The population of elderly individuals (over 64 years) in Seoul, South Korea, grew from 1.4 million to 1.7 million between 2018 and 2023. During the same period, the number of elderly taxi drivers rose from 27,739 to 35,166. Additionally, the number of fatal and severe injury (FSI) crashes caused by at-fault elderly taxi drivers has steadily increased, surpassing those caused by non-elderly taxi drivers since the onset of the COVID-19 pandemic. This shift has raised safety concerns among transportation authorities and the public. Previous studies have explored the factors influencing taxi driver crash injury severity outcomes; however, there has been little focus on investigating the stability of these factors over time and across taxi driver age groups. This study examines the stability of factors influencing taxi driver at-fault crash injury severity outcomes and the differences between elderly and non-elderly taxi driver at-fault crash severities using data from Seoul, South Korea (2017-2023). Risk factor stability across taxi driver at-fault age groups and time periods was assessed using log-likelihood ratio tests, which revealed that these factors were not stable, highlighting the need for estimating separate models. Separate statistical models were developed using the random parameters binary logit framework to examine the associations between risk factors and FSI outcomes. This approach allowed us to account for potential heterogeneity in the means of the random parameters for both elderly and non-elderly taxi driver at-fault crashes across different periods: pre-, during, and post-COVID-19. Factors such as midnight to early morning hours, dry roads, signal violations, elderly not-at-fault parties, and posted speed limits of 80 km/h increased the likelihood of FSI outcomes in most models. The results showed that the indicator for elderly not-at-fault drivers increased the probability of FSI outcomes the most when involved in a crash with elderly at-fault taxi drivers. Additionally, the probability of FSI outcomes was highest for elderly at-fault taxi drivers who violated traffic signals. Heterogeneity analysis revealed that intersection-related taxi driver at-fault crashes were likely to be more FSI on weekdays. Out-of-sample simulations demonstrated a clear difference in injury severities between elderly and non-elderly taxi drivers, with non-elderly taxi drivers predicting fewer FSI outcomes in recent years. Key measures to improve taxi safety for drivers over 64 include introducing free and mandatory assessments to ensure that taxi drivers are fit for the profession. Additionally, taxi management companies could implement fatigue and distracted driving detection systems to monitor driving behavior, especially during midnight and early morning hours. Collected data could be used to incentivize elderly taxi drivers to maintain safe driving practices. Further, introducing more flexible or reduced hours, part-time shifts, and retirement incentives for unfit taxi drivers would further reduce risks. Attracting younger drivers through incentives could also lessen reliance on elderly drivers, lowering the risk of FSI crashes. Finally, championing enhanced safety training, improved lighting and signal visibility at intersections-especially at night-stricter enforcement on high speed roads, and lower speed limits in high-risk areas would further increase safety.
