Theoretical analysis of the effects of rural revitalization talent-training programs on farmers’ social capital

Rural revitalization talent-training programs not only bear the direct mission of enhancing farmers’ human capital [14], but also exert profound influences on rural social networks and the configuration of social capital by reshaping patterns of social interaction and resource allocation. Drawing on the core dimensions of social capital theory, this study develops an analytical framework of “network restructuring–resource empowerment–norm reconfiguration” to systematically elucidate the mechanisms through which farmers’ participation in rural revitalization talent-training programs promotes social capital accumulation. Specifically, this study mainly follows Bourdieu’s perspective that social capital represents the actual or potential resources embedded in social relationships, emphasizing individuals’ ability to obtain resources through established networks [13,15]. Coleman’s view on institutional embeddedness and Putnam’s emphasis on trust and norms are further incorporated to explain the mechanisms through which training programs reshape resource access and interpersonal relationships. It should be noted that the social capital indicator used in this study—whether a farmer can borrow a certain amount of money from relatives or friends—primarily captures farmers’ capacity to access resources through kinship- and locality-based ties, thereby reflecting bonding social capital. While this measure effectively reflects mutual assistance and support within acquaintance-based networks, it cannot directly capture bridging social capital, which is characterized by connections across different social groups.

First, skill acquisition and social network restructuring. Bourdieu defines social capital as the aggregate of actual or potential resources linked to possession of a durable network of institutionalized social relationships at the individual or group level [13,15]. Rural revitalization talent-training programs, through systematic skills training and organized, intensive learning arrangements, create structural opportunities for farmers to expand the boundaries of their social interactions. During training, farmers participate in centralized instruction, technical exchanges, and experience-sharing activities, gradually moving beyond relatively closed networks dominated by kinship and locality to establish cross-group and cross-hierarchical connections with heterogeneous actors such as agricultural experts, business operators, and cooperative leaders [15]. This expansion of networks characterized by weak ties not only increases the diversity of farmers’ social relations, but also substantially enhances their capacity to obtain heterogeneous information and external resources, thereby laying a network foundation for social capital accumulation.

Second, institutional embeddedness and structural resource empowerment. Coleman emphasizes the structural nature of social capital and argues that its sustained accumulation depends on support from institutional arrangements [16]. Rural revitalization talent-training programs, through institutionalized design, embed farmers into new channels for resource access and public participation. At the policy level, the establishment of an integrated institutional system linking “education and training–qualification certification–policy support” enables participating farmers to obtain more development opportunities within formal institutional frameworks. On the one hand, farmers who obtain relevant certifications are granted priority in credit support, project applications, and policy coordination; on the other hand, talent certification functions as an institutionalized signal that enhances farmers’ credibility and visibility in market transactions and public affairs. Such institutional embeddedness not only empowers farmers in terms of resources, but also strengthens their capacity to participate in grassroots governance and public affairs, thereby expanding the institutional basis of their social capital.

Third, identity recognition and the reconfiguration of trust norms (knowledge diffusion and normative mechanisms). Putnam points out that social capital is reflected not only in network structures but also in cultural and cognitive dimensions such as trust and norms [17]. By strengthening knowledge diffusion and professional competence recognition, rural revitalization talent-training programs promote the reconstruction of farmers’ occupational identity. When farmers acquire new occupational identities—such as “rural artisans” or “local experts”—through systematic training, their professional knowledge and technical capabilities gain recognition within the community. This not only enhances individuals’ self-efficacy but also fosters, at the community level, a trust mechanism grounded in professional competence. Compared with the acquaintance-based trust typical of traditional relationship-oriented societies, this competence-based trust is more open and scalable. It helps overcome the constraints of limited trust, facilitates the formation of norms for knowledge sharing and cooperation, and substantially increases farmers’ voice in community governance and collective decision-making.

In summary, rural revitalization talent-training programs systematically promote the accumulation and upgrading of farmers’ social capital through multiple mechanisms, including social network expansion, institutional resource empowerment, and the reconfiguration of trust and norms. Among these mechanisms, the improvement of farmers’ ability to mobilize resources through existing social networks is directly reflected in the measurement of bonding social capital adopted in this study, whereas network expansion, institutional support, and trust formation provide potential explanations for this outcome. This process not only provides sustained empowerment for farmers’ individual development but also lays a solid social foundation for optimizing rural social structures and advancing the rural revitalization strategy in the long run. Based on the above analysis, this study proposes the following hypothesis:

H1: Participation in rural talent revitalization training programs enhances farmers’ bonding social capital.

Theoretical analysis of how different training modalities affect farmers’ social capital

Within the implementation of the rural revitalization strategy, differences between correspondence education and distance education in their effects on improving farmers’ social capital essentially stem from the distinct forms of social interaction and relationship-building mechanisms shaped by different training media. The formation of social capital depends not only on acquiring knowledge and skills, but also—crucially—on the gradual accumulation of trust, social identification, and stable social networks through sustained interaction. It is therefore necessary to systematically analyze how different training modalities shape farmers’ social capital formation mechanisms from the perspectives of interaction patterns, social learning processes, and social network structures.

First, differences in interaction depth shape the accumulation of relational social capital. According to social presence theory [18], correspondence education, through periodic, face-to-face intensive learning and communication, creates interaction settings with high social presence and high frequency, which facilitates multidimensional social learning among participants. Such face-to-face interaction not only promotes the transfer of explicit knowledge, but also cultivates shared cognition and professional identification through observational learning and experience exchange [19], making it easier for participants to form strong relational ties based on trust and reciprocity [20,21]. By contrast, distance education, due to its asynchronous nature and spatial dispersion, filters out social cues to some extent; interaction among participants tends to take the form of low-frequency and fragmented information exchange, thereby constraining the formation of group belonging and trust. Because social capital relies heavily on trust, norms of reciprocity, and emotional bonds formed through repeated interactions, the face-to-face communication and high-frequency interaction environment provided by correspondence education is more conducive to building stable and enduring relational social capital among farmers, whereas distance education is relatively limited in fostering deep, cumulative trust.

Second, differences in social network structure affect pathways of resource access and social capital conversion. Correspondence education typically organizes participants into fixed classes, within which members maintain stable contact over an extended period, gradually forming a high-density, strongly connected, and relatively closed network structure. In such networks, members often assume multiple roles simultaneously—such as technical exchange, information sharing, and mutual assistance—while repeated interactions strengthen informal constraints and improve the efficiency of mobilizing internal resources. In contrast, distance education more readily forms an open network structure centered on platforms or instructors. Relationships among participants are more likely to be weak ties; although this may facilitate information diffusion, the capacity for resource integration and sustained mutual assistance is relatively limited [22]. Especially in rural contexts, whether social networks can be effectively converted into social capital often depends on the stability and continuity of network relations. Accordingly, the high-density networks formed through long-term offline contact in correspondence education are more likely to translate into sustained resource mutual aid and social support, whereas the weak-tie networks formed through distance education may face constraints in their efficiency of social capital conversion due to real-world conditions [23].

Third, differences in the degree of institutional embeddedness influence social identity recognition and the formation of reputational capital. Rural revitalization talent training is not only a process of transmitting knowledge and skills, but also an important carrier of institutional empowerment. Correspondence education is often more closely linked institutionally with local governments, universities, and grassroots organizations, and its training process and outcomes are frequently accompanied by clear academic credentials or qualification certifications. Such institutional embeddedness helps strengthen farmers’ identity recognition within village society and grassroots governance systems, making them more likely to be regarded as key actors who are “technically skilled and capable,” thereby enhancing their social reputation and social recognition [24]. In comparison, distance education relies more on platform-based and technology-mediated approaches; the visibility of training outcomes and the degree of institutional recognition in grassroots communities are relatively weaker, meaning that its social-capital effects depend to a greater extent on individual differences in capability.

Based on the above analysis, given its higher interaction intensity, more stable social network structure, and stronger institutional embeddedness, correspondence education is expected to generate more pronounced effects on social capital improvement than distance education. Accordingly, this study proposes the following hypothesis:

H2: Correspondence education in rural revitalization talent-training programs increases farmers’ bonding social capital by a larger margin than distance education.

Data sources

Jiangxi Province is one of China’s major agricultural provinces and a key rice-producing region. However, for a long time, the stock of rural human capital has been relatively insufficient, and the overall quality of the labor force has remained low, making it difficult to meet the practical demand of the rural revitalization strategy for high-quality, versatile rural talent. Against this backdrop, Jiangxi Province took the lead in launching the OVOUS project in 2012, with the aim of systematically cultivating high-level and high-quality practical rural talent. The project combines correspondence education with distance education and adopts diversified teaching approaches—including on-campus study, a flexible schooling system, alternating between agricultural work and study, and delivering instruction to rural areas—to provide higher vocational education for key rural groups such as specialized large-scale farming households, heads of leading agribusiness enterprises, leaders of cooperative organizations, and members and reserve cadres of village-level “two committees.” Existing studies show that the OVOUS project has significantly strengthened participating farmers’ intentions toward entrepreneurship and remaining in agriculture, playing a positive role in promoting industrial revitalization and increasing farmers’ income [25].

With the deepening of the rural revitalization strategy, Jiangxi Province issued the Implementation Opinions on Advancing the One Village, One University Student Project in Depth in 2020. Building on earlier efforts, the province continued to expand the scope of trainees by incorporating leaders of rural social service organizations, returning migrant workers, and returning entrepreneurial youth into the training system, and gradually extended the training level from junior college and undergraduate programs to postgraduate programs. To date, the OVOUS project has been implemented continuously in Jiangxi for 12 years, enrolling more than 70,000 farmer participants. As a result, the number of farmers with higher education increased from an average of fewer than 0.4 per administrative village in 2012 to nearly 5. In terms of duration, scale, training level, and coverage, Jiangxi ranks among the leading provinces nationwide, making it highly demonstrative and representative. Therefore, selecting Jiangxi as the study area helps examine the implementation effects of rural revitalization talent training within a relatively mature and systematic policy context.

Regarding data collection, the research team conducted a household survey in December 2019. The sample included farmers who participated in the OVOUS project after 2012 as the treatment group, and farmers who did not participate as the control group. To enhance representativeness, the survey adopted a multi-stage stratified random sampling strategy. First, 12 counties (county-level cities/districts) were randomly selected from Jiangxi Province based on per capita industrial value added. Second, using per capita public fiscal revenue as the stratification criterion, three townships were randomly selected from each county. Third, three administrative villages were randomly selected from each township. Finally, ten farmers who had not participated in the OVOUS project were randomly selected from each administrative village for interviews. In total, 2,805 questionnaires were distributed; after data cleaning and screening, 2,552 valid questionnaires were obtained, yielding an effective response rate of 90.98%. This sampling design accounted for within-province heterogeneity in geography, economic development, and individual farmer characteristics, thereby providing a relatively reliable data foundation for the empirical analysis.

Ethics statement

The study protocol was approved by the Institutional Review Board of the Research Office of Jiangxi Agricultural University (Approval No. JXAULL-2018032; approved on December 24, 2018). Written informed consent was obtained from all participants before data collection. Before participation, members of the research team provided each participant with written information describing the purpose and procedures of the study, the voluntary nature of participation, confidentiality protections, the use of anonymized data for academic research, and the right to withdraw at any time without adverse consequences. All participants consented to the use, analysis, and publication of their anonymized data for academic purposes.

Selection of variables

1. Explained variable: Farmers’ bonding social capital is selected. This study refers to existing research [26,27] and divides social capital into two categories, namely, bonding social capital, referring to close social relationships formed through blood ties, ethnicity and family ties, and bridging social capital, referring to social relationships connected through colleagues and friends. In this study, the measurement of social capital focuses specifically on bonding social capital, as the available indicator reflects farmers’ ability to mobilize resources through close interpersonal networks rather than broader connections across social groups. Accordingly, the question ‘If there is an emergency at home, can you get a 5,000-yuan loan from friends and family?’ is used as a measure of farmers’ capacity to access reciprocal financial support within their bonding social networks.
2. Key explanatory variable: It involves whether farmers participate in rural revitalization talent cultivation. In this study, whether farmers participate in the OVOUS project is used as a proxy variable for the cultivation of talent in rural revitalization, focusing on the impact of participating in the OVOUS project on the enhancement of farmers’ social capital. Notably, there may be a reverse causality problem between the causal variables. On the one hand, participation in the OVOUS project may promote the enhancement of farmers’ social capital; on the other hand, farmers with rich social capital tend to be more likely to participate in the OVOUS project. Therefore, measures will be taken in the empirical analysis to address the resulting endogeneity problem.
3. Control variables: Based on existing studies [28], the following control variables were selected. First, individual control variables, including age, gender, whether they are party members and whether they are engaged in non-agricultural work. Second, household characteristics control variables, including the number of family members, actual area of arable land operated and number of people in the labor force. Third, locational characteristics control variables, including the topography of the villages, distance from villages to the county towns and economic development level of the region.
4. Instrumental variable: farmers’ knowledge of the OVOUS project policy. To address potential reverse causality and ensure model identification, this study introduces farmers’ awareness of the OVOUS policy as an instrumental variable. This variable is suitable as an instrument primarily because it operates by influencing farmers’ decision-making regarding whether to participate in rural revitalization talent-training programs. After controlling for potential confounders such as individual characteristics, household characteristics, and locational factors, policy awareness itself neither directly alters farmers’ existing social relationship structures nor significantly increases the stock of mobilizable resources within their social networks. Any potential impact on social capital is mainly realized indirectly through the mediating pathway of “participation in rural revitalization talent training.” Therefore, in theoretical terms, this instrument satisfies the exclusion restriction, i.e., it does not directly affect the dependent variable, and thus meets the conditions for a valid instrumental variable. Detailed variable definitions and descriptive statistics are reported in Table 1.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Descriptive analysis of variables.

https://doi.org/10.1371/journal.pone.0357662.t001

Descriptive statistics

The mean and standard deviation of the variables for the whole sample is provided in Table 1. Overall, the sample presents the following characteristics. First, 1502 farmers in the sample participated in rural revitalization talent cultivation, accounting for 58.9% of the sample, of which 809 were in correspondence education and 693 were in distance education. Moreover, 1,050 were non-participants, accounting for 41.1% of the sample. The mean value of social capital was 0.835, indicating that most farmers could get support from friends and relatives by borrowing 5,000 yuan in an emergency. Furthermore, the participation in non-farm work is as high as 0.650, emphasizing its importance as the main source of household income. In addition, the average age of the sample was 46.971 years, with a gender mean of 0.757, reflecting that the respondents were predominantly middle-aged and older males and had a higher proportion of party members (with a mean of 0.458). Notably, despite the high proportion of farmers participating in rural revitalization talent cultivation, the mean value of their understanding of the policy is only 1.989, indicating that the sample’s overall awareness of the policy needs to be improved.

Table 2 further compares the differences in the means of the variables for the ‘participant’ and ‘non-participant’ groups and the two modes of correspondence education and distance education. The results show that participants in the OVOUS project are younger, have a higher proportion of party members, have a larger area of family farmland and have more migrant workers than non-participants. Both correspondence education and distance education can enhance the social capital of farmers, and the difference is significant at the 1% level, which preliminarily verifies the positive effect of the two training methods. However, the difference between the mean values of social capital, although existing, is not sufficient to be attributed directly to participation in rural revitalization talent cultivation. Therefore, to assess the impact of participation in rural revitalization talent cultivation on farmers’ social capital accurately, a more rigorous econometric analysis method is needed for in-depth exploration.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Differences in the means of variables for participation in rural revitalization talent cultivation and different training methods.

https://doi.org/10.1371/journal.pone.0357662.t002

Measurement methods

In theory, farmers’ participation in rural revitalization talent-training programs can enhance their social capital; however, the level of social capital may also shape farmers’ willingness to participate, which can give rise to reverse causality and self-selection bias. Although this study controls for a wide range of factors, it remains difficult to fully rule out interference from unobserved variables. While propensity score matching can mitigate differences in observable characteristics, it cannot adequately correct biases driven by latent factors. Therefore, following Lokshin [29,30], this study employs an endogenous switching probit (ESP) model to jointly estimate the participation decision and the outcome equation, thereby identifying the causal effect of training in a more robust manner.

The ESP model simultaneously estimates the following equations:

(1)

(1) In the equation, denotes the level of farmers’ social capital; is the individual characteristics, family characteristics, location characteristics and other factors that affect the level of farmers’ social capital; is the decision term of whether farmers participate in rural revitalization talent cultivation; indicates that farmers participate in rural revitalization talent cultivation and indicates that farmers do not participate in rural revitalization talent cultivation. In addition, ^and are the coefficients to be estimated; is the random error term.

The ESP model consists of two stages. In the first stage, a probit model is used to estimate the determinants of farmers’ participation in the talent-training program as well as the choice between different training modalities. In the second stage, the outcome equation is estimated to evaluate the effects of training participation and training modality on farmers’ social capital. Specifically, the model is specified as follows:

Phase I

Behavioral equations (whether farmers participate in rural revitalization talent cultivation):

(2)

Behavioral equations (whether farmers participate in correspondence education for rural revitalization talent cultivation):

(3)

Phase II

Resulting equation 1 (treatment group 1, farmers’ level of social capital to participate in rural revitalization talent cultivation):

(4-a)

Resulting equation 2 (treatment group 2, level of social capital of farmers participating in rural revitalization talent cultivation correspondence education):

(4-b)

Resulting equation 3 (treatment group 3, level of social capital of farmers participating in rural revitalization talent cultivation distance education):

(4-c)

Resulting equation 4 (treatment group 4, level of social capital of farmers not participating in rural revitalization talent cultivation):

(4-d)

In equations (2) and (3), is a binary choice variable indicating whether farmers participate in rural revitalization talent cultivation; indicates whether farmers participate in rural revitalization talent cultivation correspondence education; is a series of factors influencing whether farmers participate in rural revitalization talent cultivation; is a series of factors affecting whether farmers participate in rural revitalization talent cultivation; is an error term, representing unobservable influences, such as whether the personality of the ‘principal’ of the sample farmer’s family is extroverted, and labor habits.

In equations (4-a), (4-b), (4-c) and (4-d), ,, and denote the level of social capital of farmers participating in rural revitalization talent cultivation, participating in rural revitalization talent cultivation by correspondence education, participating in rural revitalization talent cultivation by distance education and not participating in rural revitalization talent cultivation, respectively. Furthermore, , , and denote the parameters to be estimated in each equation, respectively; is a series of factors affecting farmers’ social capital and , , and form the error term in the resulting equation.

The ESP estimation results reveal heterogeneous effects of various factors on social capital for farmers who participated in training versus those who did not. To evaluate the overall effect of training, it is necessary to compute the average treatment effect on the treated (ATT) on social capital based on the estimated model coefficients. The calculation is as follows:

(5)

(6)

In equations (5) and (6), denotes the expected average social capital if all sample farmers participate in rural revitalization talent cultivation; denotes the expected average social capital if all sample farmers participate in rural revitalization talent cultivation correspondence education; is the expected average social capital if all sample farmers did not participate in rural revitalization talent cultivation.

In the empirical estimation, this study applies full information maximum likelihood (FIML) to jointly estimate the selection equation and the outcome equation. Specifically, the ESP model is estimated using the user-written switch_probit command in Stata 18, which implements a full-information maximum likelihood estimator for endogenous switching probit models. This approach allows the error term in the selection equation to be correlated with the error term in the outcome equation, thereby accounting for potential self-selection associated with unobserved factors. The model tests for the presence of selection bias by estimating the correlation coefficient (ρ) between the error terms: if ρ is statistically significant and differs from zero, this indicates correlation between the selection and outcome equations, implying that estimation using the ESP model is necessary and appropriate; conversely, if ρ is not statistically significant, selection bias is not pronounced.

ESP model estimation results and analysis

Table 3 reports the estimation results of the ESP model of the impact of farmers’ participation in rural revitalization talent cultivation on their social capital. The Wald test rejects the original hypothesis that the behavioral and outcome equations are independent of each other at the 1% level of significance, indicating that the model possesses strong significance overall. The estimated value of rho0 is statistically significant at the 10% level, suggesting a significant correlation between the error terms of the selection and outcome equations. This implies the presence of sample self-selection bias, making joint estimation using the endogenous switching probit (ESP) model necessary. This finding also provides structural support for the role of the instrumental variable in identifying the participation decision.

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Estimated results of the impact of participation in rural revitalization talent cultivation on farmers’ bonding social capital.

https://doi.org/10.1371/journal.pone.0357662.t003

Within the ESP framework, the validity of the instrumental variable depends on whether it satisfies two fundamental conditions: relevance and exogeneity. Regarding relevance, the selection-equation estimates show that policy awareness has a statistically significant effect on farmers’ participation in rural revitalization talent training, indicating a robust statistical association between the instrument and the endogenous explanatory variable and thus meeting the relevance requirement. Regarding exogeneity, after controlling for potential confounders such as individual characteristics, household characteristics, and locational factors, policy awareness does not directly affect farmers’ existing social relationship structures or the stock of mobilizable resources within their social networks. Its influence on social capital is primarily realized indirectly through the mediating channel of the participation decision. Therefore, this variable is theoretically expected to satisfy the exclusion restriction. Taken together, the results support using policy awareness as an instrumental variable in the ESP model, which helps improve identification validity and enhances the credibility of the estimated effects.

Column 1 of Table 3 reports the average marginal effects from the selection equation estimating the determinants of farmers’ participation in rural revitalization talent cultivation. In terms of individual characteristics, farmers who are female, younger and party members are more likely to participate. The possible reasons for this are women’s lower education level and limited off-farm employment opportunities; furthermore, education positively contributes to a higher female labor supply and household income. Therefore, this group expects to strengthen agricultural production and income by participating in rural revitalization talent cultivation. In terms of family and location characteristics, the number of migrant workers, cultivated land area of the family and regional economic development level all significantly and positively affect farmers’ participation in rural revitalization talent cultivation. The number of migrant workers, as the human capital of the family, influences farmers’ behavioral decisions [31]. The larger the area of family farmland, the more farmers tend to learn advanced technology to reduce production risks and stabilize income [32]. Therefore, farmers with abundant labor and a large area of arable land are more inclined to participate in rural revitalization talent cultivation.

Estimation results and analysis of the ESP model of the factors affecting farmers’ social capital

Columns 3 and 4 in Table 3 report the estimation results of the factors affecting the social capital enhancement of farmers who participated and did not participate in rural revitalization talent cultivation, respectively. First, the actual cultivated area significantly and positively affects the social capital of both types of farmers. This finding is consistent with existing studies that show that an increase in cultivated land area is accompanied by social capital enhancement [33]. Second, the distance from the village to the county town significantly and positively affects the social capital of both types of farmers. Furthermore, older men significantly affect social capital enhancement after participating in rural revitalization talent cultivation. This is consistent with Christoforou’s finding that age positively affects social capital [34], as older groups have more leisure time to socialize and tend to cooperate.

Results and analysis of average treatment effect estimates

Line 2 of Table 4 presents the estimation of the average treatment effect of participating in rural revitalization talent cultivation on farmers’ social capital (ATT = 0.068), which is statistically significant at the 1% level of significance. This indicates that participation in rural revitalization talent training increases farmers’ probability of obtaining emergency financial support from relatives and friends by 6.8 percentage points compared with the counterfactual scenario of non-participation. As a result, hypothesis H1 of this paper is verified, i.e., participation in rural revitalization talent cultivation has a significant positive effect on farmers’ social capital. The underlying reason may be that participation in rural revitalization talent training, by providing a systematic learning and interaction platform, enhances frequent interaction and information sharing among farmers, thereby facilitating the formation of trust-based relationships and norms of reciprocity and further promoting the accumulation of social capital. Related studies indicate that education and training not only improve individual capabilities, but also significantly foster the growth of social capital by expanding social networks, strengthening social participation, and raising levels of trust [35]. This mechanism is particularly critical in rural settings, where standardized training can effectively compensate for deficiencies in network resources and information access arising from dispersed social ties.

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. Mean treatment effect of the impact of participation in rural revitalization talent cultivation on farmers’ bonding social capital.

https://doi.org/10.1371/journal.pone.0357662.t004

Further, the third and fourth rows of Table 4 report the average treatment effect estimates of participation in rural revitalization talent cultivation correspondence education and distance education on farmers’ social capital, respectively (ATT of 0.081 and 0.058, respectively), both of which are statistically significant at the 1% level of significance. After controlling for observable and unobservable factors, although both modes of training have a significant positive effect on farmers’ social capital, the social capital effect of correspondence education is more prominent compared to distance education (8.1% > 5.8%). Therefore, hypothesis H2 was tested. The reason why correspondence education outperforms distance education in enhancing social capital may lie in its more frequent and higher–social presence face-to-face interactions, which facilitate repeated interaction, peer communication, and the formation of stable social networks during the learning process. Prior research suggests that learning environments characterized by face-to-face contact or high social presence are more conducive to deep interaction, the formation of trust, and the sustained development of social networks—mechanisms that are central to social capital accumulation [36]. By contrast, the fragmented interactions and weakened connectivity typical of distance education may constrain its ability to effectively translate social relationships into social capital.

Heterogeneity analysis

Given that demographic characteristics may shape individuals’ access to social resources and interaction opportunities, this study further conducts exploratory heterogeneity analyses by age and gender to examine whether the effects of talent training vary across different groups.

To analyses the relationship between farmers’ participation in rural revitalization talent cultivation and the growth of their social capital in depth, this paper divides farmers into an older age group (≥47 years old) and a younger age group (<47 years old) based on whether their age exceeds the mean value of 47 years and then explores the differential impact of age heterogeneity on the enhancement of their social capital after participating in rural revitalization talent cultivation. The estimation results are reported in Table 5, which shows that participating in rural revitalization talent cultivation has a significant positive impact on their social capital in both the senior and junior age groups. Specifically, the probability of social capital enhancement is 10.8% and 10.5% higher for the senior and junior groups, respectively, compared to farmers who did not participate in training for expertise in rural revitalization. This suggests that participation in rural revitalization talent cultivation can promote the accumulation of social capital among farmers of different age groups significantly. However, in comparison, the increase in social capital gained by farmers in the senior age group through participation in rural revitalization talent cultivation is slightly higher than that of the junior age group, which is consistent with existing research that older farmers have accumulated richer social capital than young and middle-aged farmers [37].

Download:

• PNG
  larger image
• TIFF
  original image

Table 5. Comparison of the social capital effects of participation in rural revitalization talent training across age groups.

https://doi.org/10.1371/journal.pone.0357662.t005

To further examine gender heterogeneity in the effects of participation in rural revitalization talent training on social capital growth, the sample is divided into a male group and a female group, and heterogeneous patterns of social capital improvement after training are analyzed across genders. Table 6 reports the corresponding estimation results. The results indicate that participation in rural revitalization talent training has a significantly positive effect on social capital for both male and female farmers, with estimates significant at the 1% level. Specifically, relative to non-participants, the probability of improved social capital increases by approximately 5.1% for male farmers after training, whereas the corresponding increase is about 12.8% for female farmers. This suggests that rural revitalization talent training effectively promotes social capital accumulation for both genders, but the magnitude of the effect differs substantially. In particular, female farmers experience significantly larger gains in social capital than male farmers, consistent with related findings that education and training often yield stronger marginal effects for groups whose pre-existing social networks are relatively constrained, helping them expand social connections and enhance their capacity for social participation [38].

Download:

• PNG
  larger image
• TIFF
  original image

Table 6. Comparison of the social capital effects of participation in rural revitalization talent training across gender groups.

https://doi.org/10.1371/journal.pone.0357662.t006

Robustness tests

To further verify the effect of farmers’ participation in rural revitalization talent training on the enhancement of their social capital, this study conducts robustness checks in the following three aspects.

Alternative model specification

To further examine whether the main findings are sensitive to alternative model specifications, this study employs a recursive bivariate probit model as a robustness check. Similar to the ESP model, this approach jointly estimates the participation and outcome equations while allowing the error terms of the two equations to be correlated, thereby accounting for potential selection on unobservables.

Table 7 reports the estimation results. The average structural discrete change associated with participation in rural revitalization talent training is 0.0708 (robust SE = 0.0426; z = 1.66; p = 0.097; 95% CI: −0.0128 to 0.1543). This estimate is close to the baseline ESP estimate of 0.068, indicating that the positive association between participation in talent training and farmers’ bonding social capital is stable in terms of direction and magnitude across alternative model specifications. However, the alternative-model estimate is only marginally significant at the 10% level, suggesting that statistical precision is somewhat weaker under this specification.

Download:

• PNG
  larger image
• TIFF
  original image

Table 7. Robustness check using a recursive bivariate probit model.

https://doi.org/10.1371/journal.pone.0357662.t007

In addition, the coefficient of policy understanding in the participation equation remains positive and statistically significant (coefficient = 0.5635, p < 0.001), and the estimated error correlation coefficient is positive and significant (, Wald test p = 0.049). These results further support the need to account for endogenous selection in estimating the effect of training participation. However, this robustness check does not independently verify the exclusion restriction of the instrumental variable.

Excluding extreme cropland-area observations.

Given the potential influence of extreme cropland-area observations, this study re-estimates the ESP model after excluding farmers whose cropland area exceeds the 95th and 90th percentiles of the sample distribution. The corresponding cutoff values are 20 mu and 11 mu, respectively.

Table 8 reports the results. After excluding observations above the 95th percentile, the ATT estimate for participation in rural revitalization talent cultivation is 0.092, which remains statistically significant at the 1% level. When observations above the 90th percentile are excluded, the corresponding ATT estimate is 0.090, also significant at the 1% level.

Download:

• PNG
  larger image
• TIFF
  original image

Table 8. Robustness check excluding extreme cropland-area observations.

https://doi.org/10.1371/journal.pone.0357662.t008

These findings indicate that the positive effect of rural revitalization talent training on farmers’ bonding social capital is not driven by extreme cropland-area observations, confirming the robustness of the baseline results.

Placebo test

To assess robustness, this study further conducts a placebo test. By randomly generating a treatment variable and repeatedly re-estimating the model, we construct the distribution of the estimator under a scenario with no true policy shock, as shown in Fig 1. The results indicate that the placebo estimates mostly fluctuate around zero and do not exhibit systematic deviation. In contrast, the baseline estimate lies clearly outside the placebo distribution, suggesting that the identified effect is unlikely to be driven by random factors or model-specification bias. This provides additional evidence that participation in rural revitalization talent training increases farmers’ social capital.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Placebo test.

https://doi.org/10.1371/journal.pone.0357662.g001