How to make more cycling good for road safety?

Accident Analysis and Prevention 44 (2012) 19–29

Contents lists available at ScienceDirect

Accident Analysis and Prevention journal homepage: www.elsevier.com/locate/aap

How to make more cycling good for road safety? Fred Wegman a,∗ , Fan Zhang a , Atze Dijkstra b a b

SWOV Institute for Road Safety Research, Delft University of Technology, The Netherlands SWOV Institute for Road Safety Research, The Netherlands

a r t i c l e

i n f o

Article history: Received 31 October 2010 Accepted 8 November 2010 Keywords: Cyclist safety Bicycle crashes Prevention State-of-the art review

a b s t r a c t This paper discusses the current level of the road safety problems of cycling and cyclists, why cyclists run relatively high risks, and why cyclists may be considered as ‘vulnerable road users’. This paper is based on peer-reviewed research which give some idea how to reduce the number of cyclist casualties. However, this research is rather limited and the results cannot (easily) be transferred from one setting or country to another: generalization of results should only be done with the utmost care, if it is to be done at all. Interventions to reduce cyclist casualties worldwide seem to be of an incidental nature; that is to say, they are implemented in a rather isolated way. In a Safe System approach, such as the Dutch Sustainable Safety vision, the inherent risks of traffic are dealt with in a systematic, proactive way. We illustrate how this approach is especially effective for vulnerable road users, such as cyclists. Finally, the paper addresses the question of whether it is possible to make more cycling good for road safety. We conclude that when the number of cyclists increases, the number of fatalities may increase, but will not necessarily do so, and the outcome is dependent on specific conditions. There is strong evidence that well-designed bicycle facilities—physically separated networks—reduce risks for cyclists, and therefore have an impact on the net safety result, for example if car-kilometres are substituted by bicycle kilometres. Policies to support cycling should incorporate these findings in order to make more cycling good for road safety. © 2010 Elsevier Ltd. All rights reserved.

1. Introduction If we compare bicycle use among different countries there are enormous differences: from near absence to widespread use in countries such as the Netherlands. The amount of cycling is often partly determined by a country’s geography (hills and mountains) and its climate (temperatures, snowfall). There are countries where cycling is practiced for recreation. And, finally, there are countries in which cycling is a substantial part of everyday life. Although cycling activities also take place in rural areas, the majority of the bicycle kilometres in such countries are travelled in towns and cities, and over relatively short distances. Differences in bicycle use can be observed in bicycle culture, purpose of bicycle use, the position of the cyclist in traffic, and the measures that have been taken to make cycling safer. Many different arguments can be used to promote cycling. An important distinction that must be made is whether cycling is recreational, or whether it is a means of transport to travel from A to B. Some of the arguments used are: cycling is healthy, cycling is good for the environment if it takes the place of motorized journeys, cycling makes a contribution to the prevention of congestion

∗ Corresponding author. E-mail address: [email protected] (F. Wegman). 0001-4575/$ – see front matter © 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.aap.2010.11.010

because cyclists take up less space than (parked) cars, and cycling is cheaper than travel by passenger car or public transport. Compared to walking, cycling increases the distances that can be covered and in developing countries it can contribute to the economic development and aid the fight against poverty. Nowadays, more and more governments, cities and villages, communities encourage their citizens to cycle. One important objection can be made against promoting cycling: it is rather dangerous. As a direct consequence of the laws of (bio)mechanics and the fragility of the human body cyclists are vulnerable in traffic. Cyclists fall easily and can sustain serious injury. In crashes, other than sometimes by a bicycle helmet, a cyclist is unprotected. Brain damage is a serious and frequent injury, often sustained by young people in particular. When a cyclist is injured in a crash with a motorized vehicle travelling at high speed, kinetic energy is processed. Furthermore, a cyclist can lose control of the bicycle, take a fall, and be injured, especially if a cyclist is inexperienced or when obstacles play a role. Often cyclists fail to obey the traffic rules and show unexpected behaviour in the eyes of other road users. The consequence is that cyclists have a relatively high crash rate compared to that of pedestrians and particularly that of drivers. Because cycling is relatively risky we have to ask ourselves whether or not it will increase injuries and fatalities if a government is successful and more people do indeed use a bicycle.

20

F. Wegman et al. / Accident Analysis and Prevention 44 (2012) 19–29

Kilometers of cycling

3 2,5

2,5 2

1,6

1,5 0,9

1 0,5

0,7 0,4 0,1

0,1

0,1

0,1

0,2

0,4

0,5

0,9

0,7

0,5

0,2

Ita ly I EU re l a -1 nd 5 ve ra g A Fi e nl an Sw d ed e Be n lg iu m G er m an D en y m N a et he rk rla nd s

U K Au st ria

Sp ai n G re ec Po e rt u ga l U SA Fr an ce

0

Fig. 1. Kilometres cycled per inhabitant per day in some European countries and in the USA (from Pucher and Buehler, 2008; 2000-data from the European Commission).

1.1. Cycling worldwide The use of a bicycle as a transport mode differs tremendously among countries: from practically non-existing up to almost 30% of all trips being made by bicycle in, for example, the Netherlands. The number of bicycle kilometres travelled per inhabitant varies as can be seen in Fig. 1. Bicycle use is increasing (slightly) in some European countries, more specifically in cities. Many countries experienced a steep decrease in bicycling after the Second World War in conjunction with an increase in motorization (FHWA, 2010). In other countries, such as China, the cycling tradition has been taken over by motorized vehicles. Bicycle use is also determined by the geography of a country and its climate. A bicycle can be used for recreation or be substantial part of the modal split. Sometimes cycling is restricted to certain age groups (young and old), to lower socio-economic classes, and in many countries worldwide cycling is a male activity. There is such a variety in bicycle traffic that to sketch a mean picture would obscure reality. A fact or finding from one country can only be transferred to other countries with great caution, if this should be done at all. The most common problem for cyclists worldwide is that our modern traffic system is designed largely from a car-user perspective, which results in a lack of coherent planning of route networks for cyclists (ETSC, 1999). Nor does the system take the main characteristics of cyclists into account: a cyclist is vulnerable (in a crash), flexible (in behaviour), instable (may fall off the bike), inconspicuous (difficult to see), has differing abilities (due to a wide range of the population), is conscious of effort (i.e., highly motivated to minimize energy expenditure), and sometimes seen as intruders in the traffic system, rather than as an integral part. These key problems also occur in combination.

1.2. Risk Different countries worldwide indicate that cyclists have a relatively high crash rate compared to car drivers and pedestrians. As will be discussed in Section 2.1, we know that bicycle crashes suffer from underreporting, especially when non-fatal injury is concerned. In general, we find higher crash rates for cyclists than for drivers. A difference of a factor of more than 10 was found in a relatively old study in Europe (PROMISING, 2001). More recent developments in the Netherlands indicate that in a 20 year period (1988–2009) the number of cyclist fatalities was halved and that of car drivers and passengers reduced by 55%. At the same time the number of

kilometres travelled by cyclists increased by about 30% and that of cars, an important crash opponent, increased by almost 80%. This means that risks of cyclists (per kilometres travelled) decreased less than risks of car occupants during this period. It is important to understand that when comparing transport modes risk factors obscure age effects. Because especially the young and the elderly have relatively high crash rates (see also Table 2), the mean crash rate for a country is strongly influenced by the share of these high risk groups in the total number of fatalities. 1.3. Combination of numbers and risks Fig. 2 shows the relation between fatality rates and bicycle usage for a number of European countries (as referred to by van Hout, 2007). The graph depicts the relationship between the distances travelled by bicycle (kilometres travelled per person per year) in several European countries with the fatality rate (number of fatalities per kilometre travelled) in those countries. A regression line can be drawn which suggests that countries with a lot of bicycle traffic have a relatively low fatality rate (e.g. the Netherlands and Denmark) while countries where inhabitants do not cycle much (a dozen of kilometres per inhabitant per year only) face relatively high fatality rates. However, if we compare Portugal, Spain, France and the UK, all counties with less than 100 km travelled per year, we observe a wide variance. We, therefore, apparently need more factors than just the number of kilometres travelled to explain differences in fatality rate. The graph in Fig. 2, however, is very important indeed in scientific literature (Jacobsen, 2003). This intriguing relationship will be discussed in Section 5. Three different groups of countries can be distinguished according to different levels of bicycle use and fatality rate: – Denmark and the Netherlands, in which the distance travelled by bicycle is relatively high and the fatality rate is relatively low. – Southern European countries like Portugal and Spain, in which the distance travelled by bicycle is relatively low, but the fatality rate is relatively high. – Countries (almost all other) in which the distance travelled by bicycle and number of trips are low, but the fatality rate for cyclists is also low. However, in the latter group, we can observe major differences; for example, between the UK and Sweden on the one hand, and Austria on the other hand. This illustrates that the distance travelled cannot explain all differences in fatality rates; an interesting observation for further research.

Fatalities per billion kms travelled (cyclists)


21

250,00

200,00 Portugal 150,00 Spain 100,00 Austria Finland

50,00

Italy

France UK

0,00 0

Belgium

Ireland 100

200

Netherlands Denmark

Germany Sweden 300

400

500

600

700

800

900

Distance travelled by bicycle per person per year (kms) Fig. 2. Relation between fatality rates and bicycle usage for European countries based on IRTAD-data (referred to by van Hout, 2007).

1.4. Cycling benefits: improved health, and reduction of air pollution, congestion, and noise Cycling advocates never tire of stating the great advantages of cycling, especially compared with travelling by car. For example, they refer to the health benefits if cycling results in incorporating exercise into someone’s life. Using active transportation modes results in lower obesity prevalence for populations and better cardiovascular health. In their review Pucher et al. (2010) suggest that ‘the combined evidence presented in these studies indicates that the health benefits of bicycling far exceed the health risks from traffic injuries, contradicting the widespread misperception that bicycling is a dangerous activity’. However, In his book ‘Pedaling revolution’ (2009) Mapes states that Pucher ‘.. makes no pretense of being the detached academic’. The sources supporting the claim that cycling is healthy need some scrutiny before it can be accepted. This can be illustrated with a frequently cited conclusion of Bassett et al. (2008) that there is an inverse association between active transportation (walking, biking and public transport/transit use) and obesity rates in the countries studied: the more walking, cycling and public transport use, the lower the obesity rates. However, these authors, and rightly so, do not claim to have established a causal relationship, because they were not able to control for other factors that could influence obesity rates such as other physical activity domains or international differences in ‘energy intake’. A Dutch study (de Hartog et al., 2010) published findings using results of international literature which were applied to the Netherlands. The study assumes that 500,000 people (ages 18–64) make a transition from driving to bicycling for short trips on a daily basis, and concludes that the beneficial effects of the increased physical activity amount to 3–14 life months gained, which is substantially higher than the potential mortality effect of the increase in inhaled polluted air (0.8–40 days lost) and the increase in traffic crashes (5–9 days lost). Therefore, it concludes that the health benefits of cycling outweigh the health risks. This study, however, excluded high risk groups (cyclists younger than 18 and older than 64) and including theses groups will definitely influence the end result (see also Section 5). The advantages of cycling are often cited in policy documents promoting cycling all over the world (Dekoster and Schollaert, 1999; CEMT, 2004; Racioppi et al., 2005; Austroads, 2005; USDOT, 2010; WHO, 1999). The supposed advantages are not only related to individual health improvements, but also to less air pollution, less congestion and less noise. Although, these claims are supported by logical assumptions (e.g. bicycle traffic will replace motor-

ized traffic), the reported results are dependent on the conditions under which the change takes place. This limits generalization, and therefore reported results are not universally valid. Designers of strategies to promote cycling should always keep that in mind. 2. Road safety problem of cycling The considerable differences among countries do not allow for a meaningful comparison of the well-known crash characteristics as reported in police or hospital reports. Three issues are problematic in drawing conclusions from police reports: underreporting, age distribution, and some fundamentals in the causes of crashes with cyclists in relation to the key problems of this transport mode. 2.1. Underreporting of crashes Reliable, accurate data can help to identify road safety problems, risk factors and priority areas. Good data are required to adequately document the magnitude and nature of the road safety problems. A survey by the World Health Organization (WHO, 2009) showed that approximately half of the 178 participating countries only use police records to collect data on fatalities, the other half use different sources, and only 2% of the countries have no data at all. International comparison of fatality data requires a standardized definition of a road traffic fatality. However, countries use a wide range of definitions. Data on severely injured lack even more in quality, reliability, and international agreement on definitions (Derriks and Mak, 2007). Underreporting must be considered a crucial problem in relation to road safety data, more specifically with crash and injury data. Underreporting not only complicates international comparisons, but it also biases problem analysis within countries. The WHO (2009) concludes that among the factors that can affect the quality of reported data are political influences, competing priorities, and availability of resources. Derriks and Mak (2007) are more specific and identify low perceived relevance of this task by the police and the weaknesses in the error-prone processes (organisational, administrative, ICT) in building and maintaining a good database of road crashes. A few countries have made initial attempts to link data sources, such as police and hospital data, in order to overcome at least part of this underreporting problem. Problems in data systems, especially underreporting, affect cyclists more than any other transport mode. In other words: underreporting is selective and biased. For most conflict types, the number of casualties registered by the police are reasonably accurate. Although underreporting is a

22


Table 1 Percentages of casualties reported in three different severity classes (fatality, severe injury and slight injury) in Great Britain and the Netherlands. Great Britain

Passenger car Motorcycle/scooter Bicycle Pedestrian Total

The Netherlands

Fatal (%)

Severe (%)

Slight (%)

Fatal (%)

Hospitalized (%)

Slight (%)

100 100 100 100 100

89 70 33 85 76

77 51 21 67 62

96 94 86 90 93

92 63 31 56 60

33 13 4 20 13

Source: ERSO, 2008a; SWOV-AVV.

well-known phenomenon, it has not been very well researched. We find incidental evidence that more severe crashes and crashes involving motorized vehicles are better reported than less severe crashes and crashes involving ‘vulnerable road users’. Crashes involving cyclists tend to have a relatively low reporting rate compared with other transport modes. These crashes, even those with serious injury, are often not reported to the police. Therefore, the real numbers of injured are grossly underestimated by the police. Comparing serious traffic injuries (inpatients with a maximum abbreviated injury score with a minimum of 2) in the hospital data base with police data in the Netherlands (Reurings and Bos, 2009) led to the conclusion that the police records only contained 59% of the seriously injured casualties in motor vehicle crashes and only about 4% of the casualties in which no motor vehicle had been involved (e.g. crashes between two bicycles). Data from Great Britain and the Netherlands clearly shows that the underreporting rate increases as the victims’ transport mode changes from passenger car to cyclist and as injury severity increases (Table 1). For all three severity classes, casualties among cyclists are reported far less frequently than casualties among other road users. Bicycle crashes in which no other vehicle was involved are heavily underreported. Examples of such crashes are accidents in which the cyclist fell, slipped, or collided with an obstacle. Underreporting is a serious problem and has not yet been thoroughly researched. The international safety data community (IRTAD, WHO, European Union) has recommended that this phenomenon be studied further and that proper approaches to reduce underreporting be developed. In the meantime, cyclists injury data should be used with extreme caution. 2.2. Age distribution of cyclist casualties According to ERSO (2008b) in 2005, 44% of all bicycle fatalities in Europe were cyclists older than 60 (see Fig. 3). In Finland and Sweden more than 60% of bicycle fatalities were older than 60.

Table 2 Casualties in the Netherlands per age group per billion kilometres travelled for bicycles and cars in 2008. Age category

Bicycle

Car

Ratio (B/C)

0–11 12–19 20–29 30–39 40–49 50–59 60–74 75+ All ages

3.6 6.6 4.5 3.1 5.9 8.3 19.4 102.4 10.6

0.5 4.9 5.5 2.1 0.9 1.3 1.9 12.3 2.3

7.9 1.4 0.8 1.5 6.4 6.6 10.4 8.4 4.7

Source: Ministry of Transport—Statistics Netherlands.

The highest proportion of cyclist fatalities among children is found in the ages between 6 and 14. About 14% of the fatalities in this age group were cyclist fatalities; twice the average percentage for all age groups combined. Children between 10 and 14 years old also have the highest proportion of cyclist casualties: 30% of the casualties in this age group were cyclist casualties. However, comparison of these figures with those from other parts of the world, might lead to an entirely different conclusion. Table 2 shows the estimated number of casualties in the Netherlands in 2008 per age category per billion kilometres travelled. As we can see from the table, for all ages there are about 4.7 times more casualties per kilometre travelled for bicycles than for cars. Except for young adults (age 20–29), cycling is riskier than driving a car for all age groups, with about 10 times more casualties among the 60–74 years old and 8 times more casualties among the 0–11 years old and among those older than 75. The differences in fatality rates between bicycle and car on short trips are probably overestimated, because long car trips on motorways are usually relatively safe and therefore they reduce the mean crash and fatality rates for cars. In 1999, a European Commission

50 45 40 35 30 25 20 15 10 5 0 0-14

15-24

25-39

40-59

Fig. 3. Bicycle fatalities for different age groups (total of all columns = 100%). Source: ERSO, 2008b.

60+


23

study compared the risk of a fatal crash for car drivers and cyclists in the Netherlands. In this report Dekoster and Schollaert (1999) used two correction factors, which meant that motorway journeys were excluded and the hazard which motorists represent for pedestrians and cyclists was included. Their calculation showed a similar risk of accident per million kilometres for cyclists (21.0), and for car drivers (20.8). 2.3. Causes of risks and injuries: vulnerability, incompatibility and behaviour A cyclist-only crash only provides a small amount of kinetic energy due to the relatively small masses and low speeds, whereas the kinetic energy of motor vehicles is much greater. The different amounts of kinetic energy when different types of road users use the same traffic area and collide, combined with the differences in protection, result in incompatibility. Elvik (2010) calculated the fatality rates that can be attributed to incompatibility between different types of vehicles and groups of road users in Norway. Elvik focused on the ratio of the number of injured road users using other modes of transport in relation to injured road users in their own mode of transport. This ratio reflects the risks for the other types of road users. For the period 1998–2005 that ratio was 0.05 for cyclists, which mean that for each cyclist who was injured in a crash, there were 0.05 people belonging to other groups of road users who were injured in that same crash. In comparison with the cyclists, the ratio for other vehicles is higher (varying from 0.27 for car occupants, to 1.45 for van and bus occupants, to 3.46 for truck occupants). In contrast, the ratio for pedestrians was much lower: 0.03. Wegman and Aarts (2006) made a similar calculation for fatalities and severely injured and found that the incompatibility factor (casualties in the weakest party divided by casualties in the strongest party) is much higher (bicycle–car ratio = 150:1). This means that for one car casualty there are 150 bicycle casualties in car-bicycle crashes. Therefore, vulnerable road users such as cyclists run an even higher risk of getting injured in a crash with a ‘disproportionally strong crash opponent’. A second major safety problem for cyclists seems to be the single-vehicle crash. It is a misunderstanding that the safety problem of cyclists is just a problem of crashes between motorized vehicles and bicycles. Falling off a bicycle, for whatever reason, hitting an obstacle or leaving the road can also result in (serious) injury. Dutch statistics illustrate this (Schepers, 2008). Very little research is available about the effects of cyclist behaviour or cyclist characteristics on crashes and the risk of cyclist crashes. We have, for example, strong evidence for car drivers on the relationship between age/experience and their fatality rate, we know quite a lot about the relationship between fatality rate and blood alcohol content (BAC), we have some crash prediction models for different road types and design characteristics, but for cyclists this type of information has not been very well researched, if it exists at all. We recommend collecting this information for consideration when policies are designed to promote cycling and improve cyclist safety. 3. Interventions to reduce risks and severity of bicycle crashes Good evidence that cycling is relatively hazardous and a growing interest in many parts of the world in promoting cycling raises the question of how to reduce the hazards of cycling and how to minimize the negative consequences of a crash. Our survey of peer-reviewed research indicated that a rather limited amount of research on cyclist safety is available. Before applying solutions, it is important, once again, to make clear that results cannot (easily)

Fig. 4. Different packages of measures for different densities of facilities. Source: Adonis (Road Directorate Denmark, 1998).

be transferred from one setting or country to another. Generalization of results should be done with utmost care, if it is to be done at all. This seems to be a major hurdle in making progress towards a better understanding of bicycle safety. Some documentation of interventions to improve safety is available for cyclists from North-Western Europe (e.g. Pucher and Buehler, 2008; Pucher et al., 2010; FHWA, 2010), mainly from countries such as the Netherlands, Denmark and Germany. Only a limited number of studies into the safety effects of interventions have been carried out and very few of them meet rigorous standards. Two subjects related to cyclist safety that have been discussed thoroughly in peer-reviewed literature are bicycle helmets and roundabouts. If countries were to be categorized into different groups according to their current bicycle usage and existing level of bicycle facilities, it has been hypothesized that these differences would result in different investment schemes and types (Road Directorate Denmark, 1998). If neither policy nor facilities do yet exist, then a start will be made by taking elementary simple isolated low-cost measures. The more cycling and the higher the density of facilities, the more advanced measures will be contemplated and implemented (Fig. 4). These more advanced packages of measures may be part of a systemic and network-wide approach. One can assume that the density of facilities is proportional to the share of cycling in the modal split. This assumption was tested in a European ‘of the art’ project on cyclists and pedestrians, called Adonis. Adonis not only included so-called best practices for cyclists, but also for pedestrians. In Fig. 4, the share in modal split is on the horizontal axis and the density of facilities on the vertical axis. Three areas have been distinguished in this figure: • area 1 with low density of facilities and a small share of cycling or walking; • area 2 with medium density of facilities and medium share of cycling or walking; • area 3 with high density of facilities and large share of cycling or walking. For each area Adonis selected a package of measures for cyclists and concluded that investments could be found at the diagonal of this matrix rather than elsewhere. This is an interesting hypothesis, which certainly needs further testing: certain intervention types seem to correlate with the development stage of cycling. If this hypothesis is confirmed, two interesting conclusions can be derived. First, if we use results from evaluation research in package 3, we cannot simply transfer

24


these results to settings characteristic of packages 1 and 2. Second, if investments in the countries which were included in this study—Belgium, Denmark, the Netherlands and Spain—could be placed in these three categories, could this be the result of a logical step-by-step approach which should be followed by other jurisdictions elsewhere? If the answer is positive, then we have a logical step-by-step approach that can generally be followed. 3.1. Interventions to reduce crash rate and injury rate for cyclists Traffic safety has three primary dimensions: exposure (to risk), risk (of having a crash given a certain amount of exposure), and consequences (e.g. injury given a crash). When it comes to cycling and cycling policy, many countries and cities seem to aim for an increase in the number of kilometres travelled by bicycle rather than a reduction. Increasing the bicycle kilometres travelled can only reduce exposure to risk if potential conflicts between cyclists and other road users are prevented. This can be done by separating different modes of transport in space or in time. If this cannot be done, motorized vehicles need to reduce their speed to a ‘impact speed’ when meeting with a cyclist. A safe impact speed is such a speed that the risk of serious injury is very limited. Although we have some indication about safe impact speeds for pedestrians (Tingvall and Haworth, 1999), we do not have specific information on impacts between cyclists and motorized vehicles. The different impact manoeuvres do not allow the simple assumption that the same impact speed is safe for both pedestrians and cyclists (Rodarius et al., 2008). Therefore, further research into safe impact speeds between cars and cyclists is recommended. Besides preventing crashes between cyclists and other road users, we must address the safety problem of single-vehicle crashes for cyclists. This crash type is characterized as follows: the more kilometres travelled, the higher the exposure to risk. Relatively little scientific literature exists about this crash type (Schepers, 2008). We can assume that the risk of such a crash increases with poor vehicle handling and hazardous conditions (bad road surface, obstacles, etc.). This crash type seems to be underestimated and better analysis of these crashes is advisable before coming up with solutions. 3.2. Reducing the injury rate: bicycle helmets and vehicle design of cars and trucks The only protective device available for cyclists is the bicycle helmet. It is widely accepted that a properly designed helmet provides very good protection for the most vulnerable part of the body, the head, against being severely injured in a crash (SWOV, 2009). SWOV has calculated the maximum effect of a bicycle helmet to be approximately a 45% reduction of the risk of head and brain injury when a good helmet is worn correctly. However, if one needs to use ‘controlled trials’ as the only way of finding proper scientific evidence, some doubts remain. Whereas the helmet generally is compulsory for participants in sporting events, wearing a helmet for bicycle touring or bicycle rides in general is still optional in most countries. This introduces another heavily debated issue: how conclusive are studies that evaluate the safety effects of introducing helmet legislation? Some cyclists are opposed to the helmet as it conflicts with their feeling of freedom that goes with riding a bicycle, or because it is unsightly, uncomfortable, or unnecessary over short distances. Others are strongly in favour of it as it provides good head protection. One of the often-heard arguments is the negative impact of compulsory helmet use on the use of bicycles, and this is supported by an unmistakable decline in the use of bicycles after the introduction of compulsory helmet use in provinces in Australia and Canada. Presently, the final conclusion must be that deciding on helmet leg-

islation is a political decision, given that such a decision will have positive and negative effects on cycling and cyclist safety. Injuries to cyclists (and pedestrians) can be reduced by better design of cars and heavy vehicles. Design measures include crash-friendly car fronts, and side-underrun protection on lorries (Wittink, 2001). The design mainly focuses on the points of the body where cyclists or pedestrians are hit by cars. Lorries can be made much safer for third parties by the application of adequate protection around the vehicle. Such protection prevents the dangerous underrun of, for instance, cyclists and riders of other two-wheeled vehicles. In some of the crashes between heavy goods vehicles and two-wheelers, injury severity can be reduced by side-underrun protection. The number of traffic fatalities in urban areas due to crashes of this type could be reduced. For cyclists, but also for moped riders and pedestrians, closed side-underrun protection (with full screens) on lorries is more effective than open protection (with two or three bars only). Both open and closed side-underrun protection appeared in the top 10 list of relevant and cost-effective measures in the Netherlands to reduce the number of casualties as a result of crashes involving lorries (van Kampen and Schoon, 1999). A study of collisions between cars and vulnerable road users such as pedestrians and cyclists (Rodarius et al., 2008) showed that not all safety measures for pedestrians are effective for cyclists. This can, for example, be observed in collisions of pedestrians and cyclists with passenger cars. The study indicated that while head injury is the most dominant (life-threatening) injury, the impact point of a cyclist’s head is higher than that of a pedestrian’s head. Equipping cars with exterior airbags should therefore be considered, and cost-benefit analyses (CBA) should be made to support such decisions. 3.3. Reducing the crash rate: infrastructure measures Crashes between motorized vehicles and cyclists are a traditional and well-known crash type. Two methods can be used to reduce the crash rate for cyclists: for road stretches methods can be developed to physically separate both modes of transport (bicycle tracks, etc.). At intersections, speed reduction of motorized vehicles must be aimed for. Research is available into two types of countermeasures: bicycle tracks and roundabouts. When separating cyclists form motorized traffic it is almost inevitable to use an area-wide approach and to relate the planning of infrastructure to the nature of bicycle trips. This introduces the level of planning and design of comprehensive packages of measure to improve cyclists’ safety. But, once more, very little research is available on this subject. Bicycle facilities (bicycle lanes and bicycle paths) on road segments of roads where cars can drive relatively fast (>50 km/h) reduce risks for cyclists (SWOV, 2008). Also cyclists benefits from traffic calming measures, reducing speeds of motorized traffic to less than 30 km/h. Converting three-leg or four-leg intersections into roundabouts is good for road safety. Although to a lesser extent than for motorized vehicles, this is also the case for cyclists. A roundabout reduces the number of potential conflict points and, if well-designed, it reduces the severity of a conflict because of reduced approach and travel speeds. Studies report substantial reductions in the number of crashes and risks as a result of the construction of roundabouts (Elvik et al., 2009). As is usually the case, reduction of more severe injuries is estimated to be higher than of less severe injuries: a two-thirds reduction of fatal crashes and a 46% reduction of injury crashes; the reductions in urban areas are lower (25%) than in rural areas (69%). There is some indication (from the Netherlands and Denmark) that these very positive results are not entirely true for


cyclists. A Belgian study even found that roundabouts increased bicycle injury crashes (Daniels et al., 2008). It is recommended to do further research into bicycle safety on the relationship between the design features of roundabouts and bicycle facilities and actual behaviour of motorists and cyclists. We may observe complicated interactions between cyclists and motorized traffic; more specifically related to how cyclists on a roundabout intersect with traffic entering or leaving a roundabout. The design of the roundabout and its bicycle facilities in relation with the priority regulations seem to be the relevant variables (Hels and Orozova-Bekkevold, 2007; Sakshaug et al., 2010). 3.4. Improving poor cyclist behaviour by education and enforcement In order to improve their understanding of the traffic rules and regulations and to improve their cycling skills, Dutch, Danish, English and German children receive some training in safe and effective cycling techniques as part of their regular school curriculum (FHWA, 2010). Because many children travel to school by bicycle, governments in these countries consider training in safe cycling essential to improving their safety. Car drivers and truck drivers sometimes also receive education and awareness programs in relation to cyclist safety. However, no effects of education on cyclist safety have been found in scientific research and it is recommended to carry out these studies using, for example, approaches proposed in the CAST-study (Delhomme et al., 2009). Another way to reduce risks for cyclists is enforcement. We found little evidence of specific police campaigns on the behaviour of cyclists. A second approach is police enforcement related to driver behaviour, for example by enforcement of speeding and drinking and driving laws. Better driver behaviour of motorists can be expected to also reduce risks for cyclists, but no evidence is available to support this. 4. Sustainable Safety: how to reduce cyclist injuries drastically? Because traditional policies were becoming less effective and less efficient in many highly motorized countries, and because the inherent risks of our road transport problems had not adequately been addressed, the Sustainable Safety vision was developed in the Netherlands (Koornstra et al., 1992; Wegman and Aarts, 2006; Wegman, 2010). The increasingly diffuse character of the road safety problems requires a different approach from that in the past. With the Sustainable Safety vision SWOV believes it has found a suitable answer. Sustainable Safety has been acknowledged as an answer suitable for many more, if not all countries worldwide, as expressed by the World Health Organization, the World Bank (Peden et al., 2004) and the OECD/ITF (OECD, 2008). The OECD uses Sustainable Safety and Vision Zero from Sweden as examples of a Safe System approach. Road traffic today is inherently dangerous. Other than for railroads and air traffic, the road traffic system is not being designed with safety as a starting point. The interventions for bicycle safety that were discussed in the previous section indicate how to avoid crashes by preventing errors and violations. The principles and planning methods behind these measures are important. On one hand, we can adjust the environment to the human measure in such a way that people commit fewer errors and, consequently, have a lower risk. On the other hand, it is necessary to deal effectively (and efficiently) with violations and unsafe human behaviour (excessive/novice behaviour). In preventing cyclist crashes we are almost fully dependent on the behaviour of the individual human being. However, human

25

Table 3 The five Sustainable Safety principles. Sustainable Safety principle

Description

Functionality of roads

Monofunctionality of roads as either through roads, distributor roads, or access roads, in a hierarchically structured road network Equality in speed, direction, and mass at medium and high speeds Road environment and road user behaviour that support road user expectations through consistency and continuity in road design Injury limitation through a forgiving road environment and anticipation of road user behaviour Ability to assess one’s task capability to handle the driving task

Homogeneity of mass and/or speed and direction Predictability of road course and road user behaviour by a recognizable road design Forgivingness of the environment and of road users State awareness by the road user Source: Wegman and Aarts (2006).

beings commit errors and violations. This is the reason why we need a paradigm shift to a sustainable safe traffic system: we do not want to hand over a road traffic system to our children in which approximately the same number of people will be injured or killed as is the case today. 4.1. Sustainable Safety The Sustainable Safety vision was launched in 1992 with the following ambition: “In a sustainably safe road traffic system, infrastructure design inherently and drastically reduces crash risk. Should a crash occur, the process that determines crash severity is conditioned in such a way that severe injury is almost excluded (Koornstra et al., 1992)”. Initially Sustainable Safety used three principles, functionality, homogeneity and predictability, but the vision was recently updated and refreshed with new scientific insights (Wegman and Aarts, 2006). The updated vision on sustainable safe road traffic (Wegman and Aarts, 2006) has five central principles: functionality, homogeneity, predictability, forgivingness and state awareness (Table 3). The key aspects of the Sustainable Safety approach that have been identified (Wegman, 2010) are: • Ethics ◦ We do not want to hand over a traffic system to the next generation with the current fatality and injury levels; these must be considerably lower. ◦ A proactive approach. • An integral approach ◦ Integrates man, vehicle and road into one safe traffic system. ◦ Covers the entire road network, all vehicles and all road users. ◦ Integrates road safety with other policy fields. • Man is the measure of all things ◦ Human capacities and limitations are the guiding factors. • Reduction of latent errors (system gaps) in the system ◦ In preventing a crash we will not fully be dependent on whether or not a road user makes a mistake or error. • Use the criterion of preventable injuries ◦ Which interventions are most effective and cost-effective? Sustainable Safety identifies three basic factors that play a role in danger, risk and harm: speed (in crashes), mass/protection (of/by vehicles) and physical vulnerability (of man). All these factors, and the five principles, are very relevant for preventing crashes and reducing injuries of cyclists. Based on this vision, many road safety measures have been implemented in the Netherlands and these measures

26


and investments have been evaluated (Weijermars and van Schagen, 2009). The conclusion is that a substantial number of traffic safety measures were implemented in the period 1998–2007, and these measures had a positive road safety effect (a 32% reduction in traffic fatalities) with a cost-benefit ratio of 1:3.6. 4.2. Sustainable Safety and cyclists As illustrated in Section 2 cyclists (and pedestrians) are unprotected in traffic and travel at low speeds and mass. This makes them vulnerable and they can suffer very severe consequences in crashes with other road users. Therefore, preventing crashes between fast and slow traffic is one of the most important requirements for sustainable safe road use by cyclists (and pedestrians). The measures that need to be sought must be aimed at the ‘disarmament’ of motorized traffic (physical separation and, if that is not possible, reducing impact speeds). Furthermore, we have to prevent cyclists from falling off their bikes, for whatever reason, because these falls can result in severe injuries. These injuries will not always be visible in the official statistics due to underreporting (see Section 2.1). One of the key aspects of Sustainable Safety is to cover ‘the entire network, all vehicles and all road users’. As was indicated earlier, ‘high risk groups’ among cyclists (the young and the old) can be identified, but it will hardly be possible to identify ‘high risk locations’ for cyclists. When they meet motorized traffic travelling at high speeds, crashes will have serious consequences. Therefore, we should eliminate this type of conflict everywhere. If that is not feasible, we can slow down motorized traffic to a safe speed. If a collision then occurs, it will not result in serious injury. Of course, this approach has far reaching consequences as it may require much physical space and large investments. But the main problem seems to be found in society: can a society accept investments in facilities for cyclists given the limited amount of physical space, given the limited resources, and given the fact that other problems (congestion relief, environmental problems) also require investments? When talking about costs and cost-benefits of investments for cyclists, it is important to understand that traditional CBA’s and impact assessments are rather motorized vehicle oriented and currently do not include all effects (Elvik, 2000). Elvik concluded that a number of factors that are relevant for cyclists had not been included (changes in amount of cycling, changes in travel time for cyclists, changes in feelings of safety and in road users’ health). Some hypothetical examples were discussed by Elvik who concluded that inclusion of relevant factors for cyclists (and pedestrians) can be decisive factor that would support these investments to encourage bicycle use and cyclist safety. Some measures that have been implemented to achieve sustainable safe road traffic and that have positive effects for vulnerable road users are: (1) separation of traffic flows that differ in speed, direction and mass on a systematic basis (separate bicycle tracks alongside roads); (2) ‘moped on the carriageway’ instead of mopeds using bicycle tracks; (3) construction of 30 and 60 km/h zones; (4) mandatory side-underrun protection on new heavy goods vehicles; (5) development of a pedestrian and cyclist-friendly car front. The first three measures are particularly aimed at preventing crashes, and the latter two measures aim to reduce the severity of crashes if they should occur.

5. More cycling and its road safety impact A growing interest in the promotion of cycling by making it more attractive can be observed in highly motorized countries. This is also the case in the plans of many European cities. Signals indicating a move in this direction are also coming from countries like the United States and Australia. Other than the odd exception like Bogota in Peru, such signals are not received from Asian, African and Latin American countries. In contrast, these countries rather have an increase in car use which results in fewer people using a bicycle. But if the numbers of bicycles and cyclists were to increase, relatively dangerous kilometres would be added to present road traffic as a kilometre travelled by bicycle is more hazardous than a kilometre travelled by car. If a car kilometre were to be replaced by a bicycle kilometre this could result in a higher casualty rate. Investigating the validity of this premise is interesting. Two approaches are being taken to do so. In the first place the ‘Safety in Numbers’ theory is being considered. In the second place, the results of scenario studies that have estimated the safety effects of replacing vehicle kilometres by bicycle kilometres are being evaluated. For road safety developments it is very interesting to learn the safety impacts of increased bicycle use. We can approach this question from different angles. The first angle is the phenomenon of non-linearity of risk: an increase of exposure (numbers, volumes, etc.) results in a less than proportional increase of the number of crashes (Eenink et al., 2007). This implies that if the number of vehicles increases, the crash rates will go down. There even is some evidence that if the number of vehicles grows any further, not only will the crash rates come down, but also the crash density (crashes per kilometre of road length) will decrease (Fig. 5). The risks of cyclists and pedestrians are also non-linear, that is to say an increase in numbers results in a non-proportional increase of crashes. Elvik summarized some of these studies (Elvik, 2009). Most of the studies used by Elvik developed so-called prediction models of the following form: NCYC = ␣ · QMV ␤2 · QCYC ␤2 in which NCYC is the number of bicycle crashes, ˛ is a scaling parameter, Q represents volumes of motor vehicles, and QCYC represents the number of cyclists. It is of interest to learn the values of the exponents. Elvik concluded that either coefficient takes a value for cyclists between 0.31 and 0.65. That means lower than 1. The same is true for an increase of the number of motor vehicles (0.46 < ˇ < 0.76). This means that the risk for each cyclist declines if QCYC increases; the risk for each driver of colliding with a cyclist declines if QMV increases; and the risk for each cyclist increases as the number of motor vehicles increases. This suggests that the total number of crashes could go down if a substantial share of the journeys by motorized transport is transferred to cycling. This non-linearity in risk entered the cycle safety literature in a study by Jacobsen in 2003, and many authors have referred to this study since. Jacobsen analyzed several different data sets from the United States (California) and European countries. He correlated the measure of injuries to cyclists with the amount of cycling. His main conclusion is that a driver is less likely to collide with a person cycling when there are more cyclists (approximately 0.4 power). Jacobsen also concludes that “Policies that increase the numbers of people bicycling appear to be an effective route to improving the safety of people bicycling’. It can be added here that Jacobsen reached the same conclusion for pedestrians. This result has become widely quoted in bicycle safety circles (Mapes, 2009). But before adopting this conclusion it is important to investigate the reasons for this non-linearity of risks.


27

5,0

Road crashes per kilometre

4,5 4,0 3,5 3,0 2,5 2,0 1,5 1,0 0,5 0,0 0

10000

20000

30000

40000

50000

AADT Fig. 5. Relationship between traffic volumes (AADT) on Dutch rural roads and accident density (accidents per kilometre road length). Source: Reurings and Janssen (2006).

From one perspective, there is the explanation of expectancy. That is to say: if a road user expects the presence of another road user, or can predict the behaviour of that other road user, one may expect lower risks (Houtenbos, 2008). Räsänen and Summala carried out an in-depth study of 188 bicycle–car accidents in Finland studying the actual movements before the crash. They found that only 11% of the drivers noticed the cyclist before the impact (and 68% of cyclists noticed the car). Under Finnish conditions with low volumes of cyclists motorists do not seem to expect cyclists. Another interesting finding is that 92% of the cyclists that had noticed the car expected right of way (as required by law) but did not get it. In the words of Hudson (1978) it is not just a matter of expectation, but also of respect: “the fact that cyclists’ rights are more respected in towns where cycling is prevalent suggests that an increase in the number of cyclists on all roads would condition car drivers to expect and allow for them”. The presence of large numbers of cyclists may also help underpin their legal use of roadways and intersection crossings and generate public and political support for more investment in bicycling infrastructure. In a country, such as the Netherlands, in which cyclists are a very common feature in everyday life, expectation also plays an important role. Although generally cyclists are expected, cyclists coming from an unexpected direction run a higher risk than cyclists from the expected direction (Schepers and Voorham, 2010) However, if numbers of cyclists are correlated with risks and these numbers are assumed to be the only explanation, we are in error. Large numbers of cyclists in countries such as the Netherlands, Denmark and Germany are associated with high densities of bicycle facilities. If not both numbers of cyclists and bicycle facilities are taken into account, the wrong conclusions may be arrived at. There is no solid evidence that the low fatality rates in Fig. 1 can only be explained by ‘numbers’. Therefore, Jacobsen’s conclusion may be wrong if we simply add numbers of cyclists to the system without adding safety quality, that is to say, risk reducing measures. From this perspective Wegman paraphrased the safety in numbers by the ‘awareness in numbers’ theory (in Mapes, 2009): expectancy/awareness is one important factor, but the other one is safe conditions for cyclists, and it is not evident yet which of these two—or perhaps another factor—has resulted and will result in lower risks. When investing in facilities to make cycling safer this is an important area for further research, in relation to the Adonis

study findings of different phases/stages of development of safety facilities for cyclists. From a different perspective researchers have tried to estimate the safety consequences of a modal shift; more specifically from a transfer of car kilometres to bicycle kilometres. Elvik (2009) tried to estimate the number of crashes by adding up the number of motor vehicle crashes, the number of single vehicle crashes and the number of bicycle–motor vehicle crashes (in his study he included pedestrians as well). Elvik proposed six different scenario’s for change (transferring trips from cars to pedestrians or cyclists, different parameters in the prediction models). Although injury rates for cyclists are considerably higher than for car drivers, if we accept the non-linearity of risks (more cyclists, lower risks), the results of Elvik show that the number of crashes could go down if a substantial share of trips by motorized transport were transferred to walking and cycling. How realistic such a shift could be, is debatable. Stipdonk and Reurings (2010) approached the same question in a different manner. In their study they substituted only a small fraction of the short car trips in the Netherlands by bicycle trips, a substitution of 10% of car trips shorter than 7.5 km, and then they estimated the number of casualties. It was a ceteris paribus analysis: all relevant parametres are assumed to remain equal. They distinguished between age groups and gender. Stipdonk and Reurings took into account not only the risks for the car driver or cyclist, but also the risks run by other road users. From their estimate it can be concluded that if the young drivers (