Credit constraints have emerged as a pivotal factor in corporate innovation and the adoption of new technologies. A substantial body of academic research currently focuses on the impact of credit constraints on corporate innovation and the uptake of new technologies (Hai et al., 2022). Market imperfections stemming from information asymmetry can lead to financing constraints for businesses, especially considering that technological innovation and adoption are long-term, high-risk investment activities, thus often severely hindering corporate innovation (Amore et al., 2013; Hall and Lerner, 2010; Khan et al., 2018). In the face of capital scarcity, relying solely on internal financing is insufficient to meet the financial needs for business survival and growth, thus making external financing a necessity (Wang et al., 2022). Companies with ample cash flow are more likely to effectively implement Corporate Social Responsibility (CSR). Notably, Amore and Bennedsen (2016) demonstrated that businesses heavily dependent on external financing tend to excel in CSR implementation and green innovation promotion.

Although farmers' cooperatives are not traditional enterprises, they engage in commercial activities similar to those of enterprises, and thus are similarly affected by credit constraints in their production and operations. The funding sources for cooperative development primarily include internal and external financing (Stiglitz and Weiss, 1981). Internally, cooperatives rely primarily on share capital and member fundraising for financing. However, most members of cooperatives are ordinary farmers who lack sufficient capital accumulation and investment funds, making it difficult for cooperatives to effectively secure internal financing. Additionally, farmers often have a short-term focus, prioritizing immediate gains over the cooperative's long-term value, which reduces their motivation to invest (Osborne, 2006). Therefore, relying solely on equity financing from internal members is insufficient to sustain the long-term survival and development of cooperatives, necessitating external financing for support (Baarda, 2007; Zang et al., 2023). Externally, financial institutions such as banks and private lenders are the main sources of external financing for cooperatives. The lack of tangible assets that can be used as collateral is a primary reason that hinders cooperatives from obtaining secured loans from formal financial institutions (Stiglitz and Weiss, 1981). Moreover, the lack of transparency and sharing of financial information, coupled with non-standardized internal management, leads to significant information asymmetry and moral hazard for financial institutions when providing credit to cooperatives (Yu and Nilsson, 2017). Although government subsidies and policy incentives can provide direct assistance to cooperative financing, a sustained reduction in fiscal support from the government can also lead to financing difficulties for cooperatives (Roxana, 2005).

Credit constraints limit farmers' access to funds, which in turn can create barriers to the adoption of green technologies. A substantial body of research on farmer technology adoption indicates that credit constraints are a major factor hindering the adoption of green production technologies by farmers (Jiang et al., 2024), and improving credit accessibility is considered a crucial means of enhancing the level of technology adoption among farmers (Yu et al., 2020). Wolde et al. (2017), in a study of 252 randomly selected households, found that participation in the credit market positively influences technology adoption, while credit constraints have a negative impact. The credit constraints faced by farmers make it difficult for them to achieve optimal resource allocation, thereby affecting the investment in agricultural technology (Shiferaw et al., 2015). Shiferaw et al. (2015), using data from nearly a thousand farmer surveys in Africa, found that dual credit constraints are the main reason for the low adoption rate of hybrid peanuts. Credit availability impacts farmers' capital endowments, influencing their adoption behavior of green production technologies (Wollni et al., 2010).

Research on technology adoption within farmers' cooperatives is still relatively scarce. Farmers' cooperatives, which are mutual-aid economic organizations composed of farmers, can foster broader participation among members, thereby facilitating the widespread adoption of new technologies. Such collective action can enhance resource efficiency and diminish the financial and operational risks associated with technology adoption for individual farmers (Odoemenem and Obinne, 2010; Ma et al., 2023). However, in China, due to systemic, operational, and policy factors, rural cooperative financial institutions commonly face issues such as poor asset quality, inadequate risk compensation capabilities, and incomplete legal governance, which constrain the development of the rural financial market and its capacity to serve agriculture, rural areas, and farmers (Chen Z. G. et al., 2022). Compared with individual farmers adopting green technologies, farmers' cooperatives often adopt one or more comprehensive technological systems that are strategic, innovative, and economical in nature. Therefore, in the pursuit of green and low-carbon development, the primary challenge faced by farmers' cooperatives is credit constraints. Cooperatives' adoption of green technologies is characterized by substantial capital requirements, extended payback periods, sophisticated organizational demands, and significant economic returns. These factors heighten members' expectations for improving cooperative performance through technology. Simultaneously, they must carefully evaluate the cost-benefit ratio and shared interests, making decisions democratically to maximize the technologies' benefits. However, constraints such as limited financing, capital shortages, and membership limitations (Odoemenem and Obinne, 2010), significantly hinder the adoption of these technologies, affecting the outcomes of technology implementation. Thus, we propose the following hypothesis:

Social networks, which have expanded from individuals to encompass businesses and organizations, play a crucial role in China's relationship-based society, influencing economic and social operations (Bian, 1997). They alleviate credit constraints, reduce information asymmetry in financing for small and medium-sized enterprises (SMEs), and reduce costs by increasing financing channels (Cheng et al., 2017; Goss and Roberts, 2011; Luo and Ying, 2014). Social capital mitigates information scarcity in lending, transmits signals of borrowers' repayment ability, and motivates repayment through trust and social homogeneity (Bhawani, 2015). Additionally, it also serves as “social collateral,” easing the lack of physical collateral and enhancing credit accessibility (Madajewicz, 2011).

In modern agricultural production, social networks are equally important. Social capital can improve farmers' access to productive resources beyond economic collateral. Cooperatives can effectively mitigate credit constraints by strengthening relationships with other organizations within the industry chain. For instance, vertically integrated production organizations (agricultural company—specialized cooperative–farmers' cooperative) enhance the operational efficiency and effectiveness of the entire agricultural industry chain by connecting production, processing, distribution, and sales processes (Ma and Abdulai, 2016). This shifts the disadvantaged market position of farmers' cooperatives. Besides, social networks can collect and aggregate information from various entities in the chain, fostering the formation of an internal information network system within the agricultural industry chain and achieving a certain degree of information sharing with external formal financial institutions (Zhou et al., 2019). Moreover, formal financial institutions can obtain dynamic production and operational information from farmers' cooperatives (Pearce, 2003). In cooperatives, increased social capital promotes stable agricultural development by overcoming market failures, reducing transaction costs, and alleviating information asymmetry, thus potentially influencing credit accessibility and the likelihood of securing loans from financial institutions (Holloway et al., 2000; Kustepeli et al., 2020). In addition, establishing network relationships with the government can also substantially enhance the financing performance of cooperatives, facilitating their access to bank credit (Yu and Nilsson, 2018).

Social networks are crucial for alleviating credit constraints and fostering technology adoption and innovation. They benefit cooperatives by improving productivity and farmers' incomes, and are divided into internal and external components (Chesbrough and Crowther, 2006; Kumar and Igdalsky, 2019; Tian et al., 2023). External networks, formed with institutions such as governments and banks, reduce transaction costs and increase access to credit (Yu and Nilsson, 2018; Qiao et al., 2017; Wang and Lu, 2015). Internal networks enhance team collaboration and resource efficiency, and trust within these networks alleviates concerns over technology adoption (Rodrigo et al., 2017). However, the formalization of internal networks into rules is challenging, which can hinder collective actions and green production (Cui et al., 2017). Considering the financial limitations of small farmers in China (Osborne, 2006), the ability of internal networks to ease credit constraints on technology adoption may be limited. Thus, we propose the following hypothesis:

The data were collected during a field survey conducted by our research team in Henan and Shaanxi provinces in 2023. Henan Province, recognized as a major grain-producing region in China, has significant advantages in agricultural development and has a well-established agricultural technology extension system. Its agricultural development exerts a substantial impact on the national economy. Shaanxi Province, as a pivotal province in the development of Western China, is rich in crop varieties and agricultural resources, and its fruit industry enjoys a prestigious reputation nationwide. The two provinces respectively represent China's “grain security-oriented” and “specialty value-added” agricultural models, providing a representative sample framework for investigating green production practices of planting-type farmer professional cooperatives.

In the sampling design, a hybrid approach integrating stratified random sampling and multi-stage random sampling was implemented to mitigate bias. The study population was first stratified into two distinct provincial layers: Henan Province as the grain production base stratum and Shaanxi Province as the specialty agriculture stratum. Subsequently, municipal sampling frames were established based on regional economic development gradients and dominant agricultural industries, encompassing eight cities in Henan (Puyang, Anyang, Xinxiang, Zhengzhou, Kaifeng, Luohe, Shangqiu, and Zhoukou) and five cities in Shaanxi (Weinan, Tongchuan, Baoji, Hanzhong, and Xianyang). Details can be found in Figure 1. The multi-stage procedure then commenced with systematic sampling of 4–5 agriculturally dominant counties per city, ranked by agricultural output value. At the final stage, planting-type production cooperatives actively operating for more than 2 years were randomly selected from rosters provided by the County Bureaus of Agriculture and Rural Affairs using random number tables, with 8–10 cooperatives sampled per county. A total of 582 organizational questionnaires were collected. After excluding non-planting, non-production and management cooperatives, and samples with significant data missing, 535 valid questionnaires were obtained, with 360 from Henan Province and 175 from Shaanxi Province. Questionnaires were distributed to the chairpersons and managers of the cooperatives, covering aspects such as the basic situation of the cooperatives, operational mechanisms, business models, adoption of GPCT, internal control within the cooperatives, government regulation, and the individual characteristics and technical cognition of the chairpersons.

In the process of technology adoption, farmers' cooperatives first decide whether to adopt technology and then determine how many technologies to adopt. For those cooperatives that have not adopted any technology, their degree of technology adoption is unobservable, leading to sample selection bias in the green technology adoption of farmers' cooperatives. This bias can be analyzed using the Heckman sample selection model. To overcome the issue of sample selection bias, an identifying variable that satisfies the exclusion restriction is incorporated into the first stage of the cooperatives' decision to adopt technology—this variable is the government's support and promotion of green technology. This variable has no direct impact on the degree of technology adoption in the second stage. The model specification is as follows:

Equation 1 serves as the selection equation. Dtechi represents whether the i-th cooperative adopts GPCT. The observed adoption decision is:

CCs_i is an indicator variable that equals one if the farmers' cooperative faces credit constraints, and zero otherwise. Z_i is the exclusion restriction variable, in this case, it indicates whether the local government has provided technical propaganda support (Govsup_i); if the local government has promoted green technology, it equals one, otherwise zero. This variable exhibits dual characteristics. First, government interventions through policy advocacy, technical training, and other administrative measures substantially enhance cooperatives' initial propensity to adopt technologies, as evidenced by the statistically significant effect of Z_i on Dtech_i in the first-stage equation. Second, the government's promotion efforts do not directly influence cooperatives' technology decisions regarding adoption intensity, meaning this variable remains uncorrelated with the error term ε_i in the second-stage.

X_i represents a series of control variables that characterize the cooperatives, such as asset turnover rate, economic output efficiency, and land area per cooperative member. v_i is the residual term. By using Equation 1, we calculate the inverse Mills ratio (IMR) to correct for the sample selection bias by incorporating the conditional expectation of u_i given selection:

Where Z_iα is the predicted values and ϕ(.), Φ(.) are the PDF and CDF of standard normal distribution from Equation 1. Incorporating IMR_i into the second-stage outcome equation:

where the dependent variable represents the degree of adoption of green pest control technologies. CCs_i continues to serve as the credit constraint variable, and ε_i is the error term. Other variables are defined as in the probit Model (1).

The inclusion of the IMR in the Equation 4 provides critical insights into sample selection bias. The statistical significance of its coefficient τ directly reflects the presence of unobserved heterogeneity: a τ value significantly different from zero indicates systematic interference from latent factors affecting both adoption decisions and intensity. Notably, the sign of coefficient β₁ for the credit constraint variable CCs_i in Equation 4 carries substantial economic implications. A significantly negative β₁ suggests that credit constraints suppress adoption intensity by elevating financing costs and restricting technological investments. Conversely, a positive β₁ may indicate a “constraint-induced innovation” mechanism, where credit limitations compel cooperatives to prioritize high-return technologies.

To ensure the validity of the estimation results, multicollinearity among the explanatory variables was assessed. The results indicated that the maximum Variance Inflation Factor (VIF) value was only 2.28, suggesting that multicollinearity among the explanatory variables is minimal, thus satisfying the criterion for variable independence. The Heckman sample selection model was estimated using Stata 17.0 software, and the results are summarized in Table 3. From the estimation results, the Wald statistic is significant at the 1% level, and the inverse Mills ratio (IMR) value is non-zero and significant at the 5% level. This finding indicates the presence of sample selection bias, confirming that there is a connection between the two stages of the cooperative's technology adoption behavior. The use of the Heckman model is appropriate for analyzing the data, and the model exhibits a good overall fit.

The identifying variable, Govpro, is significant at the 1% level in the selection equation, fulfilling the critical exclusion restriction requirement. This significance suggests that the government's promotion of green technology substantially affects the selection into the sample of technology adopters, and its influence is not directly observable in the second stage of technology adoption. This aligns with the exclusion restriction required for the Heckman model.

Table 3 presents the regression results for Equations 1, 2. Columns (1) and (2) present baseline specifications without control variables, whereas Columns (3) and (4) incorporate a comprehensive set of control variables. Columns (1) and (3) correspond to the regression results of the first stage, where credit constraints (CCs) are regressed on willingness to adopt green pest control technologies (Dtech). The analysis reveals that credit constraints have a significantly positive impact on the cooperatives' intentions to adopt GPCT. When faced with capital constraints, cooperatives are likely to pursue efficiency improvements and cost reductions through technological advancements.

Columns (2) and (4) display the second-stage regression results, which show that credit constraints inhibit the degree of technology adoption. This is because, with limited financial support, cooperatives tend to prioritize technologies that are more urgently needed and avoid indiscriminate investment in numerous new technologies. Consequently, Hypothesis 1 is confirmed.

Regarding the other control variables, in the first stage, the number of cooperative members (Members) and land area per capita (Landpc) significantly and positively influence the willingness to adopt GPCT. The exclusive variable, Govpro, which is the local government's promotion and propaganda of technology, significantly increases the cooperatives' intention to adopt green technology. In the second-stage regression, the number of cooperative members has a significantly negative relationship with the degree of technology adoption. This suggests that cooperatives with more abundant labor resources are less likely to adopt high-efficiency green technologies.

To verify the robustness of the aforementioned regression results, this section conducts robustness checks through three methods: replacing the independent variable, changing the regression model, excluding special samples, and adding provincial fixed effect.

Columns (1) and (2) in Table 4 present the two-stage regression results with the SA index as a proxy for credit constraints. Farmers' cooperatives are agricultural organizations with characteristics of enterprises, so when measuring enterprise financing constraints, we refer to the SA index constructed by Hadlock and Pierce (2010) as a proxy for enterprise credit constraints. The SA index is constructed using two variables that do not change much with events and have strong exogeneity: enterprise size and enterprise age. (The specific calculation formula is: SA = −0.737 × Size + 0.043 × Size²-0.040 × Age). The more negative the SA index, and the larger its absolute value, the higher the degree of enterprise financing constraints.

The results in Columns (1) and (2) show that the SA regression coefficient is significantly negative, suggesting that the higher the financing constraints, the lower the degree of technology adoption in cooperatives. The result is in agreement with Kou et al. (2025), who use the SA index demonstrate that firms with fewer financial resources face heightened cost constraints, significantly inhibiting green innovation. Extensive research has also established the negative effect of credit constraints on farmers' technology adoption (Ojo et al., 2020; Mishra et al., 2023; Wang et al., 2025). Thereby, the second-stage regression results are consistent with the baseline results.

Given to the advantages of Poisson two-stage regression model in handling count data, heteroskedasticity, and zero observations (Harris et al., 2012), we switch to this model with Pseudo maximum likelihood estimation for the two-stage regression analysis. Columns (3) and (4) display the results of using the Poisson regression model in the second stage. They also reveal a negative relationship between credit constraints and the degree of technology adoption.

Columns (1) and (2) in Table 5 present the regression results after excluding farmers' cooperatives in the provincial capitals of Zhengzhou and Xi'an. Provincial capitals typically possess more financial institutions and richer social resources, and local farmers' cooperatives have access to more agricultural new technologies, giving them a clear advantage over other cities. Therefore, we excluded 19 farmers' cooperatives in the provincial capitals. The results continue to demonstrate a significant inhibitory effect of credit constraints on the degree of agricultural technology adoption. This finding is consistent with the conclusions drawn by Zhang et al. (2024), who highlight the substantial impact of rural household credit access on farmers' production input. This further validates the robustness of the benchmark results.

Columns (3) and (4) in Table 5 present the regression results after incorporating provincial fixed effects. In our study, we have included two major grain-producing provinces in China: Henan and Shaanxi. Although both provinces are representative in the development of farmer professional cooperatives, they exhibit notable differences in culture, policy environment, and economic structure. Therefore, incorporating province fixed effects to control for the unique characteristics of each province is crucial for mitigating omitted variable bias. Columns (3) and (4) present the two-stage Heckman regression results after controlling for province fixed effects. The results indicate that the impact of credit constraints on the willingness to adopt technology and the extent of technology adoption remains significant, with minimal changes in the estimated coefficients. This is in line with findings in similar studies, for example, Aker et al. (2023) who find significant regional variation in an improved storage technology adoption. Our finding further confirms the robustness of our baseline results, even after accounting for unobservable provincial characteristics.

How do credit constraints influence the degree of green technology adoption in cooperatives? we analyze the intermediary mechanisms from the perspective of social networks. Social networks can be categorized into internal and external social networks (Chesbrough and Crowther, 2006). Both significantly affect the cooperatives' access to financial resources. Therefore, based on Equations 1, 2, we construct the following mechanism test models:

where the interaction term CCs_i × network_i represents the interaction between credit constraints in cooperatives and the extent of social networking. In our empirical analysis, we first utilized a dummy variable (network) to test the mechanism by which credit constraints affect technology adoption through social networks. We further distinguish between two types of social networks—internal (Internet) and external (Externet)—and use separate dummy variables to examine the impact of credit constraints on technology adoption through these different types of networks.

During the survey, we distributed questionnaires to cooperative members, asking them to rate the cooperative's relationships with government, banks, credit unions, or other financial institutions on a scale of 0 to 5, with higher scores indicating closer external social networks. Similarly, internal social networks are assessed by members rating the frequency of board meetings, general membership meetings, and the establishment of strict production, sales, and technical training management systems within the cooperative, also on a scale of 0–5. After averaging the members' scores, we obtained the total social network scores, as well as the external and internal social network scores for each cooperative. These scores were then categorized based on the median, resulting in binary variables for social network (network), as well as external (Externet) and internal (Internet) social networks, where a value of 1 indicates a close (external or internal) social network, and 0 otherwise.

Table 6 presents the regression results based on Equations 3, 4. In the first stage of the regression, both credit constraints (CCs) and the interaction term between credit constraints and social networks (CCs × network) are positive but not statistically significant. However, in the second stage, the interaction term, CCs × network, is positive and significant at the 10% level. This suggests that while social networks can enhance the willingness of cooperatives facing credit constraints to adopt new technologies, their role in mitigating the suppressive effect of credit constraints on technology adoption is even more pronounced. This finding highlights the important role of social networks in the context of technology adoption within cooperatives. Social networks act as a buffer against the restrictive impact of credit constraints, thereby facilitating a more substantial uptake of new technologies. The amplifying effect of social networks on the diffusion of new technologies aligns with existing literature, which highlights the importance of social capital in fostering innovation and technology transfer (Ma and Abdulai, 2016; Zhou et al., 2019).

Furthermore, Table 7 illustrates the impact of different types of social networks on green technology adoption. The two interaction terms, CCs × Externet and CCs × Internet represent the regression results of the interaction between credit constraints and external social networks, as well as credit constraints and internal social networks. Upon comparison, it is observed that in the second-stage regression results, the interaction term between external social networks and credit constraints (CCs × Externet) is significantly positive, while the interaction term between internal social networks and credit constraints (CCs × Internet) is positive but not significant. This suggests that external social networks can mitigate the suppressive effect of credit constraints on the adoption of new technologies. This finding is generally consistent with existing literature, which concludes that cooperatives with good external financial resources can significantly alleviate their financing performance, thereby increasing the level of new technology adoption (Yu and Nilsson, 2018; Zhou et al., 2019). On the other hand, internal social networks play a more significant role in the management process of cooperatives, and their impact on alleviating credit constraints in technology adoption is not apparent (Cui et al., 2017). Therefore, cooperatives should establish good external social networks and improve external financing channels to provide a robust source of funds for improving the agricultural technology level of cooperatives. Thus, Hypothesis 2 is validated.

In the transition from traditional to modern agriculture, farmers' cooperatives frequently encounter financial constraints, especially when adopting new agricultural technologies, which are inextricably linked to financial credit activities. This study utilizes survey data of 535 planting cooperatives in Henan and Shaanxi provinces, regions representative of China's grain production heartland and ecological transition zones. Employing a two-stage Heckman selection model to correct for potential selection bias, our methodology precisely delineates the credit constraints encountered by these cooperatives and empirically evaluates their impact on the adoption of GPCT. The empirical findings reveal that credit constraints operate as a double-edged sword: while substantially increasing cooperatives' willingness to adopt GPCT, they concurrently suppress actual adoption intensity, revealing previously undocumented behavioral complexity at the organizational level.

Crucially, our mechanism analysis identifies that the external social networks emerge as a powerful mediator that significantly alleviates credit constraints' detrimental impact on adoption scale—a mechanism notably absent in internal networks. Heterogeneity analysis further reveals the disproportionate inhibiting effect of formal (vs. informal) credit constraints, the catalytic role of product green certification in facilitating adoption, and the enhanced capacity of larger-asset cooperatives to implement GPCT.

This analysis underscores the influence of credit availability on the ability of farmers' cooperatives to invest in and adopt green technologies. The presence of credit constraints, particularly from formal financial institutions, appears to limit the cooperatives' capacity to secure the necessary funds for technological upgrades, which in turn affects their overall operational efficiency and sustainability. This finding highlights the importance of addressing credit barriers to enhance the adoption of green technologies in the agricultural sector, which may ultimately contribute to improved environmental outcomes and economic sustainability for these cooperatives. For additional requirements for specific article types and further information please refer to “Article types” on every Frontiers journal page.

To expand social networks, alleviate credit constraints, and increase the adoption rate of green production technologies among farmers, several key measures should be considered: First, governments and financial institutions should increase support for credit funds for farmers adopting green technologies, implement preferential interest rates, improve financial service levels, and reduce transaction costs in the rural financial market.

Second, rural financial reforms should be implemented by leveraging the social network characteristics of local farmers, fully considering the social network resources of farmers, providing financial support for farmers to adopt new technologies; the government should also focus on cultivating farmers' social networks, encouraging them to enhance communication and trust through social networks, improving their technical cognition, and broadening agricultural technology dissemination.

Third, the government should increase financial support for cooperatives engaged in the production of green products, support cooperatives in leveraging their resource endowments to strengthen advantageous and characteristic industries, standardize the operational and management mechanisms of cooperatives, and increase farmers' technology adoption rates.