Cognition is the foundation of behavior, as individuals’ observations and understanding of objects form their cognitive systems, which in turn guide actions (Ajzen and Fishbein, 1975; Fishbein, 1975). According to Ajzen’s Theory of Reasoned Action, behavior arises from attitudes formed through cognitive evaluation, based on the assumption that humans are rational (Ajzen, 1991). The “knowledge–affect–behavior” theory posits that behavioral change is a sequential process: individuals first acquire knowledge (“cognition”), develop emotions (“affect”), and ultimately act under the combined influence of both (Westbrook and Oliver, 1991; Frijda, 1993). Based on these theoretical foundations, Davis proposed the Technology Acceptance Model (TAM) in the field of information technology to explain and predict individuals’ adoption of new technologies, defining usage decisions in terms of behavioral intention and actual use (Davis et al., 1989). Later TAM extended into the UTAUT. UTAUT includes four core dimensions: performance expectancy (PE), effort expectancy (EE), social influence (SI), and facilitating conditions (FC), which respectively reflect perceptions of system utility, ease of use, social pressure, and organizational support (Venkatesh et al., 2003). This framework has been widely applied to study technology adoption and decision-making, and it can be extended by incorporating context-specific variables. For most farmers, participation in forestry carbon sink projects involves adopting entirely new management practices and carbon monitoring technologies. Performance expectancy refers to the degree to which an individual perceives that adopting carbon sink technologies will be beneficial to their work. Effort expectancy denotes the amount of effort perceived to be required to use these technologies. Social influence reflects the extent to which an individual feels affected by the surrounding group, while facilitating conditions indicate the degree of organizational support perceived by the individual regarding relevant technologies and equipment for carbon sink implementation. Drawing on the UTAUT framework, this study provides a detailed depiction of farmers’ perceptions of carbon sink projects and systematically examines how different cognitive dimensions influence both participation intentions and actual management behaviors.

Forest carbon projects provide ecological, social, and economic benefits. Farmers’ decisions are influenced not only by expected returns and policy support but also by cognitive perceptions. Existing research often emphasizes cost–benefit analysis, overlooking psychological factors. Farmers, as the primary implementers of carbon projects, face decisions shaped by risk perceptions, policy regulation, and potential economic and ecological outcomes. From a cognitive perspective, variations in participation behavior reflect differences in farmers’ understanding of carbon project benefits and risks. This study investigates how different cognitive dimensions influence farmers’ participation in bamboo forest carbon projects in Zhejiang Province, aiming to identify pathways to enhance engagement and inform policy and incentive design.

Behavioral intention is a prerequisite for action, as the likelihood of behavior depends on individual perception (Dodds et al., 1991). For most farmers, forest carbon projects represent new technologies; therefore, participation intention is a critical driver of actual behavior (Davis, 1989). Certain cognitive factors influence behavior indirectly by shaping intention, reflecting bounded rationality in farmers’decision-making (Venkatesh et al., 2003). Based on this, we propose an extended UTAUT framework for farmer participation (Figure 1).

This study was conducted in Zhejiang Province, located on the southeastern coast of China (118°01’–123°10’ E, 27°02’–31°11’ N), covering an area of 10.55 million hectares and comprising 11 prefecture-level cities. According to the third national land resources survey of Zhejiang Province, the forest area reached 6.0936 million hectares (91.4036 million mu) in 2021, with a forest coverage rate of 61.36%. Nearly 70% of the forests are classified as young and middle-aged, indicating considerable potential for forest quality improvement. Bamboo forests account for 906,300 hectares (13.595 million mu), representing 14.87% of the province’s total forest area and 11.99% of China’s total bamboo forest area.

Data for this study were collected through a household survey conducted by the research team in Zhejiang Province, a region with abundant bamboo resources. The survey focused on Anji, Longyou, Suichang, Kaihua, and Lin’an, areas with high forest coverage and rich bamboo plantations, making them priority sites for bamboo forest carbon sequestration projects. Fieldwork was carried out from June 2022 to August 2023 using one-on-one, face-to-face interviews. The questionnaire was developed through expert consultation and team discussion, and enumerators received comprehensive training. Respondents were contacted in advance with the support of local forestry bureaus and received participation compensation. A stratified random sampling method was used: two townships were selected per project, 2–3 villages per township, and 35 households with moso bamboo plantations per village, resulting in 850 households surveyed. After removing invalid responses, 811 valid questionnaires were retained.

This study focuses on the effects of differences in economic cognition, ecological cognition, risk perception, and social cognition on farmers’ intentions and behaviors in carbon forest management, and examines the influence of policy cognition on participation behavior. Economic cognition reflects farmers’ perceived value of participating in forestry carbon sequestration projects. Ecological cognition captures farmers’ assessments of the environmental benefits of carbon sequestration forest management. Risk cognition refers to farmers’ perceptions of project-related risks and their tolerance for such risks. Social cognition represents farmers’ perceptions of the social benefits generated by these projects. Policy cognition indicates farmers’ understanding of policies related to carbon sequestration forest management. All cognition dimensions were measured using a 5-point Likert scale, with higher scores indicating stronger cognition. Based on prior studies and data availability, the key explanatory variables are defined as Table 1 (Cammarata et al., 2024; Block et al., 2024; Wu et al., 2025). According to the literature, farmers’ personal characteristics, household attributes, and regional resource endowments play important roles in their decision to engage in carbon forest management. Following Wang et al. (2018), control variables related to carbon forest management decisions were selected from personal and household characteristics; the specific variable definitions and descriptive statistics are presented in Table 1. Considering that some policies are effective only under certain scales, and that larger-scale households often participate in land transfer, land transfer was included as a moderating variable. In the survey, farmers were asked whether they had transferred land, with “yes” coded as 1 and “no” as 0.

Before the empirical analysis, variance inflation factors (VIF) were used to check for multicollinearity. All VIF values were below 5, indicating no multicollinearity. The Probit model’s marginal effects were also calculated to assess the impact of each variable (Table 2). Results from Model (1) indicate that economic and ecological cognition both have significant positive marginal effects on farmers’ willingness and actual participation in carbon forest management at the 1% level. Specifically, a one–unit increase in economic cognition raises the probability of participation by 1.6%, while a similar increase in ecological cognition increases participation probability by 2.8%. Social cognition also shows a significant positive marginal effect at the 1% level, increasing the likelihood of participation by 3.0%. In contrast, risk cognition exhibits a significant negative marginal effect at the 1% level, with higher risk awareness reducing participation probability by 2.7%, suggesting that perceived uncertainty discourages engagement in carbon forest management. From Model (2) show that policy cognition has a strong and statistically significant positive marginal effect on farmers’ participation behavior at the 1% level. Farmers with higher awareness of carbon afforestation policies are approximately 10.0% more likely to participate in carbon forest management. Overall, these results underscore the role of different forms of cognition in shaping farmers’ participation decisions, with policy cognition showing the largest marginal effect among the examined factors.

The control variables reveal distinct patterns in farmers’ decision processes. Age has a significant negative effect on participation intentions, indicating that older farmers exhibit lower willingness to adopt novel practices. In contrast, serving as a village cadre, holding a forest tenure certificate, and being a technical demonstration household all significantly enhance participation intentions, reflecting both stronger capacities to engage with new initiatives and greater institutional or tenure security. For actual participation behavior, age, forest tenure certificates, and forestry subsidies exert significant positive effects. Apart from the consistent influence of tenure certificates across both models, the divergence between intention and behavior highlights the practical constraints farmers face when translating willingness into action. Older farmers’ accumulated managerial experience may facilitate their final decision to engage in carbon forestry, while financial and technical barriers make subsidies particularly effective in enabling participation. Overall, these results corroborate Hypotheses 1-5.

Results in Table 3 indicate that the interaction between land transfer and policy cognition is positive and significant at the 10% level, suggesting that land transfer significantly moderates the effect of policy cognition on farmers’ participation in forestry carbon projects. Farmers with land transfer are more likely to act on their policy awareness and participate in carbon forestry projects, supporting hypothesis H6. Given the scale requirements of forestry carbon projects, simply improving policy awareness may be insufficient to increase participation. Clearly defining operational standards, scale requirements, or providing dedicated land transfer policies for carbon forestry development may be more effective in promoting engagement.

To further explore the sources of cognitive differences, the sample was divided by age and village leadership status. Considering that age and village cadre status significantly influenced the results in previous regressions, farmers were grouped into older adults ( ≥ 60 years, n = 411) and middle-young adults ( < 60 years, n = 400), and into village cadres (n = 274) and non-cadres (n = 537) for heterogeneity analysis.

As shown in Table 4, Model (5) indicates that all five cognition variables are statistically significant at the 1% level for the senior group. For the younger group, ecological, social, risk, and policy cognition remain significant. While economic cognition is not. The coefficient signs are consistent with earlier results. This suggests that elderly farmers place greater weight on expected economic returns when deciding whether to participate in forestry carbon sink projects, whereas younger farmers focus more on non-economic values, making economic cognition insignificant in their decisions. Model (6) shows that, among village cadres, economic cognition has no significant marginal effect, while the other four cognition variables significantly influence participation decisions with consistent directions. For non-cadre farmers, only social cognition is insignificant, and all other cognition variables remain significant. This difference may reflect the fact that village cadres have better access to policy information and a stronger recognition of the social and public value of carbon sink projects, which weakens the role of economic cognition. In contrast, non-cadre farmers are more sensitive to short-term economic returns, increasing the importance of economic cognition.

To address potential endogeneity between cognitive differences and farmers’ participation in carbon forestry management, this study employs the widely recognized instrumental variable Probit (IV-Probit) model (Rivers and Vuong, 1988). Careful logic and creativity are required in selecting instruments. Based on the peer effect, individual socioeconomic outcomes are often influenced by characteristics of the collective to which they belong (Chen, 2012). Following Liu et al. (2023), we consider that farmers’ cognitions are strongly shaped by other residents within the same village. We use village-level averages excluding the individual as instrumental variables for re-estimation (Shen et al., 2025). Specifically, the average education level of other villagers represents economic cognition, average adoption of ecological technologies represents ecological cognition, average out-migration ratio represents social cognition, average financial literacy represents risk cognition, and the average proportion of villagers holding forest property certificates represents policy cognition. These instruments allow us to account for peer-driven variation in cognition when conducting the regression analysis.

The results are reported in Table 5. First, the first-stage regressions indicate that all instruments significantly influence the endogenous explanatory variables. Second, after accounting for endogeneity, farmers’ cognition still significantly affects participation in carbon forestry, suggesting that ignoring endogeneity would overestimate the impact of cognition on participation. Third, weak instrument tests (AR and Wald) are both significant, confirming the validity of the instruments. Overall, these results reinforce the robustness of the baseline regressions.

To ensure the reliability of the empirical results, this study follows Liu et al. (2021) and conducts robustness checks by employing alternative model specifications. Table 6 reports regressions of farmers’ willingness and behavior to participate in carbon forestry using a binary Logit model. The significance and direction of the estimated coefficients remain consistent with the baseline results, indicating robust findings.

Considering that farmers over 70 years old may differ significantly from younger farmers in physical capacity, cognitive ability, and knowledge, and that carbon forestry management requires long-term input-to-output investment, older farmers may be less able to implement such practices. Following Ren and Guo (2023), regressions were re-estimated after excluding farmers aged over 70. As shown in Table 7, the results remain largely consistent with the baseline model, further confirming the robustness of the estimates.