*Article* **The Impact of Land Transfer on Vulnerability as Expected Poverty in the Perspective of Farm Household Heterogeneity: An Empirical Study Based on 4608 Farm Households in China**

**Zheng Wang 1,2, Mingwei Yang 1,3,\* , Zhiyong Zhang 1,4, Yingjuan Li <sup>1</sup> and Chuanhao Wen <sup>5</sup>**


**Abstract:** Poverty eradication is one of the global challenges, and land transfer provides an effective path to address farmers' poverty; however, the effect of poverty reduction can show heterogeneity depending on the location, household, and head of household. This study employs the propensity value matching technique to compare the effects of the land transfer on the future alleviation of poverty among farm households, based on the vulnerability as expected poverty, using data from 4608 household tracking surveys. The findings point to the following: In general, rural land transfers can significantly lessen farm households' VEP. In terms of regional variations, the positive effects of land transfers on farm households' VEP are mainly in the west. In terms of the differences among households, it was found that land transfers contribute to lower VEP for non-poor, non-financingconstrained, and government-subsidized farm households. With regard to differences in household headship, land transfers have abating effects on the VEP of self-employed heads of farm households. The results of the study can provide a useful reference for policy-making on land management and poverty reduction among farmers

**Keywords:** land transfer; vulnerability as expected poverty; farm households; heterogeneity

### **1. Introduction**

Land is essential to the survival of farmers, as it serves multiple purposes, including production, livelihood, and social security [1,2]. However, unlike other commodities, rural land in China is collectively owned and therefore cannot be traded freely [3,4]. Historically, the dated land system has stifled agricultural productivity and created obstacles to the movement of rural labor to industries and regions that provide better economic opportunities [5]. In order to accommodate the rapid growth of China's economy and technology, the government has continuously enhanced the land property rights system [6]. In 2014, the Central Committee's No. 1 document proposed the "three rights division" for contracted land. Farmers are able to protect their land rights and interests and obtain the right to dispose of and benefit from land management rights, and land management rights can be freely traded on the market, thereby activating the asset function of land. In 2015, the "Decision of the Central Committee of the Communist Party of China and the State Council on Winning the Battle of Poverty Alleviation" proposed that farmers' cooperatives and other business entities be supported to increase the income of poor households through land transfer methods such as land trusteeship and the absorption of farmers' land management rights as shares. Through land transfer, the disadvantage of land fragmentation is eliminated, creating conditions for the realization of agricultural scale and modernization [7,8].

**Citation:** Wang, Z.; Yang, M.; Zhang, Z.; Li, Y.; Wen, C. The Impact of Land Transfer on Vulnerability as Expected Poverty in the Perspective of Farm Household Heterogeneity: An Empirical Study Based on 4608 Farm Households in China. *Land* **2022**, *11*, 1995. https://doi.org/10.3390/ land11111995

Academic Editors: Li Ma, Yingnan Zhang, Muye Gan and Zhengying Shan

Received: 5 October 2022 Accepted: 4 November 2022 Published: 7 November 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

Land transfer is a crucial component of China's efforts to deepen the reform of its rural land system [9–11]. By granting farmers more property rights, land transfer has activated the efficiency of market resource allocation, enabled the scale and modernization of agriculture, and become a crucial factor in farmers' ability to escape poverty and become affluent [12].

China's household contracted cultivated land circulation area reached 530 million mu by the end of 2018, representing 35% of the country's cultivated land area. Consequently, a large number of surplus laborers were transferred to cities, which promoted urbanization and industrialization, boosted the overall social welfare level, and improved rural conditions by coordinating the integrated development of urban and rural areas, thereby becoming an effective practice for alleviating poverty in China's rural areas [13–15]. In 2018, the per capita disposable income of rural residents in impoverished areas was CNY 10,371, which was 1.99 times higher than in 2012 and grew by 12.1% annually on average. The alleviation of poverty in China's land transfer can serve as a valuable theoretical reference for farmers in other developing countries seeking to eradicate poverty.

Land transfer in this article refers to the transfer of land use rights from farmers with contracted land management rights to other farmers or economic organizations [16–18]. The essence of rural land transfer is that, in the context of multiple factors in the agricultural economy, the land rental market transfers land use rights from those with lower land valuations to those who are more eager to increase their production value through a mechanism of price equilibrium [19,20], and it assists farmers with varying land labor endowments in re-adjusting their marginal products [21,22]. In addition, land transfer can facilitate the transfer of rural surplus labor from agriculture to other industries [23]. This is the internal mechanism for improving the income of farmers through land transfer [24] (Figure 1).

**Figure 1.** Mechanisms of the effect of land transfer on farm households' VEP.

The effect of the land transfer system on reducing poverty has attracted the attention of numerous scholars [25–27]. Land transfer has a leveling effect on marginal output, transaction income, and the Pareto effect, and it has become a crucial method for efficiently allocating rural land elements [28]. It has a significant effect on the multidimensional poverty of rural households in poor villages through a mechanism known as the "preventive saving motive" [29]. Kassie et al. believe that land transfer not only is conducive to reducing agricultural costs, but also encourages non-agricultural employment of farmers, improves the income structure of farmers, and improves the overall welfare of farmers [30]. It is well acknowledged that "preventing" poverty is far more important than "governing" poverty [31], and "preventing" poverty requires measuring farmers' susceptibility to poverty [32]. The poverty index measures only the welfare level at a static point in time; examines whether farmers are in the ex-post state of poverty, which cannot reflect the poverty risk that has not yet occurred [33]; and disregards the long-term impact of rural

land transfer [34]. Land is the most important self-owned resource of poor households [35], has the material function of providing a means of subsistence for poor households, and is deeply embedded in these families' production and living processes [36]. In comparison to the income obtained from the land transfer, the loss of land management rights will have a negative impact on the future employment, social security, and mental health of farmers, thereby increasing the likelihood of future poverty [37]. The 2002 World Development Report of the World Bank used vulnerability as expected poverty (VEP) to measure the likelihood of an individual or family falling into poverty in the future [38]. This research makes use of VEP to calculate the likelihood of a farmer falling into poverty in the future.

In recent years, a number of scholars [39,40] have examined the relationship between land transfer and rural household poverty from the perspective of poverty vulnerability. Data from a field survey of 1682 farmers in Hubei Province by Peng et al. revealed that land transfer can significantly lessen a farmer's vulnerability to poverty and that this vulnerability declines as the area of land transferred increases [41]. Sun et al. found that the poverty vulnerability of land transfer households was 5.13% lower than that of nontransfer households [42]. Nonetheless, some scholars are concerned that the loss of land management rights will increase the likelihood of poverty among farmers [43,44]. Zhang et al. conducted an empirical study based on the survey data of 1386 rural households in southern Xinjiang and discovered that land transfer can significantly increase the income level of farmers, but cannot effectively reduce their susceptibility to poverty [45].

Presently, the academic community has not reached a relatively consistent conclusion regarding the effect of land transfer on farmers' vulnerability to poverty. In light of this, the formulation and implementation of rural land policies should vary from person to person, based on how well they take into account the differences between farmers. This paper will add to the existing body of knowledge in the following ways: We explore the general findings of the impact of land transfer on the poverty vulnerability of farm households using data from a national sample survey to supplement the existing studies that are restricted to a particular province or region. We classify farm households according to the characteristics of land transfer, and we verify the impact of land transfer on poverty vulnerability under different characteristics by regression, so as to identify what groups of characteristic farm households can reduce poverty vulnerability through land transfer, carve out the groups of farm households suitable for land transfer, and provide theoretical support for the implementation policy of land transfer policy classification.

Based on previous studies, this paper empirically analyzes the impact of land transfer on the VEP of farm households using a logit model based on the 2018 Chinese Family Panel Studies (CFPS) data; uses the stepwise regression method to select variables with significant effects; and regresses each of the five dimensions of regional distribution, poverty level, financing constraints, government subsidies, and nature of work. On this basis, propensity score matching (PSM) is used to test the robustness of the study results.

#### **2. Materials and Methods**

#### *2.1. Data*

The information in this article is drawn from the most recent (2018) Chinese Family Panel Studies (CFPS) (URL: http://www.isss.pku.edu.cn/cfps/; accessed on 15 October 2021). The database survey aims to track and investigate data at three levels: individual, family, and community, reflecting societal, economic, and population changes in China [46]. The survey covers a variety of topics, including family finances, education, health, and childrearing. The 2018 CFPS database survey targets a sample size of 14,218 households across 31 provinces (excluding Hong Kong, Macao, and Taiwan). In order to acquire high-quality research data, the data were efficiently screened. First, all urban household data were eliminated and only rural household registration data were retained; second, the characteristic data of the household head corresponding to the "financial respondent" were matched and the individual data of non-heads of households were eliminated; finally, the missing values, outliers, and samples missing important variables were deleted; and finally, a valid sample

of 4608 households was obtained, including 780 households with land transfer and 3828 households without land transfer.

#### *2.2. Method*

#### 2.2.1. Vulnerability as Expected Poverty

Chaudhrui et al. proposed the concept and measurement method of vulnerability as expected poverty [47]. VEP allows for the precise identification of households that may fall into poverty in the future. In particular, this approach quantifies the likelihood that a family will either enter or remain in poverty as a result of the risk of a sudden economic shock [48]. If the likelihood exceeds the predetermined threshold for vulnerability, the family is considered to be vulnerable to poverty. The concept of poverty vulnerability proposed by the VEP measurement method is simple to comprehend, reflects the dynamic characteristics of poverty, and can be effectively applied to cross-sectional data; consequently, it is widely utilized in academia. The formula for calculating poverty vulnerability is as follows:

$$\hat{\mathbf{V}}\_{i} = \text{Prob}(\ln c\_{i} < \ln z | \mathbf{X}\_{i}) = \Phi\left[ (\ln z - \mathbf{X}\_{i} \hat{\boldsymbol{\theta}}\_{\text{FGLS}}) / \sqrt{\mathbf{X}\_{i} \hat{\boldsymbol{\theta}}\_{\text{FGLS}}} \right] \tag{1}$$

where *V*ˆ *i* is the estimated value of the probability of poverty of farmer *i* in the future, *c<sup>i</sup>* is the per capita consumption of the household, *z* is the poverty line, *Φ* is the cumulative distribution function of the normal distribution, and *β*ˆ *FGLS* and ˆ*θFGLS* denote the expected value and variance of the family's future consumption estimated by the feasible generalized least squares (FGLS) method; compared with the ordinary least squares, FGLS can effectively eliminate the heteroscedasticity of the model and improve the accuracy of the estimation results. *X<sup>i</sup>* is an observable variable, mainly including family characteristic variables (including income, population, assets, liabilities, employment, and education) and household head characteristic variables (including age, gender, marriage, health, and occupation).

In order to assess a family's VEP, this study uses per capita household consumption. With respect to poverty based on consumption, two observations can be made: First, income is easily underestimated in micro-surveys, while consumption may better reflect the family's level of well-being, and second, using income as an explanatory variable can easily lead to strong endogenous problems in the measurement model. Concerning the choice of the poverty line, there are primarily two standards of per capita daily consumption of USD 1.9 and USD 3.1 proposed by the World Bank in 2015; based on China's average purchasing power and CPI index, we convert them into CNY 2800 and CNY 4570 per capita annual consumption at the end of 2018 [49]. This paper primarily measures the farmers' VEP based on their USD 1.9 per day per capita consumption, in line with existing research. Regarding the vulnerability threshold, this paper refers to the work of Gunther and Ward and sets the vulnerability line at 0.29; i.e., if the probability of a rural household falling into poverty in the future is greater than 0.29, set to 1; less than 0.29, set to 0 [50,51].

### 2.2.2. Econometric Model

In order to comprehensively examine the impact of land transfer on the poverty vulnerability of farmers, the logit model is constructed as follows:

$$\begin{cases} \log\left(\frac{p}{1-p}\right) = \beta\_0 + \beta\_1 transfer\_i + \beta\_2 X\_i + \varepsilon\_i\\ p = prob(v\_i = 1) \end{cases} \tag{2}$$

where *v<sup>i</sup>* is the household poverty vulnerability status of farmer *i* calculated based on the per capita consumption level of the household, *v<sup>i</sup>* = 1 represents a poor and vulnerable household, and *v<sup>i</sup>* = 0 represents a non-poor and vulnerable household. *trans f er<sup>i</sup>* indicates whether farmer *i* participates in land transfer, and *X<sup>i</sup>* indicates a series of control variables, mainly including family characteristics and household head characteristics.

#### 2.2.3. Stepwise Regression

Since there are many control variables affecting the dependent variable in the model, and the influence of individual variables on the dependent variable is small, the sum of squared errors does not decrease when these variables are included, but on the contrary, the error becomes larger due to the increase in the number of variables, which affects the overall stability. To address this problem, the stepwise regression method is used to select the variables with significant effects from many control variables to establish a regression model. The specific model is as follows:

Given the set of candidate control variables *T* = {*x*<sup>1</sup> · · · *xm*}, from which a subset *T*<sup>1</sup> ∈ *T*, the sum of squared errors of the regression model constructed from *T*<sup>1</sup> and the dependent variable is *Q*, and then the square of the remaining standard deviation of the model is Equation (3).

$$T^2 = \frac{\mathcal{Q}}{n - l - 1} \tag{3}$$

X in the formula is the data sample size. The selected subset *T*<sup>1</sup> should minimize *T* as a quantitative criterion for variable selection.

Then determine an initial subset, each time from the subset outside the impact of significant variables to introduce a maximum impact on the dependent variable, and then the original subset of variables to test, from the variables that become insignificant to eliminate a minimum impact, until it cannot be introduced and eliminated. Meanwhile, there are highlights worth noting in this model. First, the significance level *a<sup>m</sup>* for the introduced variables and *aout* for the excluded variables should be selected appropriately; obviously, the larger the *a<sup>m</sup>* is, the more variables are introduced, and the larger the *aout* is, the fewer variables are excluded. The second highlight is that due to the correlation between individual variables, the introduction of a new variable will make a variable originally considered significant insignificant and thus be dropped, so we selected variables that are as independent of each other as possible.

#### 2.2.4. PSM Method

Whether land transfer reduces the poverty vulnerability of farm households is a non-randomized experimental self-selection problem that is highly susceptible to selective error, which can be effectively addressed by propensity score matching (PSM). PSM is a non-parametric analysis method of counterfactual inference, which can effectively reduce selectivity bias and endogeneity by analytically processing non-experimental and observational data, and it is commonly applied in the evaluation of policy effects. PSM is used to test the robustness of the logit regression results. The logit model is used to calculate the conditional probability fitting value of the sample farmers' land transfer, which is the propensity score (PS).

$$PS\_{\bar{i}} = \Pr[D\_{\bar{i}} = 1 | \mathbf{X}\_{\bar{i}}] = \mathbb{E}[D\_{\bar{i}} = \mathbf{0} | \mathbf{X}\_{\bar{i}}] \tag{4}$$

$$ATT = \frac{1}{N^t} \sum\_{i \in I^t \cap S} \left\{ Y\_i - \sum\_{j \in I^t \cap S} W\_{ij} Y\_j \right\} \tag{5}$$

where *N<sup>t</sup>* is the number of samples of land transfer households, *I t* is the sample set of the disposal group (participating in land transfer), *I c* is the sample set of the control group (not involved in land transfer), *Y<sup>i</sup>* is the observed value of the sample of the disposal group, and *Y<sup>j</sup>* is the sample of the control group. *S* is the common support domain set, *Wij* is the matching weight, and ATT is the average disposition effect. The main method is to match the samples of the control group and the disposal group according to the propensity value to ensure that there is no significant difference in their main characteristics. Then, the control group is used to estimate the counterfactual state of the treatment group (i.e., no participation in the transfer) and calculate the poverty caused by the land transfer and the net treatment effect of vulnerability ATT.

#### *2.3. Variables*

Since the focus of this paper is on whether the loss of land management rights increases the risk of poverty for the farmers, the key variable defined is whether farmers transferred their farmland. Farmers who transferred their farmland are controlled for using question FS2 "whether they lease their land to others", with a value of 1 if the household transferred its farmland and 0 otherwise. A significance of *p* < 0.1 is set, and 10 control variables are identified by excluding insignificant variables through stepwise regression. Descriptive statistics are provided in Table 1.


**Table 1.** Variable definition and descriptive statistics.

#### **3. Results**

*3.1. Baseline Regression and Sub-Regional Regression*

The results of the baseline regression of the impact of the land transfer on the VEP of farm households are presented in Table 2. The results indicated that across the sample, all other factors being equal, land transfer effectively decreases the VEP of farm households, in that the transfer of land management rights does not increase the likelihood of farm households falling into poverty in the future. This conclusion is generally consistent with Deng and Wang et al.'s findings [52,53].


**Table 2.** Baseline regression and sub-regional regression results.

Note: Standard errors are in parentheses; \* *p* < 0.10, \*\* *p* < 0.05, \*\*\* *p* < 0.01. Same below.

It is widely accepted in the academic literature that regional economic development is a crucial factor in determining whether or not rural households are able to reduce their poverty levels [54]. There are obvious gaps in the economic development of eastern, central, and western China, and land prices are relatively low in economically underdeveloped regions [55]. Rapidly advancing industrialization and urbanization have substantially increased urban labor compensation, causing a large number of rural laborers to migrate to cities [56]. The level of economic development, industrial structure, employment capacity, and wage level of cities have an immediate impact on the income of urban farmers [57]. Generally, the non-agricultural supply of labor pattern of Chinese agricultural families exhibits specific geographical characteristics [58], and this pattern is driven by the development of non-agricultural enterprises outside the village [59]. In other words, the growth of these industries will have an obvious effect on rural labor migration but will have no discernible effect on the mobility of agricultural property rights [60]. However, within the sub-regional sample, the effect of land transfer on household VEP is statistically insignificant in the eastern region, but statistically significant at the 5% and 1% levels in the central and western regions. There is significant heterogeneity in the estimated coefficients for land transfer between the central and western regions. This may be due to the fact that the development of tertiary industry in the western region lags behind and cannot provide enough alternative employment opportunities outside agriculture, and that the livelihood conversion costs incurred by farm households after the transfer of farmland are more expensive; therefore, the impact of the land transfer on reducing VEP of farm households is not as significant in the western region as it is in the central region.

The results for the control variables show the following: (1) For household characteristics, the estimated coefficients for the variables household income, household assets, agricultural assets, social capital, and housing property were negatively significant, while the estimated coefficients for household size and government subsidies were positively significant. This suggests that higher household income, increased assets, the accumulation

of social capital, participation in non-farm businesses, and improved housing conditions will all contribute to reducing household poverty vulnerability. The greater the household size, the greater the risk of future poverty shocks to the farm household. Land transfer would reduce the risk of poverty for households receiving government subsidies; a possible explanation is that the funds obtained from land transfers are an important source of income for such households. (2) Regarding the characteristics of the household head, the marital status and education of the household head had a significant negative effect on the VEP of the household, while the age of the household head had a significant positive effect. VEP is influenced by many factors, such as natural resources, ability, and economic cycles, and only 10 control variables were selected for our study, resulting in a regression with an R <sup>2</sup> between 0.2 and 0.4, but it is sufficient to explain the impact of land transfer.
