Beyond human error: a socio-technical analysis of escalator accidents using national census data in South Korea

This study employs a socio-technical system (STS) framework to analyze escalator accident mechanisms using a national census of 35,259 units and 605 validated accident records (1993-2024) in South Korea. This research adopts a step-by-step approach, eliminating the drawbacks of the pre-stage to achieve the highest performance. First, we conducted descriptive statistics to capture the general trends across variables. Second, we used non-parametric statistics to estimate trends in greater detail. Third, we constructed a general regression for predictive purposes. Fourth, we enhanced the predictive model by inserting 'causal factors' of the government agency. Fifth, we did an experimental trial to capture the endogenous variable by identifying the implicit causal variable. Sixth, we designed a structural equation model to address endogeneity and obtain more refined coefficients of the independent variables. Different types of accidents are driven by distinct causal mechanisms, where worker negligence has little effect on 'Falling on the steps' but is relevant for 'Entrapment,' and manufacturer negligence impacts 'Entrapment' more than 'Falling on the steps.' Additionally, transport buildings increase the risk of 'Falling on steps' but decrease the risk of 'Entrapment.' The findings support a bifurcated safety strategy: prioritizing density control and signage in transit sectors while implementing demographic-specific velocity calibration. Future research needs to consider per-unit passenger volume to produce analyses based on usage intensity and hazard signals per passenger risk.