A study by IIT-Delhi has exposed a major blind spot in how Delhi measures bicycle safety: the overwhelming majority of crashes reported by cyclists do not appear in police records. The finding is more than a discrepancy between two datasets. It suggests that official crash statistics may be capturing the most visible and severe incidents while missing the everyday collisions that shape whether cycling is safe, practical and accessible in the city.
The study, titled “Characteristics and Incidence Rates of Self-Reported Bicycle Crashes in Delhi: A Cohort Study”, examined crash experiences reported directly by cyclists. Its findings indicate that only 2.8% of bicycle crashes in Delhi were reported to police. In other words, 97.2% of the crashes recorded through the study’s cyclist-based method did not enter the police-reported system.
That gap changes the way Delhi’s cycling problem must be understood. Police records remain important for documenting serious incidents, identifying locations and establishing official accountability. But if they represent only a small share of the crashes experienced by cyclists, they cannot by themselves describe the full safety conditions on the road. A cyclist who falls, collides with a motorcycle or sustains an injury without filing a police complaint may disappear from the city’s formal road-safety picture.
The study was conducted by Srishti Agarwal and Rahul Goel of IIT-Delhi’s Transportation Research and Injury Prevention Centre. Researchers used on-road intercept surveys involving 844 cyclists at 27 locations in Delhi last year. Of these, 742 cyclists provided crash and exposure information during at least one follow-up and were included in the prospective cohort. The participants were followed for an average of 4.8 months.
This method matters because it begins with the people using bicycles rather than with incidents that reach an institution. Police data is generated when a crash is reported, recorded and classified. A cohort study can capture less visible events, including collisions that result in injury but do not lead to a police complaint or formal case. The two systems therefore describe different parts of the same road environment.
The contrast is especially clear in the types of vehicles involved. Half of the self-reported crashes involved motorcycles as the striking vehicle. Cars accounted for 25%, while electric three-wheeled rickshaws were involved in 11% of cases. By contrast, most police-reported cases in the study involved trucks. Police records also contained a higher proportion of trucks and a lower proportion of motorcycles than the self-reported data.
The difference points to a visibility problem in road-safety data. A truck-related collision is more likely to be serious, noticeable and formally reported. A conflict between a bicycle and a motorcycle may be less likely to generate a police record, even when it causes injury or discourages cycling. The study does not suggest that truck crashes are unimportant. Rather, it shows that the official record may be weighted towards particular types of incidents and vehicles.
The researchers linked the prominence of motorcycles to their role in Indian urban transport and their use of kerbside road space. The study said motorcycles often occupy kerbside lanes where cyclists usually ride, particularly during congestion. This creates a direct conflict over one of the narrowest and most contested parts of the carriageway.
The kerbside is not simply leftover road space. It is where cyclists may seek a sense of protection from faster traffic, where motorcycles filter through congestion, where autorickshaws stop and where pedestrians may cross or walk. When these movements overlap without clearly separated space, the risk is produced by the design and operation of the street, not only by individual behaviour.
The injury data reinforces the seriousness of crashes that do not necessarily enter police records. Among self-reported crashes, 78% resulted in injury, and 67% of those injured sought medical treatment. Knee and lower-leg injuries accounted for 60.4% of injuries, followed by wrist, hand and finger injuries at 27.8%. Head injuries accounted for 14.8%.
These figures describe a pattern dominated by injuries to limbs, but they also show that an unreported crash is not necessarily a minor event. Medical treatment was sought in a substantial share of injury cases. The absence of a police report therefore cannot be treated as evidence that a crash had no meaningful consequence for the cyclist.
The study found that all police-reported non-fatal crashes resulted in injury and received medical treatment. That finding is consistent with the role of police records as a selective system: incidents that reach the police may be those that are more severe, more visible or more likely to involve formal intervention. Self-reported data captures a wider range of outcomes, including crashes that are still harmful but do not proceed through the criminal justice or administrative system.
Another result sets Delhi apart from the international comparisons cited in the study. Only 2% of the reported incidents were single-bicycle crashes. These included loss-of-control events, sudden braking or manoeuvring to avoid a collision, mechanical failures and impacts with objects or potholes. The researchers noted that this was substantially different from findings in Australia, Europe and the United States, where single-bicycle crashes account for 40% to 70% of bicycle crashes.
The study attributed part of the difference to lower cycling speeds in Delhi. But the finding also places attention on the surrounding traffic environment. In Delhi, the study’s data indicates that interactions with other road users, particularly motorcycles, dominate the reported crash experience. A cyclist’s risk is therefore not limited to the condition of the bicycle or the cyclist’s control over it. It is closely connected to how different modes are arranged and allowed to move through the same road space.
The reported crash rate provides another way to understand the scale of the problem. The study found a rate of 28.6 injury-producing crashes per million kilometres cycled. It said this was 60 to 200 times higher than in Denmark, Germany and the Netherlands, countries with extensive cycling infrastructure. The comparison, as presented in the study, supports the importance of infrastructure conditions in shaping injury risk.
The evidence does not establish that infrastructure alone explains every difference between Delhi and those countries. Cycling volumes, traffic composition, reporting practices, road design and travel behaviour may differ. But the comparison does establish that Delhi’s cyclists are operating in a safety environment that the study describes as substantially more hazardous for injury-producing crashes.
This is where the study moves beyond the question of under-reporting. If the city measures cycling safety only through police cases, it may underestimate the number of conflicts and injuries taking place on ordinary streets. If it treats motorcycles, cars and three-wheelers as a general category of traffic, it may also miss the particular danger created where motorised vehicles and bicycles compete for kerbside lanes.
The policy challenge is partly institutional. Police records, medical treatment records, transport surveys and cyclist reports are generated by different systems, each with its own purpose. The study’s findings indicate that these systems are not producing a complete common picture of bicycle safety. Without a way to connect them, officials may know where serious crashes occurred but not where repeated low-visibility conflicts are making cycling unsafe.
The research also shows why cycling policy cannot be based only on demand or infrastructure length. A city may count bicycle lanes or identify cycling as a sustainable mode, but those measures do not reveal whether cyclists can use the network without entering conflict with motorcycles and other vehicles. The study’s evidence places the interaction between road users, especially at the kerbside, at the centre of the safety question.
The central conclusion is not that police data should be discarded. It is that police data should be understood as one layer of evidence. Cyclist self-reports reveal crashes that official records may miss; police records reveal the incidents that reach formal reporting channels. Together, they offer a more complete account than either source alone.
What remains uncertain from the supplied findings is how the crashes were distributed across Delhi’s roads, whether particular junctions or corridors recorded higher risks, and how reporting varied by cyclist profile or location. Those questions would require access to the full study and its underlying data. The evidence already available, however, is sufficient to show that Delhi’s bicycle-crash burden is significantly larger than police records alone suggest, and that motorcycles occupy a central place in the city’s cycling-safety challenge.
