Who is at fault? A quasi-induced exposure analysis of rider age, licensing and powered two-wheeler crash involvement in Istanbul

OBJECTIVE: Powered two-wheelers carry a large share of urban travel in Istanbul, yet no travel survey or odometer data describe how much they are used, so crash counts alone cannot show whether their riders are recorded as at fault more often than other road users. This study estimates relative crash involvement for riders and other road-user groups using quasi-induced exposure, and identifies the rider characteristics that account for the result. METHODS: The analysis is based on police-reported injury crashes in Istanbul between 2023 and 2025 and is limited to two-vehicle crashes in which exactly one party was recorded as at fault. Crashes involving parked vehicles were excluded because a parked vehicle cannot be considered part of the traffic at the time of the crash, leaving 35,400 crashes and 70,800 drivers. Not-at-fault drivers were treated as a sample of the traffic present, so that the relative involvement ratio for each group is its share of at-fault drivers divided by its share of not-at-fault drivers. The ratio was also calculated separately based on rider age, sex, license status, registered use, year, and road environment, and adjusted odds ratios were obtained from a conditional logistic regression matched on crash. RESULTS: Rider age and license status were the strongest predictors of recorded fault. Within the rider group, the ratio was 1.71 (95% CI 1.53-1.91) for riders aged 16 and 17, ranged from 0.80 to 0.85 for those aged 25 and older, and was 2.28 (2.08-2.49) for riders without a valid license, compared to 0.93 for licensed riders. After adjusting for age, sex, and license status within the same crash, the odds ratio for riders decreased from 1.12 to 0.91 (0.88-0.94), while the odds ratio for no valid license across the five motorized groups was 2.24 (2.04-2.45). At the vehicle-group level, riders were found to be modestly over-involved at 1.11 (1.09-1.13), while minibuses and buses (0.59) and heavy goods vehicles (0.63) fell below expectations, and these ratios represent crash-relevant estimates rather than traffic-stream shares, since the vehicle composition of not-at-fault drivers depends on which vehicle types they encounter. The driver composition of not-at-fault riders overlapped with that in multi-vehicle crashes within the limits these comparisons can detect. CONCLUSIONS: In Istanbul, riders' over-involvement in recorded faults is concentrated among young and unlicensed riders, and this excess disappears when these characteristics are held constant within the crash. This is an adjusted association and does not constitute evidence that riders are inherently safer. For young riders, graduated access, when evaluated in conjunction with licensing requirements, training, and enforcement, takes priority over measures that target riders as a single group, although no valid license combines multiple license statuses that the data cannot distinguish.