Exploring the Effects of Farmland Transfer on Farm Household Well-Being: Evidence from Ore–Agriculture Compound Areas in Northwest China

Abstract

Due to long-term interactions between intensive resource exploitation and rapid social development, there are multiple challenges to maintaining and improving the well-being of farm households in ore–agriculture compound areas in Northwest China. However, few studies have focused on the effects of farmland transfer on farm household comprehensive well-being. This study collected 485 valid questionnaires through a structured questionnaire technique and then compared the well-being level and its five components between farm households who participated and did not participate in farmland transfer based on an index system of well-being. Further, a propensity score matching (PSM) method was used to estimate the net effects of farmland transfer on farm household well-being and its heterogeneity. The results showed the following. (1) Overall, farm household well-being in ore–agriculture compound areas in Northwest China was at a moderate level (mean value was 0.433), but there were large differences among its five components. The orders of the five components of well-being in the three study sites were consistent, and the well-being index of farm households participating in farmland transfer was generally greater than that of those not participating in farmland transfer. (2) The results of the PSM revealed that farmland transfer only increased the levels of well-being, security, and freedom of choice and action by 4.9%, 8.8% and 6.1%, respectively. (3) The younger the household heads and the higher their education levels, the greater the effects of farmland transfer on farm household well-being. Local government sectors should continue to improve their farmland transfer system and strengthen institutional innovation. Meanwhile, venerable groups’ well-being should be paid more attention in the process of farmland transfer.

1. Introduction

The 2030 Agenda for Sustainable Development adopted by the United Nations in 2015 set out 17 ambitious goals that provided a blueprint and guidance for global sustainable development [1]. It focused on vulnerable groups in the world and directly or indirectly involved in multiple aspects of human well-being, such as poverty reduction, hunger eradication, health well-being, education quality and equity [2]. By the end of 2023, 43% of the population was in rural areas and the world’s poverty rate was 8.5% [3]. Additionally, subsistence producers and small farm wage laborers in the rural areas of low-income countries represented over two-thirds of the global poor and food insecure populations [4]. The majority’s agricultural production is highly exposed to uncertain climate change, frequent natural disasters and market shocks, and their livelihoods are vulnerable. Therefore, how to improve farmers’ well-being has become one of the research hotspots in the world.

Farmland resources belong to the natural capital that farm households depend on. Farm households obtain agricultural products from their limited resources through agricultural production activities to maintain and even improve their well-being. In the increasingly changing natural ecosystem and socio-economic system, they need to constantly adjust their livelihood strategies. At present, diversified and non-farm livelihoods have become the main strategies to maintain or enhance farm household well-being [5,6,7,8,9].

In the 1990s, with the rapid development of urbanization and industrialization in China, farmers, as “rational economic men”, migrated to urban areas to maintain or improve their household well-being. On the one hand, this alleviated the contradiction of “people and farmland” in rural areas; on the other hand, the phenomenon of abandoned farmland gradually became prominent. The farmland abandonment rate in some areas even reached more than 30%. Therefore, it was urgent to reallocate farmland resources [10]. In this context, the Chinese government adopted farmland transfer as an effective means to achieve the dual objectives of effective farmland resource allocation and agricultural modernization in rural areas. China’s farmland transfer is based on the household contract responsibility system and means that farm households with farmland contract management rights transfer their farmland management rights (use rights) to other farm households or economic organizations [11]. However, farmland for farm households has a dual function of providing economic benefits and social security, because they both pursue the “maximization of farmland economic benefits” and have deep emotional dependence on their farmland. Therefore, it is necessary to clarify the effect mechanism of farmland transfer on farm household well-being, which could provide theoretical support for formulating more effective and feasible farmland transfer schemes.

This study focused on the ore–agriculture compound areas in Northwest China. Using household survey data, we quantitatively tested the effect of farmland transfer on the farm household well-being and then further explored the group disparities in the well-being effect of farmland transfer. Finally, we proposed countermeasures to improve farm household well-being under the existing farmland transfer models. It is significant to optimize existing farmland transfer system for local government.

2. Literature Review and Theoretical Analysis

The concept of well-being was initially proposed as an important supplement to gross domestic production (GDP) indicators. With the development of sustainability science, well-being has become the ultimate goal of social development [12]. However, scholars have not reached a consensus on the concept of well-being [13]. Up until now, it is generally accepted that well-being is a multidimensional and inclusive concept. It refers to people’s evaluations of life quality and life satisfaction [14,15], including both subjective and objective well-being [16,17]. According to existing studies, there is no a universal paradigm for assessing human well-being. Many scholars constructed an indicator system to assess human well-being based on Maslow’s hierarchy of needs and Amartya Sen’s feasible capability theory [18,19,20]. The influential Millennium Ecosystem Assessment (MA) proposed by the United Nations defined human well-being as “the basic material needs for a good life, health, security and safety, good social relations and choice and freedom” [14]. This human well-being assessment framework has been widely used in the construction of indicator systems for comprehensive human well-being assessment [21,22].

Farm households, the basic units in rural areas, are one of important objects in sustainability research. Farm household well-being research not only includes farm well-being assessment but also mainly involves the fields of ecosystem services and sustainable livelihoods. In the sustainable livelihood framework, farm household well-being, being often regarded as a livelihood outcome, had received considerable attention. The existing research showed that policies, environment, livelihood capitals and livelihood strategies all affected farm household well-being. Participation in high-return sectors such as trade or salaried jobs, for instance, could improve farm household well-being [6]. Cash income from fishery production increased farmers’ subjective well-being in Vietnam [23]. It was reported that agricultural production activities throughout the year had a significant impact on farm households’ life satisfaction. Specifically, farmland preparation and harvesting could significantly increase their life satisfaction, while weeding and irrigation significantly reduced life satisfaction [24].

In recent years, the Chinese government has issued and implemented a series of policies and strategies concerning agricultural and rural development to solve the issues of agriculture, rural areas and farmers in rural areas. In this context, Chinese scholars also focused on the well-being effects of policy implementation [20,25,26,27]. Some researchers focused on the well-being effects of farmland transfer, which mainly involved income, poverty reduction, equity and subjective well-being.

Specifically, some studies have confirmed a significant promoting effect of farmland transfer on household income [28,29]. Some studies further revealed the significant heterogeneity of income increase effects and income distributive effects among different groups, such as farmland renting-in and renting-out families [10,30]. Secondly, some research explored the fairness effect of farmland transfer, but there is still no unified conclusion. Generally, the metric of income disparity is regarded as one of the important indicators of fairness. Liu et al. [31] showed that farmland transfer-out expanded the income disparity while farmland transfer-in narrowed it, which simultaneously showed significant regional heterogeneity. Other studies reported that farmland transfer could promote conventional fairness and effectiveness by flowing farmland resources from land-rich and labor-poor households to land-poor and labor-rich households [32,33], considering the potential “stress renting” by poor landlords; however, fairness should be defined along a broader set of dimensions [33].

Additionally, a few researchers explored the effects of farmland transfer on farm households’ subjective well-being. Yuan et al. [34] estimated the current situation of subjective well-being of farm households participating in farmland transfer and then found that the subjective well-being of farm households transferring farmland out was higher and that different livelihood capitals shaped the subjective well-being of different types of farm households. However, Qiu et al. [35] pointed out that transferring-out farmland did not increase farm household subjective well-being if ignoring the transaction partner types, which strengthened the importance of the channel of investing in social capitals. Tong et al. [36] explored the impacts of farmland renting-in on farm households’ life satisfaction. Their results showed that renting-in farmland indirectly and significantly enhanced farm households’ life satisfaction by increasing their income. Further, they pointed out that farmland renting-in improved low-income households’ life satisfaction but restrained that of high-income households. Hu et al. [37] also revealed the significant improvement effect of farmland transfer on farm households’ subjective well-being and its heterogeneity in farm households and regional areas.

Few studies explored the well-being effect of farmland transfer after comprehensively evaluating farm household well-being. For example, Ma et al. [29] evaluated the well-being level of rural middle-aged and elderly people from the four dimensions of family asset status, quality of life, viability and social security, and they then verified the significant improvement effect of farmland transfer on their well-being. Huang et al. [26] explored the impact of farmland acquisition for rural tourism on landless farmers’ well-being and its group disparities after assessing their well-being based on the MA’s human well-being framework.

To sum up, studies on the well-being effects of farmland transfer are insufficient and the existing relevant studies show that the impacts of farmland transfer on farm household well-being are not clear. Therefore, it is necessary to further examine the well-being effect of farmland transfer by more case studies and then provide a theoretical reference for local governments to formulate more perfect farmland transfer systems.

Given the above, this study empirically investigated the effect of farmland transfer on farm household well-being and its heterogeneity. Before the exploration, an index system was constructed from both subjective and objective aspects, so that farm household well-being could be comprehensively evaluated in the study areas. Based on this, we assumed that farmland transfer could significantly improve farm household well-being. A conceptual framework for this study was proposed, which explained the improvement of farm household well-being (Figure 1). In the context of global climate change, farmers living in ore–agriculture compound areas in Northwest China with a fragile and sensitive ecological environment face more highly agricultural risks, such as more frequent extreme weather events and heat waves. They lead to the instability of agricultural production and then decrease farmers’ income or even cause deficits. On the other hand, with the rapid development of urbanization and industrialization in China, the dualistic urban–rural structure has become increasingly prominent. In ore–agriculture compound areas, local collieries and related enterprises provide employment opportunities for rural residents. According to the theories of “Rational smallholder” [38] and “Entity smallholder” [39], farm households are more likely to engage in non-farm activities (such as drivers and miners) rather than traditional agricultural production. Simultaneously, farmland expropriation compelled rural labors to engage in non-farm activities. This leads to large-scale abandoned farmland. Rural labor transfer and low agricultural production efficiency have led to rural decline, which hindered farm household well-being and rural sustainable development. Farmland transfer can promote the formation of new agricultural management entities and develop moderate-scale management to improve agricultural productivity and modernization [40,41] and ultimately increase farm households’ sense of gain and well-being. Moreover, it can also increase farm households’ income (farmland transfer-in and transfer-out), release more surplus rural labors and increase their non-farm employment time [42,43]. In addition, farmland transfer in the ore–agriculture compound areas chosen for this study was mainly organized by village committees or rural governments. They would also conduct village governance and provide perfect infrastructures to improve the living environment of rural residents. Based on the above impact mechanism of farmland transfer on farm households’ production and lives, our main objective was testing the improvement effects of farmland transfer on farm household well-being in ore–agriculture compound areas in Northwest China.

We also assumed that there were group heterogeneities in the farm household well-being effects of farmland transfer, because the existing literature had revealed the heterogeneities of farmland transfer from the aspects of family income, fairness, subjective well-being and life satisfaction [10,30,31,32,33]. Considering the importance of a household head to his or her family livelihood decision-making, this study tested the well-being effect heterogeneities of farmland transfer in terms of household heads’ age and education levels.

Our objectives are as follows: (1) comprehensively assessing and analyzing farm household well-being based on a composite index method; (2) exploring the net effects of farmland transfer on farm household well-being and its components by the PSM method, which eliminates sample self-selection bias through constructing a “counterfactual framework”; and (3) attempting to reveal the heterogeneity of the well-being effects of farmland transfer from the perspective of household head heterogeneity.

This study makes some contributions. Firstly, we used the human well-being framework proposed by the MA to comprehensively assess farm household well-being from both subjective and objective aspects. Based on this, the effects of farmland transfer on farm household well-being were evaluated. Secondly, we focused on the ore–agricultural compound areas in Northwest China. This region has certain particularities in terms of its geographical location and social economy. On the one hand, farm households’ agricultural production faces huge pressures due to the fragile and sensitive ecological environment and strong resource exploitation. On the other hand, they are supported by ecological protection policies and agricultural development policies [44,45]. Taking this region as a case study can examine the effectiveness of existing farmland transfer policies. Thirdly, we constructed a “counterfactual framework” through the PSM method to eliminate sample self-selection bias. Compared with other relevant studies, we tested the net effect of farmland transfer on farm household well-being.

The remainder of this study is organized as follows. The third section concerns the materials and data, including the study areas, the conceptual framework, data collection and methods. The fourth and fifth parts present the results and analysis, and the discussions, respectively. The last section shows the conclusions and implications.

3. Materials and Data

3.1. Study Areas

The ore–agriculture compound areas selected for this study are north of the Great Wall and on the southern edge of the Mu Us Desert, which belongs to the desert–loess transitional zone (shown in Figure 1). Administratively, the study areas belong to Yulin City in Northwest China, accounting for 36.7% of the total area of the whole city. The total population is about 2.82 million. Among them, the rural population is 1.17 million, accounting for about 41% of the total population. According to a report by the Yulin Municipal Government [46], the added value of the primary industry of the study area in 2022 was CNY 31.225 billion, and the disposable income of rural residents was CNY 16,956, which was higher than the city’s average level. Geographically, the selected ore–agriculture compound area is a typical ecologically sensitive zone characterized by high vulnerability and low resilience [47]. Although the selected study region has large areas and a sparse population, fertile arable land resources are very limited. The climate belongs to the temperate arid and semi-arid continental monsoon climate characterized by aridity, large temperature differences and frequent sandstorms. The average annual temperature is 8 °C~10 °C, and the annual rainfall is 200~450 mm, mostly from July to October each year [47]. The annual evaporation is 2000~3000 mm, which is 4~5 times the annual precipitation [48]. Local farmers make their living mainly by animal husbandry, migrant working and growing crops such as wheat, maize, potatoes, flax and small grains. Geologically, the region is in the Ordos Basin, which is rich in energy resources like coal, oil and gas [49] and plays a huge role in the national economic development. In the 1990s, coal resources began to be mined in the study area, and intense mining activities had a profound impact on rural areas, gradually forming ore–agriculture compound areas. Coal resource exploitation not only promoted the economic development of rural areas but also brought many ecological and survival problems, like available arable land resources reduction, environmental pollution, geological disasters, ground collapse, and domestic water shortage and pollution.

In the context of global climate change, the fragile ecological environment and rapid urbanization, the sustainable development of agriculture, rural areas and farmers in the study areas faced great challenges. In recent years, with the support of the targeted poverty alleviation strategy and rural revitalization strategy formulated by the Chinese central government, farmland transfer has become an important way for local governments to develop modern agriculture, construct beautiful and livable villages and improve farm household well-being, which has made a great progress.

3.2. Sampling and Data Collection

A multistage sampling technique was applied to obtain samples from the ore–agriculture areas in Northwest China. In the first stage, we selected three administrative units, Yuyang District, Shenmu City and Fugu County, because they are rich in coal reserves and have a long history of exploitation. In the second stage, we selected two or three townships with a coal mine at least in each city or county. As a result, a total of 7 townships were selected. Using the same method, we then chose two administrative villages nearby coal mines in every township. Finally, a total of 14 administrative villages were determined (shown in Figure 2). In the last step, 36 farm households from every village were selected as the respondents by a random sampling method. Face-to-face interviews with household heads were conducted using a structured questionnaire, and then focus group discussions (FGDs) with village cadres from the given villages were performed to understand the livelihood of local farmers and the development of local farmland transfer. Before the interviews, we trained the interviewers in our team and conducted a pre-test to adjust our interview content. The survey was conducted in August 2021 and a total of 504 questionnaires were distributed. After deleting the missing and wrong questionnaires, we collected 485 valid questionnaires, with an effective rate of 96%. The study areas selected by us lay in Yulin City in Shaanxi Province, with a population of 3.86 million. According to the method proposed by Dong [50], the calculated sample size of this survey can be between 384 and 960. Therefore, the number of farm households selected in our survey was reasonable.

The questionnaire mainly included three aspects: (1) a family’s basic information, such as the ages and educational levels of family members, family size, skills and health; (2) the farmland transfer situation and livelihood capital activities, such as housing area, durable goods and income; and (3) the farm household well-being and life satisfaction.

3.3. Measurement of Farm Household Well-Being

3.3.1. Indicator System

Given the definition of human well-being proposed by the MA, this study constructed the indicator system of farm household well-being (FHWB) from five aspects, including the basic material needs for a good life (BMN), health (HEA), security (SEC), good social relations (GSRs), and freedom of choice and action (FCA). There are currently three methods of indicator system construction: subjective well-being indicator system, objective well-being indicator system and comprehensive well-being indicator system [21]. This study used an integrated method of subjective and objective indicators to construct the indicator system of farm household well-being. Here, the subjective well-being indicators were mostly measured by farm households’ satisfactions, while the objective well-being indicators were usually the capital status possessed by farm households. Specific indicators were selected based on the available literature.

(1) The basic material needs for a good life (BMN). Considering the importance of income to farm household well-being, many studies listed income and material foundation as dimensions of well-being. Huang and Li [26] used income satisfaction, total family income and number of durable goods for living to reflect the material basis of landless farmers. Gao et al. [20] adopted income satisfaction and per capita income to measure the BMN when they constructed an indicator system of anti-poverty relocated households. Feng et al. [22] and Wang et al. [51] all integrated durable goods for living into the indicator system for measurement. Tang et al. [18] also selected housing structures to construct the dimension of the BMN. In view of this, this study selected per capita net income and income satisfaction to represent the economic status of farm households, and the per capita housing area, housing structure and living durable goods to represent the material basis of farmers.

(2) Health (HEA). Li et al. [19] selected the number of patients and health satisfaction to evaluate the health of ecological migrators. Huang and Li [26] used the proportion of medical expenditure and Engel’s coefficient to measure landless farmers’ health. Liang et al. [52] adopted physical health status and medical expenditure intensity when measuring farm household well-being in the mining area of the Loess Plateau. Wang et al. [51] measured farmers’ health well-being by medical service satisfaction, distance between home and hospitals, perception of medical expenditure and social insurance. Considering the data availability and studies area characteristics, this study selected medical expenditure and number of patients to measure the health well-being of farm households.

(3) Security (SEC). This dimension has a wide definition, which includes but is not limited to resident security, resource security, environmental security, ecological security, and personal property security. Liu et al. [53] measured residents’ SEC dimension in the Loess Plateau from the three aspects of resource security, environmental security and residential security. Huang and Li [26] chose public security satisfaction and living environment satisfaction. Feng et al. [22] and Wang et al. [51] also selected public security satisfaction. In view of the environmental pollution, ecological environment degradation and more migrant workers in the ore–agricultural compound areas, this study used public security satisfaction and residential environment satisfaction to measure SEC.

(4) Good social relations (GSRs). Ciftcioglu [54] believed that GSRs should include family cohesion and community cohesion. Xiong et al. [55] integrated not only family relationships and neighborhood relationships but also family burden and a sense of community integration into the GSR dimension. In this study, neighborhood trust was selected to reflect neighborhood relationships and external help in difficult times and cash gift expenditure was selected to reflect family relationships.

(5) Freedom of choice and action (FCA). FCA means that farm households can engage in satisfying and valuable livelihood activities based on coping with current or future livelihood activities [14]. Ciftcioglu [54] defined it as “the ability to choose who, where, when and why to be”, including both economic freedom and personal development. Huang and Li [26] measured FCA by non-farm working times, household income diversity and public service satisfaction. Xiong et al. [55] measured it from the perspective of fairness and selected the two aspects of election fairness and subsidy fairness. Feng et al. [22] measured it from the two aspects of working and leisure, and rights and equity. Given the above, this study selected non-farming working time and income source diversity to reflect farm households’ economic freedom, and it applied public service satisfaction and frequency of participation in public affairs to measure their rights and equity.

Finally, a total of 17 indicators were identified to calculate farm household well-being. The indicators’ descriptions, properties and references used during construction of the indicator system are shown in

Table 1. The indicator system and their weights for farm household well-being.

Components (Weights)	Indicators	Variable Descriptions	Properties	Weights	References
The basic material needs for a good life (BMN) (0.171)	Per capita net income	Per capita net income of a farm household/10,000 yuan	+	0.484	[20,53]
	Income satisfaction	1 = very dissatisfied; 2 = dissatisfied; 3 = general; 4 = satisfied; 5 = very satisfied	+	0.146	[20,22,26,51]
	Durable goods	Number of durable goods owned by a farm household/piece	+	0.058	[22,51]
	Per capita housing area	Per capita living area of a farm household/m2	+	0.209	[18,53]
	Housing structure	1 = wigwam; 2 = civil house; 3 = brick house; 4 = concrete house; 5 = storied house	+	0.104	[18]
Health (HEA) (0.092)	Medical expenditure	Total annual medical expenditure of family members in a farm household	−	0.650	[26,51,52]
	Number of patients	Number of patients in a farm household	−	0.350	[19]
Security (SEC) (0.291)	Public security satisfaction	1 = very dissatisfied; 2 = dissatisfied; 3 = general; 4 = satisfied; 5 = very satisfied	+	0.217	[22,51,52]
	Residential environment satisfaction	1 = very dissatisfied; 2 = dissatisfied; 3 = general; 4 = satisfied; 5 = very satisfied	+	0.783	[22,26,53]
Good social relations (GSRs) (0.206)	Neighborhood trust	1 = very distrust; 2 = distrust; 3 = general; 4 = trust; 5 = very trust	+	0.066	[51,55]
	External help in difficult times	1 = very few; 2 = fewer; 3 = general; 4= more; 5 = a great many	+	0.292	[51]
	Cash gift expenditure	Cash gift expenditure of a farm household per year/yuan	+	0.642	[26]
Freedom of choice and action (FCA) (0.240)	Non-farming working time	Number of months for family labor to engage in migrant work or non-farm activities	+	0.390	[26,51]
	Income source diversity	Number of income sources owned by a farm household (mainly included 12 income sources such as wage income, operating income, property income and transfer income)	+	0.158	[26]
	Public service satisfaction	1 = very dissatisfied; 2 = dissatisfied; 3 = general; 4=satisfied; 5 = very satisfied	+	0.072	[26,51]
	Frequency of participation in public affairs	1 = very few; 2 = fewer; 3 = general; 4 = more; 5 = a great many	+	0.381	[22,51]

.

3.3.2. Calculating the Well-Being Index of Farm Households

Due to the differences in the orders of magnitude, dimensions and characters of the original indicators, the original data were normalized by the range standardization method. The standardization formula for positive indicators is:

$$Y_{ij}={\displaystyle \frac{\left(X_{ij}-X_{jmin}\right)}{\left(X_{jmax}-X_{jmin}\right)}}$$

(1)

The standardization formula for negative indicators is:

$$Y_{ij}={\displaystyle \frac{\left(X_{jmax}-X_{ij}\right)}{\left(X_{jmax}-X_{jmin}\right)}}$$

(2)

where $X_{ij}$ and $Y_{ij}$ are the original value and standardized value of i-th respondent among j-th index, respectively; and $X_{jmax}$ and $X_{jmin}$ are the max value and min value, respectively.

The next step was to calculate the weights of the indicator system. Compared with the subjective weighting methods, the entropy method, one of the objective weighting methods, is widely applied in social economy fields. This method can calculate the variable importance according to the “information” provided by each variable, so it would better reflect the effect value and reduce subjectivity [56]. The more information a variable reflects, the lower its entropy value and the greater its weight. Given this, we applied the entropy method to calculate the indicators’ weights in this study. The specific calculation formulas are:

$$P_{ij}={\displaystyle \frac{Y_{ij}}{\sum _{i=1}^m{Y}_{ij}}}$$

(3)

$$e_j=-k\sum _{i=1}^m{P}_{ij}ln{P}_{ij},k={\displaystyle \frac{1}{lnm}}$$

(4)

$$g_j=1-e_j$$

(5)

$$W_j={\displaystyle \frac{g_j}{\sum _{j=1}^n{g}_j}}$$

(6)

where i = 1, 2, …, m; j = 1, 2, …, n; $P_{ij}$ represents the proportion of the j-th indicator for the i-th study unit; $e_j$ donates the entropy value and $w_j$ is the objective weight of j-th indicator, where $g_j$ represents difference coefficient of the same indicator. It is to be noted that $P_{ij}$ should add a smallest unit to clear up the effect of 0 on the next step during the calculation.

We used the entropy method to calculate the weights of five components of farm household well-being and the indicators in every dimension, respectively, which are shown in Table 1. Further, we summed the five component values and well-being levels of farm households based on their weights.

3.4. Calculating the Net Effect of Farmland Transfer

3.4.1. Propensity Score Matching

The current study aims to estimate the effect (i.e., treatment effect) of farmland transfer on farm household well-being in ore–agriculture compound areas in Northwest China. Firstly, the following multiple linear regression model is constructed:

$${wellbeing}_i=\alpha +{\beta }_i{transfer}_i+{\gamma }_i{control}_i+{\mu }_i$$

(7)

where ${wellbeing}_i$ represents the index of farm household well-being; ${transfer}_i$ is the status of a farm household participating in farmland transfer; ${control}_i$ is other factors affecting farm household well-being; $\alpha$ represents the intercept term; ${\gamma }_i$ is the coefficient to be estimated; and ${\mu }_i$ is the random error.

However, in the actual estimation, whether farm households participate in farmland transfer and their well-being is affected by many factors can lead to sample self-selection bias in the process of model estimation. Therefore, the evaluation results for well-being may be biased.

Due to potential observable and unobservable biases, it is not simple to accurately estimate the effects of farmland transfer on farm household well-being. Theoretically, the well-being effect of farmland transfer is the well-being index of farm households participating in farmland transfer minus that of those not participating in farmland transfer. However, we can only observe one of the potential outcomes for a farm household (i.e., actual outcome). A missing data problem arises here [57], making it difficult to estimate the treatment effect of individuals. Rosenbaum and Rubin [58] proposed the propensity score matching (PSM) method to simulate a quasi-natural experiment by matching samples from the experimental group (participating in farmland transfer) and the treatment group (not participating in farmland transfer). This method eliminates the possible sample self-selection bias so that it can more accurately estimate the average treatment effect of farmland transfer on farm household well-being.

PSM is a causal inference method based on a counterfactual frame, which can alleviate endogenous problems caused by self-selection bias. Its advantage is that it reduces the dimension of the vector to a scalar and balances the observed variables between treated and control groups to obtain an unbiased estimation, because the factors prompting a farm household to make a decision are usually a multidimensional vector composed of several covariates. The steps for estimating the treatment effect through PSM are as follows [59].

(1) Estimating the conditional probability of each farm household participating in farmland transfer (identified as propensity score, ${PS}_i$) by logit model. The formula is:

$${PS}_i=P_r\left[D_i=1|X_i\right]=E\left[D_i=0|X_i\right]$$

(8)

where $D_i$ is the treatment variable indicating whether the farm household i is participating in farmland transfer. $D_i=1$ represents the farm households participating in farmland transfer; the value is 0 otherwise. $X_i$ is the characteristic vector of farm household i consisting of a set of observable covariates.

(2) Matching the treated group with the control group. We selected k-nearest neighbor matching to explore the treatment effect of farmland transfer on farm household well-being and its five components. This algorithm chooses an individual from the comparison group as a match for a treated sample in terms of the closest propensity score. In the current study, we set k to 4 and implemented one-to-four matching to minimize the mean square error. Considering the PSM method works under two assumptions (the conditional independence assumption (CIA) and the common support condition), we implemented matching using another four matching algorithms (nearest caliper matching, radius matching, kernel matching and local linear matching) according to previous studies to check the robustness of the results [60].

(3) Calculating the differences in farm household well-being between the treated group and the control group, i.e., the average treatment effect on the treated (ATT). The PSM estimator of ATT can be specified as:

$$ATT=E\left[y_{1i}|D_i=1\right]-E\left[y_{0i}|D_i=1\right]=E\left[y_{1i}-y_{0i}|D_i=1\right]$$

(9)

where $E\left[y_{1i}\right|D_i=1]$ represents the expected outcome of farm household well-being with treatment, which can be observable. On the contrary, $E\left[y_{0i}\right|D_i=1]$ denotes the counterfactual outcome, which cannot be observable.

(4) PSM assumptions tests. There exist two tests: the common support region test and the balancing property test. Firstly, the common support region judges whether the treated and control groups have a common support area and whether the value ranges of the two groups exist in a partial overlap. Specifically, according to the overlap degree analysis of the propensity score interval between the treated and control groups, the higher the overlap degree, the larger the common support region is. Second, the balancing test uses the standard deviation between matched and unmatched samples to compare the covariates between the treated and control groups before and after matching to judge the matching quality.

(5) Sensitivity analysis. In this study, the Rosenbaum bounds method was applied to estimate the sensitivity of the model results to hidden biases [61], which identifies the hidden bias by estimating the sensitivity parameter Γ value. Γ represents the odds of the observed covariates influencing the difference between the potential of the treated and untreated [62], with greater values of Γ indicating models are more robust against such bias.

3.4.2. Variable Selection

The FHWB and its five components are identified as the dependent variable in PSM. The treatment variable is farmland transfer. If a farm household participates in farmland transfer, the value is 1, while it is 0 otherwise. Whether a farm household participates in farmland transfer is a “self-selection” problem, which is affected by family characteristics and household head characteristics [10]. In addition, other external factors, such as soil erosion and droughts, are also important influencing factors regarding their farmland transfer. Therefore, we firstly selected 13 variables from the above aspects as covariates in the PSM model. Before matching, in order to achieve a good matching effect, we adopted the method proposed by Imbens and Donald [63] that was used to screen covariates, and finally 9 covariates were determined. Among them, the characteristics of the farm household include the income change, farm income, non-farm income, whether the farm household has rural cadres and education pressure. The characteristics of the household heads include their age and education level. Other external factors include the soil erosion degree and drought degree. The specific variables and their descriptions are shown in

Table 2. Variables and their descriptions in the PSM.

Variables	Variable Descriptions	Treatment Group		Control Group		Difference
		Mean	Std. Dev	Mean	Std. Dev
Treatment variables
Farmland transfer	Whether farm households transferred their farmland (1 = yes, 0 = no)
Dependent variables
FHWB	Farm household well-being	0.521	0.116	0.433	0.109	0.088 ***
BMN	The basic material needs for a good life	0.342	0.131	0.283	0.116	0.058 ***
HEA	Health	0.835	0.187	0.796	0.219	0.038 *
SEC	Security	0.691	0.248	0.533	0.266	0.158 ***
GSRs	Good social relations	0.313	0.128	0.279	0.123	0.034 ***
FCA	Freedom of choice and action	0.502	0.199	0.412	0.191	0.090 ***
Covariates
Age	Age of household head	53.28	12.014	59.48	11.525	−6.200 ***
Education level	Education level of household head (1 = primary school and below, 2 = junior high school, 3 = senior high school, 4 = junior college, 5 = university and above)	1.900	1.009	1.630	0.862	0.275 ***
Income change	Total income changes of a farm household in recent years (1 = decreased a lot, 2 = decreased a little, 3 = no change, 4 = increased a little, 5 = increased a lot)	3.310	0.705	2.760	0.768	0.554 ***
Farm income	Farm income of the farm household/yuan	1.829	2.341	1.085	1.895	0.744 ***
Non-farm income	Non-farm income of the farm household/yuan	9.594	11.706	6.443	8.339	3.151 ***
Rural cadres	Whether the farm household has rural cadres (1 = yes, 0 = no)	0.430	0.738	0.170	0.406	0.253 **
Education stress	Whether the farm household has education stress from children (1 = yes, 0 = no)	0.100	0.295	0.170	0.374	−0.072 ***
Soil erosion	1 = no, 2 = slight, 3 = general, 4 = serious, 5 = very serious	2.970	1.376	3.210	1.296	−0.248 *
Dry degree	1 = no, 2 = slight, 3 = general, 4 = serious, 5 = very serious	4.300	0.850	4.540	0.714	−0.236 ***

Note: ***, **, and * denote 1%, 5% and 10% significance level, respectively.

.

4. Results and Analysis

4.1. Statistical Descriptions of Farm Household Well-Being

The farm household characteristics from the treatment and control groups were statistically and significantly different (shown in Table 2). Specifically, farm households participating in farmland transfer were younger, had higher educational levels, and earned more family income (including farm and non-farm income) than those not participating in land transfer. Families with village cadres were more likely to transfer their farmland. Comparatively, farm households who did not transfer their farmland thought the local soil erosion and droughts were more serious, and the pressures caused by their children’s education were greater.

According to whether a farm household participated in farmland transfer, all the samples were divided into the treated group (participating in farmland transfer) and the control group (not participating in). A total of 115 respondents were in the treated group, accounting for 24%, while 370 respondents, accounting for 76%, were in the control group. Subsequently, we summed the levels of well-being and its five components of the two groups, respectively. Figure 3 shows a bar chart of the farm household well-being and its five components of two groups in three ore–agriculture compound areas. We found that the well-being and its components in the treated group were greater than those in the control group in the three study areas except HEA in Fugu County. Moreover, the rankings of the five components from the well-being framework tended to be consistent in the three regions. Specifically, HEA was the highest followed by SEC, while BMN was the lowest.

Overall, the average FHWB value of the treatment group was 0.521, and that of the control group was 0.433. In terms of its five components, the average values of BMN, HEA, SEC, GSRs and FCA in the treatment group were 0.342, 0.835, 0.691, 0.313 and 0.502, while those in the control group were 0.282, 0.796, 0.533, 0.279 and 0.412, showing statistically significant differences between the two groups (shown in Table 2). Figure 4a shows the ridge plots of the farm household well-being of the full sample, the treated and control groups. The medians of the three groups were near their peaks, and the distribution of well-being was similar between the full sample and the control group, and their medians were around 0.43. The treatment group had a higher level of well-being, with a median of 0.53.

Figure 4b is a boxplot of the five components of well-being for three groups in the whole study area. It was found that HEA was the highest, and BMN and GSRs were the lowest. Except for HEA, the data distributions were similar between the control group and the full sample, and the treatment group had greater values than the control group. Therefore, it could be preliminarily concluded that there were differences in well-being and its five components between the treatment group and the control group. An independent t-test was conducted to further test the differences in farm household well-being and its five components between the treatment group and the control group. The statistically significant differences are also marked in Figure 4b. The results also confirmed the significant differences between the two groups.

4.2. Results of Propensity Score Matching

The nearest neighbor matching algorithm (k = 4) was adopted to estimate the impact of farmland transfer on FHWB and its components. The results (

Table 3. The average treatment effects of farmland transfer on farm household well-being and its components based on the k-nearest matching.

Components	Sample	Treated	Controls	Differences	S.E.	T-Stat
FHWB	Unmatched	0.521	0.433	0.088	0.012	7.47
	ATT	0.515	0.470	0.045 ***	0.014	3.13
BMN	Unmatched	0.342	0.283	0.058	0.013	4.55
	ATT	0.336	0.324	0.012	0.016	0.74
HEA	Unmatched	0.835	0.796	0.038	0.023	1.69
	ATT	0.828	0.823	0.006	0.026	0.22
SEC	Unmatched	0.691	0.533	0.158	0.028	5.66
	ATT	0.683	0.595	0.088 **	0.033	2.68
GSRs	Unmatched	0.313	0.279	0.034	0.013	2.59
	ATT	0.311	0.299	0.012	0.016	0.71
FCA	Unmatched	0.502	0.412	0.090	0.021	4.35
	ATT	0.496	0.434	0.061 **	0.026	2.40

Note: *** and ** denote 1% and 5% significance levels, respectively. FHWB: farm household well-being; BMN: the basic material needs for a good life; HEA: health; SEC: security; GSRs: good social relations; FCA: freedom of choice and action.

) showed that the ATT values of FHWB, SEC, and FCA were 0.045, 0.088, and 0.061, respectively, and were significant at the 5% significance level, while the ATT values of the other components were not significant. This suggested that transferring farmland increased the farm household well-being, security and freedom of choice and action by 4.5%, 8.8% and 6.1%, respectively.

To ensure the robustness of the matching results, an additional four matching algorithms (such as nearest neighbor matching within caliper, radius matching, kernel matching and local linear matching) were used to estimate the effects of farmland transfer on FHWB (

Table 4. The average treatment effects of farmland transfer on farm household well-being.

Components	Sample	Treated	Controls	Differences	S.E.	T-Stat
	Unmatched	0.521	0.433	0.088 **	0.012	7.46
K-nearest neighbor matching (k = 4)	ATT	0.515	0.470	0.045 ***	0.014	3.36
CK-nearest neighbor matching within caliper (k = 4, r = 0.04)	ATT	0.515	0.464	0.051 ***	0.015	3.50
Radius matching (r = 0.04)	ATT	0.515	0.433	0.053 ***	0.014	3.81
Kernel matching	ATT	0.515	0.461	0.054 ***	0.014	3.92
Local linear matching	ATT	0.515	0.462	0.053 ***	0.017	3.04
Average	ATT	-	-	0.049	-	-

Note: *** and ** denote 1% and 5% significance levels, respectively.

). The results indicated that the average ATT of FHWB was 0.049 at a 1% significance level after matching. Furthermore, by comparing the ATT values before and after matching, we found that the ATT values decreased after matching, suggesting that the effect of farmland transfer on FHWB would be overestimated if the self-selection bias of the sample was not taken into account.

4.3. Hypothesis Test

4.3.1. Common Support Region Test

The common support region assumption is also known as the overlapping assumption, which assumes that the observed features in the treated group can also be observed in the control group. A wide common support region between the treated and control groups indicates a small sample loss, indicating a reasonable and effective matching result. The kernel density function composed of the propensity scores of farm households in the treatment and control groups before and after matching was used to test the common support region assumption. Figure 5 shows the results of the common support region test. Specifically, before matching (Figure 5a), there were differences in the kernel probability density between the treatment and control groups, and the common support region was narrow. After matching (Figure 5b), the kernel density of farm households in the treatment and control groups tended to be consistent, and the common support area between the two groups increased, which indicated that the selection bias in the samples had been eliminated. Given this, it can be concluded that it is necessary to match samples in the two groups. The matching results of this study through the k-nearest neighbor matching method (k = 4) met the requirements of the common support hypothesis.

Considering that different matching algorithms have different samples missing, we adopt five identified matching algorithms to analyze the sample losses. A total of 361 samples in the control group participated in matching, and 9 samples were lost. At least 105 samples in the treated group participated in the matching, and 10 samples were lost at most, indicating that the samples between the treated and untreated groups were well matched.

4.3.2. Balancing Test

Under the CIA and common support region assumption, a balancing test was adopted to test whether the propensity scores of the covariances were significantly different between the treatment and control groups (

Table 5. The balancing test results based on PSM.

Variables	Unmatched	Mean		%bias	t-Test
	Matched	Treated	Control		t	P > |t|
Age of household head	U	53.278	59.407	−52.2	−4.95	0.000
	M	54.299	55.028	−6.2	−0.46	0.650
Educational level of household head	U	1.904	1.659	24.6	2.31	0.021
	M	1.860	1.862	−0.2	−0.02	0.986
Income change	U	3.313	2.759	75.1	6.88	0.000
	M	3.262	3.269	−1.0	−0.07	0.941
Farm income	U	1.868	1.002	41.1	4.14	0.000
	M	1.772	1.803	−1.5	−0.09	0.931
Non-farm income	U	9.594	6.406	31.4	3.23	0.001
	M	9.124	8.631	4.9	0.33	0.740
Rural cadres	U	0.426	0.173	42.4	4.68	0.000
	M	0.308	0.301	1.2	0.09	0.925
Education stress	U	0.096	0.168	−21.5	−1.90	0.058
	M	0.103	0.131	−8.3	−0.64	0.525
Soil erosion degree	U	2.965	3.209	−18.2	−1.73	0.084
	M	3.037	2.951	6.5	0.47	0.636
Drought degree	U	4.304	4.542	4.297	2.31	0.021
	M	4.365	4.297	8.6	−0.02	0.986

). The bias ratios of all the covariances after matching were obviously reduced and the absolute values were even smaller than 10%, compared with the results before matching. In terms of the t-test results, the p-values of all the covariates were greater than 0.05, indicating that the t-values of all the covariates after matching did not pass the significant level test at a 10% level. Given the above analysis, it was found that the k-nearest neighbor matching method passed the balancing test.

Further, we tested the matching quality of different matching algorithms (

Table 6. The balance test results based on different matching methods.

Sample	Matching Method	Pseudo R2	LR chi2	P > chi2	Mean Bias	Med Bias	B Value
Before matching		0.166	88.29	0.000	38.0	31.0	103.2
After matching	K-nearest neighbor (k = 4)	0.005	1.52	0.997	5.5	5.6	16.8
	One-to-four matching within caliper (k = 4, c = 0.04)	0.005	1.43	0.998	5.2	4.2	16.4
	Radius matching (r = 0.04)	0.001	0.36	1.000	2.5	2.2	8.2
	Kernel matching	0.003	0.75	1.000	3.8	3.1	11.7
	Local linear matching	0.009	2.74	0.974	7.5	7.0	22.7

). The results showed that no matter which matching algorithm was employed, the P-R² and LR statistic decreased significantly and the p-value increased greatly after matching, compared with the regression results before matching. In addition, both the mean biases and median biases decreased significantly after matching. Moreover, the B values were reduced to less than 20%. This implied that sample matching qualities implemented by the five matching algorithms worked well, and the PSM model significantly decreased the differences in the observed covariates between the treated and control groups.

4.3.3. Sensitivity Analysis

The CIA for the PSM model means that whether a farmer participates in agricultural cooperatives does not take into account the influences of unobservable variables, so it is necessary to estimate the sensitivity to hidden biases. We performed a sensitivity analysis methodto hidden bias using the Rosenbaum bounds method. The Γ value was 1.7 according to

Table 7. Sensitivity analysis results.

Gamma (Γ)	Sig+	Sig-	t-hat+	t-hat-	CI+	CI-
1	0.000	0.000	0.04955	0.04955	0.0256	0.0740
1.1	0.000	0.000	0.0447	0.0548	0.021	0.0801
1.2	0.001	0.000	0.0406	0.0603	0.0157	0.0842
1.3	0.0027	0.000	0.0362	0.0645	0.0114	0.0883
1.4	0.0069	0.000	0.0319	0.0681	0.0064	0.0928
1.5	0.0149	0.000	0.0275	0.0719	0.0024	0.0972
1.6	0.0284	0.000	0.0244	0.0756	0.0004	0.1011
1.7	0.0491	0.000	0.0215	0.0795	0.0046	0.1045
1.8	0.0779	0.000	0.0190	0.0819	−0.0083	0.1072

, indicating that our results were moderately sensitive to the hidden biases, which indicated that the results obtained from the PSM were robust.

4.4. Heterogeneity Analysis of Effects of Farmland Transfer

Existing studies have shown that family income and the income distribution effects of land transfer were heterogeneous among farm households [64]. Therefore, we analyzed the difference in the net effect of land transfer on farm household well-being from the perspective of the heterogeneity of the age and education level of household heads.

Similarly, we used the PSM method to analyze the heterogeneity of the effects of farmland transfer on farm household well-being. The results were shown in

Table 8. Heterogeneity of the impact effects of farmland transfer on farm household well-being.

		40 Years and Below	41–50 Years	51–60 Years	More Than 60 Years	Primary School and Below	Junior High School	Senior High School
K-nearest neighbor matching (k = 4)	ATT	0.163 *	0.059 *	0.077 **	0.062 **	0.048 *	0.072 ***	0.106 *
	S.E.	0.084	0.034	0.036	0.028	0.029	0.028	0.058
K-nearest neighbor matching within caliper (k = 4, c = 0.04)	ATT	0.163 *	0.066 *	0.081 **	0.049 *	0.048 *	0.071 **	0.106 *
	S.E.	0.088	0.040	0.037	0.029	0.029	0.030	0.059
Radius matching (r = 0.04)	ATT	0.128 *	0.066 *	0.056 *	0.045 *	0.035 *	0.067 **	0.100 *
	S.E.	0.080	0.040	0.029	0.027	0.021	0.026	0.053
Kernel matching	ATT	0.142 *	0.062	0.049 *	0.055 **	0.032	0.067 ***	0.102 *
	S.E.	0.084	0.038	0.027	0.026	0.021	0.026	0.053
Local linear matching	ATT	0.138 *	0.054	0.050 *	0.055 **	0.036 *	0.077 ***	0.091 *
	S.E.	0.090	0.039	0.028	0.025	0.021	0.024	0.052
Average	ATT	0.147	0.061	0.062	0.053	0.040	0.071	0.101

Note: ***, **, and * denote 1%, 5% and 10% significance levels, respectively.

. Household heads’ ages were divided into four segments: 40 years old and below, 40 to 50 years old, 50 to 60 years old, and over 60 years old. In the four age segments, the effects of farmland transfer on farm household well-being were significant at the level of 10%. Compared with the other stages, farmland transfer led to the greatest improvement in well-being in those 40 years old and below (0.163), which might be because farmland transfer could release more young and middle-aged rural labor to pursue their non-farm employment and the overall income level of families with young household heads and be more likely to achieve livelihood transformation [26]. In addition, younger household heads are more likely to become the subjects of village collectives intended to develop new management entities.

We divided the education level of household heads into three segments: primary school and below, junior high school, and senior high school and above. Similarly to the age of household heads, the effects of farmland transfer at all the education levels on the well-being of farmers was significant at the level of 10%. Among them, household heads with senior high school or above had the highest well-being values (0.106), which may be because the higher the education level of a household head, the higher their understanding of farmland transfer benefits, and the stronger their ability to accept new knowledge and new ideas. It can enrich farm households’ livelihood activities and improve their well-being levels. Secondly, education is essential to human capital. A higher education level contributes to farmers mastering new labor skills and improving their livelihood resilience and adaptability for higher well-being.

5. Discussion

5.1. The Well-Being Effects of Farmland Transfer

This study assumed that participating in farmland transfer could improve farm household well-being in ore–agriculture compound areas in Northwest China. According to whether farm households transferred their farmland, we divided all the samples into the treatment group and control group. The improvement effects of farmland transfer on farm household well-being and its five components were confirmed in the comparative analysis and via the PSM method.

After the comparative analysis and independent sample t-test, it was found that the farm household well-being and its five components in the treatment group were higher than those in the control group. This verified the improvement effects of farmland transfer on farm household well-being to some degree.

In order to verify the net effects of farmland transfer on farm household well-being, the PSM method was performed. Different from the results of the comparative analysis and t-test, we only found the effects of farmland transfer on farm household well-being and its two components, namely SEC and FCA. However, the effects of farmland transfer on BMN, HEA and GSRs was not statistically significant. This might be because there existed sample selection bias. By constructing a counterfactual framework, the PSM method eliminated the influences of sample self-selection bias on the results and made the results more accurate. We believe that the improvement effects of farmland transfer on farm household well-being and its two components might be related to the farmland transfer modes in ore–agriculture compound areas in Northwest China.

There are mainly two types of farmland transfer in ore–agriculture compound areas, namely the land leasing and subcontracting mode, and the land stock cooperative system mode. Land leasing and subcontracting is a popular farmland transfer mode where the township government or rural collective economic organization rents the farmland back from farm households at a certain rent and then subcontracts it to scaled growers or farming enterprises. On the one hand, township governments and village committees play the role of land banks, which can earn a certain amount of net rent for rural environmental governance. On the other hand, farmers who are willing to engage in agricultural production can contract a certain amount of farmland from their township government or village committee to develop optimum-scale farm management. Therefore, many new agricultural management entities like professional growers, family farms and farmer specialized cooperatives have developed, among which farmer specialized cooperatives are the most common. According to the Department of Agriculture and Rural Affairs of Shaanxi Province [65], by April 2022, there had been 122 demonstration cooperatives, mainly managing fruits, vegetables, small grains and livestock products. There was no doubt that new agricultural management entities developed scale management and mechanization. In addition, land leasing and subcontracting promotes non-farm employment of farm households transferring farmland out to a certain extent. Therefore, land leasing and subcontracting could promote farm households’ livelihood transformation.

As for the land stock cooperative system mode, farmers voluntarily set up joint-stock companies or specialized cooperatives to engage in agricultural production by investing with their farmland management rights. The farmland is unified or entrusted in terms of management by specialized cooperatives or companies, and at the end of the year, farm households are entitled to receive dividends based on their shares [66]. In the study areas, some village collectives built the agricultural production mechanism of “the party branch + collective stock economic cooperatives + leading enterprises + farm households”, with the backing of farmland share mode, and developed ecological agriculture of “alfalfa planting-feed processing-breeding-manure returning to the field”, combining the regional advantages to strengthen the village collective economy. Some village committees developed rural complexes with the backing of land joint-stock cooperatives. Therefore, they provided some jobs for older farmers who had rich agricultural skills and those who lost their ability to engage in migrant work, which could protect some farm households’ income and facilitate off-farm labor transfer. Tang et al. [67] also confirmed through empirical study that farmland stock cooperatives promoted the labor transfer rate and extended the off-farm employment time in rural regions.

In addition, no matter what kind of farmland transfer mode, the system has played a huge role in the governance of rural settlement environment and the overall well-being of most village residents. In recent years, farmland transfer in the ore–agriculture compound areas has become an important way for local governments to implement targeted poverty alleviation and rural revitalization strategy. During our field survey, farmers in the villages organizing large-scale farmland transfer commonly felt that their living environment had been improved significantly in recent years. Particularly, many problems caused by long-term coal mining exploitation, such as groundwater, coal powder pollution, domestic water shortage and pollution, and unused land, have been effectively solved. Village committees used part of the profits to build infrastructure and expanded collective industries for increasing farm households’ income and year-end dividends [68]. Therefore, farm households’ senses of gain and happiness were significantly enhanced.

5.2. Group Heterogeneities of Well-Being Effects

We also found the heterogeneity of farmland transfer effects from the perspective of the ages and education levels of household heads. This was also confirmed by some research. The research by Abebaw et al. [69] showed that age had a significant non-linear influence on cooperative participation and the possibility of participating in cooperatives increased with age until 37 years old, which was consistent with our conclusion to some extent. Naidoo et al. [62] reported that mothers’ education significantly improved their children’s health well-being. Huang and Li [26] also revealed a similar conclusion that the impacts of farmland acquisition for rural tourism on landless farmers’ well-being had differences in terms of the ages and educational levels of household heads. However, through analysis of data from China Family Panel Studies (CFPS), He et al. [42] found that farmland transfer could improve the quality of life of the older generation but reduce that of the newer generation in rural areas in China. Our findings highlighted that young and well-educated household heads could achieve livelihood transformation and enhance their adaptability, thus they were better able to enjoy the benefits of farmland transfer. In this case, local governments should focus on improving the well-being of the elderly and low-educated farmers for fairness among farm households.

5.3. Policy Implications

Farmland transfer is the carrier of new agricultural management entities. Based on farmland transfer, China currently has a variety of farmland transfer modes, that is, a variety of agricultural management entities. Our research revealed that government-led farmland transfer had a significant and positive effect on farm household well-being. The conclusions obtained in this study are applicable to the widely ore–agriculture compound areas in Northwest China. On the one hand, because their ecological environments are similar. On the other hand, they have the inherent advantages of carrying out agricultural modernization due to many local enterprises. Large-scale farmland transfer has the functions of farmland resources reallocation, ecological environment protection and rural revitalization. Ore–agriculture compound areas are distributed widely in Northwest China, where the contradictions between economic development and ecological protection are prominent, and the rural revitalization faces multiple challenges. Therefore, it is necessary to strengthen the farmland transfer in these areas. Local governments need to cultivate new types of agricultural entities; meanwhile, they should actively mobilize multiple entities in rural areas, such as the leading role of local governments, the economic driving role of coal-related enterprises, and the organizational role of village cadres. This should stimulate the endogenous power of farmers.

Secondly, local governments should improve the existing farmland transfer system and the construction system for new agricultural entities, especially the benefit distribution mechanism, because this is directly related to the economic benefits obtained by farm households. However, the effects of farmland transfer on other aspects of household well-being were not found. The unfair distribution of economic benefits may reduce their subjective well-being, widen the income gap among the groups and lead to the unfairness of rural society. Finally, administrators should pay more attention to the well-being of the vulnerable populations in rural areas. The fairness effect of farmland transfer has been unclear, and there are group heterogeneities in the income effects among farm households. Some studies have confirmed that farmland transfer could widen the income gap among farm households. Our conclusions confirmed this to some extent. Therefore, in the process of farmland transfer, we should fully consider the vulnerable groups, such as the elderly and the people with low education levels. Local governments can carry out training in agricultural skills and create more employment.

5.4. Limitations

Overall, this study explored the effect of farmland transfer on farm household well-being. It has the following limitations. (1) The external effects of the conclusions in this study are relatively poor. On the one hand, the study areas are ore–agriculture compound areas in Northwest China with a large area with a sparse population and flat terrain. This region is suitable for large-scale farmland transfer. On the other hand, the farmland transfer modes in this study are dominated by local governments. Different modes of farmland transfer may have different effects on farm household well-being. Therefore, in future work, it is necessary to consider the impact mechanism of different farmland transfer modes on farm household well-being. (2) Although this study conducted a comparative analysis of the three selected study areas due to their homogeneity in terms of natural and social environments, this prevented us from conducting an in-depth analysis of the regional heterogeneity of the well-being effects of farmland transfer. The mechanism of regional heterogeneity should be further revealed in the future. (3) Due to the limitation of the available data, we did not analyze the mechanism of the promoting effects of farmland transfer on farm household well-being.

6. Conclusions

Northwest China possesses rich mineral resources, which are mainly distributed in the ecologically fragile rural areas. With the long-term and intense resource exploitations, farmers’ production–living–ecological spaces in the ore–agriculture compound areas are more complicated than those in the peripheral areas. In the process of rural governance and agricultural modernization, the Chinese government usually regards farmland transfer as a key and effective measure. However, the research on the effects of farmland transfer on farm household well-being is not sufficient. This study mainly examined the effects of farmland transfer on farm household well-being and its five components in ore–agriculture compound areas in Northwest China through the PSM method. On this basis, we also verified the heterogeneities of the well-being effects of farmland transfer. The evaluation results of farm household well-being showed that farm household well-being in the study areas was at a moderate level, but there were large differences among its five components. And the well-being index of farm households participating in farmland transfer was generally greater than that of those not participating in farmland transfer. The results of the PSM method revealed that farmland transfer could significantly improve farm household well-being, especially in terms of security and freedom of choice and action in ore–agriculture compound areas in Northwest China. The well-being effects of farmland transfer varied according to the age and education level of a household head. At the end of this study, we believe that farmland transfer should be regarded as an important means of rural governance and agricultural modernization in the ore–agriculture compound areas in Northwest China. In the process of implementing farmland transfer, however, local governments need to further improve their local farmland transfer system so as to comprehensively increase farm household well-being. Meanwhile, they should pay more attention to vulnerable rural groups.