Drawing on existing research, this study conceptualizes environmental literacy as a multidimensional capability that supports individuals in understanding environmental systems and making informed decisions in contexts involving ecological trade-offs. In pastoral settings, herders’ environmental literacy refers to their capacity to interpret ecological information, evaluate resource-use choices, and consider potential environmental consequences when managing grassland resources.

Literacy involves a systemic understanding of an issue, and scholars have proposed various frameworks for categorizing individual literacy dimensions. We adopt the classification system of Maurer and Bogner (2020), which conceptualizes environmental literacy as an integration of knowledge, values, and behavior. Environmental knowledge provides the cognitive basis for recognizing ecological processes and constraints, including concepts such as biodiversity, carrying capacity, and climate variability. Such knowledge can influence how individuals perceive environmental change and assess the implications of resource-use decisions. Environmental values refer to an individual’s moral, ethical, and emotional orientations toward environmental issues. These values influence how people prioritize environmental goals relative to other considerations such as income generation or cultural traditions. Environmental behaviors encompass actions that actively promote environmental protection that actively promote environmental protection, ranging from daily habits such as waste sorting and energy conservation to deliberate adaptive strategies such as livestock reduction during droughts and avoidance of overgrazing.

Importantly, these three dimensions of environmental literacy are interrelated and operate in a mutually reinforcing manner. Environmental knowledge serves as a foundation for the formation of pro-environmental values, which in turn drive the adoption of environmentally responsible behaviors. In pastoral regions such as Qinghai and Gansu, where ecological fragility intersects with strong livelihood dependence on natural resources, this perspective highlights environmental literacy as a potential micro-level factor in grazing governance.

Building on the conceptual foundation of environmental literacy established in Section 2.1, this section develops a formal theoretical model to explicate its influence on herders’ grazing decisions. The model posits that herders, as rational economic agents, make grazing choices aimed at maximizing their utility, which comprises both short-term economic returns and long-term ecological benefits. Drawing on the work of Le Coent et al. (2021), we construct a utility maximization model that explicitly incorporates grazing intensity (G) and environmental literacy (EL). The model assumes that herders derive utility from two main sources: First, economic returns from livestock production, which are positively related to grazing intensity due to increased output, but subject to diminishing marginal returns. Second, ecological utility, which reflects the long-term benefits of maintaining grassland health. This component declines as grazing intensity rises, because overgrazing degrades the ecosystem, but its weight in the herder’s utility function depends positively on environmental literacy. The utility of herder reducing grazing intensity is denoted as μ, and the utility maximization function can be expressed as:

where a ′ ( G ) > 0 , a ″ ( G ) < 0 ) capture the concave economic return function; h ′ ( G ) < 0 , h ″ ( G ) ≤ 0 ) describe the ecological health function, and b ′ ( EL ) > 0 indicates that higher environmental literacy increases the relative importance herders assign to ecological well-being in their overall utility.

The herder chooses G to maximize μ ( EL , G ) . The first-order condition is:

Given h ′ ( G ) < 0 , an increase in EL raises b ( EL ) , which magnifies the marginal disutility from additional grazing. To satisfy the first-order condition, a ′ ( G ) must increase, which requires a reduction in G .

Applying comparative statics yields:

Since b ′ ( EL ) > 0 , h ′ ( G ) < 0 , and a ″ ( G ) < 0 ) . This unambiguously implies that an increase in environmental literacy leads to a lower optimal grazing intensity, in other word, EL is negatively associated with G .

Substituting the optimal grazing intensity into the utility function reveals two reinforcing effects of environmental literacy:

The total derivative with respect to EL is:

At the optimum, the FOC ensures μ G = 0 , so the expression simplifies to:

Where μ E > 0 captures the direct effect of environmental literacy on utility. This result demonstrates that environmental literacy increases the weight placed on ecological well-being within the utility function, thereby enhancing overall utility even in the absence of changes in production output. Moreover, through its influence on optimal grazing intensity, higher environmental literacy reduces ecosystem degradation, sustaining elevated ecological utility over the long term.

Building on the utility maximization framework, we incorporate the Theory of Planned Behavior (TPB) to capture the psychological and social drivers that may not be fully reflected in the economic model. According to TPB, behavior is guided by three key components: attitudes, perceived social norms, and perceived behavioral control (Ajzen, 2011). Variations in herders’ environmental literacy levels lead to significant differences in their attitudes toward sustainable versus unsustainable grazing behaviors. This integrated approach enables us to investigate both the direct and indirect pathways through which environmental literacy influences grazing behavior (Figure 1), providing a behavioral science lens for interpreting the statistical relationships revealed in the empirical analysis.

From an indirect perspective, environmental literacy is expected to operate primarily through two interrelated dimensions: ecological cognition and information processing capacity. First, environmental literacy may enhance herders’ ecological cognition. This involves not only the accumulation of ecological knowledge and practical experience but also the gradual formation and internalization of pro-environmental values (Yu et al., 2022). With stronger ecological cognition, herders may place greater importance on the long-term productivity and sustainability of local rangelands, which in turn can influence how they evaluate grazing intensity, seasonal timing, and resource use strategies. In this sense, environmental literacy contributes to reshaping attitudes toward the relationship between livelihoods and ecological conditions. Second, environmental literacy may improve herders’ ability to acquire, interpret, and utilize environmental and policy-related information. Herders with higher literacy are more likely to understand forage growth cycles, climate variability, and early signals of pasture degradation, and to connect these factors with production outcomes. They may also be more capable of interpreting policy guidelines, technical recommendations, and institutional incentives related to grassland management (Yang et al., 2021). Through this enhanced information-processing capacity, environmental literacy can influence how herders assess risks, anticipate environmental change, and adjust grazing arrangements. These cognitive and informational processes are mutually reinforcing: stronger ecological awareness can motivate further information acquisition, while accumulated information may deepen ecological understanding.

From a direct perspective, environmental literacy may strengthen herders’ decision-making capacity and facilitate the integration of ecological considerations into everyday production practices (Li et al., 2022). More environmentally literate herders may be better positioned to evaluate trade-offs between short-term economic returns and long-term ecological sustainability, and to explore adaptive management strategies such as adjusting stocking rates, adopting improved forage varieties, or modifying feeding structures during ecologically sensitive periods. Environmental literacy may also influence how herders interpret and respond to policy instruments. A clearer understanding of ecological principles and regulatory objectives can enhance policy comprehension and perceived legitimacy, thereby affecting the likelihood of voluntary compliance and participation in grassland conservation initiatives.

These mechanisms suggest that environmental literacy may influence grazing governance along multiple dimensions: by shaping attitudes toward ecological protection, strengthening the capacity to process environmental and policy information, and enhancing decision-making under ecological constraints. This framework highlights a set of theoretically grounded pathways through which environmental literacy can be translated into behavioral adjustments in grazing practices. The subsequent empirical analysis is designed to examine whether and to what extent these pathways are supported by the data. Building upon the conceptual framework and theoretical discussion above, this study proposes the following two testable hypotheses:

Gansu and Qinghai provinces are situated in the arid and semi-arid ecological zone of northwest China, where grassland covers approximately 13.5% of the country’s usable grassland area and supports 12% of the national livestock population, highlighting their critical ecological role. In recent years, under the combined pressures of climate change and intensified human activities, over 45% of the grassland in these two provinces has experienced degradation, characterized by vegetation decline, the spread of toxic weeds, and soil desertification, posing a threat to ecological security at both regional and broader scales. Qinghai and Gansu were selected as study areas due to their typicality and representativeness: both are located in the transition zone between the Tibetan Plateau and the northern grasslands, encompassing major grassland types such as alpine meadow and temperate steppe. These regions exhibit pronounced ecological vulnerability, deep-rooted pastoral traditions, and a long history of implementing grassland ecological compensation policies, making them well-suited to reflect the common governance challenges faced in China’s pastoral regions. Through scientifically designed stratified sampling and in-depth interviews, the sample obtained in this study demonstrates strong representativeness in terms of both grassland types and socio-ecological processes, thereby providing a solid basis for the research conclusions.

A multi-stage sampling approach was employed to gather household-level data from these areas, in 2020. In the initial stage, 2–3 counties with extensive grassland coverage and 2–3 with limited coverage were randomly selected from each province. Subsequent stages included: (1) random town selection within counties, ensuring proportional grassland type representation; (2) random village selection within towns; and (3) random selection of 5–8 herder households per village. Structured interviews were conducted with household heads using detailed questionnaires, which gathered comprehensive data on household demographics, livestock production, grassland management practices, income sources, and other relevant socioeconomic and ecological variables. After excluding entries with missing values or serious logical inconsistencies, the final dataset comprises 273 herder households from 10 counties, 30 towns, and 59 villages (Figure 2).

To investigate the relationship between herders’ environmental literacy and their grazing practices, we implement the following regression model:

Where Y i denotes either grazing intensity (Grazing intensity) or overgrazing behavior (Overgrazing) of household i , Env _ literac y i represents herders’ environmental literacy. Control s i includes control variables at the individual and household level. ε i is an error term. α 0 , α 0 and α 2 are (vectors of) coefficients to be estimated.

The OLS estimation may cause biased results due to potential endogeneity between environmental literacy and grazing intensity, as unobserved confounding factors may simultaneously affect both through the error term. To address this issue and achieve unbiased identification of the causal effect, we employ an instrumental variable (IV) approach using two-stage regression methods. Specifically, we apply 2SLS and LIML estimators.

The OLS regression results reported in Table 3 explore how environmental literacy among herders influences their grazing practices, with particular attention to grazing intensity and overgrazing. Model 1–3 show that Env_literacy has a significantly negative effect on grazing intensity across all specifications. In the full model (Column 2), each one-unit rise in environmental literacy corresponds to a 0.123-unit decrease in Grazing intensity, statistically significant at the 5% level. When regional dummies are aggregated to the province level in Model 3, the estimated effect remains stable (−0.182, p < 0.01). Similarly, Model 4–6 demonstrate a significant negative relationship between Env_literacy and Overgrazing. Coefficient estimates range from −1.259 to −0.740 and are statistically significant at conventional levels, specifically the 5 and 1% thresholds. The analysis reveals that improved environmental literacy among herders is associated with reduced grazing intensity and a lower likelihood of exceeding the sustainable carrying capacity of grassland ecosystems. Among the control variables, evidence from Models 2 and 5 indicates that Education is significantly and positively correlated with Grazing intensity and Overgrazing, with coefficients of 0.036 (p < 0.10) and 0.229 (p < 0.10), respectively. This finding indicates that better-educated herders may intentionally intensify livestock production due to enhanced capacity to access commercial livestock markets, and adopt productivity-enhancing technologies. Grassland area, by contrast, shows a significant negative effect on both outcome variables (−0.008, p < 0.01; −0.123, p < 0.01). The potential explanation for this is that larger grazing areas provide greater spatial flexibility, which facilitates the implementation of rotational grazing systems and allows adequate vegetation recovery between grazing cycles.

The initial OLS findings suggest a favorable influence of environmental literacy; however, these estimates may suffer from endogeneity concerns, stemming from unobserved heterogeneity or potential reverse causation. To address this concern, Table 4 presents the results of IV regressions using both 2SLS and LIML estimators. In the first stage, both instruments—Env_literacy_village and Informal—are strong predictors of individual environmental literacy. The results show positive and statistically significant coefficients—at the 10 and 1% levels—which provide empirical support for the validity of the instruments employed. This result is further supported by the Kleibergen-Paap rk LM statistics (66.344 and 75.923, p < 0.01), which provide strong evidence against the null hypothesis that the instrumental variables are exogenous to the endogenous regressor. The second-stage estimates (columns 2–5) corroborate and strengthen the initial OLS findings. The Durbin and Wu–Hausman tests (p < 0.10) reject the null hypothesis, supporting the use of an IV estimation for consistency. Furthermore, The Cragg-Donald Wald F-statistic (50.274) exceeds conventional thresholds, alleviating concerns over weak instruments. And the Hansen J-statistics do not reject the overidentifying restrictions, supporting the validity of the instruments. After addressing endogeneity concerns, environmental literacy persists as a statistically significant negative determinant of both grazing intensity and overgrazing. After instrumenting for endogeneity, the magnitude of the estimated effects grows considerably. Specifically, a one-point increase in environmental literacy leads to a reduction of 0.415–0.429 units in grazing intensity and 2.654–2.725 units in overgrazing, reaching statistical significance at the 5 and 1% thresholds. The pronounced amplification of these coefficients suggests that the original OLS estimates may have been downwardly biased, potentially due to attenuation caused by measurement error or omitted variable bias. This discrepancy highlights the importance of addressing endogeneity to obtain more accurate and reliable estimates of the true effect of environmental literacy on grazing practices.

To conclude, evidence from both OLS and IV estimations consistently highlights the significant role of environmental literacy in facilitating the sustainable use of grassland resources. The IV estimates further highlight that failing to account for endogeneity would underestimate the true impact of environmental literacy on grazing practices.

To investigate the pathways through which environmental literacy influences herders’ grazing behavior, we examine two potential mediating factors: (1) herders’ recognition of environmental importance, and (2) their frequency of seeking and processing ecological information. As shown in Table 6 (Models 11–12), environmental literacy significantly enhances both mediators. Specifically, a one-unit rise in environmental literacy is linked to a 0.166-unit increase in environmental importance assessment (p < 0.01), and a 0.370-unit increase in ecological information attention frequency (p < 0.01). These findings indicate that environmentally literate herders demonstrate greater appreciation of grasslands’ ecological and productive value, while exhibiting more proactive information-seeking behaviors through channels like village announcements, technical workshops, and peer networks. This dual enhancement of environmental awareness and knowledge acquisition facilitates the adoption of sustainable grazing practices, including strategic supplemental feeding and rotational grazing systems. Our results therefore identify environmental cognition and information engagement as critical transmission channels through which literacy translates into improved grazing management.