What began as a small survey in 2022 [in Czech – eds.] turned two years later into an analysis that fundamentally reshaped the picture of bicycle-related accidents—one that we had previously viewed almost exclusively through the lens of official statistics.
However, official statistics and the real-life experience of people on bikes often stand in stark contrast—or miss each other entirely. Many of us know, either firsthand or secondhand, situations where a cyclist falls, gets injured, receives treatment from emergency medical services, and is then fined by the police called to the scene for “causing a traffic accident.” But let’s take things in order.
The need to examine cycling accidents from the perspective of cyclists themselves arose in connection with how official statistics were being used in safety analyses. The initial motivation was simple: to find out how many crashes actually make it into the statistics. We all know that in many bicycle accidents the police are never called. These accidents are not investigated and therefore never recorded in official datasets. Understanding the ratio between investigated and uninvestigated accidents is a key piece of information. Without it, no one has any idea how many, and what kind of, cycling accidents actually happen in the Czech Republic. Evaluating safety becomes difficult if not impossible, because the only truly reliable indicator is the number of cyclists killed.
I myself have worked with accident statistics, using them in GIS analyses to calculate relative accident risk across Prague’s transport network, depending on the type of cycling infrastructure [read here in Czech – eds.]. Relative accident risk is the ratio between traffic performance and number of accidents—in simpler terms, how many kilometers you can expect to ride before a crash occurs. Based on official statistics, relative risk looks very favorable: between 100,000 and 1,000,000 kilometers per accident, depending on road type. Everyone knows reality is different—but how different? In a previous report, I speculated offhand what the results might look like if only one in five bicycle accidents were actually investigated.
Official accident data are also used by many colleagues working on cycling safety. Each spring, media outlets publish the previous year’s statistics, always pointing out that cyclists “caused” most accidents (without distinguishing between collisions and solo crashes), and how many of them were not wearing helmets—especially among the fatalities. Official data are essential to professional and public debates about cycling safety in the Czech Republic, and they will continue to be. But it is necessary to describe how they are biased, and what they can—and cannot—tell us.
For these reasons, in the second half of 2022 I initiated a survey that gave people the opportunity to report bicycle accidents that had happened to them in the past five years. The survey of self-reported accidents is a tool for obtaining information about bicycle crashes regardless of whether or not the police were called to investigate them. At the same time, this survey served as a pilot study to test the feasibility of this type of research. It is not a finished matter, but rather the first important step.
The data collection tool was designed to allow data gathering in a structure corresponding to the attributes of official statistics. This ensures comparability with the characteristics of officially recorded accidents. At the same time, however, the tool was set up to capture nuances beyond the resolution of official statistics. These include, for example, whether another road user contributed to the accident, whether road conditions played a role, or whether the crash occurred on a street with or without a bike lane. Respondents were also required to provide a written description of the accident event, which later proved to be an immensely valuable source of data during the analysis.
The pilot study has its limitations. The first is its scope, consisting of records of 313 accident events. Yes, that’s a small number—if we had three thousand, it would allow for more in-depth and reliable analyses. Yet, paradoxically, by the end of 2024 this represents the most extensive and comprehensive dataset available in the Czech Republic on the relationship between reported and hidden bicycle accidents. The findings of this pilot should therefore be taken as hypotheses for further verification.
The second limitation is self-selection of respondents, a red flag in every research methodology textbook. I will skip the lecture on representativeness, sampling error, population, and sampling methods. In short, if for example someone whilst being drunk and riding a bike, hit a child, they would not report it, and the dataset would be structurally missing these observations. Self-selection biases the data in ways unknown to us. Nevertheless, self-selection is the only method by which we can obtain data on hidden accidents. There are some ways to address this, such as comparing self-reported accidents by level of anonymity in data collection (face-to-face vs. online), or conducting qualitative interviews (individual or focus groups) to determine what types of events people would prefer not to disclose.
Given the limited size of the dataset, the results of the pilot study should be treated as hypotheses for further verification. In 11% of the more than three hundred recorded accidents, respondents reported that the police investigated the incident. Nearly nine out of ten of the recorded accidents in the pilot study thus represent hidden accidents. This rate of investigation is not uniform, but depends on the characteristics of the accident. Accidents involving serious injuries and collisions with cars were most frequently investigated (both around 20%). Accidents involving women were more often investigated (17%), and if the person was older, the rate was higher as well (24% for those over 50). In general, the likelihood of police involvement is related to the severity of the accident—the more serious the consequences for health and property, the more likely someone will call the police. This also explains the low rate of investigation (around 5%) for accidents on bike paths or solo crashes, which can likely be attributed to their lower severity.
What is surprising, however, is the differing rate of police investigation on main roads depending on the presence of a bike lane (no distinction was made between protected and dedicated lanes). Without a bike lane, the investigation rate for recorded accidents was 13%, but when a bike lane was present on a main road, the rate rose to 20%. This represents a relative increase of 54%. If the same difference were confirmed in a larger dataset, it would mean that bike lanes contribute to capturing roughly half more accidents in official statistics—accidents that would otherwise remain hidden on roads without bike lanes.
Each respondent was required to provide a written description of how the accident occurred. This created a unique database of more than three hundred accident narratives, offering insights into cycling safety that were previously unavailable. Official statistics lack such descriptions; while many attributes of each accident are recorded, allowing some inference about how it happened, there are frequent cases where only speculation is possible. The pilot study thus significantly fills this blind spot in cycling safety in the Czech Republic.
Accident descriptions thus portray crashes in a remarkable light. From the perspective of official statistics, it is difficult to fully understand accidents, which are typically recorded simply as cyclist-caused incidents without another party involved. However, when examining the descriptions, it becomes clear that in the vast majority of cases, accidents result from a combination of several adverse circumstances. For example, in Prague and Brno, crashes often occur due to the combination of cobblestones, tram tracks, and wet, slippery surfaces. Curbs and potholes also play a role. Some accidents occur in connection with motor vehicle traffic, when a cyclist is trying not to obstruct vehicles, or as a result of close passing or attempts to avoid collisions with cars.
Thanks to the accident narratives, it was possible to identify several events that occurred on roads in the context of vehicle traffic, but were not unintended incidents. Instead, they were the result of deliberate actions, where a driver used a car as a weapon against a cyclist. Such events are indistinguishable in official statistics; if investigated, they would likely be classified simply as a collision with a non-rail motor vehicle. In the pilot study, these accounted for one percent of incidents. If the same proportion holds in reality, there could be roughly four hundred such attacks on cyclists per year in the Czech Republic.
Overall, accident descriptions provide better understanding and empathy than dry statistics. Riding in mixed traffic with motor vehicles is inherently risky, with collisions occurring from all directions, at various speeds, and in diverse situations. Riding over uneven surfaces, potholes, obstacles, cobblestones, and tram tracks inevitably generates accidents. Some types of incidents, however, remain completely invisible in official statistics.
The presented study is a pilot project—meaning a test of the feasibility of a specific type of research, while also providing the first data illustrating the problem under examination. In other words, this feasibility check paves the way for carrying out a full-scale study with proper funding. Thanks to the pilot, we now have well-founded suspicion that official statistics on bicycle accidents are significantly biased—both numerically and structurally. Official statistics remain crucial and valuable, but it is now clear they must be approached differently than by merely listing them.
This new approach lies in calibrating existing statistics through a comprehensive survey of self-reported accidents. The pilot describes the theoretical assumptions for such calibration and also provides a concrete example of how it could be implemented. The goal is to carry out a large, nationwide survey of self-reported cycling accidents to obtain precise data on reporting rates depending on region and accident characteristics, and to describe a representative internal structure of this accident rate. With these data, it will then be possible to calibrate the official statistics each year, thereby gaining access to the true state of bicycle accident rates in the Czech Republic.
The pilot also tested the data collection tool, including its limitations. Differentiation by type of road and cycling infrastructure—particularly the presence of bike lanes—proved highly valuable. It also became clear that contraflow cycle lanes and different variants of one-way streets need to be distinguished. For each type of infrastructure, it is very helpful to provide a photographic example, since some respondents used the term “bike lane” generically for different measures (pictograms, contraflow lanes, cycle paths, bike lanes). Another distinction that deserves attention is the type of vehicle (bicycle, e-bike, scooter, e-scooter, shared bike or scooter), which was not included in this pilot.
I presented the results of this study at a number of meetings and conferences. Some of the reactions were particularly interesting. Some people immediately form an opinion, explaining the low rate of police investigation by saying that if a cyclist causes an accident, they wouldn’t call the police on themselves anyway, so it’s better not to dwell on it. This perspective is essentially based on resistance to data, because such trivialization is in direct contradiction with the accident descriptions available in Chapter Six of the research report. At the same time, this reaction points to a possible direction for further research: investigating what happens immediately after an accident and under what circumstances the police are called. Currently, not calling the police after an accident is the norm, and it is rather the one-tenth of cases where the police are called that warrant explanation.
The second type of reaction involves reflecting on what accident statistics actually mean. Until now, accident numbers have been presented as pure facts, as “hard” numbers. Hard numbers suggest something certain and solid that people can rely on. However, those familiar with how human knowledge is created and formalized become alert when “hard” numbers are presented, because it indicates that this hardness is primarily a rhetorical maneuver used to avoid discussing how the numbers were produced and what they actually mean. Accident statistics are in fact soft, and that’s not a bad thing. We just need to understand what that softness entails and how to work with it.
This is an adjusted machine translation using Automat’s CycleLingo Translator (ChatGPT) of this article: https://mestemnakole.cz/2024/12/jedna-nehoda-z-deseti-nehody-lidi-na-kolech-a-oficialni-statistiky/
Líbil se Vám článek? Chcete v Česku lepší podmínky pro cyklodopravu?
Podpořte nás, stačí i 100 korun. 💙🚲
Vyberte prosím částku, poté budete přesmerováni na darujme.cz
komentář