Internet use and higher farmer participation in domestic waste sorting: micro-survey data from 2126 farming households in rural China

The paper addresses how the rapid growth of internet access in rural China affects farmers’ participation in domestic waste sorting, a pressing environmental challenge given rising rural waste volumes and limited treatment capacity. Despite national policies promoting rural waste management and classification, mixed disposal remains common and contributes to pollution and carbon emissions. Prior work has emphasized economic incentives, subsidies, and policy advocacy, and largely relied on urban or macro-level data, with limited micro-evidence on farmers’ decisions and limited attention to digital technology’s behavioral effects. Grounded in planned behavior theory, the authors argue that internet use can mitigate information asymmetry, reduce decision and transaction costs, reshape environmental values and risk perceptions, and promote rational long-term choices such as waste sorting. The study aims to (1) quantify the impact of internet use on waste sorting participation among rural households and (2) examine heterogeneity across gender, age, income, and education using micro-survey data from 2126 households across eastern, central, and western China.

The literature highlights the pervasive effects of ICT and internet diffusion on economic and social outcomes, but micro-level cognitive and behavioral impacts are less explored. Traditional determinants of waste sorting include economic incentives, policy guidance, and internal perceptions; however, cognition is multifaceted and hard to measure uniformly. Internet use can enhance information access, learning, and social networking (e.g., WeChat, QQ), reducing information asymmetry and transaction costs, and potentially strengthening pro-environmental norms and reputational incentives. Conversely, pluralistic and low-quality online information may foster confusion, path dependence, and rumor amplification, possibly deterring participation. Given these mixed channels, effects likely vary by group and context. Based on this framework, the authors state the research hypothesis: Internet use positively influences farmers’ participation in waste sorting behavior.

Design: The study uses an Ordered Probit Model (OPM) because the dependent variable—domestic waste sorting (DWS)—is an ordered categorical choice with four levels: (1) no sorting; (2) sorting valuable waste only; (3) sorting valuable and food waste; (4) sorting valuable, food, and hazardous waste. The latent utility formulation relates DWS to internet use (core explanatory variable) and a set of controls, with thresholds mapping the latent variable to observed categories. Parameters are estimated via maximum likelihood, and average marginal effects quantify impacts on each outcome category. Data: A structured questionnaire was developed (pilot in March 2020) and fielded April–July 2020. Trained enumerators (undergraduate/postgraduate students) collected data across 11 provinces representing eastern, central, and western China. Stratified sampling selected counties and villages. After cleaning, 2126 valid household observations remained (over 2200 collected). In the sample, 79.16% report internet use and 44.45% engage in some waste separation. Variables: - Dependent variable (DWS): 1=no sorting; 2=sort valuable only; 3=sort valuable and food; 4=sort valuable, food, and hazardous. - Core explanatory variable (IU): Internet adoption based on the question “Do you use devices such as mobile phones and computers to access information?” Yes=1; No=0. - Controls: Individual and household characteristics (gender, age, education, village cadre, party member, main income from farming, net income), village/external environment (security cameras, distance to nearest junior high school), perceptions (environmental quality satisfaction, environmental publicity satisfaction), and region dummies (East, Middle, West). Estimation strategy and robustness: Baseline OPM includes region dummies with/without controls; OLS included as comparison. Marginal effects are computed for interpretability. Robustness checks include alternative measures for internet conditions (presence of home internet cable; home network signal quality), alternative model (ordered logit), and binary recoding of DWS (“classified or not”) estimated via probit. Heterogeneity analyses split samples by gender, age (<40 vs ≥40), income (below vs above average), and education (below high school vs high school and above).

- Baseline effects: Internet use significantly and positively relates to higher waste sorting levels. OPM coefficient for IU is 0.556 (p<0.01) without controls and 0.427 (p<0.01) with controls; OLS coefficient is 0.253 (p<0.01), corroborating direction and significance. Model fit improves with controls (higher pseudo R-squared and LR chi-square). - Marginal effects (IU): Relative to non-users, internet users are 14.9 percentage points less likely to not sort (y=1), and more likely to sort into y=2 by 3.86 pp, y=3 by 5.99 pp, and y=4 by 5.04 pp (all p<0.01), indicating shifts toward more comprehensive sorting. - Controls: Education (DEG), party membership (PM), village security cameras (SC), distance to school (NJS, positive association with higher sorting categories), and satisfaction with environmental publicity (SEP) show significant associations; main income from farming (CFP) is negatively associated. Gender, age, net income, and environmental quality satisfaction are generally insignificant. - Robustness: Results remain significant using alternative internet measures (home internet cable; network signal quality), alternative estimator (ordered logit), and binary probit using a recoded DWS (classified vs not classified). - Heterogeneity: - Gender: Stronger positive effect for females than males (e.g., IU coefficient ~0.612 vs 0.529 without controls; 0.510 vs 0.359 with controls; all significant). - Age: Stronger effect for older generation (≥40) than new generation (<40); for younger group, effect weakens with controls (becomes insignificant), while older group remains strongly significant (e.g., 0.605 to 0.442 with controls). - Income: Stronger impact among low-income households than high-income (e.g., 0.665→0.551 vs 0.416→0.320 with controls; all significant). - Education: Stronger and more robust effects among low-education group; for high-education group, significance weakens after adding controls (0.409→0.217, latter not significant).

The findings support the hypothesis that internet use enhances farmers’ participation in domestic waste sorting. Mechanistically, internet access likely reduces information asymmetry and transaction costs, improves access to environmental knowledge and policy information, and elevates the salience of long-term benefits and reputational considerations via online social interactions. These changes promote more rational, pro-environmental decisions, shifting households from non-sorting to increasingly comprehensive sorting categories. The heterogeneity patterns indicate that internet diffusion particularly benefits groups traditionally facing information constraints—women, older individuals, low-income, and low-education households—by lowering barriers to information and learning, thereby narrowing participation gaps. These results suggest that digital inclusion policies can complement environmental governance by strengthening behavioral foundations for waste sorting in rural areas.

The study demonstrates, using micro-survey data from 2126 rural Chinese households and an ordered probit framework, that internet use significantly increases participation in domestic waste sorting, with robust effects across alternative specifications. Marginal effects show sizable shifts away from non-sorting toward more comprehensive sorting. Heterogeneity analyses reveal stronger impacts among women, older farmers, low-income, and low-education groups, underscoring the role of internet access in reducing information frictions for marginalized populations. Policy implications include investing in rural digital infrastructure and digital literacy, integrating online environmental education and publicity, and tailoring interventions to disadvantaged groups to amplify environmental and social benefits. The results contribute to understanding how digital technologies shape environmental behavior and offer guidance for leveraging internet adoption to advance rural environmental governance in developing regions.

The study is based on cross-sectional, broad survey data that may not capture fine-grained behavioral mechanisms or establish causality beyond the controls and robustness checks used. Measurement of internet use relies on adoption rather than detailed usage patterns or content exposure. Further research should unpack causal pathways, explore longitudinal or experimental designs, and examine external validity and policy transferability to other countries and contexts.