Modeling Adoption Behavior of Organic Agriculture in Shabestar County: Comparing Ordered Logit and Random Forest Approaches

Background & Objectives: Organic farming, as a sustainable approach to mitigating environmental impacts and enhancing ecosystem health, plays a pivotal role in sustainable development. Given the predominance of dryland cultivation and traditional farming practices in Shabestar County, together with the strategic importance of wheat and barley for the region's food security, this study set out to analyze the factors influencing the adoption of varying levels of organic farming among wheat and barley growers in the county. In doing so, it seeks to compare two methodological approaches—ordered logit and random forest—so as to offer a more comprehensive account of both the causal relationships and the predictive capacity of the factors shaping adoption.
Materials & Methods: Data were gathered from 175 wheat and barley farmers in Shabestar County, selected through simple random sampling in 2023, by means of questionnaires and face-to-face interviews, and were subsequently analyzed using ordered logit and random forest models. The dependent variable was the level of adoption of organic practices, ranging from non-adoption to the adoption of more than four practices, while the independent variables comprised awareness, education, age, farming experience, participation in extension classes, income, and the number of dependents. Ordered logit was employed to identify causal relationships and marginal effects, whereas random forest served to provide accurate prediction and to rank the relative importance of the variables.
 
Results: The ordered logit model indicated that 51% of farmers had adopted at least one organic practice. Awareness, education, and participation in extension training programs exerted significant positive effects, while age and farming experience showed significant negative effects on adoption. The random forest model, achieving 78% accuracy, confirmed awareness and education as the most influential positive factors and performed better in predicting lower levels of adoption. Both models revealed that 69% of farmers would adopt organic practices were guaranteed purchase arrangements in place. Moreover, the highest proportion of adoption was concentrated at the initial levels.
 
Conclusion: Targeted education and economic incentives are essential to strengthening adoption; random forest proved more accurate in predicting lower levels of adoption, while ordered logit offered a more precise account of causal relationships. It is recommended that extension programs focus on younger and more educated farmers and that markets for organic products be developed through guaranteed purchase mechanisms.