Social Capital, Crop Differences, and Farmers’ Climate Change Adaptation Behaviors: Evidence from Yellow River, China

Abstract

Against the backdrop of global climate change, enhancing farmers’ adaptive capacity to reduce crop production risks has emerged as a critical concern for governments and researchers worldwide. Drawing on social capital theory, this study develops a four-dimensional measurement framework comprising social networks, social trust, social norms, and social participation, utilizing survey data from 1772 households in the Yellow River Basin. We employ factor analysis to construct comprehensive social capital scores and apply ordered Probit models to examine how social capital influences farmers’ climate adaptation behaviors, with particular attention to the moderating roles of agricultural extension interaction and digital literacy. Key findings include: (1) Adoption patterns: Climate adaptation behavior adoption remains low (60%), with technical adaptation measures showing particularly poor uptake (13%); (2) Direct effects: Social capital significantly promotes adaptation behaviors, with social trust (p < 0.01), networks (p < 0.01), and participation (p < 0.05) demonstrating positive effects, while social norms show no significant impact; (3) Heterogeneous effects: Impact mechanisms differ by crop type, with grain producers relying more heavily on social networks (+, p < 0.01) and cash crop producers depending more on social trust (+, p < 0.01); (4) Moderating mechanisms: Agricultural extension interaction exhibits scale-dependent effects, negatively moderating the relationship for large-scale farmers (p < 0.05) while showing no significant effects for smaller operations; digital literacy consistently demonstrates negative moderation, whereby higher literacy levels weaken social capital’s promotional effects (p < 0.01). Policy recommendations: Effective climate adaptation strategies should integrate strengthened rural social organization development, differentiated agricultural extension systems tailored to farm characteristics, and enhanced rural digital infrastructure investment.

1. Introduction

Climate change has emerged as a preeminent global challenge in recent decades, profoundly impacting both ecological systems and socioeconomic development worldwide [1]. Agriculture, being the economic sector most intrinsically dependent on bioclimatic conditions, stands among the most vulnerable to climate perturbations [2,3]. Empirical evidence indicates that climate change accounts for 32–39% of annual yield variability in major global food crops (maize, rice, wheat, and soybean). In key breadbasket regions such as the corn belt in the U.S. and China’s North China Plain, climatic factors contribute to over 60% of observed yield fluctuations [4]. As one of the world’s foremost agricultural producers, China confronts severe climate-induced challenges. The China Meteorology Administration Climate Change Centre [5] has designated China as both a climate-sensitive country and a significantly affected area, with extreme weather events occurring at frequencies substantially exceeding global averages, imposing considerable pressure on agricultural systems and rural livelihoods.

While adaptive agricultural strategies, technological innovations, and risk transfer mechanisms such as agricultural insurance can effectively mitigate adverse climate impacts [6,7], China’s agricultural landscape—characterized predominantly by smallholder farming—exhibits heightened vulnerability to climate stressors. This vulnerability manifests through limited production scale, fragmented operations, low educational attainment, and suboptimal technology adoption rates [8], particularly among farmers with high natural resource dependence [9]. Consequently, a comprehensive analysis of farmers’ climate response behaviors and their determinants carries significant theoretical and practical implications for safeguarding China’s food security and promoting agroecological sustainability.

As an important agricultural production area in China, the Yellow River Basin accounts for one-third of the national rural population [10]. Nevertheless, the agricultural sector remains predominantly reliant on conventional practices, rendering it particularly vulnerable to climate perturbations while demonstrating limited adaptive resilience [11]. Local governments have implemented various adaptation initiatives, including promoting water-efficient irrigation technologies and establishing shelterbelts and agroforestry systems [12]. However, these efforts face dual constraints. First, the institutional infrastructure in rural China remains underdeveloped, with low levels of digitalization, and top-down “one-size-fits-all” intervention policies often fail to account for community-specific characteristics and farmer heterogeneity [13,14,15]; second, farmers’ reliance on governmental agricultural extension resources remains limited when addressing climate-related challenges, with households predominantly depending on endogenous resources [16]. This suggests that examining climate adaptability from farmers’ perspectives may yield more contextually nuanced insights than government-centric approaches.

Therefore, this study examines the role of social capital—specifically social networks, social trust, social participation, and social norms—in farmers’ adoption of climate adaptation behaviors. We further incorporate agricultural extension interactions and digital literacy into our analytical framework to examine their moderating effects in the relationship between social capital and adaptive behaviors, aiming to understand how farmers’ social capital influences their adoption of climate change adaptation behaviors in the Yellow River Basin. This provides valuable insights for developing and implementing effective climate change adaptation strategies in climate-sensitive communities across developing countries, while offering important empirical evidence for research on social capital’s role in climate adaptation. The remainder of this paper is organized as follows. Section 2 reviews the literature on social capital and climate adaptation behavior adoption and develops research hypotheses. Section 3 details data sources, variable operationalization, and research methodology. Section 4 presents empirical results. Section 5 discusses and analyzes the findings. Section 6 summarizes research conclusions and policy implications.

2. Literature Review and Research Hypothesis

2.1. Literature Review

2.1.1. Social Capital

Social capital serves as a critical institutional resource in rural societies, complementing formal institutions while influencing farmers’ access to resources, technologies, and information networks [17]. The conceptual framework of social capital, initially proposed by Bourdieu in his 1980 work “Le capital social: notes provisoires,” was subsequently elaborated by Coleman [18] and Putnam et al. [19] to encompass social networks, social trust, social participation, and social norms. This conceptualization facilitates understanding community dynamics and collective action within vulnerability and resilience contexts [20]. In the context of rural China’s relationship-based social structures, social capital plays a particularly crucial role in farmers’ decision-making processes and resource mobilization, given the prevalence of informal institutions and interpersonal networks in agricultural communities [21].

However, the measurement of social capital exhibits considerable variation across studies, reflecting both its multidimensional nature and contextual embeddedness. While most research incorporates core dimensions of networks, trust, and participation, the specific operationalization differs substantially. Some studies focus on structural social capital, emphasizing network density and scale [22,23,24,25], often measured through indicators like frequency of social interactions, membership in cooperatives, or connections with local government and external stakeholders. Others prioritize cognitive social capital, concentrating on trust and normative systems [20,26], with metrics such as perceived social cohesion or trust in community members and institutions. Functional social capital, encompassing cooperation and mutual assistance mechanisms, represents another measurement approach adopted by researchers examining African mountainous regions [27] and in studies of Chinese rural contexts [28,29], where reciprocity and resource sharing are key indicators. Beyond these traditional dimensions, several studies introduce nuanced categorizations tailored to specific contexts. For instance, distinctions between bonding and bridging social capital are explored in regionalized governance settings [30], highlighting the role of cross-group ties in fostering environmental behavior. Similarly, embedded versus disembedded social capital is used to capture the transition from traditional kinship ties to modern, professional networks in rural China [28]. Other innovative approaches include separating social capital into social networks and training participation to assess climate adaptation intensity [31] or integrating it with psychological constructs like social pressure within the Theory of Planned Behavior to study technology adoption [32]. Notably, Paul et al. [16] and Fletcher et al. [33] have highlighted the “dark side” of social capital, measuring its potential negative effects, including exclusionary practices and inequality exacerbation during adaptation processes. This critical perspective is echoed in studies showing that high social cohesion can constrain flexibility and negatively impact agricultural yields for vulnerable groups like women and landless farmers during environmental stress [26]. These findings underscore the complex, context-dependent nature of social capital, where positive and negative outcomes often coexist depending on socio-economic disparities and regional dynamics [29].

Social capital is a crucial institutional resource in rural societies, significantly influencing farmers’ access to resources and technology adoption. However, existing studies reveal notable limitations. Firstly, the multidimensional nature of social capital leads to significant variations in measurement methods, lacking unified standards and limiting the comparability of results. Secondly, its impact is highly dependent on socio-cultural contexts. Additionally, research on the interaction between social capital and other factors remains insufficient. Therefore, this study disaggregates social capital into four dimensions: social networks, social trust, social participation, and social norms. Based on existing research [24,34,35], we employ 14 indicators and use factor analysis to calculate comprehensive scores. Focusing on the Yellow River Basin, this study introduces digital literacy and agricultural technical assistance as moderating variables, offering a new perspective for research on climate adaptation behaviors.

2.1.2. Climate Adaptation Behaviors

Climate adaptation research emphasizes stakeholder agency, focusing on responses to specific climate stressors or anticipatory actions addressing projected impacts [36,37]. The selection of adaptation pathways constitutes a multifaceted process influenced by both subjective and objective factors, including natural resource endowments and economic, social, cultural, and political conditions [38]. Existing literature identifies two primary categories of determinants affecting farmers’ climate adaptation behaviors: household-level characteristics and external social environmental factors. Household-level factors include farmer demographics (age, gender, education), farm characteristics (land size, crop diversity), and economic conditions (income, credit access) [27,39,40,41,42]. External factors encompass government policies, institutional support, market access, and community characteristics that shape the broader context for adaptation decision-making [43,44,45].

Scholars have operationally defined and categorized farmers’ climate adaptation behaviors from multiple dimensions to deeply reveal their connotations and implementation approaches. Firstly, Smit et al. [46] classified farmers’ behaviors based on adaptation purposes into proactive adaptive behaviors and passive adaptive behaviors. The former includes preventive measures before disasters, such as adjusting planting structures and improving infrastructure, while the latter encompasses coping strategies after disasters, such as adjusting farming schedules. Secondly, based on the timing of adoption, Aryal et al. [47] further subdivided these behaviors into ex ante preventive behaviors (e.g., mulching before planting) and ex post remedial behaviors (e.g., increasing irrigation to mitigate losses). Additionally, Wang et al. [48] categorized adaptation behaviors from the perspective of implementation forms into engineering measures (e.g., constructing risk-prevention projects) and non-engineering measures (e.g., policy support and economic instruments). Meanwhile, Akinyi et al. [49] refined the classification from a resource management perspective into specific measures such as crop management, risk management, land management, water resource management, and livestock management. In summary, through multidimensional classification frameworks, scholars have comprehensively unveiled the complexity and diversity of climate adaptation behaviors, providing a solid theoretical foundation for understanding farmers’ strategies in responding to climate change.

In light of this, this paper adopts a factor-bias perspective and, drawing on the research of Tong et al. [50], further categorizes climate adaptation behaviors into capital-biased behaviors, labor-biased behaviors, and technology-biased behaviors, aiming to systematically explore the influence mechanism of social capital on farmers’ choices of adaptation behaviors.

2.1.3. Social Capital and Adaptive Behavior

Social capital serves as a critical resource for farmers in addressing climate change, playing a vital role in enhancing their climate adaptation behaviors. First, social capital facilitates the dissemination of climate-related information through social networks, helping farmers understand climate risks and adopt adaptive measures. For instance, de Brito et al. [51] found that isolated smallholders learn and adapt to climate change through social interactions, with social networks serving as key channels for sharing information and experience. Second, social capital enables farmers to access external resources such as technical support, credit, and market information, thereby strengthening their adaptive capacity. Martinez-Baron et al. [52] demonstrated in Latin America that farmers adopting climate-smart agriculture (CSA) exhibited lower vulnerability when supported by strong social capital, highlighting the role of social networks in resource acquisition. Additionally, social capital enhances community resilience by fostering collective action, particularly in response to climate-related disasters. Mondal et al. [53], through social network analysis in India’s Sundarbans, showed that internal and external support networks played a crucial role in managing health risks induced by heat stress. Finally, social capital supports diversified adaptation strategies, including crop diversification, adjusted planting schedules, and water management. Khan et al. [54] observed in Pakistan that social capital facilitated access to credit, agronomic services, and climate forecasts, enabling farmers to implement multiple livelihood adaptation strategies.

However, under heterogeneous social capital conditions, farmers cultivating different crop types exhibit varying perceptual frameworks and decision-making processes when confronting climate stressors [55]. Farmers cultivating food crops demonstrate strong enthusiasm and flexibility in adopting climate adaptation technologies. A meta-analysis by Bourgeois et al. [56] revealed that cover crops increased yields of corn and small-grain cereals by 13% and 22%, respectively. This significant yield enhancement motivates food crop growers to adopt cover cropping more readily. Additionally, Handschuch and Wollni [57] found in their study on finger millet in Kenya that smallholder farmers growing traditional food crops relied heavily on social networks and extension services for technology adoption, reflecting a strong willingness to learn and collaborate. Furthermore, Midega et al. [58] demonstrated in East Africa that climate-adapted push-pull intercropping systems increased maize yields by 2.5 times, with food crop farmers showing high acceptance of such integrated adaptation techniques. In contrast, cash crop farmers exhibit more conservative behavior in adopting climate adaptation measures, largely influenced by economic risk considerations. Nong et al. [59] found in northwest China that cash crop farmers were more reluctant to modify existing cropping systems, displaying strong risk aversion toward new technologies. The root of this conservative attitude lies in the high-input nature of cash crops. As Kursun [60] highlighted in their study on Turkey, cash crop production (e.g., rice, wheat, sunflower) is inherently unsustainable due to its heavy reliance on externally purchased inputs (fertilizers, diesel, non-renewable water resources). This high-input model makes cash crop farmers more cautious in adopting new technologies.

Digital literacy, defined as the ability to access and use digital tools like smartphones and agricultural apps, is crucial for modern farming and climate adaptation [61]. Beyond direct impacts, digital literacy may moderate the relationship between social capital and climate adaptation behaviors. Research by Yu et al. [62] shows that digital literacy moderates the effect of social interactions on ICT adoption, enabling rural farmers to leverage networks for technology use. Similarly, Mao et al. [63] found that digital skills amplify the impact of digital extension on adopting climate adaptation technologies among Chinese farmers. Gong et al. [64] further suggest that digital literacy boosts green production efficiency by enhancing market cognition through social networks.

Additionally, agricultural extension services play a vital role in bridging the gap between research and farmers, facilitating the adoption of innovative practices and enhancing resilience to challenges like climate change [65]. Research by Alam et al. [66] in Bangladesh shows that Agricultural Extension Services moderate the effect of wealth on technology adoption, reducing production risks by 2.4% and increasing adoption by 4.2% among rice farmers. Manda et al. [67] reveal that Agricultural Extension Services in Tanzania, combined with social learning, accelerate postharvest technology adoption by 49–61%, strengthening community ties. Aremu and Reynolds [68] in Nigeria indicate that Agricultural Extension Services moderate the impact of asset ownership on welfare, reducing food insecurity by 16%. These studies underscore the role of Agricultural Extension Services in amplifying the benefits of social capital.

The existing literature provides valuable insights into the multifaceted nature of farmers’ climate adaptation behaviors, yet several research gaps and limitations warrant attention. First, while studies have extensively documented the individual effects of social capital, digital literacy, and agricultural extension services on adaptation behaviors, there remains limited understanding of their interactive mechanisms and synergistic effects. Second, although scholars have recognized the heterogeneity between food crop and cash crop farmers in adaptation behavior adoption, existing research lacks in-depth analysis of the underlying mechanisms behind this heterogeneity, particularly the differentiated manifestations of social capital’s role mechanisms. Therefore, building on the exploration of social capital’s impact on farmers’ climate adaptation behaviors, this study further distinguishes the behavioral differences between food crop and cash crop farmers, and separately examines the moderating effects of digital literacy and agricultural extension services among different crop-type farmers, aiming to provide new theoretical perspectives for understanding the heterogeneous mechanisms of farmers’ climate adaptation behaviors.

2.2. Research Hypothesis

2.2.1. The Direct Impact of Social Capital on Farmers’ Climate Change Adaptation Behaviors

The socioeconomic transformation of rural China has precipitated significant shifts in the structure, function, and manifestation of farmers’ social capital endowments. First, robust social networks encourage farmers to adopt climate change adaptation behaviors through facilitating information sharing and mutual support among community members [69]. Through these networks, farmers access climate-relevant information and adaptive technologies, thereby increasing their propensity to implement adaptive practices. Second, social trust mechanisms have transcended traditional dyadic relationships within acquaintance societies and expanded toward institutionalized trust and generalized trust orientations [24]. Farmers’ trust in proximate individuals, village administrative cadres, and governmental entities enhances their ontological security, mitigates information asymmetry risks, and facilitates cooperative action—all of which increase farmers’ willingness to adopt climate adaptation strategies.

Third, social participation modalities have undergone considerable diversification, with farmers’ capacity for engagement in collective action through formal institutional platforms (e.g., agricultural cooperatives, community-based organizations) exhibiting significant enhancement [70]. Participation in collective deliberative processes enables farmers to develop a more nuanced understanding of climate change dynamics and formulate more effective adaptive responses. Fourth, social norms function as behavioral regulators through normative code establishment [71]. Village-level collective constraints and internalized self-discipline mechanisms potentially enhance farmers’ environmental stewardship ethic and ecological consciousness, encouraging proactive adoption of climate-smart agricultural practices. Based on these theoretical considerations regarding how distinct dimensions of social capital may differentially influence adaptive behavior adoption, we propose the following hypotheses:

H1. 

Social capital has a positive impact on farmers’ adoption of climate change adaptation behaviors.

H1a. 

Social networks have a positive impact on farmers’ adoption of climate change adaptation behaviors.

H1b. 

Social trust has a positive impact on farmers’ adoption of climate change adaptation behaviors.

H1c. 

Social participation has a positive impact on farmers’ adoption of climate change adaptation behaviors.

H1d. 

Social norms have a positive impact on farmers’ adoption of climate change adaptation behaviors.

2.2.2. The Impact of Crop Differences on Farmers’ Climate Change Adaptation Behaviors

Staple food crops and commercial cash crops exhibit fundamental differences in production systems, management protocols, economic valuation, and policy frameworks [72]. Empirical evidence indicates that climate variability exerts disproportionately severe impacts on cash crop production systems compared to staple grain cultivation [73]. Consequently, farmers specializing in different crop typologies demonstrate heterogeneous adaptation propensities and strategy preferences. For instance, food crop cultivators, embedded within traditional agricultural production regimes, typically gravitate toward familiar, labor-biased adaptation practices [74]. Moreover, due to comparative profitability constraints in grain production, these farmers generally exhibit diminished incentives for implementing technologically sophisticated climate adaptation strategies [75]. Conversely, cash crop producers, motivated by enhanced yield maximization and quality preservation imperatives, frequently implement multifaceted adaptive strategies to mitigate climate-induced economic losses [76]. These observations lead to our second hypothesis:

H2. 

The influence of social capital on the adoption of climate change adaptation behaviors by food crop farmers and cash crop farmers is different.

2.2.3. The Moderating Effect of Digital Literacy on Farmers’ Climate Change Adaptation Behaviors

Digital literacy can significantly reduce individuals’ dependence on traditional social networks for resource acquisition [77]. Li et al. [78] indicate that the protective effects of digital literacy are reinforced by cognitive ability but weakened by social trust. Castilla et al. [79] demonstrate that digital literacy training enables elderly users to independently access information systems, reducing their reliance on interpersonal support networks. Therefore, for farmers with higher digital literacy, independent access to climate information through digital platforms may reduce their dependence on face-to-face social interactions. This phenomenon creates a substitution effect—digital channels partially replace the information transmission functions of social capital. Accordingly, this study proposes Hypothesis 3:

H3. 

Digital literacy negatively moderates the relationship between social capital and farmers’ adoption of climate adaptation behaviors, such that the positive effect of social capital is more pronounced among farmers with lower digital literacy levels.

2.2.4. Scale-Dependent Moderating Effect of Agricultural Extension Services on Farmers’ Climate Change Adaptation Behaviors

Due to differences in resource endowments and operational characteristics, the moderating effect of agricultural extension services on the relationship between social capital and adaptive behavior may vary significantly across different farm scales. Existing studies provide multidimensional evidence for this scale-dependent moderating effect. Moore & Niles [80] found in Madagascar that smallholder farmers with lower social connectivity rely more on agricultural extension services to complement social capital deficiencies, showing a complementary relationship, while farmers with stronger social connectivity already obtain sufficient support through social networks, leading to diminishing marginal utility of extension services. Darge et al. [81] confirmed in Ethiopia that smallholder farmers require synergistic effects between social networks and extension services due to resource constraints, while large-scale farmers with existing technical and financial foundations are more likely to selectively use either resource. Thuo et al. [82] revealed in Uganda and Kenya that smallholder farmers have a 26% higher probability of obtaining information through weak ties compared to those without external support, indicating additive effects between extension services and social capital, while medium-scale farmers tend to substitute informal social networks with formal extension services, weakening the effectiveness of social capital. Therefore, this study proposes Hypothesis 4:

H4: 

The moderating effect of agricultural extension services on the relationship between social capital and adaptive behavior varies significantly across different farm scales, with substitution effects primarily occurring among large-scale farmers.

3. Materials and Methods

3.1. Overview of the Study Area

China places great emphasis on addressing climate change, having formulated the National Plan for Addressing Climate Change and the National Strategy for Climate Change Adaptation 2035, which clearly define adaptation goals and key areas. Provinces in the Yellow River Basin have actively responded. Ningxia aims to establish a climate adaptation policy system by 2025, Inner Mongolia has proposed nine major actions to enhance resilience in key sectors, and Gansu, Shanxi, and Henan have also developed targeted policies. However, policy implementation at the grassroots level remains insufficient, lacking effective monitoring mechanisms, and support for farmers’ social capital through guidance and incentives is limited [83,84].

The Yellow River Basin constitutes a critical economic zone and ecological barrier in China [85]. However, precipitation is unevenly distributed spatially and temporally, decreasing from upstream to downstream, with summer flooding risks and drought in other seasons in certain areas [86]. Temperatures are on a rising trend, with Ningxia experiencing an increase of 0.37 °C per decade, higher than the national average. Agricultural mechanization levels vary, with major crops including wheat, maize, and cotton, and the industrial structure dominated by traditional farming [84]. Ecological fragility and frequent disasters pose significant challenges to production.

Gansu, Ningxia, Inner Mongolia, Shanxi, and Henan exhibit notable differences in natural conditions, with Ningxia facing significant temperature rises and Gansu experiencing severe drought [87]. Climate change impacts crop yields differently. Henan faces grain production declines, while Inner Mongolia’s forage crops are affected by livestock industry changes. Socioeconomically, Henan has a larger economy with stronger agricultural support, whereas Ningxia and Gansu lag behind, with agriculture constituting a higher economic share but receiving less investment [88]. Adaptation measures include Ningxia’s promotion of water-saving irrigation, Inner Mongolia’s optimized rotational grazing, Shanxi’s enhanced water infrastructure, Henan’s smart agriculture development, and Gansu’s focus on ecological restoration [89]. However, the coverage and effectiveness of these measures need improvement, and the role of social capital has not been adequately considered. Therefore, this study selects Gansu, Ningxia, Inner Mongolia, Shanxi, and Henan in the Yellow River Basin as the primary research areas to explore the impact of social capital on farmers’ adoption of adaptive behaviors.

3.2. Data Sources

This study utilizes primary data collected through a comprehensive household survey administered in the middle and upper reaches of the Yellow River Basin. The data collection was conducted by the “Yellow River Basin Ecological Protection and High-Quality Agricultural and Rural Development” research team from Northwest A&F University’s College of Economics and Management between June and August 2023.

For sampling purposes, we selected six provinces in the middle and upper basin region—Ningxia, Inner Mongolia, Shaanxi, Shanxi, Henan, and Gansu—as research sites to investigate the relationship between social capital and farmers’ climate adaptation behaviors. Employing a multi-stage stratified random sampling methodology, our survey encompassed 6 provinces, 14 counties (districts), 65 townships, and 171 administrative villages, yielding 2549 household-level questionnaires. We subsequently processed the sample data through a rigorous three-stage protocol. First, owing to substantial missing values in the Shaanxi province subsample, we conducted comprehensive missing data analysis, including multiple imputation procedures. Despite these efforts, we could not ensure data representativeness and reliability, necessitating the exclusion of Shaanxi observations from the analytical sample. Second, we eliminated households that did not engage in agricultural production activities during the 2022–2023 cultivation period. Finally, we removed observations with internal inconsistencies in response patterns, ultimately yielding 1772 valid household observations for subsequent empirical analysis. The geographic distribution of the survey sites across the middle and upper reaches of the Yellow River Basin is illustrated in Figure 1, highlighting the topographic complexity and ecological vulnerability of the study region.

3.3. Description and Definition of Variables

3.3.1. Dependent Variable

Drawing on the research by Tong et al. [50], we classify farmers into three categories based on their factor input preferences at the microeconomic level: technology-biased, labor-biased, and capital-biased. This taxonomic approach facilitates enhanced policy targeting by enabling more precise factor-matching between heterogeneous farmer groups and specific adaptation pathways, ultimately improving the allocative efficiency and implementation effectiveness of climate adaptation interventions. Based on an extensive literature review and contextual considerations, we identified and examined nine distinct adaptation behaviors across these three categories.

Capital-biased adaptation behaviors (CAB) encompass farmers’ investment-driven efforts to enhance crop resilience to climate perturbations, operationalized through two binary indicators: “agricultural insurance adoption” and “implementation of engineered adaptation measures requiring capital investment”. The latter category includes infrastructure investments such as irrigation system construction or modernization, drainage facility enhancement, greenhouse installation, and similar capital-biased interventions. These engineering measures directly improve on-farm water resource management efficiency, optimize crop microenvironments, and strengthen agroecosystem resilience to extreme climate events, thereby enhancing adaptive capacity. Concurrently, agricultural insurance functions as a risk transfer mechanism, providing economic safeguards against climate-induced losses and potentially incentivizing the adoption of additional adaptation strategies by reducing financial vulnerability.

Labor-biased adaptation behaviors (LAB) refer to flexible adjustments in agronomic management practices that respond to climate change impacts through modifications to cropland environments and temporal labor allocation patterns. We operationalized this dimension through three indicators: “adjustment of planting/harvesting calendars”, “modification of input application rates (pesticides, fertilizers, seeds, plastic films, etc.)”, and “changes in fertilization and irrigation timing”. These adaptations leverage farmers’ experiential knowledge or externally acquired information to implement rapid tactical adjustments in production strategies, thereby mitigating adverse climate impacts on crop physiological development and yield formation.

Technology-biased adaptation behaviors (TAB) involve the adoption of novel agricultural technologies and germplasm resources to address climate challenges, typically requiring farmers to possess certain technical literacy and innovation capabilities. This dimension was measured through four indicators: “alternative crop introduction”, “crop variety/structural adjustments”, “cultivation method modifications”, and “adoption of water-efficient irrigation technologies”.

Since farmers frequently implement multiple adaptation strategies simultaneously in actual production systems, we constructed an adaptation intensity index based on the number of adaptation categories (technological, labor-biased, and capital-biased) adopted by each household, with values ranging from 0 to 3. A value of 0 indicates non-adoption of any adaptation strategy, while a value of 3 signifies adoption across all three adaptation categories.

Descriptive statistics reveal substantial heterogeneity in adaptation behavior adoption among sampled households. As illustrated in Figure 2, approximately 40% of sampled farmers had not implemented any adaptation strategies, indicating significant barriers to climate adaptation within the study region. Among households that had adopted adaptation measures, capital-biased strategies exhibited the highest implementation rate, followed by labor-biased and technology-biased adaptations, respectively. This adoption pattern potentially reflects farmers’ preference for adaptation strategies that provide immediate economic risk reduction benefits while requiring minimal technical knowledge acquisition.

3.3.2. Core Independent Variable

This study operationalizes social capital through Putnam et al.’s [19] conceptual framework, measuring it along four distinct dimensions: social networks, social trust, social participation, and social norms. The specific variable descriptions are shown in

Table 1. Definition and measurement of social capital variables.

Variable Name	Dimension	Measurement Item	Assignment Criteria	Factor Loading	Cumulative%
Social Capital	Social Trust	Degree of trust in family members	Very distrustful— very trustful: 1~5	0.785	40.684%
		Degree of trust in relatives, neighbors, and friends		0.749
		Degree of trust in village cadres		0.751
		Degree of trust in government information release		0.744
	Social Network	The closeness of interpersonal relationships	Never—Frequently: 1~5	0.715	12.515%
		The frequency of family gatherings		0.892
		The frequency of gatherings with friends		0.872
		The number of people who can offer help in times of difficulty	Very few—Very many: 1~5	0.645
	Social Participation	Level of participation in group activities	Never—Frequently: 1~5	0.826	10.541%
		Degree of attention to national and social affairs		0.802
		Degree of interest in village committee elections		0.812
	Social Norm	Do you consider this village’s rules and policies to be operating effectively?	Strongly Disagree— Strongly Agree: 1~5	0.667	7.091%
		Do the actions of other villagers affect your personal conduct?		0.782
		Are you willing to correct your own inappropriate behavior?		0.745

.

(a) Social networks. This study follows Li et al.’s [90] research and selects the following indicators to reflect farmers’ social network characteristics: “interpersonal relationship closeness”, “frequency of gatherings with relatives and friends”, and “number of potential assistance providers during difficulties”. The first two indicators capture network tie strength, reflecting the robustness of farmers’ interpersonal connections that facilitate information diffusion and resource mobilization within the community. The third indicator quantifies farmers’ access to instrumental support during adversity, directly measuring the breadth of their social safety nets. Strong interpersonal relationships typically correlate with expanded support networks, enhancing farmers’ socioecological resilience when confronting climate-related risks and challenges.

(b) Social trust. Within rural China’s context, the “acquaintance society” predicated on consanguineous and geographical proximity remains prevalent, with social capital formation fundamentally dependent on mutual trust and reciprocal recognition among community members [91]. Accordingly, we measured farmers’ social trust through two primary dimensions: “trust levels in family members, relatives, neighbors, and friends” and “trust levels in village cadres and government information dissemination” [92]. The former represents informal trust based on affective bonds, while the latter reflects institutional trust, indicating rural governance effectiveness, governmental credibility, and farmers’ acknowledgment of formal institutional authority.

(c) Social participation. As a critical component of social engagement, participation in collective activities significantly enhances farmers’ information acquisition capabilities and collective interest recognition [93]. We evaluated farmers’ social participation through three indicators: “group activity participation level”, “attentiveness to national and societal affairs”, and “village committee election participation enthusiasm” [94]. These dimensions capture not only farmers’ community embeddedness and social network depth but also provide objective measures of civic initiative, participation consciousness, and social responsibility within the community context.

(d) Social norms are shared guidelines that constrain behavior. Ostrom [95] defines social norms as “shared understandings about actions that are obligatory, permitted, or forbidden”. In rural communities in China, this shared understanding is primarily manifested through “soft rules” such as village regulations, local customs, and traditions, which are used to guide villagers in adhering to environmental rules [96]. Based on this theoretical framework and drawing on the research methodology of Zhao & Xia [97], this study evaluates farmers’ compliance with and perception of general social norms through three dimensions: “perceived effectiveness of village rules and policies”, “influence of other villagers’ actions on personal conduct”, and “willingness to self-correct inappropriate behavior”. These indicators facilitate a comprehensive understanding of normative constraint mechanisms operating within rural communities, including institutional effectiveness perceptions, peer influence dynamics, and internalized moral self-regulation.

We employed factor analysis using SPSS 27 to quantify farmers’ social capital levels and systematically examine dimensional performance variations. Prior to analysis, we conducted a rigorous psychometric assessment of the measurement instrument. Reliability analysis yielded Cronbach’s alpha coefficients of 0.853, 0.822, 0.861, and 0.690 for social networks, social trust, social participation, and social norms dimensions, respectively, indicating satisfactory internal consistency. The Kaiser–Meyer–Olkin (KMO) measure of sampling adequacy was 0.869 (exceeding the 0.6 threshold), and Bartlett’s test of sphericity yielded a chi-square value of 12,394.183 (p = 0.000), confirming the data’s suitability for factor analytic procedures. Employing orthogonal rotation (varimax method) on the factor loading matrix, we extracted four distinct common factors with a cumulative variance explanation of 70.759%, demonstrating robust construct validity.

3.3.3. Moderating Variables

1. Agricultural technical support is measured by the number of contacts farmers had with agricultural technical personnel in the previous year.

2. Digital literacy is comprehensively assessed through four dimensions. First, whether farmers have joined agriculture-related WeChat/QQ groups (Yes = 1, No = 0); second, for farmers who have joined such groups, whether they engage in discussions about agricultural production techniques (Yes = 1, No = 0), exchange information about agricultural input purchases (Yes = 1, No = 0), and share information about agricultural product sales (Yes = 1, No = 0). The scores of these four indicators are summed to obtain the farmer’s digital literacy level, with a total score ranging from 0 to 4.

3.3.4. Control Variables

In addition to the variables mentioned above, certain observable variables may also influence farmers’ climate change adaptation behaviors. Therefore, to prevent bias from omitting important variables, this study controlled for other variables across four dimensions: individual characteristics, household characteristics, land and location characteristics, and external characteristics. Regional dummy variables and crop dummy variables were also included. The specific variable descriptions are shown in

Table 2. Variable description and descriptive statistics.

Variable Classification	Variable Name	Variable Definition	Min	Max	Mean	S.D.
Dependent variable	Farmers’ climate change adaptation behaviors	Number of adaptation behavior types (technological, labor-biased, and capital-biased) adopted	0	3	0.845	0.812
Core independent variable	Social capital	Calculated based on factor analysis	−2.279	1.228	0	0.628
Moderating variables	Agricultural technical support	Number of contacts with agricultural technical personnel in the previous year	0	25	0.984	1.901
	Digital literacy	Farmers’ digital literacy score	0	4	0.865	1.542
Control variable
Individual characteristics	Age	The actual age of the respondent	25	89	57.148	10.109
	Education level	The actual number of years of education of the respondent	0	21	7.700	3.673
	Farming experience	The number of years the respondent has been engaged in farming	1	65	33.297	12.297
	Gender	Male = 1, Female = 0	0	1	0.939	0.240
Family characteristics	Land scale	Log of farm operational area (Mu)	0.262	6.804	2.520	0.871
	Total income	Family total income in 2022 / 10,000	0.220	157.233	9.129	10.903
	Agricultural Labor Force	Household agricultural labor force in 2022	0	6	1.802	0.796
Land and location characteristics	Farmland soil fertility	Very Poor—Very Good: 1~5	1	5	3.384	0.821
	Distance to agricultural supplies store	Distance from home to nearest agricultural supplies store (km)	0.1	50	6.866	6.941
External characteristics	Climate disaster support policies	Is there local support for measures to deal with climate disasters? 1 = Yes; 0 = No.	0	1	0.144	0.352
	Climate change perception	Have you experienced the impact of climate change in the past five years? 1 = Yes; 0 = No.	0	1	0.564	0.496
	Agricultural irrigation infrastructure	Agricultural irrigation infrastructure improvement level (Very poor—Very excellent: 1~5)	0	5	2.743	1.351
Region (Henan as Reference Group)	Ningxia	Household Village in Ningxia (1 = Yes; 0 = No)	0	1	0.068	0.252
	Inner Mongolia	Household Village in Inner Mongolia (1 = Yes; 0 = No)	0	1	0.219	0.414
	Gansu	Household Village in Gansu (1 = Yes; 0 = No)	0	1	0.288	0.453
	Shanxi	Household Village in Shanxi (1 = Yes; 0 = No)	0	1	0.337	0.473
Crop (Cash Crop as Reference Group)	Food crop	Crop Type: Food crop (1 = Yes; 0 = No)	0	1	0.793	0.405

.

3.4. Method and Model Specification

3.4.1. Basic Regression Model

This paper focuses on examining the impact of social capital on farmers’ climate change adaptation behaviors. Since the dependent variable “climate change adaptation behavior” is assigned values from 0 to 3, representing an ordinal categorical variable with progressive characteristics, we selected the Ordered Probit model for baseline regression. The empirical model is specified as:

$${adaption}_i^{*}={\beta }_0+{\beta }_1{SC}_i+{\beta }_2{C}_i+{\epsilon }_i$$

(1)

${adaption}_i^{*}$ is a latent variable that cannot be directly observed. ${SC}_i$ is the explanatory variable in this paper, which is the social capital score of farmers calculated by the factor analysis method; $C_i$ is a set of control variables, including farmers’ individual characteristics, household characteristics, geographical and location characteristics, external characteristics, as well as provincial dummy control variables and crop dummy control variables. $\beta$ is the parameter to be estimated; ${\epsilon }_i$ is the random disturbance term. Using the latent variable, we can construct the selection model for the dependent variable $adaption$:

$${adaption}_i=\left\{\begin{array}{c}0, if {adaption}_i^{*}\le r_0\\ 1, if r_0<{adaption}_i^{*}\le r_1\\ 2, if r_1<{adaption}_i^{*}\le r_2\\ 3, if r_2<{adaption}_i^{*}\end{array}\right.$$

(2)

In Equation (2), ${adaption}_i$ represents the number of climate adaptation strategies adopted by the farmer, ranging from 0 to 3. $r_0$, $r_1$, $r_2$ and $r_3$ are the threshold values of the latent variable ${adaption}_i^{*}$. When ${adaption}_i^{*}\le r_0$, i.e., ${adaption}_i=0$, it indicates that the farmer does not adopt any type of adaptation behavior and has weak adaptability. When $r_0<{adaption}_i^{*}\le r_1$, that is, ${adaption}_i=1$, it means that the farmer adopts one type of adaptation strategy. When $r_1<{adaption}_i^{*}\le r_2$, that is, ${adaption}_i=2$, it implies that the farmer adopts two adaptation strategies. Finally, when $r_2<{adaption}_i^{*}$, that is, $adaption=3$, the farmer adopts all three strategies, reflecting a high level of adaptability.

3.4.2. Moderating Effect Model

To examine the moderating effects of digital literacy and agricultural technical support on the influence of social capital on farmers’ climate adaptation behaviors, this study introduces interaction terms between digital literacy and social capital, as well as between agricultural technical support and social capital, into the model to test the moderating effects. ${DL}_i$ and ${TS}_i$ represent the digital literacy level and agricultural technical support level of the i-th farmer, respectively.

$${adaption}_i^{*}={\beta }_0+{\beta }_1{SC}_i+{\beta }_2{DL}_i+{\beta }_3\left({{SC}_i\ast DL}_i\right)+{{\beta }_4{C}_i+\epsilon }_i$$

(3)

$${adaption}_i^{*}={\delta }_0+{\delta }_1{SC}_i+{{\delta }_2{TS}_i+\delta }_3\left({{SC}_i\ast TS}_i\right)+{\delta }_4{C}_i+{\epsilon }_i$$

(4)

4. Analysis and Discussion of Results

4.1. Baseline Regression Analysis

Prior to model estimation, we conducted multicollinearity diagnostics using Variance Inflation Factor (VIF) analysis. Results yielded a maximum VIF of 3.99 and a mean VIF of 1.82—both substantially below the conventional threshold of 10—indicating the absence of problematic multicollinearity among predictor variables. Subsequently, we employed STATA 18.0 statistical software to estimate Ordered Probit regression models examining the relationships between our dependent variable (climate adaptation behavior adoption), primary explanatory variables (social capital dimensions), and control variables. We implemented a two-stage modeling strategy, first incorporating the composite social capital score derived from factor analysis to assess its aggregate impact on farmers’ adaptation behavior adoption (Model 1), then disaggregating social capital into its constituent dimensions to examine the differential effects of social networks, social trust, social participation, and social norms (Model 2). Additionally, we computed average marginal effects (AMEs) to quantify the magnitude of social capital’s influence on adaptation behavior adoption probabilities. The empirical findings are presented in

Table 3. Baseline regression results.

Variable Name		Model (1)	Model (2)	The Marginal Effect of Model (1)
				None	1 Type	2 Types	3 Types
Core independent variable	Social capital	0.227 ***		−0.079 ***	0.020 ***	0.046 ***	0.013 ***
		(0.048)		(0.016)	(0.004)	(0.010)	(0.003)
	Social trust		0.102 ***
			(0.030)
	Social network		0.114 ***
			(0.028)
	Social participation		0.068 **
			(0.029)
	Social norm		0.020
			(0.028)
Control variable
Individual characteristics	Age	0.003	0.003	−0.001	0.000	0.001	0.000
		(0.004)	(0.004)	(0.001)	(0.000)	(0.001)	(0.000)
	Education level	−0.002	−0.004	0.001	−0.000	−0.000	−0.000
		(0.008)	(0.008)	(0.003)	(0.001)	(0.002)	(0.000)
	Farming experience	−0.007 **	−0.007 **	0.002 **	−0.001 **	−0.001 **	−0.000 **
		(0.003)	(0.003)	(0.001)	(0.000)	(0.001)	(0.000)
	Gender	−0.163 *	−0.187 **	0.057 *	−0.014 *	−0.033 *	−0.009 *
		(0.093)	(0.093)	(0.033)	(0.008)	(0.019)	(0.005)
Family characteristics	Land scale	0.122 ***	0.123 ***	−0.042 ***	0.011 ***	0.025 ***	0.007 ***
		(0.037)	(0.037)	(0.013)	(0.003)	(0.007)	(0.002)
	Total income	−0.001	−0.001	0.000	−0.000	−0.000	−0.000
		(0.002)	(0.003)	(0.001)	(0.000)	(0.000)	(0.000)
	Agricultural labor force	−0.070 **	−0.076 **	0.024 **	−0.006 *	−0.014 *	−0.004 *
		(0.036)	(0.035)	(0.012)	(0.003)	(0.007)	(0.002)
Land and location characteristics	Farmland soil fertility	−0.049	−0.051	0.017	−0.004	−0.010	−0.002
		(0.032)	(0.032)	(0.011)	(0.003)	(0.007)	(0.002)
	Distance to agricultural supplies store	0.003	0.003	−0.001	0.000	0.001	0.000
		(0.004)	(0.004)	(0.001)	(0.000)	(0.001)	(0.000)
External Characteristics	Climate disaster support policies	0.334 ***	0.324 ***	−0.116 ***	0.029 ***	0.068 ***	0.019 **
		(0.068)	(0.068)	(0.024)	(0.007)	(0.014)	(0.004)
	Climate change perception	0.495 ***	0.493 ***	−0.172 ***	0.044 ***	0.101 ***	0.028 ***
		(0.058)	(0.059)	(0.019)	(0.006)	(0.012)	(0.004)
	Agricultural irrigation infrastructure	0.105 ***	0.101 ***	−0.036 ***	0.009 ***	0.021 ***	0.006 ***
		(0.024)	(0.024)	(0.008)	(0.002)	(0.005)	(0.002)
Province dummy variable		Yes	Yes	Yes	Yes	Yes	Yes
Crop dummy variable		Yes	Yes	Yes	Yes	Yes	Yes
Wald χ2		259.26 ***	271.69 ***
LR chi-square		0.068	0.070
Observations		1772	1772	1772	1772	1772	1772

Note: *, **, *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively; robust standard errors are reported in parentheses.

.

4.1.1. Analysis of Social Capital Impact

Model 1 results demonstrate that aggregate social capital exerts a statistically significant positive effect on farmers’ climate adaptation behavior adoption, underscoring social capital’s critical role in mobilizing adaptive action among agricultural households. Decomposing social capital into its constituent dimensions in Model 2 reveals heterogeneous effects across dimensions: social trust, social networks, and social participation each exhibit significant positive influences on adaptation behavior adoption, while social norms fail to demonstrate a statistically significant effect.

These findings suggest differential instrumental pathways through which social capital dimensions facilitate adaptation. Social networks and social trust both show positive and statistically significant effects at the 1% level, while social participation demonstrates a positive and significant effect at the 5% level. These findings collectively indicate that higher levels of social connectivity, interpersonal trust, and community engagement are associated with a greater likelihood of adopting more comprehensive climate adaptation practices.

The non-significant effect of social norms merits particular attention, potentially attributable to limited norm internalization within rural communities regarding climate adaptation. Rural households may prioritize experiential knowledge and direct interpersonal communication over normative guidance when formulating adaptation strategies, thereby attenuating social norms’ behavioral influence in this domain. These empirical findings substantiate hypotheses H1, H1a, H1b, and H1c, while failing to support hypothesis H1d.

Analysis of average marginal effects from Model 1 reveals that each standard deviation increase in social capital corresponds to a 7.9 percentage point decrease in the probability of non-adoption of any climate adaptation strategy (p < 0.01), while increasing the probabilities of adopting one, two, and three adaptation types by 2.0, 4.6, and 1.3 percentage points, respectively (all p < 0.01). The strongest marginal effect magnitude for moderate adaptation intensity (two adaptation types) suggests potential threshold effects in social capital’s influence on adaptation portfolios.

Several potential explanations may underlie these observed effects. First, social capital may enhance information fluidity and trust networks, potentially enabling farmers to acquire climate change impact knowledge through community and neighborhood relationships, thereby facilitating proactive adaptation. Second, farmers with higher social capital typically demonstrate greater engagement in collective activities and cooperative initiatives, which could provide platforms for knowledge exchange and experiential learning while strengthening environmental awareness and collective action capabilities. Finally, robust social capital endowments may expand farmers’ access to critical resources—including technical guidance, financial support, and policy information—potentially enhancing their self-efficacy and perceived behavioral control regarding adaptation implementation.

4.1.2. Analysis of Control Variables’ Effects

Farming experience negatively influences farmers’ adoption of climate change adaptation behaviors at the 5% significance level. The marginal effects show that each additional unit of farming experience increases the probability of adopting no adaptation measures by 0.2%, while decreasing the probability of adopting 1–2 types of adaptation by 0.1% each. This suggests that experienced farmers exhibit path dependency, being more likely to maintain traditional cultivation methods rather than adopt new adaptation strategies. Farming scale positively influences the adoption of climate change adaptation behaviors at the 1% significance level, indicating that farmers with larger cultivation scales are more likely to adopt adaptive behaviors. This may stem from the fact that large-scale farm operators rely on agriculture as their primary income source, have heightened sensitivity to climate risks, and can more easily distribute the costs of adopting new technologies across their operations. The number of agricultural laborers in a household negatively impacts the adoption of adaptive behaviors at the 5% significance level, suggesting that households with more agricultural laborers are less motivated to adopt adaptive behaviors. This may be because labor-abundant households tend to rely more on traditional farming methods, such as increasing labor input, to address climate change. Government support for local meteorological disasters shows a significant positive effect at the 1% level, indicating that government subsidies, training, and other support measures for disasters caused by meteorological changes effectively encourage farmers to adopt adaptive behaviors. Policy support may enhance farmers’ confidence in responding to climate change by reducing perceived economic risks (e.g., post-disaster compensation) or providing technical guidance (e.g., promotion of drought-resistant technologies). Climate change perception also shows a significant positive effect at the 1% level, indicating that farmers’ climate change perception strongly motivates adaptation adoption, with the strongest effect on 2-type strategies. Direct lessons from climate change losses increase farmers’ risk awareness, prompting them to proactively adjust production strategies to avoid future risks. The level of water conservancy facilities has a significant positive effect at the 1% level, indicating that in areas with better infrastructure conditions, farmers are more likely to adopt adaptive behaviors. Well-developed water conservancy facilities can directly enhance farmers’ capacity to respond to climate change, while potentially indirectly promoting the adoption of other adaptive technologies by reducing resource constraints.

4.2. Endogeneity Analysis

Since the Ordered Probit model may have endogeneity issues, which could lead to bias in the estimation of social capital coefficients and marginal effects, this study adopted the Conditional Mixed Process (CMP) estimation method following Roodman [93]. Given that the dependent variable is an ordinal variable, the CMP method more accurately reflects data characteristics than the traditional 2SLS approach. The CMP method requires selecting appropriate instrumental variables. Considering that reasonable instrumental variables need to satisfy both relevance and exogeneity requirements, we selected the average social capital of other farmers in the same village (excluding the individual farmer) as the instrumental variable. The rationale for this selection is based on peer effect theory: the average social capital in a farmer’s environment influences the formation of their personal social capital to a certain extent, but an individual farmer’s social capital level is insufficient to change the overall social capital situation, thus ensuring the exogeneity of this instrumental variable. Before formally conducting CMP estimation, we first verified the effectiveness of the selected instrumental variable using the weak instrument test from the traditional 2SLS method. The test results showed an F-statistic of 14.28, exceeding the empirical threshold of 10, further confirming the effectiveness of the selected instrumental variable.

According to the results in

Table 4. Conditional Mixed Process estimation results.

Variable Name	Phase I	Phase II	The Marginal Effect
			None	1 Type	2 Types	3 Types
Social capital		1.168 ***	−0.355 ***	0.040 ***	0.167 ***	0.148 *
		(0.272)	(0.066)	(0.014)	(0.012)	(0.087)
Mean value of social capital	0.264 ***
	(0.069)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes
province dummy variable	Yes	Yes	Yes	Yes	Yes	Yes
crop dummy variable	Yes	Yes	Yes	Yes	Yes	Yes
Phase I F-value	10.15
Observations	1772	1772	1772	1772	1772	1772
atanhrho_12		−0.673 **
		(0.271)

Note: *, **, *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively; robust standard errors are reported in parentheses.

, the first-stage estimation shows that the instrumental variable has a positive effect on social capital at the 1% significance level, with an F-value exceeding 10, indicating that the instrumental variable is effective. In the second stage, the atanhrho value is significantly different from zero at the 5% significance level, indicating a significant correlation between the two-stage equations, thus validating the effectiveness of the CMP estimation method. Moreover, the absolute values of the marginal effects of social capital on farmers’ adoption of climate change adaptation behaviors are higher than those in Table 3, suggesting that endogeneity issues may have led to an underestimation of social capital’s influence on farmers’ adoption of climate change adaptation behaviors.

4.3. Robustness Test

a. Alternative estimation model. When the dependent variable is an ordinal categorical variable, it can be treated as an interval variable and estimated using ordinary least squares (OLS model) [98]. Accordingly, we employed the OLS model to re-examine the impact of social capital on farmers’ adoption of climate change adaptation behaviors, resulting in Models (3) and (4). Model (3) shows that social capital positively and significantly influences farmers’ adoption intentions at the 1% level; Model (4) demonstrates that social trust, social networks, and social participation positively and significantly influence farmers’ adoption intentions at the 1%, 1%, and 5% levels, respectively, while the influence of social norms is not significant. The results estimated using the OLS model are consistent with the baseline regression results.

b. Alternative dependent variable. Instead of categorizing adaptation behaviors into capital-biased (CAB), technology-biased (TAB), and labor-biased types (LAB), we used the total number of adaptive behaviors adopted by farmers to replace the number of types of adaptive behaviors adopted, to further examine the impact of social capital on farmers’ adoption of adaptive behaviors. As shown in Models (5) and (6), the results are consistent with those obtained from the baseline model and OLS model. The results are presented in

Table 5. Estimation results of the OLS model and redefined dependent variable.

Variable Name	Model (3)	Model (4)	Model (5)	Model (6)
	AB	AB	AB *	AB *
Social capital	0.154 ***		0.221 ***
	(0.032)		(0.046)
Social trust		0.071 ***		0.108 ***
		(0.020)		(0.029)
Social network		0.073 ***		0.081 ***
		(0.018)		(0.029)
Social participation		0.046 **		0.058 **
		(0.019)		(0.029)
Social norm		0.009		0.031
		(0.018)		(0.026)
Control variables	Yes	Yes	Yes	Yes
province dummy variable	Yes	Yes	Yes	Yes
crop dummy variable	Yes	Yes	Yes	Yes
F-value/Wald χ2	18.49 ***	16.63 ***	261.12 ***	262.15 ***
R2/Adjusted R2	0.150	0.155	0.051	0.051
Observations	1772	1772	1772	1772

Note: AB stands for adaptation behaviors. *, **, *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively; robust standard errors are reported in parentheses.

.

c. Alternative explanatory variable indicators. In this section, the measurement of social capital was changed from factor analysis to the entropy weight method. The results show that social capital still positively influences farmers’ adoption of climate change adaptation behaviors at the 1% significance level. According to the marginal effects analysis of Model (7), increased social capital significantly reduces the probability of farmers not adopting any adaptive behaviors, while significantly increasing the probabilities of adopting one, two, and three types of adaptive behaviors. Furthermore, social networks and social trust calculated using the entropy weight method positively influence farmers’ adoption of adaptive behaviors at the 1% and 5% significance levels, respectively, while social participation and social norms show no significant effects. The results are presented in

Table 6. Estimation Results with Alternative Social Capital Measurement Indicators.

Variable Name	Model (7)	Model (8)	The Marginal Effect of Model (7)
			None	1 Type	2 Types	3 Types
Social capital	0.962 ***		−0.335 ***	0.085 ***	0.196 ***	0.055 ***
	(0.191)		(0.065)	(0.018)	(0.039)	(0.013)
Social trust		0.462 ***
		(0.152)
Social network		0.465 **
		(0.214)
Social participation		0.204
		(0.157)
Social norm		−0.130
		(0.162)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes
province dummy variable	Yes	Yes	Yes	Yes	Yes	Yes
crop dummy variable	Yes	Yes	Yes	Yes	Yes	Yes
F-value/Wald χ2	267.24 ***	270.33 ***
R2/Adjusted R2	0.068	0.070
Observations	1772	1772	1772	1772	1772	1772

Note: **, *** indicate statistical significance at the 5%, and 1% levels, respectively; robust standard errors are reported in parentheses.

.

4.4. Heterogeneity Test

4.4.1. Differences in the Impact of Social Capital on Climate Adaptation Behaviors Among Farmers Growing Different Crop Types

Given that different crop types are associated with distinct adaptation strategies, this section divides the sample into two groups—food crop farmers and cash crop farmers—for separate empirical testing. This allows for a comparative analysis of the effects of social capital across the two types of farmers. The detailed estimation results are presented in

Table 7. Subsample Regression Results.

Variable Name	Food Crops		Cash Crops
	Model (9)	Model (10)	Model (11)	Model (12)
Social capital	0.212 ***		0.307 **
	(0.054)		(0.106)
Social trust		0.083 **		0.180 ***
		(0.033)		(0.063)
Social network		0.140 ***		0.017
		(0.031)		(0.070)
Social participation		0.070 **		0.065
		(0.033)		(0.055)
Social norm		0.013		0.044
		(0.032)		(0.061)
Control variables	Yes	Yes	Yes	Yes
province dummy variable	Yes	Yes	Yes	Yes
crop dummy variable	Yes	Yes	Yes	Yes
Wald χ2	165.73 ***	180.60 ***	127.73 ***	130.94 ***
R2/Adjusted R2	0.059	0.064	0.128	0.129
Observations	1405	1405	367	367

Note: **, *** indicate statistical significance at the 5%, and 1% levels, respectively; robust standard errors are reported in parentheses.

. Overall, social capital exerts a statistically significant influence on the adoption of climate adaptation behaviors by both grain and cash crop farmers at the 1% significance level. This finding further reinforces the incentive role of social capital in promoting farmers’ adoption of adaptive behaviors in response to climate change.

The relationship between social capital and farmers’ climate adaptation behaviors exhibits pronounced heterogeneity across agricultural production systems. Model (10) reveals that social networks exert a statistically significant positive effect on food crop farmers’ adaptation behavior adoption, consistent with aggregate sample findings. Conversely, for cash crop farmers, social networks demonstrate no significant effect. Regarding social trust, both farmer categories exhibit significant positive effects, though with differential magnitudes—food crop farmers versus cash crop farmers. These findings suggest differential causal pathways: cash crop farmers primarily leverage social trust mechanisms in adaptation processes, while food crop farmers predominantly mobilize social network resources. Furthermore, social participation yields significant effects exclusively for food crop farmers, while social norms fail to demonstrate significant influence across either agricultural system.

This observed heterogeneity can be attributed to several interrelated factors. First, divergent economic return expectations and risk profiles associated with different cropping systems likely drive differential social capital utilization. Cash crops typically generate higher market value and potential returns but face greater market volatility and price sensitivity [99,100], incentivizing farmers to cultivate robust trust relationships as mechanisms for accessing reliable market information and economic support amid heightened uncertainty during adaptation processes. Second, structural differences in social network characteristics play a crucial role in this differentiation. Grain cultivation frequently embeds within traditional agrarian sociotechnical regimes and family-centered production systems, fostering densely interconnected homophilous networks [55]. Consequently, grain producers exhibit greater reliance on experiential knowledge circulation and reciprocal support mechanisms when confronting climate stressors, elevating the instrumental value of network-based resources within this agricultural system. Finally, fundamental differences exist in social capital formation and mobilization pathways. Cash crop farmers typically adopt market-biased livelihood strategies, predominantly leveraging external knowledge sources and formal institutional arrangements rather than traditional community-based support structures [55]. This strategic orientation renders trust-based relationships particularly critical to their adaptation portfolios, while normative compliance and participatory engagement remain relatively subordinate in their decision-making calculus.

Marginal effects analysis (

Table 8. Marginal Effects of Social Capital.

	Food Crops		Cash Crops
	Marginal Effect	Robust SE	Marginal Effect	Robust SE
No adaptive behavior adopted	–0.075 ***	0.019	–0.096 ***	0.032
One category adopted	0.020 ***	0.005	0.021 ***	0.007
Two categories adopted	0.043 ***	0.011	0.061 ***	0.022
All three categories adopted	0.013 ***	0.004	0.014 **	0.006

Note: **, *** indicate statistical significance at the 5%, and 1% levels, respectively.

) reveals that social capital exerts a more pronounced improvement in adaptation behavior adoption propensity among cash crop farmers compared to food crop farmers. Specifically, a one-unit increase in social capital reduces the probability of non-adoption of any adaptation behavior by 7.5% for food crop farmers and 9.6% for cash crop farmers. Simultaneously, for food crop farmers, a one-unit increase in social capital elevates the probability of adopting one and two categories of adaptation behaviors by 2.0% and 4.3%, respectively, whereas for cash crop farmers, these probabilities increase by 2.1% and 6.1%, respectively. This disparity likely stems from the superior market value and production potential of cash crops, which incentivizes farmers to leverage social capital for information, resource, and support acquisition, thereby facilitating more proactive adaptation behavior. Moreover, cash crop cultivation typically necessitates greater external inputs and collaboration; thus, robust social capital facilitates information exchange and cooperation among farmers, enhancing their capacity to respond to environmental changes. Conversely, food crop farmers, characterized by relatively traditional production modes and greater self-sufficiency orientation, may experience attenuated social capital effects on adaptation behavior adoption.

4.4.2. Heterogeneous Effects of Social Capital on Various Types of Climate Adaptation Behaviors

Agricultural production factors possess distinct characteristics, and social capital may exert heterogeneous effects on different types of climate adaptation behaviors across varying crop types. Therefore, this study categorizes farmers’ climate adaptation behaviors into three major types—capital-biased, labor-biased, and technology-biased—for separate analysis. Since these three categories correspond to binary (0–1) variables, we employed a Probit model for regression analysis.

For food crop farmers (

Table 9. Heterogeneity Test: Climate-Adaptation Behaviors Among Food-Crop Growers.

Variable	Technology-Biased		Labor-Biased		Capital-Biased
	Model (13)	Model (14)	Model (15)	Model (16)	Model (17)	Model (18)
Social Capital	–0.006		0.259 ***		0.235 ***
	(0.065)		(0.074)		(0.060)
Social Trust		–0.067		0.152 ***		0.110 ***
		(0.042)		(0.045)		(0.037)
Social Network		0.154 ***		0.025		0.130 ***
		(0.045)		(0.043)		(0.036)
Social Participation		0.093 **		0.053		0.039
		(0.040)		(0.042)		(0.037)
Social Norms		–0.009		0.038		–0.004
		(0.040)		(0.041)		(0.036)
Control Variables	Yes	Yes	Yes	Yes	Yes	Yes
Province Dummies	Yes	Yes	Yes	Yes	Yes	Yes
Wald χ2	93.25 ***	109.81 ***	143.92 ***	144.84 ***	118.35 ***	126.50 ***
Pseudo R2	0.081	0.096	0.119	0.119	0.069	0.073
N	1405	1405	1405	1405	1405	1405

Note: **, *** indicate statistical significance at the 5%, and 1% levels, respectively; robust standard errors are reported in parentheses.

), social capital demonstrated significant positive effects on labor-biased (LAB) and capital-biased adaptation behaviors (CAB), while its impact on technology-biased adaptation behaviors (TAB) was not significant. In contrast, for cash crop farmers (

Table 10. Heterogeneity Test: Climate-Adaptation Behaviors Among Cash-Crop Growers.

Variable	TAB		LAB		CAB
	Model (19)	Model (20)	Model (21)	Model (22)	Model (23)	Model (24)
Social Capital	0.084		0.554 ***		0.097
	(0.155)		(0.134)		(0.135)
Social Trust		0.024		0.276 ***		0.071
		(0.106)		(0.083)		(0.081)
Social Network		–0.081		0.214 **		–0.084
		(0.099)		(0.087)		(0.084)
Social Participation		0.067		0.123		0.028
		(0.104)		(0.078)		(0.078)
Social Norms		–0.099		0.098		0.036
		(0.086)		(0.078)		(0.079)
Control Variables	Yes	Yes	Yes	Yes	Yes	Yes
Province Dummies	Yes	Yes	Yes	Yes	Yes	Yes
Wald χ2	87.39 ***	90.72 ***	53.89 ***	59.04 ***	125.84 ***	125.83 ***
Pseudo R2	0.190	0.196	0.128	0.134	0.248	0.251
N	362	362	367	367	362	362

Note: **, *** indicate statistical significance at the 5%, and 1% levels, respectively; robust standard errors are reported in parentheses.

), social capital only significantly influenced labor-biased adaptation behaviors, with no significant effects on technology-biased or capital-biased behaviors. This difference may be attributed to the strong seasonality and high risk-aversion characteristics of grain production, which is typically family-based, leading farmers to prefer “experience-verified” adaptation measures. Conversely, cash crop cultivation involves higher input costs and experiences greater vulnerability to climate change impacts [73], prompting farmers to adopt labor-biased adaptation behaviors (LAB) with lower learning and implementation costs.

Examining the four dimensions of social capital, we observed that social participation and social norms exhibited relatively weak effects for both grain and cash crop farmers. This may be because food crop production is constrained by rigid natural conditions, while cash crop cultivation is dominated by market fluctuations, causing farmers’ decision-making to be more heavily influenced by their own planting experience and market characteristics. Notably, social trust showed significant effects for both types of farmers, validating Putnam et al.’s [19] social-capital theory that trust acts as an essential enabling mechanism in agricultural production.

From a marginal effects perspective (

Table 11. Marginal Effects of the Social Capital Composite Index.

	Food Crops		Cash Crops
	Marginal Effect	Robust SE	Marginal Effect	Robust SE
Technology-biased	–0.003	0.018	0.019	0.037
Labor-biased	0.062 ***	0.018	0.155 ***	0.036
Capital-biased	0.085 ***	0.021	0.027	0.038

Note: *** indicate statistical significance at the 1% levels.

), social capital demonstrated a greater improvement effect on cash crop farmers’ adoption of labor-biased adaptation behaviors (LAB) compared to food crop farmers. Each unit increase in social capital increased the probability of cash crop farmers adopting labor-biased adaptation behaviors (LAB) by 15.5%, while for food crop farmers, this increase was only 6.2%. Conversely, regarding capital-biased adaptation behaviors (CAB), social capital had a greater improvement effect on food crop farmers than on cash crop farmers. In summary, these findings validate Hypothesis 2, confirming that social capital exerts differential effects on food crop farmers and cash crop farmers in their adoption of climate change adaptation behaviors.

4.5. Moderation Analysis

This section employs moderation effect models by introducing interaction terms between social capital and agricultural extension interaction frequency, as well as between social capital and farmers’ digital literacy, to empirically examine the differential moderating effects of agricultural extension services on the relationship between social capital and climate adaptation behaviors across farmer groups of varying operational scales, and the moderating effect of digital literacy on the relationship between social capital and climate adaptation behaviors (

Table 12. Results of Moderation Analysis.

Variable	Model (25)			Model (26)
	Small-Scale	Medium-Scale	Large-Scale
Social Capital	0.109	0.167 **	0.404 ***	0.225 ***
	(0.089)	(0.080)	(0.082)	(0.048)
Extension Contacts	0.088 **	0.052 **	0.037 ***
	(0.044)	(0.025)	(0.015)
SC * Extension Contacts	−0.025	−0.038	−0.052 **
	(0.047)	(0.059)	(0.021)
Digital Literacy				0.030 *
				(0.018)
SC * Digital Literacy				−0.099 ***
				(0.030)
Control Variables	Yes	Yes	Yes	Yes
Province Dummies	Yes	Yes	Yes	Yes
Crop Dummy Variable	Yes	Yes	Yes	Yes
Wald χ2	96.32 ***	69.12 ***	161.40 ***	278.88 ***
Pseudo R2	0.062	0.064	0.099	0.071
N	656	526	590	1772

Note: SC stands for social capital. *, **, *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively; robust standard errors are reported in parentheses.

).

4.5.1. Heterogeneous Moderating Effects of Agricultural Extension Services

Model (25) reveals significant scale-based heterogeneity in the moderating effects of agricultural extension interactions, providing new empirical evidence for understanding the interplay between formal and informal institutions in agricultural climate adaptation.

In the large-scale farmer subsample, the interaction term between social capital and agricultural extension interaction frequency exhibits a significantly negative coefficient (p < 0.05), indicating that agricultural extension interactions attenuate the positive impact of social capital on climate adaptation behaviors. This finding supports the resource substitution hypothesis, suggesting that formal agricultural extension services partially substitute for the functions of informal social capital. Notably, despite the negative moderating effect, social capital itself maintains a significantly positive influence (p < 0.01), demonstrating that its role is weakened but not entirely replaced. In contrast, the moderating effects for small- and medium-scale farmers are non-significant, suggesting that agricultural extension interactions and social capital may function independently or form complementary relationships among these farmer groups.

The findings indicate that for large-scale farmers, agricultural extension services can effectively substitute for the informational functions of social capital; however, for smallholder farmers, merely increasing the frequency of technical services may not achieve the desired outcomes and requires coordinated advancement with social network development. This provides scientific evidence for formulating differentiated agricultural adaptation policies. Therefore, Hypothesis 4 is supported.

4.5.2. Moderating Effect of Digital Literacy

Model (26) introduces the interaction term between social capital and farmers’ digital literacy to empirically examine the moderating effect of farmers’ digital literacy on the relationship between social capital and their adoption of climate-adaptive behaviors.

The results show that the interaction term between social capital and digital literacy presents a significant negative relationship (p < 0.01), indicating that digital literacy plays a negative moderating role in the process of social capital influencing climate adaptive behaviors. This finding reveals a substitution effect between digital literacy and social capital: when farmers have higher levels of digital literacy, the promoting effect of social capital on climate adaptive behaviors weakens; conversely, for farmers with lower digital literacy levels, the role of social capital is more prominent.

This indicates the trend of diversified channels for farmers to access resources and information in the digital age. Farmers with high digital literacy can independently obtain climate change-related information and learn adaptive technologies through digital channels such as the internet and mobile applications, reducing their dependence on traditional social networks. Meanwhile, farmers with lower digital literacy rely more on interpersonal relationships and experience sharing within social capital as traditional ways to obtain information and support. This suggests that the popularization of digital technology provides farmers with new channels for information acquisition, which can supplement or even substitute certain functions of traditional social capital to some extent. Therefore, Hypothesis 3 is supported.

5. Discussion

This study employs ordered Probit models and mediation effect analysis to comprehensively examine the mechanisms and heterogeneous characteristics of farmers’ climate adaptation behaviors from a social capital perspective: (1) how farmers’ composite social capital and its constituent dimensions—social trust, social networks, social participation, and social norms—influence climate adaptation behaviors; (2) the heterogeneous manifestations of these effects across different cropping systems; and (3) the moderating role of digital literacy and the differential moderating effects of agricultural extension services across different farming scales.

The study finds that the adoption rate of adaptation behaviors among farmers in five provinces of the Yellow River Basin is 60%, showing a significant gap compared to other regions. This adoption rate is relatively low when compared with relevant domestic and international studies. Li et al. [22] found that farmers’ adaptation behavior adoption rate in China’s Dazu District reached 82.69%, mainly concentrated on crop adjustment measures; studies on the Qinghai-Tibet Plateau showed that over 80% of farmers adopted at least one climate adaptation strategy [21]; Baffour-Ata et al. [101] found in Ghana that 88% of yam growers adjusted planting periods and 86% adopted early-maturing varieties. These comparative results indicate that farmers in the Yellow River Basin still have considerable room for improvement in climate adaptation, particularly with technology-biased strategy adoption rates of only 13%, significantly lower than levels in other regions.

Our research also finds that social trust, social networks, and social participation significantly promote farmers’ adoption of climate adaptation strategies (p < 0.01, p < 0.01, and p < 0.05, respectively), while social norms have limited influence. This finding shows both consistency and differences with domestic and international studies. Domestic research provides strong support. Wang et al. [21] confirmed the significant positive impact of social trust on adaptation strategy selection in the Qinghai-Tibet Plateau, highly consistent with this study. Li et al. [22] found in Chongqing’s Dazu District that trust in agricultural extension service information was the largest factor influencing adaptation behaviors. Chen et al. [23] found, based on a six-province survey, that farmers with higher social capital significantly improved drought adaptation capacity, with farmers adopting adaptation measures having an average of 0.27 government-employed relatives, twice that of non-adopters. International research provides comparative perspectives. Paul et al. [16] found dual effects of social trust in Ethiopia—both encouraging participation and potentially leading to complacency. Petzold and Ratter [20] emphasized that high trust promotes collective adaptation actions. Nigerian studies showed that social networks and participation have significant positive impacts on adaptation decisions [102], while studies in Australia and Canada revealed limitations of social capital’s role in developed countries [33,103].

Notably, the role of social norms exhibits clear measurement sensitivity characteristics. Wang et al. [21] found significant positive impacts of social norms from reciprocity norms and moral constraint dimensions. Lo [103] similarly found promotional effects through group and individual behavior perception measures, while this study found limited effects based on three-dimensional measures of institutional norms, social influence, and self-norms, consistent with Zhang et al.’s research results based on prohibitive social norms [24]. This indicates that the effectiveness of social norms is highly dependent on measurement methods and cultural backgrounds, requiring future research to reach a greater consensus on the conceptual definition and measurement methods of social norms.

This study categorizes farmers’ adaptation behaviors into capital-biased, technology-biased, and labor-biased types. This classification framework, based on production factors, provides a beneficial complement to the existing literature’s classification based on specific measures, offering a new theoretical perspective for understanding social capital mechanisms. The study finds significant differences in the dependence of different types of adaptation behaviors on social capital. For technology-biased adaptation behaviors, we find that social capital has no significant impact, echoing studies by Burnham & Ma [104], Jin et al. [105], and Ogunleye et al. [102], indicating that technological complexity, high investment risks, and external support needs make technology-biased adaptation behaviors face challenges beyond traditional social network functions. In contrast, social capital has significant positive impacts on both labor-biased and capital-biased adaptation behaviors, consistent with studies by Di Falco et al. [106], Naz et al. [107], Sanfo et al. [108], and Kakumanu et al. [109]. Labor-biased adaptation behaviors reduce adaptation costs through knowledge transfer, experience sharing, and labor mutual assistance in interpersonal networks, and capital-biased adaptation behaviors reduce coordination costs, promote information sharing, and risk sharing through social capital’s trust mechanisms and network functions. However, labor-biased strategies may exacerbate labor burdens, particularly for women (Djoudi & Brockhaus [110]), requiring policy design to fully consider social equity and inclusive development.

We further clarify the heterogeneity of adaptation behavior adoption across different cropping systems. Food crop growers mainly utilize social networks (p < 0.01) to obtain information, exchange knowledge, and mobilize resources, promoting the adoption of labor and capital-biased measures; cash crop growers rely more on social trust mechanisms (p < 0.01), enhancing confidence in implementing labor-biased practices through stable cooperative relationships and reliable information channels. This difference reflects the impact of different crops’ climate sensitivity and market characteristics on adaptation strategy selection. Food crops mainly face yield stability challenges, requiring diverse technical and resource information through networks, while cash crops focus more on quality and market value, requiring stable cooperative relationships based on trust to ensure effective implementation of adaptation measures.

An important innovation of this study is the moderating effect analysis of digital literacy and agricultural extension services. Digital literacy shows a significant negative moderating relationship with social capital (p < 0.01), revealing diversified information acquisition trends in the digital age. Farmers with high digital literacy can independently access climate information through digital channels, reducing dependence on traditional social networks, while farmers with low digital literacy still rely on interpersonal relationships and experience sharing. This indicates that digital technology provides new information channels that can supplement or replace certain social capital functions, but may exacerbate digital divide issues. Agricultural extension services show clear scale differentiation: for large-scale farmers, extension services negatively moderate social capital (p < 0.05), supporting resource substitution where formal services effectively replace social capital’s information function; for small and medium-scale farmers, the moderating effect is insignificant, indicating extension services and social capital function independently or complementarily. This reveals that large-scale farmers have stronger resource integration capabilities to utilize formal extension channels, while smallholder farmers need coordinated development of extension services and social networks, as simply increasing extension frequency has limited effects.

Based on these findings, this study provides important policy insights for climate adaptation. A differentiated support framework is needed: technology-biased adaptation requires strengthened institutional support through fiscal subsidies and technical training to lower adoption thresholds, while labor and capital-biased strategies should leverage social capital through cooperative and community development. Policy design must consider scale differences—large-scale farmers benefit from formal extension services, while smallholders need coordinated extension-social network development, emphasizing community capacity building. Digital inclusive development should balance improving farmers’ digital literacy with maintaining traditional social networks to prevent digital divides from exacerbating inequality. Given that smallholder farming dominates the Yellow River Basin with extremely low technical adaptation rates (13%), policymakers should strengthen institutional support through fiscal incentives and capacity-building initiatives, while promoting rural social network construction to facilitate labor mobility, capital allocation, and diversified adaptation strategies that balance advantages and risks across different approaches.

This study has several limitations requiring future improvement. First, the older age distribution of respondents may limit external validity, necessitating broader age representation in future samples. Second, the general social norm measures used may introduce bias, requiring more precise and culturally sensitive measurement tools. Third, the sample’s geographic limitation to the middle and upper Yellow River Basin, while regionally representative, requires expansion to verify result robustness and generalizability. Future research should explore interactive effects among adaptation strategies, institutional moderating mechanisms on social capital, digital transformation of social capital functions, and long-term sustainability assessments. This study contributes valuable empirical evidence from China to global climate adaptation research, providing important references for understanding smallholder adaptation behaviors in developing countries.

6. Conclusions

In the context of accelerating global climate change, this study analyzed 1772 agricultural households across five provinces in the middle and upper reaches of the Yellow River Basin using factor analysis, Ordered Probit modeling, and moderation effect frameworks to examine social capital’s multidimensional influence on farmers’ climate adaptation behaviors.

Our findings reveal five key conclusions. First, 60% of households have implemented climate adaptation measures, with only 13% adopting technology-biased strategies, indicating significant deficiencies in adaptive capacity. Second, composite social capital significantly enhances adaptation behavior adoption (p < 0.01), with social trust, networks, and participation contributing positively, while social norms show negligible effects. Third, pronounced heterogeneity exists across cropping systems. Grain farmers leverage social networks (p < 0.01), while cash crop producers mobilize social trust mechanisms (p < 0.01). Fourth, social capital differentially influences adaptation portfolios, more strongly incentivizing labor- and capital-biased adaptations among grain farmers versus labor-biased measures among cash crop farmers. Fifth, digital literacy negatively moderates social capital’s effects (p < 0.01), while agricultural extension services show scale-differentiated moderation, negative for large-scale farmers (p < 0.05) but insignificant for smallholders.

These findings suggest several policy implications. Climate adaptation policies should prioritize technology-biased strategy diffusion and incorporate social capital enhancement mechanisms within agricultural communities. Differentiated intervention strategies are needed, including trust-building mechanisms for cash crop producers and network-biased approaches for grain farmers. Resource allocation should recognize heterogeneous adaptation effectiveness across farming systems, with targeted financial support for grain farmers’ capital-biased interventions. Extension programs should incorporate social context navigation strategies while balancing digital literacy development with traditional social network maintenance to prevent digital divides from exacerbating inequality, particularly focusing on smallholder farmers who need coordinated extension/social network development.