Intentions and actions in household waste separation: A machine learning approach on the gap and determinants

Municipal solid waste management now relies on individuals to dedicate time to sorting and separating their garbage from recyclable waste. However, despite good intentions, not all households follow through. Therefore, this study measures individuals' intentions and actual behavior regarding waste separation using a donation task as a proxy for behavior. Random Forest (RF) models are employed to identify various factors motivating intentions and behaviors, using ordinary least squares regression and Tobit models for robustness checks. Among the participants (N = 985), 56.55% made smaller donations (actual behavior) than intended, with the intention–behavior gap ranging from 0 to 5; moreover, 55.93% behaved consistently with their intentions, including 31.56% environmentally-inclined individuals with strongly intentions for pro-environmental behavior and high actual contributions. Interestingly, feelings of guilt are positively correlated with both waste separation intentions and actual behavior, while hope increases intentions but lacks significant impact on behavior. Furthermore, behavioral traces, social interaction, policy acceptance, and government trust are significant drivers of both intentions and actual behavior. Notably, behavioral traces have the largest impact on behavior, while guilt is the strongest motivation for those intending to separate waste. Our research addresses overlooked aspects of the intention-behavior gap and the different impacts of emotions and behavioral traces on waste separation intentions and behaviors, offering practical insights for waste management policy formulation.