Temporal Stability Analysis of the Factors Affecting Freeway Crash Injury Severity Based on a Generalized Ordered Logit Model

This study investigates the determinants of freeway crash injury severity and examines whether their effects remain stable over time. Using crash records collected on the Guangzhou-Shenzhen Coastal Freeway in China from 2014 to 2019, we estimate generalized ordered logit models that explicitly consider the ordinal structure of injury outcomes. Two groups of likelihood-ratio tests are then applied to evaluate temporal transferability among yearly models. The estimation and marginal-effect results show clear time-varying influences rather than stable relationships. Variables associated with vehicle category, crash pattern, season, time of day, horizontal alignment, grade, and bridge segments display noticeable changes in their effects across years. These findings suggest that temporal instability should be incorporated into freeway injury-severity analysis so that transportation agencies can avoid drawing biased conclusions from pooled data.