Effects of land recuperation on farmers’ social capital: a Chinese field analysis

ABSTRACT

Social capital is an integral part of farmers’ life, which can be exogenously affected by land recuperation. Based on 1240 farmer field survey data in Gansu Province, this paper used the Logit model to analyse the influencing factors of farmers’ participation in land recuperation, and used the entropy method to measure social capital from the three dimensions of social network, social trust and social norms, and further used the propensity matching score method to measure the effect of land recuperation on farmers, and then compared the effects under different fixed ages and education groups. The following factors significantly affected farmers’ participation in land recuperation: age, years of education, migrant workers’ relationships with family and friends, relationship between migrant workers and friends and colleagues in the workplace, number of migrant workers away from home, cultivated land area, and family living standards. Land recuperation had the greatest promotion effect on farmer’ social network (163.9%), followed by social trust (28.0%) and social norm (11.3%). According to the results of group differences, land recuperation most significantly affected the social capital of farmers aged 45–55 years and household heads educated for 9–12 years compared to other age and education groups.

KEYWORDS:

• Land recuperation
• social network
• social trust
• social norm
• propensity score matching

1. Introduction

At present, the world is generally faced with the coexistence of overcapacity and insufficient food, and the degradation of cultivated land, which directly threatens the sustainable development of human beings (Hoang-Khac et al., 2022). Given the challenges posed by cultivated land degradation, it is more important than ever to increase food production and restore cultivated land using sustainable agricultural methods (Mishra et al., 2021). As healthy soils play a key role in supporting sustainable agricultural development (Zhang et al., 2021), countries around the world are actively seeking solutions to cultivated land degradation. Land recuperation policies are commonly implemented in many countries and regions such as the United States, Canada, Nigeria, Japan, and Australia (Deibert et al., 1986; Hiernaux et al., 2009; Karamanos et al., 2012; Lester et al., 2009; Mano et al., 2015). The scope of policy implementation varies greatly, from small river basins to the whole country, and land recuperation methods and compensation standards also show diversified characteristics. For China, China’s per capita cultivated land area is less than half of the world’s per capita level, and the cultivated land quality has deteriorated compared to earlier. China is facing the dual challenge of quantity reduction and quality decline. Also, the ecological environment and the challenges facing cultivated land protection are grim (Li et al., 2019). Land recuperation is an effective measure to protect farmland fertility and restore the ecological environment (Zhang et al., 2017). Land recuperation refers to measures taken by landowners or users to adopt a certain period of no farming, in order to protect, nurture, and restore land fertility with the aim of improving future farming efficiency and realizing sustainable and effective utilization of land (Luo & Zou, 2015). In 2022, the NO. 1 Document of the Central Committee of China made it clear that it is necessary to extend trials of crop rotation and the land recuperation system. Land recuperation is an sustainable agricultural development path, which has an important impact on the food security of farmers and is an important institutional guarantee for the realization of green agricultural development (Liu et al., 2019).

As for social capital, Bourdieu first formally proposed the concept of social capital. Putnam further developed the concept of social capital. Since then, the role of social capital in sustainable development, especially in rural development, has been paid more and more attention. In rural areas, social capital, as a resource that can bring convenience and economic benefits to individuals and organizations (Wan & Qin, 2011), is an important medium for farmers to build individual and family relationship networks (Fei, 2012). Farmers communicate with the government, traders and farmers in the same village through social capital (Kansanga, 2017). Studies have found that social capital can reduce poverty (Meinzen-Dick et al., 2009), improve income (Xie & Wang, 2016), protect property rights (Peng, 2004), ensure sustainable food systems (Bauermeister, 2016; Gramzow et al., 2018) and resist to natural disasters (Carter & Maluccio, 2003).

Some studies have shown that land recuperation has an impact on farmers’ social capital (Ros-Tonen & Derkyi, 2018; Yang & Gong, 2018). However, there is no consensus on the impact of land recuperation on social capital. Guo and Wang (2020) found through empirical research that in the process of gradual improvement of ecological environment, farmers’ social capital would be significantly reduced. Xiang et al. (2020) also proposed that land recuperation makes it harder for farmers to form social capital. But Wei and Bai (2019) believed that land recuperation strengthened the ‘peer effect’ and ‘complementary effect’ in the rural social network, which would enhance the social network of farmers. Xie et al. (2021) also proposed that land recuperation, as a policy that forces farmers to quit agricultural production, would promote farmers’ non-agricultural employment, broaden farmers’ social network, and further promote the stock and quality of farmers’ social capital. In other words, whether land recuperation has a positive or negative influence on social capital, what is the influence path, and what is the specific impact effect, these questions need to be studied.

To sum up, land recuperation and social capital have received extensive attention from scholars in recent years, but there is no consensus on the impact of land recuperation on farmers’ social capital, and few scholars have quantified the impact of land recuperation on social capital. Therefore, this study used a total of 1240 field survey data in Gansu province, and used propensity score matching (PSM) to study the impact effect of land recuperation on farmers’ social capital. In doing so, this study provides theoretical and practical reference for improving land recuperation policy and sustainable agricultural development.

2. Theoretical analysis and research hypotheses

This study selects social network, social trust, and social norm as three dimensions to measure social capital. Every individual is embedded in the social network, which makes the information obtained by the actor more accurate and detailed than otherwise (Granovetter, 1985) and helps to transform the intention of individual participation into actual action (Gao, 2012). The existence of social trust enhances the willingness of actors to share information resources, which is conducive to rectifying information asymmetry (Zhang, 2014). Social norm can restrain and guide individual behaviour and inhibit the creation of opportunistic behaviour. Under the constraints of social norm, such as morality and customs, individual behavioural choices should not only consider personal interests, but also conform to social value recognition.

2.1. Impact of land recuperation on social network

Land recuperation expands farmers’ social network by influencing their ability to acquire information and resources. Social network is a relatively stable social system formed by interaction among individual members of society, which emphasizes the interaction and connection between people (Shi et al., 2018). According to the perspective of embeddedness, individuals are not completely independent when making decisions, and the social network they are in has a certain impact on their behavioural decisions (Granovetter, 1985). At present, the social network in rural areas is still a social structure of differential sequence pattern formed by consanguineous, geographical and kindred ties (Bian, 1997). Mutual exchanges and communication among farmers closely connect individual farmers in a relationship network through social network. This kind of relationship network enables farmers to obtain information quickly and efficiently through the social network, and can also mobilize the land resources of other members of the relationship network to form effective connection and resource incentive.

After land recuperation, rural labour and capital flow to non-agricultural industries, and farmers’ production and life will gradually leave the traditional village, which increases the communication and contact with non-agricultural employees, and expands the social network and social interaction embedded in the social network. In the process of non-agricultural employment, land recuperation farmers accelerate the acquisition of information about non-agricultural industries, and further integrate their own resources to adapt to the changes in production structure and living habits. In this process, through the expansion of its own social network, it not only obtains more resources, but also further disseminates its own known information, thus forming a virtuous cycle and promoting the scope and quality of its social network. Based on this analysis, the present study proposes the following research hypothesis.

H1: Land recuperation has a positive effect on social network.

2.2. Influence of land recuperation on social trust

Land recuperation enhances farmers’ social trust by influencing farmers’ cooperation and value identification. Social trust refers to the subjective probability that a social individual evaluates that other individuals will take a specific action in the future, which will have an impact on the actions of a social individual and can be further subdivided into interpersonal trust and institutional trust (Uslaner & Conley, 2003). Interpersonal trust reflects farmers’ expectations about the verbal commitment of the contact objects and the reliability of written or oral statements. Institutional trust reflects farmers’ trust in systems, contracts and other aspects (Coleman, 1988). Once social trust is formed, farmers are more likely to listen to others’ opinions, generate awareness of resources sharing while giving trust, prevent information asymmetry, and then correct their decision-making behaviours.

After the implementation of the land recuperation policy, land recuperation farmers will have the homogeneity of future production modes due to land recuperation. In addition, land recuperation farmers will communicate with a variety of different people according to their chosen occupations and geographical locations, so as to broaden their horizons and understand the current national policies. At the same time, land recuperation also works through the formation of unified social norm and value identity within the village. Under the leadership of rural elites and village cadres, intangible social norm and value identification will guide farmers to have more interest demands, which will help strengthen the communication between farmers and the collective, thus strengthening interpersonal trust and institutional trust. Based on this analysis, the present study proposes the following research hypothesis.

H2: Land recuperation has a positive effect on social trust.

2.3. Influence of land recuperation on social norm

Land recuperation influences social norm by influencing internal supervision and constraints. Social norm is formed in a certain social environment over a long period of time, and the public’s common understanding of compulsory, permissible or prohibited behaviours (Lyon, 2000; Ostrom, 2000). Farmers’ behaviour choice should not only consider individual rationality, but also consider the collective rationality around them and in a wider range, and take reciprocal actions in line with social value identification, so as to obtain recognition and support from others and maintain their good reputation. In rural areas, once farmers do not fulfil social norm, they will suffer potential losses due to reputation damage within the village (Niu, 2007).

After the implementation of land recuperation policy, under the guidance of herding effect and favour preference, farmers in the village will make them choose behavioural decisions with strong convergence for face and reputation, and constantly modify their behaviours in the process of continuous self-decision-making adjustment, which can be regarded as a potential behavioural constraint and internal supervision system. Farmers will form behavioural constraints and institutional norms within the village due to their participation in land recuperation and the change of production mode after land recuperation. At the same time, the migration of labour to cities will make farmers more aware that individual norms will be affected by social norm, and they will also be aware of the constraints of social customs, rules and regulations, conventions and other behavioural norms on individual subconscious, which will further promote farmers’ social norm. Based on this analysis, the present study proposes the following research hypothesis (Figure 1).

H3: Land recuperation has a positive effect on social norm.

Figure 1. Research framework.

3. Research design

3.1. Data sources and basic information of samples

This study collected data in October 2019, from field survey of Chinese second land recuperation pilot counties in Gansu province. Gansu Province, located in the middle of the Loess Plateau, is not only an important grain-producing area in northwest China, but also a typical dry farming area in China. At the same time, the ecosystem of Gansu province is carrying the important ecological function, which is typical ecological fragile area and ecological severe degradation area in China. Considering the ecological degradation based on typicality and land recuperation pilot scale, this study finally selected Huan, Jingning, Tongwei, and Yongjing counties for the research area. These are not only Chinese second land recuperation pilot counties, but are also poverty-stricken counties at the state level. According to the situation of each county, this study selected one to three pilot towns from each pilot county; randomly selected three to four land recuperation and non-land recuperation administrative villages from each pilot town; randomly selected three to four villages from each administrative village; and finally, randomly selected six to seven farmers from each village. For this survey, the study issued a total of 1300 questionnaires, and recovered 1240 valid questionnaires (effective rate of 95.38%).

The basic information of the interviewed farmers is as follows. The sample size of farmers participating in land recuperation is 605 (48.79%), while that of farmers not participating is 635 (51.21%). The individual characteristics were as follows: there were 910 men (73.39%) and 330 women (26.61%) among the interviewed farmers; their average age was 54 years; and the highest education level of most was primary school education (64.44%). From the perspective of social capital, the average values of farmers’ social network, social trust, and social norm are 0.322, 0.564, and 0.060, respectively.

3.2. Variable selection

3.2.1 Dependent variable

The dependent variable analysed in this study is social capital, which this study measured from the three dimensions of social network, social trust, and social norm. To ensure data continuity, this study dealt with the entropy value of the measurement results of social network, social trust, and social norm scale.

3.2.2 Core independent variable

This study derived the core independent variable from the PSM key variables in the model, and used it to distinguish between the control group (not land recuperation, denoted by 0) and the treatment group (land recuperation, denoted by 1). This process determined the land recuperation policy effect on the farmer social capital model and whether the variables for the farmers participating in land recuperation were independent.

3.2.3 Control variables

Covariates can express more detailed characteristics of farmers, and the more precise the propensity score, the more accurate the matching. The selection of control variables follows two basic principles. First, these variables should affect both independent and dependent variables. Second, dependent variables cannot affect these control variables (Li, 2013). On the basis of these two principles, this study selected the following control variables: sex, age, fixed number of years of education, whether a village cadre or not, years spent in the village, home in the village, traffic condition around the house, participation in a cooperative, ability to use a smart phone, number of family members, number of migrant workers away from home, cultivated land area, condition of family life, migrant workers’ relationships with family and friends, and relationship between migrant workers and their friends and colleagues in the workplace. Table 1 shows the assignment description and descriptive statistics of the above variables.

Table 1. Description of variable assignment and descriptive statistics.

Variable categories	Variable	Meaning and assignment	Mean	Standard deviation
Dependent variable	Social network	Treated by the entropy method	0.322	0.109
	Social trust	Treated by the entropy method	0.564	0.166
	Social norm	Treated by the entropy method	0.060	0.068
Core independent variable	Whether participating in land recuperation	Farmers participating in land recuperation, yes  = 1, no  = 0	0.488	0.501
Control variable	Sex	Male = 1, female = 0	0.734	0.442
	Age	Actual age of farmers (years)	54.288	11.517
	Years of education	Actual education of farmers in number of years	5.377	4.106
	Whether a village cadre	Yes = 1, no = 0	0.053	0.223
	Years spent in the village	Actual number of years in the village	48.618	17.072
	Overall traffic condition around the house	Very inconvenient = 1, inconvenient = 2, moderate = 3, convenient = 4, very convenient = 5	3.801	0.796
	Whether joined a cooperative	Yes = 1, no = 0	0.225	0.418
	Number of people in the farmer who are able to work	Actual number of people capable of working in the family (persons)	2.801	1.152
	Number of migrant workers away from home	Actual number of family members working outside the village	1.213	0.968
	Agricultural acreage	Area of actual cultivated land at home (acres)	10.629	8.779
	Family living standards	Very low = 1, low = 2, moderate = 3, high = 4, very high = 5	2.907	0.799
	Relationship between migrant workers and their relatives and friends	No contact = 1, occasional contact = 2, moderate = 3, frequent contact = 4, constant contact = 5	3.413	0.975
	Relationship between migrant workers and their friends and colleagues in the workplace	No contact = 1, occasional contact = 2, moderate = 3, frequent contact = 4, constant contact = 5	3.315	0.973

3.3. Statistical analysis of sample differences

By comprehensively comparing the differences in the values of various variables between farmers in the treatment group (land recuperation group) and those in the control group (control group) (see Table 2), it can be observed that farmers in the treatment group were younger, had a higher education level, lived longer in villages, had better access to transport, more actively participated in cooperatives, had less arable land, and had better family living standards than did those in the control group. Meanwhile, the treatment group had fewer people with labour capability but more people working outside the home and more friends and colleagues working outside the home. Without considering other influencing factors, farmers in the treatment group had significantly higher social network, social trust, and social norm than did those in the control group.

Table 2. Comparison of farmers’ characteristics.

Variable	Treatment group		Control group		Difference value
	Mean	Standard deviation	Mean	Standard deviation
Social network	0.324	0.108	0.320	0.110	0.004***
Social trust	0.567	0.163	0.561	0.169	0.006***
Social norm	0.062	0.071	0.059	0.066	0.003**
Sex	0.751	0.433	0.716	0.451	0.035***
Age	54.106	11.497	54.345	11.544	−0.239***
Education in years	5.384	4.067	5.369	4.151	0.015***
Whether a village cadre	0.056	0.230	0.049	0.216	0.007**
Years spent in the village	48.731	16.820	48.510	17.321	0.221***
Overall traffic condition of house	3.833	0.826	3.770	0.766	0.063***
Whether joined a cooperative	0.260	0.439	0.192	0.394	0.068***
Number of people in the farmer who are able to work	2.764	1.144	2.836	1.159	−0.072***
Number of migrant workers working away from home	1.246	0.975	1.181	0.962	0.065***
Agricultural acreage	8.578	8.904	12.583	8.201	−4.005***
Family living standards	2.915	0.755	2.899	0.845	0.016**
Relationship between migrant workers and their relatives and friends	3.365	1.013	3.458	0.936	−0.093***
Relationship between migrant workers and their friends and colleagues in the workplace	3.374	0.986	3.258	0.958	0.116***

Note: *, ** and *** indicate significant at the 10%, 5% and 1% levels respectively.

3.4. Test of variables

This study selected different observation variables for measurement when measuring social network, social trust, and social norm. Table 3 shows the specific questionnaire measurement items. This study used the Likert scale to measure the overall scale, with 1–5 indicating the range from complete disagreement to complete agreement, respectively; the larger the score, the greater the agreement.

Table 3. Measurement of variables.

Variable categories	Measuring item	Code
Social network	During the Spring Festival, I visited, called, sent messages, etc. to convey New Year’s greetings to many relatives	SN1
	During the Spring Festival, I visited, called, sent messages, etc. to convey New Year’s greetings to many friends	SN2
	I have a lot of contacts on my phone	SN3
	I have many siblings and cousins	SN4
	I keep in touch with many relatives	SN5
	I have a lot of close friends	SN6
	I can turn to many people for help when I am in great financial trouble	SN7
Social trust	I have great trust in my friends and neighbours	ST1
	I have great faith in the ordinary farmers of this village	ST2
	I have great faith in the nobility of the village	ST3
	I have great faith in the village cadre	ST4
	I have great faith in the rules and regulations of the village	ST5
	I have great trust in strangers	ST6
	I have great faith in township/town leaders	ST7
	I have a lot of trust in the county/district leaders	ST8
Social norm	My family and I often take part in group activities in the village	SN1
	My family and I often give suggestions or opinions for public affairs decisions in the village	SN2
	My family and I often pay attention to state affairs and social news	SN3
	Relations among the farmers in the village are very united	SN4
	The village has a better social atmosphere than that of surrounding villages	SN5
	There is good implementation of the village’s rules and regulations	SN6

To confirm the validity and reliability of the scale, this study used SPSS 23.0 for the reliability and validity test analysis of the questionnaire. Table 4 shows the test results. This study used Cronbach’s alpha to test the reliability of the questionnaire. The reliability test results show that the Cronbach’s alpha of the total scale is 0.743, and the Cronbach’s alpha of the social network, social trust, and social norm scale are all above the acceptable level of 0.6. Therefore, the questionnaire passed the reliability test. This study used the Kaiser–Meyer–Olkin (KMO) value of the Bartlett’s sphericity test and factor loading coefficient to test the validity of the questionnaire. The results of the validity test showed that the KMO value of the total scale was 0.827, the KMO value of each subscale was higher than the ideal level of 0.7, and the Bartlett’s sphericity test was significant at the 1% level. In addition, the factor loading coefficient of each observed variable was higher than 0.6, indicating that the questionnaire had good structural validity. Therefore, the questionnaire passed the validity test.

Table 4. Results of reliability test and validity test.

Latent variable	Observational variable	Factor loading	Cronbach’s alpha	KMO	Bartlett’s sphericity test	Significance
Social network	SN1	0.697	0.679	0.804	3433.041	0.000
	SN2	0.748
	SN3	0.728
	SN4	0.801
	SN5	0.660
	SN6	0.767
	SN7	0.675
Social trust	ST1	0.614	0.851	0.815	6805.349	0.000
	ST2	0.662
	ST3	0.768
	ST4	0.766
	ST5	0.775
	ST6	0.804
	ST7	0.820
	ST8	0.747
Social norm	SN1	0.661	0.691	0.709	2726.405	0.000
	SN2	0.659
	SN3	0.655
	SN4	0.803
	SN5	0.866
	SN6	0.813
Total scale			0.743	0.827	13928.035	0.000

3.5. Measurement of social capital

The entropy method is an objective and weighted multi-indicator comprehensive evaluation method. The weight of each indicator is determined according to the degree of association of each indicator or the amount of information provided. It can effectively avoid subjective factors affecting the evaluation results and is a common tool in multi-indicator evaluation (Xie et al., 2019). For measuring the social capital of farmers, this study adopted the entropy method to assign weight to the indicators. The specific steps are as follows.

The first step is the standardization of indicators. Considering the dimensional differences among indicators and eliminating the difference of evaluation index units in the impact evaluation matrix of extreme values, it is not possible to conduct a unified calculation directly, and the attribute values of each indicator need to be standardized (Liu et al., 2020). Let x_ij be the j (j = 1,2 … n) index of the i (i = 1,2 … m) dimension, and the standardized treatment is $\mathrm{I}\mathrm{n}\mathrm{d}\mathrm{i}\mathrm{c}\mathrm{a}\mathrm{t}\mathrm{o}\mathrm{r}\mathrm{s}:Z_{ij}={\displaystyle \frac{x_{ij}-min{x}_j}{max{x}_j-min{x}_j};\mathrm{I}\mathrm{n}\mathrm{v}\mathrm{e}\mathrm{r}\mathrm{s}\mathrm{e}\mathrm{i}\mathrm{n}\mathrm{d}\mathrm{i}\mathrm{c}\mathrm{a}\mathrm{t}\mathrm{o}\mathrm{r}:Z_{ij}={\displaystyle \frac{max{x}_j-x_{ij}}{max{x}_j-min{x}_j}}}$where Z_ij is the standardized index value, x_ij is the original data of the j index of i dimension, and min x_j and max x_j are the minimum and maximum values of the j index, respectively.

The second step is to calculate the information entropy index e_j: $e_j=-k\sum _{i=1}^m{y}_{ij}ln{y}_{ij}$, where k = 1/lnm

The third step is to calculate the difference coefficient of the $j$ index $dj$: $dj=1-ej$

The fourth step is to calculate the index weight W_j: $Wj=dj/\sum _{i=1}^m{d}_j$

The fifth step is to calculate social capital S: $S=\sum _{i=1}^n{y}_{ij}{W}_j$

3.6. Model construction

This study considered land recuperation farmers as the treatment group and non-land recuperation farmers as the control group. In the estimation of the effect of grouping policy, the problem of sample selectivity arose; that is, the distribution of samples in the treatment group and the control group was not random (Xu & Li, 2015), and the actual distribution process often followed certain criteria.

In realistic conditions not affected by the land recuperation policy group, it is not possible to observe directly the results of land recuperation compensation policy; to obtain a controlled result affected by the land recuperation compensation policy framework, it is necessary to build on the notion that no event would occur under the condition of potential results or events. Robins (1999) proposed a PSM method based on a counterfactual framework. Similarly, it is necessary for this study to establish a counterfactual framework and to measure the impact of land recuperation on farmers’ social capital using the PSM method.

Figure 2. Steps of counterfactual analysis framework.

The research steps of the counterfactual analysis framework in this study are as follows (Figure 2).

The first step is to select the control variable. By referring to relevant literature, the model includes factors that affect farmers’ participation in land recuperation and social capital.

The second step is to calculate the propensity score. This study used the Logit model to calculate the propensity score of farmers to participate in land recuperation.

The third step is PSM, which is a statistical method that uses non-experimental data or observational data to evaluate intervention effects. Its core is the calculation of the propensity score, which refers to the probability that independent variables under the control of observable covariates affect the samples (Hu, 2012).

To ensure the matching effect, first, to balance the test samples (i.e. whether the overlap hypothesis holds), the treatment group and the control group in various collaborator variables no longer have different systems, and the principle is to control the matching of the variable difference, to eliminate the differences of covariate influence on outcome variables, in order to assess the net effect (Xu & Li, 2015). Generally, the closer bias is to 0 after matching, the better the matching effect. When bias is less than 10%, the equilibrium of variables between groups is better (Yang et al., 2008).

As for the matching method, owing to certain measurement deviations of different matching methods, even when processing the same sample data, the measurement results generated are heterogeneous. There is still no consensus in academia as to which method to choose for matching to achieve the optimal results. However, if the results obtained by using multiple matching methods are similar or even consistent, it means that the matching results are robust, and the sample validity is good. Therefore, to enhance the reliability of the research conclusions, this study used the following four mainstream methods for matching: (1) k-nearest neighbour matching, that is, matching by looking for k individuals of different groups with the nearest propensity score; in this study, k is set as 4 to carry out one-to-four matching so as to minimize the mean square error; (2) inside the caliper match, that is, the absolute distance to limit the propensity score; after calculation, the caliper range is set as 0.07; (3) core match; this study used the default kernel and bandwidth; and (4) spline matching; this study used the spline command for default regression (Wang & Kong, 2019).

The fourth step is to calculate the average treatment effect. There are three types of average treatment effects: the average treatment effect of the treatment group (ATT), that is, the average value of changes in the social capital of farmers participating in land recuperation. The second is the average treatment effect of the untreated group, that is, the average value of social capital changes of farmers who did not participate in land recuperation. The third is the average treatment effect of the whole sample, that is, the average change in social capital of all sample farmers (Wang & Kong, 2019). Because this study explored the impact of land recuperation on farmers’ social capital and focuses on the changes in the social capital of farmers participating in land recuperation, it selected ATT for the analysis. $\mathrm{A}\mathrm{T}\mathrm{T}={\displaystyle \frac{1}{N}{\sum }_{i:D_i=1}\left(y_i-{\hat{y}}_{0i}\right)}$where N represents the number of farmers in the treatment group, that is, the number of farmers participating in land recuperation farming; $\sum _{i:D_i=1}$ means the inclusion of only the land recuperation farmers;$y_i$ is farmer, $i^{\mathrm{\prime }}$s social capital; and ${\hat{y}}_{0i}$ is a hypothetical state, or the actual participation in land recuperation farming if not involved in the estimates of the social capital of land recuperation.

4. Analysis of influencing factors of farmers’ participation in land recuperation

To achieve sample matching, this study analyses the factors influencing farmers’ participation in land recuperation, and Table 5 shows the estimated results. Determining the key factors can identify the key population, clarify the aspects to encourage farmers to participate in land recuperation, promote the improvement of farmers’ social capital, and then improve societal benefits.

Table 5. Estimated results of farmers’ participation in land recuperation based on Logit model.

Variable name	Regression coefficient	Standard error
Sex	0.098**	0.047
Age	0.325*	0.190
Education in years	0.445***	0.153
Whether a village cadre	0.062***	0.009
Years spent in the village	0.114***	0.042
The overall traffic condition of your house	0.092*	0.052
Whether joined a cooperative or not	0.058***	0.021
Number of people in the farmer who are able to work	−0.164	0.118
Number of migrant workers away from home	0.157***	0.047
Agricultural acreage	−0.138**	0.067
Family living conditions	0.138*	0.080
Relationship between migrant workers and their relatives and friends	−0.319***	0.077
Relationship between migrant workers and their friends and colleagues in the workplace	0.339***	0.081
Pseudo R^2	0.0843 144.92 1240
LR chi^2
Sample size

Note: *, ** and *** indicate significant at the 10%, 5%, and 1% levels, respectively.

As Table 5 shows, age, years of education, the relationship between migrant workers and their relatives and friends, and the relationship between migrant workers and their friends and colleagues in the workplace are important factors affecting farmers’ participation in land recuperation. Age is positively correlated with participation in land recuperation; that is, older farmers have a stronger tendency to participate in land recuperation. This may be because, on the one hand, older farmers have a strong attachment to land and understand that land needs to lie land recuperation to maintain its fertility. On the other hand, older farmers, constrained by their own labour capacity and physical strength, may prefer to allow the land to lie land recuperation. In addition, there is a positive correlation between years of education and participation in land recuperation, indicating that farmers with higher education levels have a stronger willingness to participate in land recuperation. There is a significant (at 1%) negative impact between the migrant workers’ relationship with their friends and family and the intention to participate in land recuperation, showing that closer relationships with friends and family is not conducive to improving farmers’ willingness to participate in land recuperation. They tend to live in their village, and cultivated land is their biggest source of income. The relationship between migrant workers and their friends and colleagues in the workplace has a significant positive influence on the land recuperation intention at the statistical level of 1%, indicating that the closer the relationship between migrant workers and their friends and colleagues in the workplace, the more favourable it is for farmers to participate in land recuperation, and the more willing they are to move to other places for employment after land recuperation, so as to maintain the relationship with their friends and colleagues in the workplace. In addition, whether a village cadre, years of life spent in the village, traffic conditions, participation in cooperatives, number of migrant workers away from home, and family living standards all have significant positive effects on farmers’ participation in land recuperation, while the cultivated land area has a significant negative effect on farmers’ willingness to participate in land recuperation.

5. Measurement of the effect of land recuperation on farmers’ social capital

5.1. Analysis of matching results

To ensure the matching quality of the sample data, this study draws a test of common support domain. According to the matching results of samples under four different matching methods (see Table 6), this study retains 1234 matching samples in the treatment group and the control group after the removal of six samples, indicating a good matching effect.

Table 6. Results of common support domain.

	Unmatched sample	Matched sample	Total
Control group	2	633	635
Treatment group	4	601	605
Total	6	1234	1240

5.2. Balance test

To ensure the reliability of the PSM results, this study tests the balance of covariables. After matching, there were no significant systematic differences in covariables except for differences in social network, social trust, and social norm between farmers in the control group and the treatment group. According to the balance test results (see Table 7), after sample matching, the bias of explanatory variables decreased from 13.3% to 2.9%, 5.4%, and 8.6%, and the bias significantly reduced to less than the 10% standard specified in the balance test. The pseudo R² decreased from 0.068 before matching to 0.004–0.017 after matching. The LR statistics decreased from 117.38 before matching to 5.89–28.80 after matching. According to the analysis of these test results, application of PSM can effectively reduce the difference in the distribution of explanatory variables between the control group and the treatment group and eliminate the estimation bias caused by sample self-selection.

Table 7. Results of balance test of explanatory variables before and after propensity score matching.

Matching methods	Pseudo R^2	LR	Bias (%)
Before the match	0.068	117.38	13.3
K-nearest neighbour match	0.007	11.60	5.4
Inside the caliper match	0.007	11.55	5.4
Core match	0.004	5.89	2.9
Spline match	0.017	28.80	8.6

5.3. Effect of land recuperation on social capital of farmers and analysis of group differences

5.3.1 Measurement of impact effect

This study estimates the land recuperation on social network, social trust, and social norm. The ATT estimation results (see Tables 8) show that after using the four different methods for measuring results, the sample data have good robustness. Therefore, this study selects the arithmetic mean characterization effect for later analysis.

Table 8. Processing effect of propensity score matching.

Matching methods	Social network	Social trust	Social norm
K-nearest neighbour match	1.757**	0.274***	0.126**
Inside the caliper match	1.742**	0.275***	0.126**
Core match	1.527*	0.289***	0.104*
Spline match	1.528*	0.281***	0.097*
Mean	1.639	0.280	0.113

Note: *, ** and *** indicate significant at the 10%, 5%, and 1% levels, respectively.

Table 8 shows that, after the counterfactual estimation of PSM, land recuperation has a significant positive impact on farmers’ social network, with a net effect of 1.639, indicating that after considering the farmers’ selectivity bias, participation in land recuperation can significantly improve farmers’ social network by 163.9%, thereby verifying H1. From the perspective of the social trust dimension, the ATT was 0.280, indicating that, excluding other factors, participation in land recuperation significantly increased the social trust of farmers by 28.0%, thereby verifying H2. From the perspective of social norm, the ATT was 0.113, indicating that, excluding other factors, participation in land recuperation significantly increased the social norm of farmers by 11.3%, thereby verifying H3.

5.3.2 Group difference analysis

Even within the same region, the participation of different types of farmers in land recuperation varied greatly. Although this study selects ATT to measure the net effect of land recuperation on social capital, ATT can reflect only the average change of social capital of farmers participating in land recuperation and cannot reflect the structural difference of the influence effect of sample farmers, namely, group differences. Therefore, it would be helpful to explore the group differences of different types of farmers to enrich the research of land recuperation on farmers’ social capital.

As important human capital variables, age, and years of education can influence farmers’ choice of land recuperation. Therefore, it is necessary to analyse group differences based on age and years of education, focusing on the effect of land recuperation on farmers’ social capital. Based on this, this study uses age and years of education as markers to group the samples and to test the inter-group differences in the effect of land recuperation on farmers’ social capital. Table 9 shows a comparison of the results of the influence effect of land recuperation on farmers’ social capital based on the k-nearest neighbour one-to-four matching method.

Table 9. Group differences of land recuperation effect on farmers’ social capital.

Variable	Classify standard	Social network		Social trust		Social norm
		ATT	S.E.	ATT	S.E.	ATT	S.E.
Age (years)	$\le$35	2.391	3.390	0.061	0.366	−0.191	0.230
	35–45 (inclusive)	0.649	1.629	0.145	0.175	0.147	0.107
	45–55 (inclusive)	4.102**	1.768	0.497**	0.196	0.440***	0.125
	>55	−0.696	1.624	0.196	0.221	0.070	0.133
Education years (years)	<6	0.311	2.100	0.208	0.145	0.129	0.090
	6–9 (inclusive)	2.524	1.506	0.302	0.176	0.070	0.109
	9–12 (inclusive)	3.412***	1.282	0.388*	0.212	0.239*	0.138
	$\ge$12	3.948	2.875	0.552*	0.292	−0.343	0.216

Note: *, ** and *** indicate significant at the 10%, 5%, and 1% levels, respectively.

Age not only has a significant positive effect on land recuperation participation, but also indirectly affects farmers’ social network, social trust, and social norm after land recuperation participation. Table 9 shows that farmers aged between 45 and 55 years significantly improved their social network (410.2%), social trust (49.7%), and social norm (44.0%) after participating in land recuperation. A possible reason is that farmers aged 45–55 years are the main component of the current rural labour force, and their experience farming arable land has made them aware that relying only on farming for a living would cause their income to become unstable; they might consider themselves to be advancing in age, and in the next few years might not have the strength for engaging in several more years of cultivation, making them more willing to choose land recuperation. At the same time, to maintain or even improve quality of life, these farmers might actively explore and obtain more work information and network resources, thereby enriching their social capital.

In addition to age, years of education act on land recuperation, which in turn affects farmers’ social capital. Table 9 shows that social network (341.2%), social trust (38.8%), and social norm (23.9%) of farmers with a schooling period of 9–12 years significantly increased after they participated in land recuperation farming. Farmers with 12 or more years of education had a significant 55.2% increase in social trust after land recuperation participation. A possible reason is that the level of education affects farmers’ knowledge and foresight to a certain extent. The higher the level of education, the higher the understanding of national policies of the household head, and the stronger the implementation of the land recuperation policy, making farmers more likely to choose to seek employment outside the home, so as to improve their social capital.

6. Conclusions and recommendations

This study draws results from four cities in Gansu province, employing field survey data collected from 1240 farmers. This study used the entropy method to measure social capital, the Logit model to analyse the factors influencing farmers’ participation in land recuperation, and PSM to calculate the land recuperation effects on farmers’ social capital effect. Then, this study compared the effect under different ages and levels of education and reached the following core conclusions.

1. Such factors as age, years of education, relationship between migrant workers and their relatives and friends, relationship between migrant workers and their friends and colleagues in the workplace, number of migrant workers away from home, cultivated land area, and family living standards significantly affected farmers’ participation in land recuperation. Therefore, China should strengthen the implementation and popularization of its land recuperation policy and improve compensation performance. Adopting various forms of land recuperation publicity methods and enriching land recuperation compensation methods based on monetary compensation would encourage farmers with higher education levels, smaller cultivated land areas, and better family living standards to participate actively in land recuperation farming.

2. Land recuperation has a significant positive impact on farmers’ social capital. Among the dimensions of social capital, the net effects of land recuperation on farmers’ social network, social trust, and social norm were 163.9%, 28.0%, and 11.3%, respectively. After empirical test, this study is consistent with the conclusions of Wei and Bai (2019) and Xie et al. (2021). At the same time, both of the above two papers are qualitative analysis of the relationship between land recuperation and social capital theoretically, while this study has quantitatively measured the effect of land recuperation on social capital through PSM, to some extent, which further fills the gap in the current research field. Based on these findings, this study proposes the following three suggestions. Firstly, while encouraging farmers to actively go out for employment, China government should also provide farmers with non-agricultural employment technical training and a platform for exchange of employment experience among farmers, improve the stability of non-agricultural employment and increase income effect, and avoid idle labour after land recuperation, so as to effectively expand the social network of farmers brought by land recuperation. Secondly, village cadres should actively respond to and implement the land recuperation policy, so that farmers can effectively understand the contents of the land recuperation policy, and increase farmers’ trust in the national policy and village committee. At the same time, village cadres can also hold village cultural festivals regularly to promote communication among farmers and increase social trust among farmers. Thirdly, the village committee should clarify the provisions of rural culture within the village, strengthen farmers’ cognition of land recuperation and their responsibility consciousness, promote the typical cases of obvious recovery of cultivated land quality and improvement of farmers’ livelihood through lectures and exhibition boards, improve farmers’ perception and identity of land recuperation, and further enhance farmers’ social norm.

3. The results of group differences showed that land recuperation had a more significant impact on the social capital of farmers aged 45–55 years and with years of education of 9–12 years than on groups of other ages and with other years of education. This finding further refines the range of effects of land recuperation on farmers’ social capital. This certainly gives us some insight, goverment should focus on key groups and implement differentiated land recuperation compensation programmes. In view of the heterogeneity of farmers, differentiated land recuperation compensation could significantly improve the efficiency of compensation for the key population group of farmers left behind in rural areas, those with junior middle school or above education level, and those who are middle-aged.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This research was supported by the Fundamental Research Funds for the Central Universities, and the Research Funds of Renmin University of China (23XNH031).