How Does Poverty Alleviation Relocation Affect the Non-Agricultural Employment of Women’s Labor Forces? Evidence from Southern Shaanxi Province

Abstract

With the completion of the relocation task, fully promoting the non-agricultural employment of relocated women is of great significance for improving the livelihood sustainability of relocated peasant households’ and consolidating the results of poverty alleviation. Based on the sample data of 1616 rural households in southern Shaanxi Province, China, using the IV-Probit model, IV-Tobit model, and mediation effect model, this study empirically analyzes the impact and mechanism of participation in relocation on women’s non-agricultural employment. The results show that by introducing an instrumental variable to solve endogeneity, participation in relocation has a significant positive impact on the behavior and intensity of women’s non-agricultural employment. A mechanism analysis shows that training, public services, and land abandonment play a mediation role in the relationship between participation in relocation and women’s non-agricultural employment. In terms of women’s non-agricultural employment behavior, the mediation effects are 56.9%, 15.0%, and 11.0%, respectively. In terms of women’s non-agricultural employment intensity, the mediation effects are 58.5%, 15.7%, and 11.6%, respectively. Based on the above findings, this study puts forward policy implications for follow-up relocation support, to further release the surplus women’s labor forces and promote women’s non-agricultural employment.

1. Introduction

For a long time, relocation programs have been implemented in many countries around the world as an effective means to solve natural ecological disasters, reduce poverty, and promote the process of urbanization, which has had a profound impact on the economy, society, culture, and other dimensions [1,2,3,4]. Among them, the poverty alleviation relocation implemented in China not only has a long history but also has the largest relocation scale and capital investment in the world. It has played an important role in eliminating absolute poverty and promoting rural development, which has attracted widespread attention. This relocation project can be traced back to the 1980s, carried out in Gansu province and Ningxia province to address the agricultural production difficulties caused by water shortage, which not only effectively alleviates the contradiction between peasant households and land, but also improves the income level of peasant households. In recent years, under the call of the Chinese government to fight poverty, to address “each place cannot support its inhabitants”, a larger scale of poverty alleviation relocation has been further implemented. During China’s “13th Five-Year Plan” period from 2016 to 2020, the country invested a total of about 600 billion yuan in various types of funds and built 35,000 centralized relocation areas. More than 9.6 million poor peasants had moved out of ecologically fragile and remote areas and broken the predicament of low-level livelihood equilibrium by participating in relocation.

At present, with the completion of relocation tasks, China has entered the post-relocation era, with follow-up support as the core. Although the poverty alleviation relocation has effectively improved the welfare and endowment of peasant households, the overall livelihood space of the relocated peasants has been changed, social capital has been damaged [5], the degree of vulnerability is high, and they still face greater livelihood difficulties and risks of returning to poverty [6]. In this context, promoting non-agriculture employment for peasant households is an important step to improve the sustainable livelihood of peasant households and consolidate the results of poverty alleviation. Additionally, the No. 1 Central Document in 2022 clearly proposed to increase capital support for centralized relocated areas and actively carry out non-agriculture employment assistance actions for relocated peasants. At the same time, it is particularly important to promote non-agriculture employment for female relocated peasants. Women’s non-agricultural employment can be used as an insurance mechanism against the risk of income uncertainty for peasant households [7]. It also has a gender dividend effect, which not only improves individual women’s livelihood capital and empowerment and exerts a driving effect on other household members’ non-agriculture employment, but also contributes to increased investment in the education of the next generation and family nutrition and health [8]. This plays an important role in interrupting the intergenerational transmission of poverty and long-term development of the regional economy [9]. Overall, promoting non-agricultural employment for relocated women has become the focus of work in the post-relocation era.

In view of this, from the perspectives of non-agricultural employment behavior and intensity, this paper uses the micro-survey data of 1616 peasants in southern Shaanxi Province to explore the impact and mechanism of participation in relocation on women’s non-agricultural employment. On this basis, this paper aims to propose policy implications for the follow-up relocation support, to further release the surplus women’s labor forces, enhance women’s empowerment, and optimize peasant households’ livelihood.

The marginal contribution of this paper mainly includes: Firstly, by focusing on relocated women’s labor force in non-agricultural employment, the paper expands the research on the policy effects of poverty alleviation relocation and enriches the research paradigm, mainly focusing on relocated peasant households in existing research from a gender perspective. Secondly, by focusing on non-agricultural employment days, the paper adds to the scarce empirical evidence on the effect of participation in relocation on the intensity of non-agricultural employment. Finally, the paper not only comprehensively and deeply analyzes how women’s non-agricultural employment is affected by poverty alleviation relocation from capital constraints, time constraints, and role constraints, which enriches the theoretical framework of women’s non-agricultural employment research, but also compares the differential characteristics of the mediation effects worked by different paths.

The paper continues by presenting a literature review in Section 2 and theoretical hypotheses in Section 3. Section 4 explains the methodology of this paper, including the study area, variables, and model specification used. Section 5 offers the model estimation results and analysis. Section 6 presents discussions of this study. Section 7 draws conclusions and discusses policy implications.

2. Literature Review

2.1. Research on Poverty Alleviation Relocation on Non-Agricultural Employment

The effect of participation in relocation on the transition of peasants’ livelihood strategies to non-agricultural employment has attracted the attention of many scholars. Scholars have pointed out that by reducing the degree of dependence on the ecosystem and improving the level of the capital endowment of peasants [10,11], participation in relocation has effectively promoted the non-agricultural employment of peasants [12]. Non-agricultural employment not only promotes the improvement of the feasible ability and overall income level of peasant households but also effectively reduces the level of income gaps within peasants and alleviates relative poverty [13]. However, scholars have also noted that relocated peasants face unique barriers to labor market entry due to language barriers, lack of local experience and references, discrimination, and limited knowledge of the local labor market. This greatly affects the stability of peasants’ livelihoods [14]. From the perspective of non-agricultural employment characteristics, the employment direction of relocated peasant households is mainly concentrated in the fields of industry and services, and the employment area shows the transfer tendency of both within-the-county and out-of-county [15].

However, scholars’ research on non-agricultural employment of relocated peasant households pays less attention to gender characteristics. The impact mechanism of participation in relocation on the non-agricultural employment of peasant households is not to promote the participation of new relocated peasant households in non-agricultural employment, but to realize the reallocation of peasants within the households, in which women play an important role. Compared with the men’s labor forces, the increases in the women’s labor force supply are greater [16], which has a huge potential for improving the living standards of peasant households as well as increasing labor productivity. As Bandiera et al. [17] proposed, the asset transfer and skill improvement brought about by policy support to the poor women’s labor forces not only effectively promote them to engage in higher-paying non-agricultural activities and increase the duration of non-agricultural employment, but also significantly increase the overall value of household-durable goods. At the same time, the promotion effect of women’s non-agricultural employment will gradually appear with the implementation of the project with a time lag effect [18].

2.2. Research on Non-Agricultural Employment of Women

Women are still faced with many difficulties in non-agricultural employment. The specific manifestations are that the non-agricultural employment time of women’s labor forces is insufficient, and the quality of non-agricultural employment is generally low and has not significantly improved. The reasons are that, firstly, there are significant gender differences and inequality in the capital endowment of labor forces [19]. The “investment value” of the women’s labor force in the family is lower than that of the men’s labor force. Compared with the men’s labor force, the women’s labor forces are more impoverished in the dimensions of human and social capital endowment [20], which causes greater obstacles in non-agricultural employment. Secondly, household responsibilities such as childcare and elderly support are taken up more by women [21]. Such labor-intensive unpaid informal home care activities can significantly reduce the participation and working time of non-agricultural employment for women’s labor forces. The negative impact of household responsibilities on non-agricultural employment of women’s labor forces increases with the number of children, the age of children, and the intensity of elderly support [22,23]. When it is difficult to balance family and work at the same time, the women’s labor forces are passively withdrawn from the labor market. Finally, women’s non-agricultural employment faces the constraints of agricultural production [24]. Based on sampling data from 931 villages across the country, Luo [25] found that the proportion of women engaged in agricultural production was as high as 69.89%. Women have become the main labor force of agricultural production. The traditional Chinese household division mode of “men plow and women weave” has been transformed into “men employment and women plow”. Although this transformation of feminization of agricultural production is a rational choice by women based on factors such as household livelihood endowment level, traditional gender concepts, and village public opinion, it also greatly occupies the non-agricultural employment time of women’s labor forces and hinders the free flow of labor resources [26].

In summary, the discussions by scholars provide useful references for this paper and help to grasp the characteristics and general trends of non-agricultural employment behaviors of relocated peasant households from the overall level, but the following three shortcomings remain. Firstly, the existing literature mainly focuses on the policy effects of poverty alleviation relocation at the household level and rarely discusses the relocation between participation in relocation and women’s non-agricultural employment from a gender perspective. Secondly, the existing literature mainly focuses on non-agricultural employment of relocated peasant households in terms of non-agricultural employment behavior, with insufficient attention to the intensity of non-agricultural employment. Thirdly, the existing literature explores the constraints of non-agricultural employment of women’s labor forces mainly from a single dimension such as household responsibility and capital endowment, but under the background of agricultural production feminization, the crowding-out effect of agricultural production on women’s non-agricultural employment has not been paid enough attention. At the same time, the degree of influence and mechanisms of different factors on women’s non-agricultural employment are different, and few comprehensive studies combine them.

3. Theoretical Hypotheses

From the perspective of breaking the constraints of non-agricultural employment of women’s labor forces, this paper identified three mechanisms of promoting women’s participation in training, improving the accessibility of public services, and increasing the possibility of land abandonment. On this basis, a theoretical analysis framework for the impact of poverty alleviation relocation on the non-agricultural employment of women’s labor forces was constructed (Figure 1).

3.1. The Perspective of Insufficient Capital

Training plays an important role in promoting the non-agricultural employment of relocated women. On the one hand, participation in training can directly improve women’s human capital level through the accumulation of knowledge and mastery of labor skills, thereby improving women’s non-agricultural labor productivity and income [27]. Additionally, it can indirectly enhance women’s human capital by improving cognitive ability and learning ability, thereby effectively improving the possibility and stability of women’s non-agricultural employment. Moreover, training that is more frequent, instructive, and technical has a greater positive impact on the non-agricultural employment of female labor forces [28].

On the other hand, participation in training can also broaden the social capital of the women’s labor forces through communication among peasants. The other participants and trainers that women meet during the training effectively improve their social network [29]. Different from the traditional social capital based on blood and geography, this new type of social network based on employment can reduce information asymmetry and job search costs in the labor market through the information transmission mechanism, so that it can enable women to obtain more employment information and opportunities. In addition, this new social network forms a social exchange mechanism that helps women obtain more matching and advantageous employment resources or jobs when other capitals are relatively weak [30]. Therefore, training can improve women’s non-agricultural employment.

The poverty alleviation relocation is a project that combines both resource input and opportunity supply. Participation in relocation directly enhances women’s access to training opportunities. At the same time, by matching trainers with community factories and industrial parks and publicizing typical cases of training outcomes, women’s income expectations and future development prospects brought about by participation in training have been greatly improved [31]. The training implemented in the relocation area is mainly led by local government departments with sufficient funding supply. This not only improves women’s recognition level of training at the psychological level but also reduces the opportunity cost of participating in training through training subsidies [15]. As a result, the relevant relocation follow-up policies have effectively enhanced the women’s willingness and motivation to participate in training. Based on this, this paper argues that participation in relocation can promote the accumulation of capital endowments through the training mechanism, thereby promoting the non-agricultural employment of women’s labor forces.

3.2. The Perspective of Role Conflicts

Women not only face the heavy burden of childcare and elderly support but also shoulder the responsibility of accumulating development resources and promoting family reproduction. There is a dual-role conflict between family and work [32]. Public services, including children’s education and social support for the elderly, are important alternative resources for family care services.

As a powerful supplement to the traditional family care model, adequate infrastructure can alleviate the restrictions on female labor resources imposed by labor division mode and family roles of “men lead the outside, women lead the home”, thus breaking the barriers to labor mobility and improving the efficiency of female labor resource allocation [33]. In addition, the accessibility of public services can effectively improve women’s willingness to stay [34], reduce the possibility of returning after relocation, and improve women’s adaptability and future development expectations in relocation areas. This positive subjective development will form a new goal of family reproduction and strengthen women’s social production functions, which promote women to enter the non-agricultural employment market.

To achieve the relocation policy goal of “stable living and prosperity”, a large number of resources brought about by the relocation project are invested in the construction of infrastructure and public services in the relocation areas. It is conducive to making up for the shortcomings in the allocation of public service resources and effectively improving the accessibility and satisfaction of relocated peasants with public services [35]. At the same time, poverty alleviation relocation has broken the original scattered spatial living layout of peasants and realized the relative concentration of the population. This change in geographic space not only effectively solves the long-standing dilemma of insufficient allocation of public services, but also facilitates the scale effect and enhances the efficiency of matching resources with the population [36]. It is particularly important in reducing the inequality in the ability and opportunity to access public service resources for disadvantaged relocated peasants. Based on this, this paper argues that participation in relocation can alleviate the role conflict of the women’s labor forces by improving the accessibility of public services, thereby promoting the non-agricultural employment of the women’s labor forces.

3.3. The Perspective of Time Constraints

The feminization of agricultural production is a rational result of the optimal allocation of labor resources within the family [24], especially in economically backward and labor-exporting areas. As a scarce resource, time has an obvious substitution effect in the allocation of different production and operation activities. In this context, land abandonment is more beneficial to women’s labor forces, which not only directly reduces the labor time demand of the women’s labor forces in agricultural production, but also frees up a lot of time [37]. At the same time, land abandonment not only reduces the self-sufficiency rate of grain and increases the living cost of peasant households, but also hinders the social security function of land and aggravates the livelihood risk [38]. Therefore, it is urgently required that women’s surplus labor forces engaged in pure agriculture and part-time employment change their livelihood strategies and allocate released agricultural labor time and leisure time to non-agricultural activities [39], to realize the effective allocation of labor resources and ease the family economic burden.

Participation in relocation is different from the general rational decision-making of peasant households; that is, gradually reducing agricultural production activities after obtaining relatively stable non-agricultural employment. On the one hand, the sudden change in spatial residence location directly raises the cost of agricultural production and reduces the net return of agricultural production. On the other hand, participation in relocation promotes the transformation of peasant households’ lifestyles to urbanization and citizenship. Traditional agricultural production is not conducive to the adaptation of peasants’ lifestyles in the relocation areas. Thus, these two factors work together to increase the possibility of land abandonment by peasant households [12,40,41], and the longer the relocation period, the less the relocated households rely on the land. Based on this, this paper argues that participation in relocation can alleviate the time conflict of women’s labor forces by land abandonment, which will greatly promote women’s motivation to find non-agricultural employment opportunities and increase their non-agricultural employment time.

Based on the above analysis, the following hypotheses are proposed:

Hypothesis 1.

Participation in relocation can promote the non-agricultural employment of women’s labor forces.

Hypothesis 2.

Participation in relocation can promote the non-agricultural employment of women’s labor forces by promoting women to participate in training, improving the accessibility of public services, and increasing the possibility of land abandonment.

4. Methodology

4.1. Study Area

The data used in this paper are from a farmer survey conducted by the research group in southern Shaanxi Province in February 2018. The southern Shaanxi Province, an administrative area of 70,488 km², consists of the three cities of Hanzhong, Ankang, and Shangluo. It is located in the Qinba Mountain, with a fragile ecological environment and frequent floods and geological disasters. Additionally, it is part of the upstream watersheds of the Yangtze River, a primary water resource conservation region. As early as 2011, Shaanxi Province promoted the implementation of the Southern Shaanxi Province Relocation Project, including various types of ecological protection, disaster avoidance, and poverty alleviation, in combination with local conditions, and planned to complete the relocation of 2.4 million peasants. During the Five-Year Plan period from 2016 to 2020, the total number of relocated peasants in the 3 cities of southern Shaanxi Province accounted for more than 70% of the province, involving 197,800 peasant households and 665,800 relocated peasants, which occupied an important position in the national practice of poverty alleviation relocation. In the post-relocation era, the Shaanxi Province government has formulated a series of follow-up support policies centered on promoting non-agricultural employment for relocated peasants. Consolidating the results of poverty alleviation has achieved remarkable results, which have also been recognized and encouraged by the National Development and Reform Commission. Therefore, overall, the poverty alleviation and relocation work in Shaanxi Province has always been at the forefront of the country with important typicality and representativeness and can provide important references for improving the follow-up relocation support in the country.

In the farmer survey, the study areas included Danfeng county and Zhen’an county of Shangluo city, Baihe county, Hanbin district, and Hanyin county of Ankang city, and Liuba county, Xixiang county, and Lueyang county of Hanzhong city (Figure 2), involving a total of 57 villages in 33 townships. To ensure the representativeness of the sample, the research group adopted a combination of stratified sampling and random sampling to determine the interviewed households: 2–3 sample counties were randomly selected from each city, 3–5 sample townships were randomly selected from each county, 1–3 sample villages were randomly selected in each township, and 20–40 sample peasants were randomly selected from each village for a face-to-face questionnaire survey. The main content of the survey involved the basic information and employment situation of peasants and household members, relocation characteristics, income, expenditure, the level of livelihood capital in different dimensions, subjective evaluation of relocation policies, and participation in follow-up support policies. After data processing and cleaning and excluding invalid questionnaires such as unclear expressions and logical errors, a total of 1680 peasant household samples were obtained. This paper mainly discusses the impact of participation in relocation on the non-agricultural employment of women’s labor forces. Therefore, the peasant household samples with adult women peasants over 18 years old who are not in education were selected. Finally, a total of 1616 samples were obtained, including 1074 samples participating in the poverty alleviation relocation, accounting for 66.5% of the total samples.

4.2. Variables

4.2.1. Dependent Variable

This paper describes the non-agricultural employment of women’s labor forces from two aspects. On the one hand, the non-agricultural employment behavior of women’s labor forces was selected. This is a binary discrete variable. The value of participating in non-agricultural employment is 1, otherwise, it is 0. On the other hand, the non-agricultural employment intensity of women’s labor forces was selected. To mitigate the impact of extreme values and reduce heteroscedasticity, the natural log of women’s non-agricultural employment days was used. Among the sample farmers, 669 women were participating in non-agricultural employment, with an average of 287.749 days of non-agricultural employment.

4.2.2. Core Independent Variable

Participation in the relocation. When conducting farmer surveys, the Shaanxi Province’s relocation and construction tasks have been fully completed, with the actual occupancy rate and old homestead exit rate both exceeding 90%. This variable was set as a binary discrete variable in this paper. The value is 1 if the peasant households where the women belong participate in the poverty alleviation relocation and have already lived in the relocated house, otherwise it is 0.

4.2.3. Instrumental Variable

The instrumental variable of this paper is represented by the proportion of relocated peasant households in the village. Considering that the behavioral decision-making of peasant households has group-normative effects such as demonstration, conformity, and comparison, peasants will refer to, obey, and imitate the behavior and the performance of others because of their trust and dependence on the group or their internal need to avoid being isolated from the group [42]. That is, the participation behavior of peasant households in the relocation project is significantly and positively affected by the relocation behavior of relatives, friends, neighbors, and other subjects within the village. Therefore, the higher the proportion of peasant households involved in relocation in the same village, the greater the probability of peasant households participating in relocation. In addition, the proportion of peasant households participating in relocation in the same village is not directly related to the non-agricultural employment behavior and intensity of women. It can be considered that this variable is exogenous. Therefore, it is theoretically feasible to use this variable as an instrumental variable for participation in relocation.

4.2.4. Moderators

To verify the mechanism of the effect of participation in relocation on the non-agricultural employment of women’s labor forces, this paper selected moderators from three aspects: training, public services, and land abandonment. If women have participated in training, facilities and services such as schools and nursing homes are available in the relocation areas, and the peasant households have the behavior of land abandonment, then the values of moderators will be respectively assigned as 1. If not, the values will be 0.

4.2.5. Control Variables

Referring to the existing research, control variables were selected from the characteristics of women, peasant households, and regions in this paper. In terms of individual female characteristics, seven variables, including age, age square, household head, marriage, party member, health, and education, were selected [22,23]. In terms of household characteristics, variables including living with parents, average health of parents, average age of parents, children in different age stages (0–3 years old, 4–6 years old, 7–16 years old), household size, land area, and other members’ income were selected [43]. In terms of regional characteristics, factors such as geographical location, the level of regional economic development, and topographical features will affect the livelihood choices of peasants [6]. Therefore, distance to town and dummy variables of counties were selected to control the regional fixed effect.

The definitions of all variables introduced in this paper and the corresponding descriptive analysis are shown in

Table 1. Variable definitions and descriptive statistics.

Variable	Definition	Mean	SD
Non-agricultural employment behavior	Whether women participate in non-agricultural employment/yes = 1 no = 0	0.414	0.493
Non-agricultural employment intensity	Women non-agricultural employment days, take the natural log	2.311	2.767
Participation in relocation	Yes = 1, no = 0	0.665	0.472
Proportion of relocated peasant households in the village	Proportion of relocated peasant households to total households in the village (%)	0.591	0.220
Training	Whether women have participated in training in relocation area/yes = 1, no = 0	0.313	0.464
Public service	Whether facilities and services such as schools and nursing homes are available in the relocation areas/yes = 1, no = 0	0.778	0.416
Land abandonment	Whether the peasant households have the behavior of land abandonment/yes = 1, no = 0	0.201	0.401
Age	Women’s actual age (years)	42.710	12.047
Age square	Women’s actual age (years2)	1969.154	1117.811
Household head	Whether the woman is the household head or not? Yes = 1, no = 0	0.044	0.205
Marriage	Married = 1, unmarried = 0	0.974	0.159
Party member	Yes = 1, no = 0	0.018	0.133
Health	Bad = 1, poor = 2, general = 3, good = 4, very good = 5	3.473	0.991
Education	No education experience = 0, primary school = 1, junior high school = 2, high school or technical secondary school = 3, junior college = 4, undergraduate and above = 5	1.375	0.970
Live with parents	Yes = 1, no = 0	0.416	0.493
Average health of parents	Bad = 1, poor = 2, general = 3, good = 4, very good = 5	2.733	0.976
Average age of parents	(Years)	63.459	11.289
Children aged 0–3	Number of children aged 0–3	0.190	0.436
Children aged 4–6	Number of children aged 4–6	0.137	0.361
Children aged 7–16	Number of children aged 7–16	0.470	0.679
Household size	The total number of family members	4.264	1.258
Land area	Currently actual operating land area (mu, the Chinese version of acre, which is commonly 666.7 square meters)	2.186	8.838
Other members’ income	The income of other members in peasant households, take the natural log	10.204	1.899
Distance to town	Distance from peasant households to the nearest town (km)	2.126	3.474

.

4.3. Model Specification

4.3.1. IV-Probit Model

To explore the impact of participation in relocation on women’s non-agricultural employment behavior, referring to existing literature, this paper established the following Probit model:

$$Prob\left(Y_{1i}=1\right)=\phi \left({\alpha }_0+{\alpha }_1{D}_i+{\alpha }_2{X}_i+e_i\right)$$

(1)

In Equation (1), $Y_{1i}$ represents the non-agricultural employment behavior of rural women i, which is a 0–1 binary selection variable, $Y_{1i}=1$ indicates that women participate in non-agricultural employment, $D_i$ represents the participation in relocation of women i, $X_i$ represents a series of control variables, and $e_i$ is the random error terms.

The participation in the relocation of the peasant households where the women belong may be endogenous. On the one hand, participation in poverty alleviation relocation is a self-selected behavior. Whether or not rural households participate in relocation will be affected by factors such as the original housing characteristics, family economic conditions, and capital endowments [44]. There is a certain sample selection bias. On the other hand, while participation in relocation affects women’s non-agricultural employment, women’s non-agricultural employment may also affect the relocation behavior of peasant households. The fuller the non-agricultural employment of women’s labor forces, the stronger the “non-agricultural” characteristics of the household livelihood strategy, and the greater the possibility of participating in the relocation. This two-way causality may lead to endogeneity. Therefore, this paper introduced the instrumental variable of the proportion of relocated peasant households in the village and used the IV-Probit model for two-stage estimation to solve bias caused by potential endogeneity problems.

4.3.2. IV-Tobit Model

Considering that non-agricultural employment days have a value of 0, which is limited continuous censored data, this paper established the following Tobit model to explore the impact of participation in relocation on women’s non-agricultural employment intensity:

$$\{\begin{array}{c}{Y}_{2i}{}^{\ast }={\beta }_0+{\beta }_1{D}_i+{\beta }_2{X}_i+e_i\\ Y_{2i}=\max \left(0,Y_{2i}{}^{\ast }\right) \end{array}$$

(2)

In Equation (2), $Y_{2i}{}^{\ast }$ is a latent variable, and $Y_{2i}$ represents the non-agricultural employment intensity of women i. $D_i$ represents the participation in relocation of women i, $X_i$ represents a series of control variables, and $e_i$ is the random error terms. Similarly, by introducing an instrumental variable, this paper adopted the IV-Tobit model to solve potential endogeneity problems.

4.3.3. The Mediation Effect Test Model

To explore whether the three mechanisms of participation in relocation on the non-agricultural employment of the women’s labor force are valid or not, this paper used the stepwise regression analysis proposed by Baron et al. [45] to test the mediation effect, and established the following model:

$$Y_i=c_1{D}_i+c_2{X}_i+e_{1i}$$

(3)

$$M_i=a_1{D}_i+a_2{X}_i+e_{2i}$$

(4)

$$Y_i=d_1{D}_i+d_2{M}_i+d_3{X}_i+e_{3i}$$

(5)

In Equations (3)–(5), $Y_i$ represents the non-agricultural employment of women i (including the behavior and intensity of non-agricultural employment). $D_i$ and $X_i$ are the same as the previous equations, $M_i$ are a series of moderators, $c$,$a$, $d$ are the estimated coefficients, and $e_{1i}, e_{2i}, e_{3i}$ are random error terms. Substituting Equation (4) into Equation (5), the mediation effect $d_2{a}_1$ can be obtained, that is, the effect of participation in relocation on the non-agricultural employment of women’s labor forces through moderators. The proportion of the mediation effect in the total effect can be obtained by $d_2{a}_1/c_1$.

5. Model Estimation Results and Analysis

Based on the model specification, the STATA statistical software was used for the IV-Probit model, IV-Tobit model, and the mediation effect test model. Manual calculations were used to account for the proportion of mediation effect.

5.1. Estimated Results of Participation in Relocation on Women’s Non-Agricultural Employment

Table 2. Results of the impact of participation in relocation on women’s non-agricultural employment.

	Non-Agricultural Employment Behavior			Non-Agricultural Employment Intensity
Variable	Probit	IV-Probit		Tobit	IV-Tobit
		The First Stage	The Second Stage		The First Stage	The Second Stage
	(1)	(2)	(3)	(4)	(5)	(6)
Participate in relocation	0.190 **		0.726 ***	0.743 **		3.357 ***
	(0.076)		(0.203)	(−0.324)		(0.980)
Age	0.122 ***	0.002	0.116 ***	0.563 ***	0.002	0.554 ***
	(0.030)	(0.007)	(0.029)	(−0.114)	(0.007)	(0.116)
Age square	−0.002 ***	−0.000	−0.001 ***	−0.007 ***	0.000	−0.007 ***
	(0.000)	(0.000)	(0.000)	−0.001	(0.000)	(0.001)
Household head	0.645 ***	0.0887 *	0.567 ***	2.758 ***	0.089 *	2.460 ***
	(0.166)	(0.049)	(0.166)	(−0.685)	(0.053)	(0.708)
Marriage	−1.195 ***	0.146 *	−1.258 ***	−2.135 **	0.146 *	−2.612 ***
	(0.396)	(0.080)	(0.389)	(−0.942)	(0.077)	(0.981)
Party member	0.728 **	−0.131	0.775 **	2.137 **	(0.131)	2.501 **
	(0.320)	(0.088)	(0.315)	−1.019	(0.083)	(1.053)
Health	0.214 ***	0.014	0.202 ***	0.964 ***	0.014	0.938 ***
	(0.043)	(0.013)	(0.042)	(−0.178)	(0.013)	(0.182)
Education	0.210 ***	−0.017	0.214 ***	0.901 ***	−0.017	0.951 ***
	(0.046)	(0.013)	(0.046)	(−0.185)	(0.014)	(0.190)
Live with parents	0.205 **	−0.033	0.234 **	0.903 **	−0.033	1.082 **
	(0.099)	(0.031)	(0.098)	−0.421	(0.031)	(0.434)
Average health of parents	0.211 ***	0.007	0.203 ***	0.753 ***	0.007	0.753 ***
	(0.058)	(0.018)	(0.057)	(−0.238)	(0.018)	(0.244)
Average age of parents	−0.010 *	−0.000	−0.010 *	−0.046 *	0.000	−0.044 *
	(0.006)	(0.002)	(0.006)	(−0.025)	(0.002)	(0.026)
Children aged 0–3	−0.476 ***	0.015	−0.461 ***	−2.075 ***	0.015	−2.064 ***
	(0.104)	(0.032)	(0.102)	−0.441	(0.031)	(0.448)
Children aged 4–6	−0.268 **	−0.015	−0.252 **	−1.146 **	−0.015	−1.109 **
	(0.107)	(0.033)	(0.105)	(−0.470)	(0.033)	(0.479)
Children aged 7–16	−0.080	−0.030	−0.061	−0.266	−0.030	−0.185
	(0.061)	(0.019)	(0.061)	(−0.255)	(0.019)	(0.261)
Household size	0.048	0.009	0.041	0.140	0.009	0.108
	(0.044)	(0.013)	(0.044)	(0.184)	(0.013)	(0.187)
Land area	−0.019	0.001 **	−0.019	−0.105 **	0.001	−0.105 **
	(0.014)	(0.001)	(0.013)	(−0.044)	(0.001)	(0.044)
Other members’ income	−0.175 ***	0.015 **	−0.1776 ***	−0.477 ***	0.0152 ***	−0.511 ***
	(0.028)	(0.007)	(0.028)	(−0.071)	(0.006)	(0.074)
Proportion of relocated peasant households in the village		0.779 ***			0.779 ***
		(0.054)			(0.054)	y
Regional characteristics	Yes	Yes	Yes	Yes	Yes	Yes
_cons	−0.938	−0.068	−1.167 *	−9.088 ***	0.779 ***	−10.491 ***
	(0.703)	(0.191)	(0.696)	(−2.75)	(0.054)	(2.850)
F		11.08			14.06
Wald		7.03 *** (p = 0.008)			8.10 *** (p = 0.004)

Note: ***, **, and * show significance levels of 1%, 5%, and 10%. Robust standard error in parentheses.

reports the estimated results of the impact of participation in relocation on the behavior and intensity of women’s non-agricultural employment. As shown in Regressions (1) and (4) of Table 2, the estimation results of the Probit model and Tobit model showed that participation in relocation significantly improved women’s non-agricultural employment behavior and intensity at the 5% significance level. Regressions (2) and (3) are the estimated results of the IV-Probit model. Wald endogeneity results rejected the hypothesis of no endogeneity at the 1% significance level. The estimated coefficient of the proportion of relocated peasant households in the village was significantly positive at the 1% significance level. The F statistic was 11.08, which is greater than the critical value. This rejects the null hypothesis of “the existence of weak instrumental variables” and the selection of instrumental variables in this paper was appropriate. Similarly, the Wald endogeneity results and F statistic in Regressions (5) and (6) led to the same conclusion. After introducing instrumental variables to address the potential endogeneity, the influence significance and influence degree of the estimated coefficient of participation in relocation increased. This result indicates that participation in relocation has a significant positive impact on the behavior and intensity of women’s non-agricultural employment. Hypothesis 1 is verified.

From the estimated results of the control variables, in terms of individual characteristics of women, age had a significant positive impact on non-agricultural employment behavior and intensity, while the age square had a significant negative impact, that is, there was an inverted U-shaped relationship between age and women’s non-agricultural employment. Middle-aged women are the main forces of non-agricultural employment. Household head, party members, health, and education had a significant positive impact on the non-agricultural employment of women, which are consistent with the relevant theoretical hypotheses of human capital and political capital [43]. Marriage was significantly and negatively correlated with non-agricultural employment. Unmarried women face fewer constraints on family care and agricultural production, and they are more likely to transition to non-agricultural livelihoods [46].

In terms of characteristics of peasant households, the average age of parents had a significantly negative impact on non-agricultural employment of women, while the average parental health level had a significantly positive impact. This shows that younger and healthier parents can take better care of themselves, so there is a lower demand for women to sacrifice their working time to support the elderly. Parents who live together can significantly improve the non-agricultural employment of women by relieving the pressure of raising children alone through intergenerational care. The number of children aged 0–3 and 4–6 significantly negatively affected women’s non-agricultural employment, while the negative effect of the number of children aged 7–16 was not significant. Younger children need more care from women. As children grow older, their need for care decreases, and they can also enter boarding schools to study, which greatly eases the conflicting roles of women in raising children. The income of other members was significantly and negatively correlated with women’s non-agricultural employment. The higher the income of others in the households, the less necessary for women to be employed. In addition, compared with the non-agricultural employment intensity, the negative impact of the land area on women’s non-agricultural employment behavior was not significant. The possible reason is that the relocated women are limited by a lower own capital endowment and inadequate labor skills. The agricultural female labor forces released by land abandonment cannot immediately complete the transfer of non-agricultural employment, but it will facilitate additional working days for women who have already been engaged in non-agricultural employment.

5.2. Robustness Test

To check the robustness of the above results, this paper further applied the propensity score matching (PSM) model using nearest-neighbor matching (NNM) and kernel-based matching (KBM), respectively. The propensity score values of relocated and non-relocated households had a relatively large common support area, and the data balance of the two matching methods was good. From the estimated results shown in

Table 3. Robustness test by PSM model.

Matching Method	Non-Agricultural Employment Behavior			Non-Agricultural Employment Intensity
	ATT	S.E.	T-Stat	ATT	S.E.	T-Stat
Before match	0.051	(0.026)	1.970 *	0.281	(0.146)	1.930
NNM	0.065	(0.030)	2.160 **	0.356	(0.168)	2.120 **
KBM	0.066	(0.029)	2.300 **	0.371	(0.163)	2.280 **

Note: ***, **, and * show significance levels of 1%, 5%, and 10%.

, the T-stats after matching were all bigger than the empirical critical value of 1.96, which indicates that the average treatment effect (ATT) was significantly positive at the 5% significant level. Therefore, participation in relocation can significantly improve women’s non-agricultural employment. The empirical results of this study were robust.

5.3. Estimated Results of Mediation Effect

As shown in

Table 4. Estimated results of the mediation effect in the effect of participation in relocation on non-agricultural employment behavior.

Moderators	The Effect of Participation in Relocation on Moderators	The Effect of Moderators on Non-Agricultural Employment	Mediation Effect	Sobel Test	Bootstrap Test	The Proportion of Mediation Effect (%)
				Z-Stat P-Stat	95% Confidence Interval
	(7)	(8)	(9)	(10)	(11)	(12)
Training	0.271 ***	0.123 ***	0.033 ***	Z = 4.514	[0.020, 0.047]	56.9
	(0.024)	(0.025)	(0.007)	P = 0.000
Public service	0.122 ***	0.072 ***	0.009 **	Z = 2.402	[0.002, 0.016]	15.0
	(0.022)	(0.027)	(0.004)	P = 0.016
Land abandonment	0.077 ***	0.084 ***	0.006 **	Z = 2.306	[0.001, 0.012]	11.0
	(0.021)	(0.028)	(0.003)	P = 0.021
Control variables	yes	yes	yes	yes	yes	—

Note: ***, **, and * show significance levels of 1%, 5%, and 10%. Robust standard error in parentheses. The number of repetitions of the deviation correction non-parametric percentile Bootstrap method is 5000. The values in square brackets are the confidence intervals at the 95% confidence level.

, Regressions (7) and (8), respectively, are the estimated results of the impact of participation in relocation on moderators and the impact of moderators on women’s non-agricultural employment behavior after controlling for the effect of participation in relocation, corresponding to $a_1$ and $d_2$ in Equations (4) and (5). Regression (9) is the mediation effect $d_2{a}_1$ obtained by multiplying the coefficients. To further test the robustness of results of the mediation effect, the Sobel method and the deviation correction non-parametric percentile Bootstrap method were used to test the significance of the coefficient $d_2{a}_1$, the results of which are listed in Regressions (10)–(11).

$a_1$ and $d_2$ were significantly positive at the 1% significant level in Regressions (7) and (8), which indicates the significant mediation effects of the 3 moderators. In terms of the Sobel test and Bootstrap test, the mediation effects of training, public service, and land abandonment were significantly positive at the 1% and 5% significance levels, respectively, and the confidence intervals did not include the value of 0 at the 95% confidence level. This indicates that the mediation effects of the 3 moderators were all significant. Hypothesis 2 is partly confirmed.

In addition, in terms of the proportion of the mediation effect to the total effect in Regression (12), the mediation effect of training was the largest, accounting for 56.9% of the total effect. Public service followed, accounting for 15.0% of the total effect, and land abandonment played the smallest role, accounting for 11.0% of the total effect. It can be seen that participation in relocation promotes women’s non-agricultural employment behavior mainly through training mechanisms.

Table 5. Estimated results of the mediation effect in the effect of participation in relocation on non-agricultural employment intensity.

Moderators	The Effect of Participation in Relocation on Moderators	The Effect of Moderators on Non-Agricultural Employment	Mediation Effect	Sobel Test	Bootstrap Test	The Proportion of Mediation Effect (%)
				Z-Stat/ P-Stat	95% Confidence Interval
	(13)	(14)	(15)	(16)	(17)	(18)
Training	0.271 ***	0.686 ***	0.186 ***	Z = 4.519	[0.105, 0.267]	58.5
	(0.024)	(0.140)	(0.041)	P = 0.000
Public service	0.122 ***	0.412 ***	0.050 **	Z = 2.441	[0.010, 0.090]	15.7
	(0.022)	(0.152)	(0.021)	P = 0.015
Land abandonment	0.077 ***	0.482 ***	0.037 **	Z = 2.346	[0.007, 0.068]	11.6
	(0.021)	(0.157)	(0.016)	P = 0.019
Control variables	yes	yes	yes	yes	yes	—

Note: ***, **, and * show significance levels of 1%, 5%, and 10%. Robust standard error in parentheses. The number of repetitions of the deviation correction non-parametric percentile Bootstrap method is 5000. The values in square brackets are the confidence intervals at the 95% confidence level.

lists the estimated results of the mediation effect of the effect of participation in relocation on the intensity of women’s non-agricultural employment. The same as in Table 4, Regressions (13) and (14) correspond to the effect of participation in relocation on the moderators and the effect of the moderators on the non-agricultural employment intensity. The corresponding coefficients of variables were significantly positive at the 1% significance level, which indicates the significant mediation effects of the 3 moderators. In the results of Regressions (15)–(17), the mediation effects of the 3 moderators were significantly positive, and the confidence intervals did not include the value of 0 at the 95% confidence level. This verifies that training, public service, and land abandonment had a significant positive mediation effect on the impact of participation in relocation on women’s non-agricultural employment intensity. Hypothesis 2 is verified.

The results in Regression (18) are similar to those in Table 4. Compared with the other two mechanisms, participation in training played the largest mediation role in the effect of participation in relocation on women’s non-agricultural employment intensity, accounting for 58.5% of the total effect.

6. Discussions

Promoting the non-agricultural employment of relocated women’s labor forces is not only an important way to solve the development dilemma caused by the change of spatial living location and enhance the livelihood sustainability of the relocated peasant households, but also to improve women’s economic status in the households and alleviate their inequality in the allocation of household resources. Therefore, from the dimensions of non-agricultural employment participation and non-agricultural employment, we explored the impact and mechanisms of participation in relocation on women’s non-agricultural employment. The study pointed out that participation in relocation has a positive impact on women’s non-agricultural employment by breaking the constraints from three dimensions. It showed that the traditional gender division within the relocated households has been changed, and the large number of surplus women’s labor forces released by the poverty alleviation relocation has been effectively transferred and reallocated.

The result is particularly important in many critiques of the relocation project. In fact, some scholars believe that the relocation project is not completely voluntary, and it leads to the loss of peasants’ homes, a reduced sense of belonging, social and psychological marginalization, and an increasing level of inequality within the relocated communities [4,47,48,49]. However, consistent with Liu et al. [10] and Zhang et al. [12], these results provide important theoretical evidence of the positive effects of the relocation project, which not only effectively validates the necessity of the relocation project’s implementation but also complements the empirical research on the effect of participation in relocation on increasing non-agricultural employment intensity.

At the same time, unlike the prevailing household-level studies [11,12,13], this paper explored the impact of the relocation project on women from an individual perspective. This is not only a further refinement and decomposition of peasant household-level research but also a more innovative way to deeply analyze the effects of the relocation project on women who are in a more disadvantaged position of household resource allocation. Compared with treating peasant households as homogenized wholes, it better reflects the important role of relocation projects in improving the livelihood status and the welfare of vulnerable female peasants, which represents an important theoretical addition to existing research.

Moreover, in terms of the results on the mechanism, consistent with existing research [22,24,32], insufficient capital, role conflict, and time constraints have important effects on women’s non-agricultural employment. More innovatively, in the context of poverty alleviation relocation, this research further compared the contribution extent of different mechanisms to improving women’s non-agricultural employment. The results pointed out that compared to alleviating role conflicts and time constraints, improving training is the biggest driving factor for the non-agricultural employment of the relocated women’s labor forces. The possible reason is that, on the one hand, even if spare time is increased due to the shortening of agricultural labor time, women will not give priority to the allocation of time in non-agricultural employment [50]. On the other hand, although public service relieves women’s family pressure to some extent, it also brings regional restrictions on women’s non-agricultural employment due to the actual need of balancing work and the household. This increases the difficulty and competitive pressure for female laborers to obtain non-agricultural employment opportunities near relocation areas.

In addition, this paper has some limitations, which can be addressed in future studies. Specifically, (1) this paper focused on the effect of poverty alleviation relocation on women’s non-agricultural employment from the gender perspective. Future studies can further enrich the way of dividing labor force characteristics. For example, from an intergenerational perspective, we can deeply analyze the impact of participation in relocation on peasants’ non-agricultural employment of different age groups. (2) Women’s non-agricultural employment is an important reflection of the improved economic status and empowerment level of relocated women peasants. Based on the important role of women’s welfare enhancement, future studies can strengthen the focus on female peasants and explore the effects of participation in relocation on different gender peasants in terms of other aspects, such as human capital and livelihood resilience.

7. Conclusions and Policy Implications

In the post-relocation era, it is important to fully promote the non-agriculture employment of relocated female peasants. However, existing studies have not paid enough attention to it. In this context, from the perspective of gender, based on the sample data of 1616 rural households in southern Shaanxi, this paper empirically analyzed the impact and mechanisms of participation in relocation on the non-agricultural employment of women. The research conclusions are as follows.

Firstly, participation in relocation had a significant positive impact on women’s non-agricultural employment behavior and intensity.

Secondly, training, public services, and land abandonment played significant positive mediation roles in the effects of participation in relocation on women’s non-agricultural employment. Further analysis showed that, in terms of women’s non-agricultural employment behavior, the mediation effects were 56.9%, 15.0%, and 11.0%, respectively. In terms of women’s non-agricultural employment intensity, the mediation effects were 58.5%, 15.7%, and 11.6%, respectively.

Based on the above conclusions, to further promote the non-agricultural employment of relocated women, this paper provides the following policy implications. Firstly, based on the remarkable promotion effect of non-agriculture employment of female peasants played by poverty alleviation relocation, in the context of the new era of rural revitalization, governments should continue to implement relocation projects in ecologically fragile areas and pay more attention to multi-dimensional follow-up support for female relocated peasants. Secondly, on the one hand, the governments in relocation areas should increase the inclination of training resources for women and improve the applicability and relevance of training; on the other hand, they should increase women’s motivation to participate in training through training subsidies and other means. Thirdly, while raising multiple funds to improve the shortcomings in the construction of public services such as schools and elderly care in relocated areas, the level of utilization of public services by relocated peasant households should also be improved. At the same time, the new concept of gender equality and the new culture of sharing family responsibilities between men and women should also be promoted to reduce the pressure on women to take care of children and support the elderly alone. Finally, to further release surplus female labor forces, the government should accelerate, and guide relocated peasant households to participate in land transfer and land shareholding by accelerating the construction of trading platforms, improving the social security system, and cultivating new agriculture business entities.