Multidimensional Drivers of Green Production in Non-Timber Forest Products: A Cross-Validation of Econometrics and Machine Learning

Based on survey data from 579 farmer households in major non-timber forest product (NTFP) regions of Zhejiang Province, this study comprehensively employs binary logit/ordered probit and machine learning methods for cross-validation. It systematically examines the effects of multiple factors, including perceived property rights security, technical training, village rules and regulations, ecological awareness, and economic incentives, on forest farmers’ adoption of green production technologies. The cross-validation between econometric and machine learning approaches enhances the reliability of the findings. Results show that perceived property rights security is robustly and positively associated with green production behavior, while village rules and regulations and ecological awareness emerge as the two most critical driving factors. These associations exhibit significant NTFP-type heterogeneity: the impacts of technical training and forestry subsidies vary in direction depending on the crop cultivated, rendering traditional one-size-fits-all policies ineffective. This study highlights the crucial role of informal institutions and environmental awareness in the green transition, offering empirical evidence for designing differentiated training programs, optimizing penalty gradients, and implementing targeted subsidy policies.