90% > Non-agricultural income share ≥ 50%
Non-agricultural famers	Non-agriculture-dominated	Non-agricultural income share > 90%

), household livelihood strategies are categorized into four types based on income structure: pure agricultural, I-part-time agricultural, II-part-time agricultural, and non-agricultural farmers [34,35,36]. For pure agricultural farmers, agricultural income accounts for over 90% of total income, with family members primarily engaged in crop cultivation and livestock breeding; for non-agricultural farmers, non-agricultural income exceeds 90%, with family members mainly involved in off-farm employment, retail businesses, or factory operations. I-part-time and II-part-time farmers engage in both agricultural and non-agricultural activities, differing only in the proportion of income from each source. The proportion of each farmer type within a province reflects its income structure characteristics. This study uses the proportions of pure agricultural (LS₁), I-part-time agricultural (LS₂), and II-part-time agricultural (LS₃) farmers as explanatory variables to analyze the impact of part-time farming structure on LUT.

Additionally, considering the pivotal role of economic and social development in promoting LUT [16,72], this study uses per capita GDP, the industrial structure upgrading index, and urbanization level (the ratio of the urban population to the total population) as control variables for the external economic environment. The industrial structure upgrading index is defined as the ratio of the value of the tertiary industry to that of the secondary industry.

2.4. Research Methods

2.4.1. Entropy Weight Method

The entropy weight method is an objective evaluation approach used for weighting. Its core principle involves standardizing various evaluation indicators and employing Shannon entropy to compute the information utility value of each indicator, thereby deriving the evaluation weights. This study adopts the entropy weight method, building on existing methods for measuring livelihood capital, to ascertain the weights of each indicator. The procedure includes the following:

Step 1: Perform dimensionless processing on various livelihood capital indicators.

For positive indicators, we have the following:

$$X_{ab}^{\prime }={\displaystyle \frac{X_{ab}-X_{min}}{X_{max}-X_{min}}} \mathrm{a}=1,2,\dots ,\mathrm{m}; \mathrm{b}=1,2,\dots ,\mathrm{n}$$

(1)

For negative indicators, we have the following:

$$X_{ab}^{\prime }={\displaystyle \frac{X_{max}-X_{ab}}{X_{max}-X_{min}}} \mathrm{a}=1,2,\dots ,\mathrm{m}; \mathrm{b}=1,2,\dots ,\mathrm{n}$$

(2)

In this formula, $m$ is the number of sample households in a given year, and $n$ is the number of measurement indicators for livelihood capital ($n$ = 6). $X_{ab}^{\prime }$ is the normalized value of the b-th indicator for the a-th sample, $X_{ab}^{\prime }$ is the variable value of the b-th indicator for the a-th sample, $X_{max}$ is the maximum value of the b-th indicator, and $X_{min}$ is the minimum value of the b-th indicator.

Step 2: Calculate the proportion $P_{ab}$ of the indicator value for the a-th sample household with respect to the b-th indicator.

$$P_{ab}={\displaystyle \frac{X_{ab}^{\prime }}{{{\displaystyle \sum }}_{j=1}^m{X}_{ab}^{\prime }}} \mathrm{i}=1,2,\dots ,\mathrm{m}; \mathrm{j}=1,2,\dots ,\mathrm{n}$$

(3)

Step 3: Calculate the entropy value $e_b$ for the b-th indicator. When dealing with the logarithm of zero, which is undefined, replace the value 0 with 0.01, per established research methodologies.

$$e_b=-k{\displaystyle \sum }_{i=1}^m{P}_{ab}\mathit{ln}{P}_{ab}\hspace{1em}\hspace{1em}\left(k={\displaystyle \frac{1}{\mathit{ln}m}}\right)$$

(4)

Step 4: Obtain the weight $w_b$ of the b-th indicator.

$$w_b={\displaystyle \frac{\left(1-e_b\right)}{{{\displaystyle \sum }}_{j=1}^n\left(1-e_b\right)}}$$

(5)

Step 5: The sustainable livelihood capital index for the sample households is determined using a weighted summation method based on the calculated indicator weights.

$$P_{ab}={\displaystyle \frac{X_{ab}^{\prime }}{{{\displaystyle \sum }}_{j=1}^m{X}_{ab}^{\prime }}} \mathrm{i}=1,2,\dots ,\mathrm{m}; \mathrm{j}=1,2,\dots ,\mathrm{n}$$

(6)

2.4.2. Land Use Transfer Matrix

• Land use dynamic degree

The land use dynamic degree quantifies changes in land use quantity over a specified time period and is categorized into single and comprehensive land use dynamic degrees. The single land use dynamic degree indicates the rate of change for a specific land use type, used to characterize the speed and magnitude of change during the study period. In contrast, the comprehensive land use dynamic degree characterizes the change magnitude across all land use types [73,74]. This paper employs the single land use dynamic degree as a key indicator to reflect the quantitative structure of land use change, as provided in the following formula:

$$K={\displaystyle \frac{S_{T_2}-S_{T_1}}{S_{T_1}}}\times {\displaystyle \frac{1}{T}}\times 100\%$$

(7)

In the formula, $K$ is the dynamic degree of a specific land use type; $S_{T_1}$ and $S_{T_2}$ are the area of that land use type at the beginning and end of the study period, respectively; and T is the duration of the study. When $T$ is expressed in years, $K$ is the annual change rate of the specified land use type within the study area.

2.

Contribution rate of land use conversion

The land use transfer matrix is an essential tool for dynamically analyzing and quantitatively describing regional land use patterns [75]. It enables the determination of net transfer-in and net transfer-out areas of land use types. However, because net changes in land use may overlook portions offset by transfers, potentially underestimating the true extent of land use change, it is necessary to extract the exchange volumes of land use changes to accurately reflect conversion directions [76]. This facilitates calculating the proportion of land use type “conversion acquisition” or “conversion reduction” over a period, relative to the net gains or losses [14]. The formula is as follows:

$$P_{loss\left(c\right),\mathrm{d}}=\left[S_{cd}/\left(S_{\mathrm{c}+}-S_{cc}\right)\right]\times 100\%$$

(8)

$$P_{gain\left(c\right),d}=\left[S_{dc}/\left(S_{+c}-S_{cc}\right)\right]\times 100\%$$

(9)

In the formula, $S_{cd}$ and $S_{dc}$ are the areas where land use are mutually converted in the transfer matrix during a certain period. $S_{\mathrm{c}+}-S_{cc}$ is the net losses of land use type $c$, and $S_{+c}-S_{cc}$ is the net gains of land use type $c$. $P_{loss\left(c\right),\mathrm{d}}$ refers to the proportion of land use type $c$ converted to $d$ within the total area transferred out of land use type $c$ during the period, which is the contribution rate of conversion reduction. $P_{gain\left(c\right),d}$ refers to the proportion of land use type $d$ converted to $c$ within the total area transferred into land use type $c$, which is the contribution rate of conversion acquisition.

2.4.3. Exploratory Spatial Data Analysis

Exploratory Spatial Data Analysis (ESDA) encompasses a range of techniques and methods for classifying spatial data. Within the ESDA framework, spatial autocorrelation indicators measure the spatial distribution characteristics of geographic variables and their influence on neighboring areas. These indicators are generally categorized into Global Moran’s I (GMI) and Local Moran’s I (LMI). GMI primarily evaluates the correlation and aggregation of attributes in spatially adjacent units. This paper utilizes GMI to assess the global spatial autocorrelation of major land use conversions in the study area from 2010 to 2022. The formula is as follows:

$$GMI={\displaystyle \frac{n{{\displaystyle \sum }}_{i=1}^n{{\displaystyle \sum }}_{j=1}^n{w}_{ij}\left(X_i-\overline{X}\right)\left(X_J-\overline{X}\right)}{S^2{{\displaystyle \sum }}_{i=1}^n{{\displaystyle \sum }}_{j=1}^n{W}_{ij}}}\left(i\ne j\right)$$

(10)

$$S^2={\displaystyle \frac{1}{n}}{\displaystyle \sum }_{i=1}^n{\left(X_i-\overline{X}\right)}^2$$

(11)

In the formula, n is the number of provincial samples. $X_i$ and $X_J$ are the major land use conversions area for provinces i and j, respectively. $\overline{X}$ and $S^2$ are the mean and standard deviation. $W_{ij}$ is the spatial weight between provinces i and j. This study constructs the spatial weight matrix using the first-order Queen contiguity rule, considering two provinces as adjacent if they share a boundary or a vertex, with matrix elements $w_{ij}$ = 1; otherwise, $w_{ij}$ = 0. The range of GMI values is [−1, 1]; GMI > 0 indicates positive spatial correlation, GMI < 0 indicates negative spatial correlation, and greater absolute values signify stronger spatial correlation. GMI = 0 indicates that attributes are randomly distributed spatially, with no spatial correlation.

2.4.4. Spatial Econometric Regression Analysis

Based on the analysis of national LUT, selected livelihood capital indicators serve as independent variables, while socio-economic factors function as control variables, and the conversion area of major land use conversions in the study area is designated as the dependent variable. Stata 18.0 is employed to develop an Ordinary Least Squares (OLS) model and the Spatial Durbin Model (SDM) for regression analysis. Ultimately, a comprehensive analysis identifies the correlation between LUT and its driving factors.

• OLS regression

This paper formulates the following global linear regression model to estimate the impact of all independent variables on land use change:

$$LUC{C}_i=\mathsf{\alpha }+{\mathsf{\alpha }}_1{X}_i+{\mathsf{\alpha }}_2{C}_i+{ϵ}_i$$

(12)

In the formula, $LUC{C}_i$ is the land use change value for province i; $X_i$ is the livelihood capital indicator value for province i; $C_i$ is the control variables; ${\mathsf{\alpha }}_1$ and ${\mathsf{\alpha }}_2$ are the regression estimation coefficients for the explanatory and control variables, respectively; $\mathsf{\alpha }$ is the constant term; and ${ϵ}_i$ is the error term.

2.

Spatial econometric model

This paper utilizes spatial econometric models to address the endogeneity issues arising from omitted spatial variables, focusing on the Spatial Autoregressive Model (SAR), Spatial Error Model (SEM), and Spatial Durbin Model (SDM). The SDM estimator uniquely reveals both the effects of land use change in adjacent areas and the influencing factors on land use change within the study area. Additionally, it can, under specific conditions, simplify into the SAR and SEM estimators, rendering it more suitable than the other two models for a comprehensive examination of various influencing factors on land use change. It can be written as follows:

$$LUC{C}_{it}=\mathsf{\alpha }+\rho WLUC{C}_{it}+{\alpha }_1{X}_{it}+\beta W{X}_{it}+{\alpha }_2{C}_{it}+\xi W{C}_{it}+{\pi }_i+{\delta }_t+{\mu }_{it}$$

(13)

In the formula, $W$ is the spatial weight matrix; and subscripts i and t indicate the sample province and year, respectively. $LUC{C}_{it}$ is the land use change value for province i in year t, $X_{it}$ is the livelihood capital indicator value for province i in year t, and $C_{it}$ signifies the control variables. $\rho$ is the spatial lag term, reflecting the degree of spatial dependence of the explained variable among provinces. $\beta$ and $\xi$ are the coefficients of the spatial lag terms for the explanatory and control variables, respectively. ${\mathsf{\alpha }}_1$ and ${\mathsf{\alpha }}_2$ are the regression estimation coefficients for the explanatory and control variables, respectively, whereas $\mathsf{\alpha }$ is the constant term. ${\pi }_i$ and ${\delta }_t$ denote the province fixed effect and the year fixed effect, respectively, and ${\mu }_{it}$ is the random disturbance term.

3. Results

3.1. Spatio-Temporal Patterns of Household Livelihood Capital

Using the regional household livelihood capital measurement model and CFPS survey data, this study derives the livelihood capital index of households in the study area and examines its spatiotemporal characteristics. Temporally, the household livelihood capital level in the study area trended upward from 2010 to 2022. By 2022, significant improvements were observed in both the eastern and central and western regions compared to 2010 (Figure 3). Notably, human, physical, and financial capital increased annually, with the most substantial gains in physical and financial capital. Conversely, natural and psychological capital showed a fluctuating downward trend. From 2010 to 2022, structural differences in livelihood capital emerged within the study area, characterized by an uneven distribution across different dimensions. Among the household livelihood capitals, psychological, natural, and human capital levels were higher, whereas social, financial, and physical capital levels were lower (Figure 4) in the study area. In 2010, the order was as follows: psychological capital (0.687) > natural capital (0.492) > human capital (0.440) > social capital (0.292) > financial capital (0.173) > physical capital (0.067). In 2016, the order was psychological capital (0.635) > natural capital (0.476) > human capital (0.474) > financial capital (0.357) > social capital (0.342) > physical capital (0.177). In 2022, the order was psychological capital (0.605) > social capital (0.577) > human capital (0.518) > financial capital (0.494) > natural capital (0.420) > physical capital (0.218).

Over the 12-year period, significant improvements in household education and health led to a continuous increase in human capital. The growth in physical capital was primarily demonstrated by a substantial rise in the value of durable goods and agricultural machinery owned by households. Financial capital expanded due to a considerable increase in cash and deposits, coupled with a reduction in household debt, indicating a marked enhancement in household asset inventory and financial capacity. The rise in social capital was reflected in yearly increases in transportation and communication expenses, suggesting closer outside connections. With economic and social development, the quality of households’ drinking water and domestic fuel improved annually, while the value of the land asset decreased, resulting in an overall decline in natural capital. The drop in life satisfaction and confidence primarily reduced the psychological capital stock (

Table 3. The weight and summary statistics of household livelihood capitals.

Variable	2010		2012		2014		2016		2018		2020		2022
	Mean	w	Mean	w	Mean	w	Mean	w	Mean	w	Mean	w	Mean	w
N1	2.827	0.710	3.105	0.642	2.861	0.726	2.771	0.776	2.606	0.791	2.608	0.838	2.592	0.865
N2	3.344	0.158	3.463	0.192	3.474	0.148	3.545	0.108	3.585	0.103	3.660	0.068	3.709	0.055
N3	1.310	0.132	1.461	0.165	1.459	0.126	1.525	0.116	1.600	0.106	1.669	0.093	1.714	0.080
H1	0.622	0.254	0.586	0.289	0.569	0.313	0.555	0.312	0.532	0.340	0.553	0.356	0.549	0.361
H2	1.174	0.541	1.281	0.485	1.302	0.421	1.357	0.400	1.411	0.359	1.657	0.344	1.682	0.334
H3	1.964	0.204	2.091	0.226	2.284	0.265	2.334	0.288	2.431	0.301	2.267	0.300	2.366	0.306
M1	1.166	0.333	1.229	0.352	1.327	0.352	1.436	0.386	1.358	0.428	1.413	0.432	1.390	0.461
M2	1.165	0.434	1.302	0.404	1.480	0.415	1.763	0.357	1.867	0.351	2.068	0.324	2.176	0.294
M3	1.443	0.233	1.511	0.243	1.754	0.233	1.834	0.257	2.047	0.221	2.039	0.244	2.156	0.245
F1	1.217	0.823	1.621	0.752	1.656	0.750	1.891	0.707	2.010	0.690	2.216	0.669	2.446	0.636
F2	0.357	0.177	0.318	0.248	0.349	0.250	0.340	0.293	0.328	0.310	0.330	0.331	0.313	0.364
S1	3.829	0.083	3.732	0.154	3.846	0.171	3.726	0.194	3.743	0.310	3.716	0.341	3.786	0.317
S2	1.923	0.917	1.992	0.846	2.367	0.829	2.468	0.806	2.625	0.690	2.789	0.659	2.981	0.683
P1	3.931	0.374	3.894	0.353	3.599	0.386	3.692	0.395	3.220	0.458	3.321	0.439	3.476	0.448
P2	3.629	0.626	3.705	0.647	3.218	0.614	3.449	0.605	3.074	0.542	3.091	0.561	3.255	0.552

).

In the spatial regions of the study area, the livelihood capital stock across provincial-level administrative units generally exhibits a growth trend, influenced by factors such as resource endowment, economic development levels, and modes of production and living. The eastern region consistently exhibits higher levels of livelihood capital compared to the central and western regions. Comparing different regional livelihood capital levels with the study area’s average reveals significant improvement in the central and western regions, where the difference from the average decreased from −1.60% to −0.40%. Conversely, in the eastern region, the difference from the study area’s average narrowed from 2.66% to 0.66%. Sub-item livelihood capital also shows significant regional differences (

Table 4. Regional comparison of household livelihood capital.

Year	Eastern Region		Central and Western Region		Study Area
	Capital	Difference	Capital	Difference	Capital
2010	2.210	2.66%	2.118	−1.60%	2.153
2012	2.437	3.71%	2.297	−2.23%	2.350
2014	2.506	3.56%	2.369	−2.13%	2.420
2016	2.505	1.81%	2.434	−1.09%	2.461
2018	2.406	2.33%	2.318	−1.40%	2.351
2020	2.724	0.95%	2.683	−0.57%	2.698
2022	2.850	0.66%	2.820	−0.40%	2.832

).

From 2010 to 2022, owing to the higher economic development, the eastern region maintained superior levels of human, financial, and physical capital. In contrast, the central and western regions, rich in natural resources, especially arable land, possess much higher natural capital than the eastern region. As development and economic progress unfolded in the central and western regions, the gap in livelihood capital levels with the eastern region gradually narrowed, evidenced by diminished differences in material and social capital. Moreover, the central and western regions displayed relatively higher psychological capital than the eastern region (Figure 5).

Specifically, Jiangsu and Zhejiang provinces in the eastern region consistently rank among the top nationwide in household livelihood capital levels, predominantly due to the abundant human and financial capital accumulation. In 2022, Shanghai’s average household livelihood capital level was the highest in the country, primarily due to its accumulations of human, financial, and psychological capital, surpassing other provinces. Over the span of 12 years, provincial-level administrative units such as Yunnan, Guangxi, and Hubei in the central and western regions have seen their livelihood capital levels advance to the forefront nationwide. This improvement is mainly attributed to their natural, material, social, and psychological capital surpassing the national average. Conversely, Chongqing’s livelihood capital has consistently remained at the lowest level because of deficiencies in natural, material, and financial capital, significantly below the national average (Figure 6).

3.2. Spatio-Temporal Patterns of LUT

An analysis of the land use transfer matrix and inter-type conversions across 24 provincial-level administrative units in China between 2010 and 2022 reveals significant changes in the quantities of various land use types within the study area (

Table 5. Change matrix of each compared land use type in 2000 and 2020 (unit: ×10⁴ ha).

2010	2022									Total
	Cropland	Construction	Woodland	Shrub	Grassland	Water	Snow/Ice	Wetland	Barren
Cropland	14,422.17	528.62	1117.98	14.63	334.71	83.38	0.00	0.38	3.55	16,505.42
Construction	80.54	1428.01	4.72	0.00	1.14	22.67	0.00	0.00	0.14	1537.22
Woodland	1083.06	27.57	19,335.54	50.86	58.45	8.31	0.00	0.07	0.09	20,563.96
Shrub	37.63	0.01	107.59	113.47	22.05	0.02	0.00	0.00	0.01	280.79
Grassland	474.48	20.68	293.12	16.35	4330.95	7.95	1.32	1.68	80.80	5227.33
Water	138.04	36.00	8.20	0.04	5.77	592.44	0.15	0.01	3.69	784.34
Snow/Ice	0.00	0.00	0.03	0.00	1.66	0.26	16.13	0.00	10.17	28.25
Wetland	2.65	0.00	0.22	0.00	1.23	0.07	0.00	3.87	0.00	8.04
Barren	19.41	9.02	0.14	0.00	72.80	12.46	4.83	0.01	1719.74	1838.42
Total	16,257.98	2049.91	20,867.55	195.36	4828.74	727.55	22.43	6.03	1818.20	46,773.76
Loss	2083.24	109.21	1228.42	167.32	896.38	191.90	12.12	4.16	118.68
Gain	1835.81	621.89	1532.01	81.89	497.79	135.11	6.31	2.15	98.46
K (%)	−0.12	2.78	0.12	−2.54	−0.64	−0.60	−1.71	−2.08	−0.09

). Cropland, shrub, grassland, water, snow/ice, wetland, and barren land exhibited a decreasing trend, whereas construction land and woodland displayed an increasing trend. In terms of conversion volume, there was substantial loss of cropland, woodland, and grassland, coinciding with a notable increase in cropland, woodland, construction land, and grassland. Specifically, cropland was extensively converted to construction land. Furthermore, significant interconversion occurred among cropland, woodland, and grassland, with major land use conversions from cropland to construction land, from cropland to woodland and grassland, and from woodland and grassland back to cropland in the study area.

Regarding the change intensity, the land use dynamic degrees for impervious surfaces, shrub, and wetlands were the most pronounced, each exceeding 2.00%. Over the span of 12 years, cropland, shrub, grassland, water areas, snow/ice, wetlands, and barren land decreased annually on average by 0.12%, 2.54%, 0.64%, 0.60%, 1.71%, 2.08%, and 0.09%, respectively. In contrast, construction land and woodland increased annually on average by 2.78% and 0.12%. Notably, construction land significantly expanded its share of the regional land area, rising from 3.29% to 4.38% with a dynamic degree of 2.78%, substantially surpassing other land types. This expansion was primarily fueled by the conversion of cropland (528.62 × 10⁴ ha), which constituted 85.00% of the total area converted from other land types to construction land. Shrub significantly reduced its share of the regional land area, decreasing from 0.60% to 0.42% with a dynamic degree of −2.54%, primarily losing to woodland (1083.06 × 10⁴ ha), which represented 64.30% of the total area converted from shrub to other land types. In the conversion of wetlands to other types, cropland and grassland together accounted for over 90.00%, with some grassland converting back to wetlands, indicating an overall reduction trend. Although the expansion of forest areas and reduction in grassland areas were relatively minor, large-scale conversions between land types occurred. Agricultural expansion led to the substantial conversion of woodland and grassland into cropland, resulting in losses of 1083.06 × 10⁴ ha and 474.48 × 10⁴ ha, accounting for 88.17% and 52.93% of the losses in forest and grassland, respectively. During this period, “Grain for Green” had a significant effect on vegetation restoration, leading to a substantial conversion of cropland back to woodland and grassland [77,78], comprising 70.97% and 67.24% of the increases in forest and grassland, respectively. Due to construction land expansion and ecological protection policies, the cropland area decreased by 247.43 × 10⁴ ha over 12 years, reducing its share in the regional land area from 35.29% to 34.76%, mainly converting to forest, construction land, and grassland. The dynamic degree of snow/ice was second only to wetlands, decreasing by 5.81 × 10⁴ ha, mainly converting between barren land and grassland. The reduction in water areas was primarily due to cropland encroachment, followed by construction land expansion, both contributing over 90.00% to the water area reduction. Barren land primarily converted to grassland or resulted from grassland degradation, continuing the overall reduction trend (Table 5 and Figure 7).

Over a 12-year period, the intensity of the major land use conversions has increased, showing a spatiotemporal differentiation characterized by an east–west divergence (Figure 8 and Figure 9). Specifically, the conversion of cropland to construction land followed a distribution pattern of “high in the east, low in the central and western regions”, with economically developed eastern provinces such as Shandong, Hebei, Henan, Zhejiang, and Jiangsu experiencing much greater scales of cropland conversion compared to the less economically developed western provinces like Gansu, Yunnan, Guizhou, and Guangxi. Conversely, the mutual conversion between cropland and woodland or grassland followed a pattern of “high in the central and western regions, low in the east”, predominantly occurring in provinces with abundant forest and grassland resources such as Gansu, Sichuan, Guizhou, Yunnan, Guangxi, Hunan, and the northeastern province of Heilongjiang. Over time, the scale of mutual conversion between cropland and woodland/grassland has significantly increased. However, due to ecological protection policies, the overall scale of woodland and grassland being converted to cropland is slightly lower than the scale of converting cropland back to woodland and grassland.

3.3. Influencing Factors of LUT

3.3.1. Validation of Spatial Econometric Models

Referring to the existing research, the study employs the spatial econometric regression analysis method to analyze the correlation between major land use conversions and driving factors within the study area. To avoid bias in the results caused by interactions among the various sub-variables of household livelihood capital and socio-economic characteristics, the study conducts multicollinearity tests on all explanatory variables. The results indicate that the variance inflation factors (VIFs) among sub-variables are all below 10, suggesting that there is no multicollinearity problem among the variables.

The results of the ESDA (

Table 6. Global Moran’s I test for major land use conversions.

Year	Cropland–Construction Land		Cropland–Woodland and Grassland		Woodland and Grassland–Cropland
	Moran’s I	p	Moran’s I	p	Moran’s I	p
2010	0.3139	0.0126 **	0.6367	0.0000 ***	0.5200	0.0001 ***
2012	0.4111	0.0014 **	0.6686	0.0000 ***	0.5057	0.0002 ***
2014	0.3018	0.0132 **	0.5979	0.0000 ***	0.4930	0.0002 ***
2016	0.3416	0.0064 ***	0.7288	0.0000 ***	0.6445	0.0000 ***
2018	0.2983	0.0154 **	0.6375	0.0000 ***	0.4738	0.0003 ***
2020	0.2625	0.0097 ***	0.6169	0.0000 ***	0.4015	0.0019 ***
2022	0.3366	0.0066 ***	0.6295	0.0000 ***	0.5521	0.0001 ***

Note: *** and ** indicate statistical significance at 1% and 5% levels.

) indicate that the Moran’s I for the major land use conversions in the study area from 2010 to 2022 are consistently above zero and statistically significant. This suggests that changes in the principal land use conversions across each province, municipality, and autonomous region exhibit a notable spatial positive correlation. Moreover, the significant Moran’s I (error) index from the OLS regression model suggests a potential bias or invalidity in the model’s estimation results. To investigate the impact of livelihood capital on the changes in major land use conversions, it is crucial to construct a spatial econometric model that considers spatial spillover effects and accounts for endogeneity due to the omitted variable bias and reverse causality. The first step is to resolve endogeneity issues stemming from the spatial dependence of the dependent variable. The presence of spatial lag terms implies that using least squares for the spatial model estimation can lead to inconsistent parameter estimates [79]. While a maximum likelihood estimation can address the endogeneity linked to spatial lag terms [80], it may result in biased parameter estimates for models with fixed effects. To mitigate this, we employ Quasi-maximum Likelihood Estimators (QML) to achieve consistent parameter estimates in the spatial fixed-effects model [81]. Furthermore, addressing the potential reverse causality between livelihood capital and land use change is essential. The use of lagged endogenous variables instead of current values, as recommended by the existing studies, helps control endogeneity [82,83]. In the regression analysis phase, replacing explanatory variables with their lagged values helps alleviate the endogeneity problem caused by the bidirectional causality between livelihood capital and land use change.

To determine the specific estimation form of the spatial econometric model, this study conducts the Lagrange Multiplier (LM), the Likelihood Ratio (LR), and the Wald tests on traditional OLS estimates (

Table 7. Spatial panel econometric model test results.

Test	Cropland–Construction Land	Cropland–Woodland and Grassland	Woodland and Grassland–Cropland
LM-Lag	24.789 ***	97.387 ***	60.990 ***
LM-Error	12.602 ***	47.769 ***	23.331 ***
Robust LM-Lag	13.677 ***	59.870 ***	57.755 ***
Robust LM-Error	1.490	10.252 ***	20.095 ***
LR Ind	46.640 ***	19.610 ***	26.180 ***
LR Time	207.880 ***	94.910 ***	115.55 ***
Wald (SAR)	40.750 **	71.520 ***	60.810 ***
Wald (SEM)	30.510 **	41.270 ***	68.640 ***

Note: *** and ** indicate statistical significance at 1% and 5% levels.

). The results from the LM and Robust-LM tests for the spatial lag and spatial error lag terms suggest that the econometric model includes spatial effects, thereby allowing the use of a more general SDM estimator for spatial econometric estimation. Due to distinct provincial conditions and the presence of time-invariant omitted variables, the use of a fixed-effects model is warranted. Considering the short time dimension of the panel data and LR test results, an individual fixed-effects model is selected to avoid potential multicollinearity issues and ensure stable estimates [84,85]. The Wald test results further suggest that the SDM estimator will not degenerate into SAR or SEM estimators. Additionally, model fit indicators such as R-squared, Log likelihood (Log-L), Akaike information criterion (AIC), and Schwarz criterion (SC) are compared, with a better model fit indicated by larger Log-L and smaller AIC and SC values (

Table 8. Estimation of OLS regression and Spatial Durbin Model.

Variable	Cropland–Construction Land			Cropland–Woodland and Grassland			Woodland and Grassland–Cropland
	OLS	SDM	QML	OLS	SDM	QML	OLS	SDM	QML
	Model 1	Model 2	Model 3	Model 5	Model 6	Model 7	Model 9	Model 10	Model 11
R2	0.556	0.633	0.633	0.589	0.826	0.826	0.556	0.822	0.822
Log-L	−289.723	−143.357	−133.976	−592.436	−434.474	−383.505	−594.275	−460.035	−405.415
AIC	625.447	378.713	359.952	1230.872	960.947	859.01	1234.550	1012.071	902.830
SC	697.298	522.416	496.563	1302.723	1104.65	995.621	1306.401	1155.773	1039.441
Obs.	168	168	168	168	168	168	168	168	168

). Compared to the OLS model results, the individual fixed-effects model, accounting for time-invariant factors such as elevation and slope, exhibits superior Log-L, AIC, and SC values. The individual fixed-effects model based on the QML method further enhances the traditional SDM estimator, validating the effectiveness of the spatial econometric model. In conclusion, this study employs the QML method to estimate the individual fixed-effects model, revealing the spatial effects of each explanatory variable on land use change.

3.3.2. Spatial Durbin Model Results

According to the regression results of livelihood capital on major land use conversions under the SDM, the direct effects and spatial lag terms of various influencing factors demonstrate that livelihood capital impacts LUT (

Table 9. Estimation results of the Spatial Durbin Model of major land use conversions.

Variable	Model 3		Model 4		Model 7		Model 8		Model 11		Model 12
	Main	Wx	Main	Wx	Main	Wx	Main	Wx	Main	Wx	Main	Wx
N1	1.122	1.034	1.004	0.115	−7.061	5.551	−18.128 *	2.007	5.507	21.818	−7.168	31.446
N2	−7.732	−10.799	6.591	−8.420	−46.547	1.610	6.325	20.524	18.875	−203.933 *	−7.121	−5.976
N3	7.418	8.594	16.274	−45.758 **	0.359	−115.972	51.203	−15.634	−24.318	28.041	78.867	−420.414 ***
H1	−14.445 **	−10.997	−26.259 ***	−14.640	−33.827	46.605	28.872	−62.487	3.268	150.395 **	−5.507	319.340 ***
H2	0.980	5.368	8.051	−1.750	−5.158	135.033 *	−53.880	457.677 ***	4.834	133.025	−88.573 *	233.795 **
H3	−6.559	−5.812	−14.404 ***	−59.016 ***	−30.599	−109.064 *	−53.608 *	−72.934	−54.115 *	−215.481 ***	43.840	−134.540
M1	3.297	−9.264	1.230	7.596	107.994 ***	99.004	112.602 **	146.558	53.736	158.295	210.026 ***	−141.840
M2	−1.339	5.572	−1.142	−5.730	51.549 *	90.465 *	15.930	207.329 ***	59.377 *	183.512 ***	46.135	79.093
M3	8.707	−21.412	−6.742	0.803	42.682	−2.149	39.086	−8.019	39.010	−115.276	10.249	192.137 *
F1	−3.895 *	−0.605	0.178	23.712 ***	−10.293	22.980	−7.263	20.028	−6.028	−32.329	−20.122	−7.444
F2	−0.742	−5.690	−4.556	1.221	26.454	68.655 **	19.550	14.278	60.433 ***	90.334 *	2.709	60.787
S1	−17.320 **	18.463 **	−9.887	18.009	−181.783 ***	184.433 ***	−0.450	43.230	−74.452	77.342	−215.177 ***	210.736 ***
S2	0.015	1.346	4.361	−8.428 *	−3.030	−23.540	−14.593	−15.222	−6.389	−35.220	−44.674 **	−8.959
P1	−9.779 ***	−3.313	1.709	−2.082	43.125 **	43.558	42.378 *	−16.474	40.745 *	168.928 ***	56.888 **	−133.359 **
P2	3.241	3.538	2.712	−2.655	−11.573	6.503	−39.047 ***	21.917	−13.351	−13.990	−16.307	69.627 **
LS1	−3.854 *	−6.742 *	−3.323	0.805	−14.392	−41.525 **	−10.096	−12.860	−36.401 **	−41.670	−8.805	−43.295
LS2	0.726	0.535	4.008 *	−7.531	18.189 *	−5.613	−12.239	8.536	−4.831	30.937	13.037	−25.419
LS3	−1.889	4.631 *	2.187	−5.003	4.297	−2.490	28.438 ***	9.771	12.422	−24.700	16.871	−3.712
$\rho$	0.489 ***		0.432 ***		0.747 ***		0.720 ***		0.476 ***		0.577 ***
CV	Yes		Yes		Yes		Yes		Yes		Yes
IND	Yes		Yes		Yes		Yes		Yes		Yes

Note: ***, **, and * indicate statistical significance at 1%, 5%, and 10% levels.

). The spatial autoregressive coefficients ($\rho$) for conversions from cropland to construction land, cropland to woodland and grassland, and woodland and grassland to cropland are significantly positive at the 1% significance level, regardless of whether the explanatory variables are lagged. Assuming other factors remain constant, a 1% increase in the conversion levels of neighboring provinces’ cropland to construction land, cropland to woodland and grassland, and woodland and grassland to cropland corresponds to increases of 0.489%, 0.747%, and 0.476% in the LUT levels within the province, respectively. This suggests that the three major land use conversions exhibit significant positive spatial spillover effects, whereby the expansion of major land use conversions in a province considerably increases the scale of land use change in neighboring provinces through geographical connections.

3.3.3. Model Effect Decomposition

Due to the presence of spatial lag terms, the regression coefficients of the Spatial Durbin Model cannot directly reflect the actual impact of the explanatory variables. According to LeSage and Pace [79] and Elhorst [80], this study uses the partial differentiation method to decompose the effects into direct, indirect, and total effects. The direct effect includes the direct impact of changes in local explanatory variables on the local dependent variable, as well as the feedback effect transmitted back from spatially related areas. The indirect effect reflects the spillover impact of the changes in local explanatory variables on dependent variables in spatially related areas. Robustness tests show that, regardless of whether explanatory variables are lagged by one period, the direction of influence of most livelihood capital indicators on the three major land use conversion types remains consistent, indicating that the direct effect results are relatively robust (

Table 10. Spatial Durbin Model effect decomposition of major land use conversions.

Variable	Model 3		Model 4		Model 7		Model 8		Model 11		Model 12
	Direct	Indirect	Direct	Indirect	Direct	Indirect	Direct	Indirect	Direct	Indirect	Direct	Indirect
N1	1.470	3.007	1.141	0.919	−6.469	2.544	−21.649	−36.707	10.119	44.515	−0.373	61.360
N2	−11.029	−28.540	5.155	−10.657	−64.493	−141.727	10.284	66.622	−18.343	−355.484 *	−13.949	−33.474
N3	11.048	24.580	12.249	−61.106 *	−37.308	−411.685	69.996	99.503	−12.635	37.108	2.651	−819.420 **
H1	−17.756 ***	−34.422	−30.240 ***	−44.621 *	−25.466	76.068	12.202	−156.421	29.948	275.801 *	67.473	700.134 ***
H2	2.263	12.159	8.473	3.939	47.825	494.018 *	98.324 *	1415.092 ***	27.992	241.668	−47.760	394.299
H3	−7.893	−16.995	−23.911 ***	−109.123 ***	−80.710 *	−483.898 **	−90.308	−358.183	−92.203 **	−427.063 ***	22.332	−228.667
M1	1.982	−13.201	2.427	14.160	179.325 **	673.143 *	196.799 **	787.052 *	84.669	331.145	206.164 ***	−35.125
M2	−0.308	12.103	−1.866	−6.726	105.100 **	510.788 **	98.076 *	774.296 **	95.364 **	392.450 ***	70.772	251.365
M3	5.965	−33.251	−6.625	−5.217	56.110	101.743	48.721	54.046	25.204	−180.344	58.869	429.710 *
F1	−4.387 *	−5.720	3.677	38.773 ***	−5.187	44.511	−2.333	44.121	−12.333	−70.253	−24.726	−47.457
F2	−2.021	−11.114	−4.942	−1.273	61.772 **	339.989 **	26.084	76.989	79.667 ***	216.786 ***	13.767	127.648
S1	−15.207 **	17.583 **	−7.437	22.160 **	−161.023 ***	172.783 ***	17.450	139.583 *	−64.989	70.499	−193.491 ***	179.445 **
S2	0.067	2.394	3.142	−10.950	−14.565	−99.190	−26.546	−96.206	−14.187	−69.662	−54.890 ***	−81.321
P1	−11.112 **	−14.913	1.664	−1.792	74.858 **	292.462	47.130	39.158	73.706 **	340.045 ***	34.135	−224.545
P2	3.993 *	8.902	2.265	−3.012	−13.854	−16.870	−43.156 **	−28.440	−17.819	−39.906	−4.031	129.488 *
LS1	−5.424 **	−16.574 *	−3.403	−1.101	−35.229 *	−197.406 **	−17.581	−70.232	−46.099 ***	−106.062 **	−19.973	−108.315
LS2	1.047	2.105	3.207	−9.415	22.067	31.176	−12.511	−4.089	0.645	51.155	9.002	−40.039
LS3	−1.300	6.462	1.547	−6.851	4.093	−0.581	38.907 ***	99.519	8.906	−35.159	17.649	9.959
CV	Yes		Yes		Yes		Yes		Yes		Yes
IND	Yes		Yes		Yes		Yes		Yes		Yes

Note: ***, **, and * indicate statistical significance at 1%, 5%, and 10% levels.

).

A spatial econometric regression analysis reveals significant differences in the direction and magnitude of household livelihood capital’s influence on the three primary land use transition types within the study area. Conversion from cropland to construction land is chiefly inhibited by human capital, financial capital, social capital, psychological capital, and the proportion of pure agricultural farmers, while social capital exhibits positive spillover effects and the proportion of pure agricultural farmers shows negative spillover effects. Conversion from cropland to woodland and grassland is predominantly promoted by physical, financial, and psychological capital, and inhibited by human and social capital; positive spillover effects arise from human, physical, financial, and social capital, whereas negative spillovers stem from the proportion of pure agricultural farmers. Conversion from woodland and grassland to cropland is largely driven by physical, financial, and psychological capital, impeded by human capital and the proportion of pure agricultural farmers, with positive spillover effects from human, physical, financial, and psychological capital, and negative spillover effects from natural capital and the proportion of pure agricultural farmers.

The factors influencing the three main land use transition types are analyzed as follows:

• The conversion from cropland to construction land

The Model 3 results indicate that the conversion from cropland to construction land is significantly related to human, financial, social, and psychological capital, as well as the proportion of pure agricultural farmers. It is negatively correlated with the household labor force share at a 1% confidence level, household deposits at a 10% confidence level, and household social status, life satisfaction, and the proportion of pure agricultural farmers at a 5% confidence level. Conversely, it is positively correlated with future confidence at a 10% confidence level. Regarding the spatial spillover effects, household social status shows a positive effect at a 5% confidence level, while the proportion of pure agricultural farmers shows a negative effect at a 10% confidence level. This suggests that, within the province, human and financial capital exert an inhibitory effect; psychological capital has both inhibiting (life satisfaction) and promoting (future confidence) influences; social capital inhibits within the province but promotes in neighboring regions; and the proportion of pure agricultural farmers has an inhibitory effect both within and beyond the province.

2.

The conversion from cropland to woodland and grassland

Model 7 results indicate that the conversion from cropland to woodland and grassland is significantly associated with human, physical, financial, social, and psychological capital, as well as the proportion of pure agricultural farmers. This conversion is negatively correlated with household social status at a 1% confidence level, and household health status and the proportion of pure agricultural farmers at a 10% confidence level, and positively correlated with the value of agricultural machinery, durable goods, household debt, and life satisfaction at a 5% confidence level. Regarding spatial spillover effects, the education status of the labor force and the value of agricultural machinery exhibit positive spillover effects at a 10% confidence level; the value of durable goods and debt level show positive spillover effects at a 5% confidence level; household social status presents a positive spillover effect at a 1% confidence level; and labor health status and the proportion of pure agricultural farmers display negative spillover effects at a 5% confidence level. This suggests that, for the conversion from cropland to woodland and grassland, human capital has an inhibitory effect within the province and exerts both an inhibitory effect (health status) and a promoting effect (education status) on neighboring provinces. Physical and financial capital promote both within the province and beyond, while social capital inhibits within the province but benefits neighboring regions. Psychological capital promotes within the province, and the proportion of pure agricultural farmers inhibits both within and beyond the province.

3.

The conversion from woodland and grassland to cropland

The Model 11 results reveal that the conversion from woodland and grassland to cropland is significantly influenced by natural, human, physical, financial, and psychological capital, along with the proportion of pure agricultural farmers. This conversion is negatively correlated with household health status at a 5% confidence level and the proportion of pure agricultural farmers at a 1% confidence level, while positively correlated with household debt at a 1% confidence level, and with durable goods and life satisfaction at a 5% confidence level. In terms of spatial spillover effects, drinking water sources exhibit a negative impact at a 10% confidence level, household health status at a 1% confidence level, and the proportion of pure agricultural farmers at a 5% confidence level. Conversely, labor force share shows a positive impact at a 10% confidence level, whereas durable goods, household debt, and life satisfaction all demonstrate positive impacts at a 1% confidence level. This analysis suggests that, for the conversion from woodland and grassland to cropland, natural capital exerts an inhibitory effect on neighboring provinces, and human capital has an inhibitory effect within the province and exerts both inhibitory (health status) and promoting (labor force share) effects on neighboring provinces. Physical, financial, and psychological capital encourage development both within the province and beyond, while the proportion of pure agricultural farmers inhibits growth both locally and in the surrounding provinces.

3.4. Regional Heterogeneity Analysis

To thoroughly analyze the direct and spatial spillover effects of household livelihood capital on LUT, this study categorizes the research area into eastern and central and western regions based on urban geographical locations and economic development levels to explore their heterogeneous impacts. The results indicate that the influence of household livelihood capital on land use change varies across different regional classifications (

Table 11. Spatial Durbin Model results on regional heterogeneity of major land use conversions.

Variable	Cropland–Construction Land				Cropland–Woodland and Grassland				Woodland and Grassland–Cropland
	Eastern Region		Central and Western Region		Eastern Region		Central and Western Region		Eastern Region		Central and Western Region
	Direct	Indirect	Direct	Indirect	Direct	Indirect	Direct	Indirect	Direct	Indirect	Direct	Indirect
N1	11.503	−13.175	0.031	6.828 **	−2.173	15.715	6.973	2.850	18.608	−2.107	5.335	59.605
N2	−157.742 ***	48.388	−5.268	−24.798 *	77.488	332.869 **	−77.964	−310.165	−36.057	385.552 ***	14.682	−626.277 ***
N3	187.191 ***	−47.826	4.792	13.744	−159.493	−422.580 ***	10.631	−218.452	−9.466	−488.541 ***	234.485 **	−30.355
H1	33.281	44.985	−20.088 ***	−7.059	−136.615 **	−145.809	88.048	101.913	−112.769 **	−128.170 **	63.905	100.277
H2	1.313	−23.512	−0.003	−3.662	85.578 *	241.358 *	251.443 ***	1035.544 **	76.045 **	95.710	−11.482	125.728
H3	2.593	−21.049	−2.509	10.653	−43.276 *	−147.499 **	−96.205	−162.834	13.478	−77.438 **	−148.534 *	−241.203
M1	−56.696	18.983	−5.048	7.022	263.204 ***	304.342 *	118.170	129.967	20.622	154.221	85.021	−16.327
M2	−17.485	−21.094	6.514	25.316 **	40.503	76.423	137.334	449.447	17.932	72.848	0.235	256.241
M3	5.309	−37.955	8.215	4.150	25.712	123.427	97.284	147.636	39.211	12.982	78.232	−356.535
F1	−3.625	−21.380 *	−0.088	−6.088	12.769	37.605	−40.170	−132.190	16.468 **	16.638	11.534	−42.817
F2	−18.361 **	42.813 *	−1.872	1.626	48.093 **	71.014	112.335 **	268.063	18.479	116.950 ***	94.037 **	144.231
S1	−35.854 **	37.135 **	−14.159 *	14.394 *	8.040	27.646	−221.402 ***	161.044 *	−15.368	34.509	−28.271	33.841
S2	−7.501	12.681	1.787	−5.519	−21.182	−5.783	3.382	86.819	−16.814	6.844	−29.165	−9.957
P1	−39.141 ***	14.233	1.071	−0.036	1.131	−47.935	13.276	45.324	3.443	57.846 **	93.791	419.324 ***
P2	16.835 ***	−3.488	2.841 *	1.884	1.873	45.085 *	−10.263	52.152	2.071	−4.712	−9.252	−24.547
LS1	30.155	25.113	−0.406	−9.282 ***	−30.502	−74.362	−69.416 **	−193.453 *	−10.797	−78.224 **	−48.878 **	−145.472 ***
LS2	10.726	59.190 **	−3.344 *	−13.834 ***	17.972	32.316	59.725	189.951	−16.414	0.995	14.027	104.821
LS3	−16.215 ***	−8.822	1.995	7.803 ***	18.482 *	27.549	−3.649	−42.478	−4.738	25.702 **	24.491	−94.666 **
CV	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Obs.	63		105		63		105		63		105
R2	0.855		0.850		0.910		0.838		0.931		0.878
Log−L	−41.289		−29.630		−87.187		−250.816		−74.488		−258.785
AIC	262.579		239.261		354.373		681.632		328.976		697.570
SC	441.587		464.244		533.382		906.615		507.984		922.552

Note: ***, **, and * indicate statistical significance at 1%, 5%, and 10% levels.

).

In the eastern region, the conversion from cropland to construction land is significantly inhibited by drinking water sources, household debt status, household social status, life satisfaction, and the proportion of II-part-time agricultural farmers. Conversely, domestic fuel and confidence in future life have significant promotive effects. In the central–western region, labor force share, household social status, and the proportion of I-part-time agricultural farmers significantly inhibit conversion, while confidence in future life has a promotive effect. Therefore, natural capital, financial capital, and the proportion of II-part-time agricultural farmers significantly inhibit conversion in the eastern region but are not significant in the central–western region. The inhibitory effect of social capital and the promotive effect of confidence are stronger in the eastern region compared to the central–western region. Meanwhile, human capital and the proportion of I-part-time agricultural farmers have significant inhibitory effects in the central–western region but are not significant in the eastern region. Overall, the central–western region is less affected by spillover effects from neighboring provinces. Additionally, in the eastern region, the spatial spillover effect of household deposits is significantly negative, while debt status, household social status, and the proportion of pure agricultural farmers have significantly positive spillover effects. In the central–western region, the spatial spillover effects of drinking water sources, the proportion of pure agricultural farmers, and the proportion of I-part-time agricultural farmers are significantly negative. In contrast, land assets, the value of durable goods, household social status, and the proportion of II-part-time agricultural farmers have significantly positive spatial spillover effects.

Concerning the conversion from cropland to woodland and grassland, in the eastern region, the education status of the labor force, agricultural machinery, debt status, and the proportion of II-part-time agricultural farmers significantly promote this conversion within the province, whereas labor force share and health status significantly inhibit it. In the central–western region, the education status of the labor force and debt status significantly enhance the conversion, while household social status and the proportion of pure agricultural farmers notably restrict it. Overall, in the eastern region, human capital, physical capital, and the proportion of II-part-time agricultural farmers have a more significant impact on conversion compared to the central–western region. Conversely, in the central–western region, social capital, the proportion of pure agricultural farmers, and financial capital exert a stronger influence compared to the eastern region. Additionally, in the eastern region, the spatial spillover effects of drinking water sources, education status of the labor force, value of agricultural machinery, and confidence in future life are significantly positive, whereas the spatial spillover effects of labor health status and domestic fuel are significantly negative. In the central–western region, the spatial spillover effect of household social status is significantly positive, the positive spatial spillover effect of the education status of the labor force is stronger than that in the eastern region, and the spatial spillover effect of the proportion of pure agricultural farmers is significantly negative.

Regarding the conversion from woodland and grassland to cropland, in the eastern region, the education status of the labor force and household deposits significantly promote this conversion, while labor force share significantly inhibits it. In the central–western region, domestic fuel and debt status significantly promote conversion, while the health status of the labor force and the proportion of pure agricultural farmers significantly inhibit it. Overall, natural capital and social structures in the central–western region show significant effects, in contrast to the eastern region. Human capital and financial capital have significant effects in both regions, with financial capital and labor force quality acting as promotive factors, and labor capacity acting as an inhibitory factor. Additionally, in the eastern region, the spatial spillover effects of the drinking water source, debt status, transport and communications expenditure, life satisfaction, and the proportion of II-part-time agricultural farmers are significantly positive, while the spatial spillover effects of domestic fuel, labor force share, health status, and the proportion of pure agricultural farmers are significantly negative. In the central–western region, the spatial spillover effect of the confidence status in future life is significantly positive, while those of the drinking water source, the proportion of pure agricultural farmers, and the proportion of II-part-time agricultural farmers are significantly negative, and the spatial spillover effect of the drinking water source in the central–western region is the opposite of that in the eastern region.

4. Discussion

With the rapid economic growth and changes in the social environment, drastic fluctuations in household livelihood capital levels have led to significant changes in production decisions, essentially affecting land use behavior. A quantitative analysis of household livelihood capital and its influence on LUT is crucial for the sustainable use of land resources. Based on the relationship between livelihood capital and LUT, this paper quantitatively examines the spatiotemporal evolution of household livelihood capital and LUT across 24 provincial-level administrative units in China from 2010 to 2022, as well as the mechanism by which household livelihood capital affects LUT. This analysis provides a scientific foundation for policy formulation on regional sustainable development, territorial spatial planning, and land development and ecological management. Given the research scope, the driving mechanism of LUT influenced by household livelihood capital is discussed as follows:

• Household livelihood capital demonstrates spatiotemporal evolution with annual increases, reflecting both regional differences and a tendency toward equilibrium.

China has consistently prioritized issues related to “agriculture, rural areas, and farmers” as a primary focus of the party’s agenda, promoting the prioritization of agricultural and rural development. The implementation of rural revitalization strategies, poverty alleviation policies, and other measures has resulted in continuous improvements in household livelihood levels, as evidenced by the overall upward trend in household livelihood capital [38,86,87]. Additionally, notable structural differences and regional imbalances exist within household livelihood capital. Psychological, natural, and human capital dominate the livelihood capital structure. The level of household livelihood capital is higher in the eastern region compared to the central and western regions, consistent with previous research findings [71,88]. However, in the long term, the relative advantage of the eastern region has been diminishing annually, while the central and western regions have shown a clear catch-up trend. With the rise of the central region and the implementation of the western development strategy, industries, policy resources, and other elements have shifted toward the central and western regions, promoting the rapid growth of rural livelihood capital in these areas. Chinese household livelihood capital is beginning to move toward regional balanced development [89].

2.

Household livelihood capital and livelihood strategies demonstrate differentiated driving effects on LUT.

Natural capital has no significant direct effect on provincial LUT, but has a negative spatial spillover effect on the conversion from woodland and grassland to cropland. Water resources, as a key resource endowment for agricultural production, are interdependent with cropland. Water resource constraints can drive the permanent abandonment of agricultural land and habitat restoration [90]. In China, the utilization efficiency of cropland and water resources exhibits spatial clustering, but competitive effects can sometimes lead to inter-regional exclusion [91]. When neighboring provinces have similar water resource endowments, favorable water conditions generate positive ecological benefits for the surrounding environment, thereby reducing the likelihood of converting woodland and grassland to cropland.

Human capital has an inhibitory effect on provincial LUT. The education status of the labor force and labor force share have positive spatial spillover effects on the conversion between cropland and woodland and grassland, while the health status of the labor force shows a negative spatial spillover effect. In addition to providing an adequate labor force for the local region, human capital may also supply cheap labor to neighboring provinces, significantly influencing households to maintain the current state of cropland use [29], thereby inhibiting the expansion of construction land in both the province and neighboring provinces. High-quality labor, as a crucial resource in constructing ecological civilization, generates spillover effects in neighboring provinces during cross-regional flow [92], promoting the expansion of conversion from cropland to woodland and grassland. However, when households face health shocks, high financial vulnerability and capital risks [88] drive both provincial and neighboring households to engage in conservative agricultural economic activities on cropland [88,93], increasing the scale of conversion from woodland and grassland to cropland. Conversely, good health may encourage households to engage in higher-value-added agricultural economic activities, thus reducing the scale of conversion from woodland and grassland to cropland.

Physical capital significantly promotes the conversion between cropland and woodland and grassland, exhibiting positive spatial spillover effects. The level of agricultural mechanization directly influences land use patterns [94] by replacing agricultural labor with labor-saving technologies, thereby freeing up labor resource [95]. This reduction in farmers’ dependence on agricultural income enhances livelihood stability [96] and facilitates the continuation of agricultural operations or participation in “Grain for Green” initiatives [29]. The pursuit of ecological civilization and sustainable development has become a key component of government officials’ performance evaluations. In implementing “Grain for Green,” local governments engage in policy competition, fostering constructive interactions [97,98]. These projects drive technological upgrades and industrial transformation, resulting in physical capital accumulation and establishing regional development demonstration effects [99]. Additionally, the accumulation of physical capital and technological progress enable technical demonstration and diffusion through cross-regional agricultural machinery operations [100,101], further promoting “Grain for Green” in neighboring provinces via spatial spillover effects.

Financial capital discourages the conversion of cropland to construction land while encouraging the exchange between cropland and woodland and grassland, with added positive spatial spillover effects. Households with substantial financial capital generally maintain higher savings and lower debt levels, which decreases their dependency on land acquisition compensation. This, in turn, reduces the incentive for local governments to expand construction land through land finance. Additionally, these households have enhanced capital allocation capabilities, making the cultivation of economic forests or livestock farming a rational choice due to their comparative returns [102,103,104]. Such efforts can further adjust planting structures towards forestry and animal husbandry, boosting farmers’ income [105,106]. This influence extends beyond local areas, creating a demonstrative effect across provinces, thereby facilitating the conversion of cropland to woodland and grassland both within and beyond the province. Conversely, financial capital accumulation and its inter-provincial mobility can also drive deforestation and reclamation within the province and neighboring areas. While “Grain for Green” offers favorable ecological benefits in the long term, short-term economic incentives may lead farmers to clear forests for agricultural expansion [60,107,108], thereby increasing the scale of conversion from woodland and grassland to cropland.

Social capital hinders the conversion of cropland to construction land and to woodland and grassland, while promoting positive spatial spillover effects. Farmers’ land use decisions are not purely driven by economic rationality, but are heavily influenced by social relationships. According to status-power theory, farmers adopt certain land use practices to appease their reference groups (such as family, community, or government) [109]. Consequently, farmers or village elites with high social status are unlikely to deviate from the social norm that views cropland as the “lifeline of farmers,” as they aim to protect their social reputation; thus, they prefer maintaining traditional land use. Furthermore, local elites, often key beneficiaries and protectors of the current land system, possess strong incentives and capabilities to resist changes like reforestation that could disrupt their vested interests [110]. However, in the context of inter-provincial interactions, converting cropland to construction land or to woodland and grassland signifies land use transitions with potential economic or ecological advantages. Farmers with high social capital can engage in sharing experiences, acquiring technologies, and securing preferential access to policy resources through inter-regional social networks [111]. When conditions are favorable, their technical and management expertise can spill over [112], reducing barriers to transition and boosting confidence.

Psychological capital exerts both inhibitory and promotive effects on the conversion from cropland to construction land and facilitates the mutual conversion between cropland and woodland and grassland, with a positive spatial spillover effect on the conversion from woodland and grassland to cropland. Higher life satisfaction suggests that farmers lack the motivation and willingness to change the status quo, and land attachment will inhibit the conversion from cropland to construction land [113]; conversely, stronger confidence in the future reflects an optimistic and positive attitude toward change, and farmers react more positively to new information, tending to raise their expectations of future returns [114], thereby promoting the conversion. As people increasingly seek better ecological and living environments, life satisfaction may encompass the recognition of long-term ecological benefits. Existing studies have shown that subjective well-being can effectively enhance farmers’ pro-environmental behaviors [115] and significantly affect their willingness to participate in an agricultural environmental program [116], prompting farmers to sacrifice short-term economic gains for “Grain for Green” while also showing a high acceptance of deforestation. An optimistic mindset equips farmers to overcome adversity and raise income expectations, spreading to neighboring provinces through inter-provincial mobility, thereby promoting deforestation in neighboring provinces.

In livelihood strategies, the proportion of pure agricultural farmers hinders land use transition and contributes to negative spatial spillover effects. Since cropland is the most vital production resource for agriculture, pure farming households depend heavily on it for their livelihoods [117]. Therefore, they tend to protect cropland and sustain its current productivity [29], thus preventing the loss of cropland through conversion to construction land and woodland or grassland. Furthermore, because pure farmers generally prefer maintaining the status quo and are risk-averse [118], they oppose land use practices such as establishing new cropland, which could disrupt existing livelihood models, thereby suppressing woodland and grassland conversion to cropland. Their cautious stance towards altering land use spreads to neighboring provinces through inter-provincial population movements, thereby imposing negative spatial spillover effects on LUT.

3.

Household livelihood capital and livelihood strategy have spatial heterogeneity in their driving effects on LUT.

In the eastern region, high urbanization levels complicate land use decisions. The conversion from cropland to construction land is primarily driven by natural, financial, social, and psychological capital. Each province demonstrates a “co-opetition effect” during its development process [119], striving for leadership or avoiding lagging in competition with neighboring provinces while transforming its advantageous livelihood capital factors into development drivers for nearby areas. The advanced economic level also provides a robust foundation of human, physical, and psychological capital for grain-for-green initiatives [120], allowing the province’s livelihood capital to exert spillover effects and be significantly influenced by neighboring provinces’ overlay effects. In the central–western region, lower economic development levels result in insufficient endogenous momentum for urban development. The expansion of construction land is primarily suppressed by natural capital and agricultural production, whereas physical capital, social capital, and non-agricultural household employment play a promotive role. Development is constrained by economic development levels [121], basic productive factors [122], and social structure path dependence [123]. The central–western region is more influenced by improvements in labor quality than the eastern region, and, together with physical capital endowment, demonstrates a spillover effect on neighboring provinces. Additionally, due to the imbalanced spatial and temporal distribution of water resources in China, the central–western region has lower water resource reserves than the eastern region, making drinking water sources a crucial factor restricting “cropland reclamation” in the central–western region [124,125]. Furthermore, deforestation for cropland reclamation contradicts the principles of China’s ecological civilization construction. As the country enters an accelerated phase of green and low-carbon development, greater emphasis is placed on ensuring that ecological civilization construction and food security advance concurrently. As a crucial aspect of government performance evaluation across regions, the governance of cropland reclamation involves policy interactions among regional governments [97,98]. Consequently, natural capital, human capital, and social structures play a significant role in curbing cropland reclamation in both the eastern and central–western regions.

This study proposes the following policy recommendations. First, enhance the overall level of household livelihood capital by strengthening physical, financial, and social capital, and reduce pure farming households’ dependence on cropland. Second, utilize the spatial spillover effects of human capital by facilitating the dissemination of ecological farming techniques through cross-provincial skill training. Third, enhance ecological compensation and financial support mechanisms, innovate green financial products, and curb deforestation. Fourth, implement differentiated regional policies: promote intensive land use in the eastern regions while prioritizing drinking water safety and strictly controlling new cropland in the central and western regions. Fifth, strengthen the performance assessments of ecological civilization and establish coordinated provincial ecological protection redlines and cross-regional compensation linkage mechanisms.

This paper examines the spatiotemporal differentiation patterns of household livelihood capital and LUT, as well as the mechanism through which livelihood capital influences LUT. It reveals that LUT is affected by multidimensional livelihood capital and exhibits heterogeneous effects across different regions. However, certain research limitations persist: First, human social systems and natural environmental systems collectively form an interacting macro-system. This paper only addresses the unidirectional driving mechanism of livelihood capital on LUT and lacks insights into their interaction. Second, there is an objective necessity to study household livelihood capital levels and LUT from a multi-scale perspective. Due to data availability limitations, this paper is confined to a provincial-level analysis. Future research should enhance the objective understanding, theoretical exploration, and investigation of the interaction mechanisms between livelihood capital and LUT from a multi-scale spatiotemporal perspective.

5. Conclusions

This study indicates that LUT is a socio-economic process significantly influenced by household livelihood capital and involves complex spatial interactions. Different types of household livelihood capital exert varying direct and spatial spillover effects on LUT, exhibiting regional heterogeneity. The specific conclusions are as follows:

From 2010 to 2022, the level of household livelihood capital in the provincial-level administrative units of the study area demonstrated an upward trend. Household psychological capital, natural capital, and human capital levels were relatively high, while social capital, financial capital, and physical capital levels were relatively low. Although the eastern region’s livelihood capital level is higher than that of the central and western regions, its relative advantage diminishes yearly. The central and western regions show a distinct catch-up trend, with exceptionally high values concentrated in eastern coastal provinces like Jiangsu and Zhejiang, while Chongqing’s livelihood capital consistently remains at a low national level.

Between 2010 and 2022, the extents of cropland, shrub, grassland, water, snow/ice, wetland, and barren land have decreased, whereas the extents of construction land and woodland have increased in the study area. LUT is predominantly characterized by the conversion from cropland to construction land, from cropland to woodland and grassland, and from woodland and grassland to cropland. The intensity of LUT has intensified over time, exhibiting a west-to-east spatial gradient. The conversion from cropland to construction land reveals a spatial differentiation pattern of “high in the east, low in the central–west,” while the mutual conversion between cropland and woodland or grassland reveals a pattern of “high in the central–west, low in the east.”

LUT is influenced by household livelihood capital, and the direct and spatial spillover effects generated by different types of livelihood capital vary. Natural capital demonstrates a negative spatial spillover effect on the conversion from woodland and grassland to cropland. Human capital inhibits provincial LUT, with labor force education status and labor force share showing positive spatial spillover effects for the mutual conversion between cropland and woodland and grassland, while labor health status has negative effects. Physical capital and financial capital promote the mutual conversion between cropland and woodland and grassland exhibiting positive spatial spillover effects, while financial capital inhibits the conversion from cropland to construction land. Social capital inhibits the conversion from cropland to construction land and promotes the conversion from cropland to woodland and grassland, with positive spatial spillover effects on both. Psychological capital has both inhibitory and promotive effects on the conversion from cropland to construction land, promotes the mutual conversion between cropland and woodland and grassland, and demonstrates a positive spatial spillover effect on the conversion from woodland and grassland to cropland. Regarding the livelihood strategy, the proportion of pure agricultural farmers inhibits LUT and shows a negative spatial spillover effect.

Livelihood capital’s impact on LUT exhibits significant regional heterogeneity. In the eastern region, construction land expansion is primarily inhibited by the natural, financial, and social capital of adjacent provinces, whereas psychological capital encourages it. In contrast, the central–western region’s expansion is driven by physical capital, social capital, and non-agricultural household employment. Labor force quality and labor capacity, respectively, facilitate and constrain the mutual conversion between cropland and woodland and grassland across both regions. In the eastern region, the spatial spillover effects of livelihood capital are more pronounced, influencing conversion from cropland to woodland and grassland by overlay effects from neighboring provinces. Conversely, in the central–western region, financial capital promotes this conversion. The conversion from woodland and grassland to cropland in the central–western region is strongly encouraged by natural capital and discouraged by pure agricultural farmers. Furthermore, the scarcity of water resources imposes a negative spatial spillover effect, distinguishing the central–western region from the eastern region.