The food delivery platform labor market has been continuing to grow rapidly in Korea, which resulted in traffic accident increases of delivery riders. To prevent traffic accidents while delivery, this study conducted a survey for 462 delivery riders and analyzed the statistical relationships of delivery characteristics and risk perception with delivery accidents. The results of this study revealed that riders with young age (20s: 46.6%) and/or low delivery experience (less than 1 year: 50.6%) had significantly higher proportion of accident experience than other age groups (over 40s: 36.2%) and high delivery experience (more than 2 years: 36.4%). In addition, side job riders (61.5%) showed significantly higher proportion of accident experience than main job riders (39.1%). The riders with accident experience had more number of deliveries per hour (weekday: 3.56, weekend: 3.91) than the riders without accident experience (weekday: 3.29, weekend: 3.68). Lastly, the riders with accident experience rated significantly higher perceived level of risk on weather, violation of traffic laws, uninspected motorcycle, receiving calls while driving, missing safety training, missing personal protective equipment than the riders without accident experience. This study suggested four aspects based on the study results to prevent traffic accidents for delivery riders.

PURPOSES : This study drew factors affecting motorcycle accidents in Seoul by severity using an ordered probit model and aimed to analyze and verify the drawn influence factors. METHODS: As the severity of the accidents could be classified into three types (fatal injury, serious injury and minor injury), this study drew the factors affecting accidents by a comparative analysis employing an ordered probit model, removed the variables that would not secure significance sequentially to construct a model with high explanatory power regarding the factors affecting the severity of motorcycle accidents, and calculated the marginal effect of each factor to understand the degree of each factor’s impact on the severity. First, Model 1 put in all variables; Model 2 was constructed by removing the variables of the road surface conditions that could not meet the level of significance (p=0.608); Model 3 was constructed by removing gender variable (p=0.423); and Model 4 was constructed finally by removing age variable (p=0.320). RESULTS : As a result of an analysis, statistically significant variables were time of occurrence, type of accident, road alignment and motorcycle displacement, and it turned out that the impacts on the severity were in the following order: a road alignment of left downhill, the type of motorcycle-to-vehicle accidents and a road alignment of a flatland on the left. The significance of the models was tested using the likelihood ratio, the level of significance and suitability statistics about them, and as a result of the test, the significance level and suitability of the constructed models were all excellent. In addition, the model accuracy indicating the accuracy of a predicted value compared to that of the value actually observed was 70.3% for minor injury; 70.1% for serious injury; and 68.6% for fatal injury, and the overall accuracy was 70.2%, which was very high. CONCLUSIONS: As a result of an analysis of motorcycle accidents in Seoul through the ordered probit model and the marginal effect, it turned out that their severity increased in nighttime accidents as compared to daytime ones and gradually increased in the order of motorcycleto- vehicle accidents, motorcycle-to-person ones and the ones involving motorcycle only. As a result of an analysis, the severity of accidents in road alignments of left downhill, left flatland and straight downhill increased as compared to those in a road alignment of straight flatland and that the severity of accidents of motorcycles with a displacement larger than 50cc was higher than that of those with a displacement smaller than 50cc.