Discrepancies between intention, self-reported behavior, and actual behavior in e-bike helmet wearing: Evidence from child passengers with machine learning analyses

Helmets are a critical passive safety device that can dramatically reduce fatalities for e-bike riders and passengers. Most existing studies focused on improving intentions or self-reported behavior rather than actual behavior. The gap between intentions, self-reported behavior and actual behavior suggests that policies based on the former alone will have a limited impact on increasing helmet use. This study conducted 1035 questionnaires and follow-up field observations of parents putting helmets on their child passengers before riding e-bikes in Zhenjiang, China. This approach overcame the challenge of observing the intentions, self-reported behavior, and actual behavior of the same respondents. The results reveal substantial discrepancies: 84.5% of parents' intentions did not align with their actual behavior, and 72.6% of self-reports were inconsistent with observed actions. We used an XGBoost model interpreted via SHAP values to analyze these gaps. The model performed well in identifying gaps, achieving accuracy and precision rates above 80%, and recall and F1 scores above 90%. The number of helmets carried is a key factor contributing to the gap between intentions, self-reported behavior, and actual behavior. Ownership of dedicated child-helmets and parents’ helmet wearing behavior could also reduce the gaps. Psychological factors, such as subjective norm, perceived behavioral control, and attitude, also have significant impacts on the gaps. Practical recommendations include redesigning e-bikes to accommodate multiple helmets, deploying public helmet storage lockers at key destinations, and tailoring enforcement and education efforts to overcome specific behavioral obstacles.