Bicycle helmets – A case of risk compensation?

Abstract

Several studies have shown that bicycle helmets have the potential of reducing injuries from accidents. Yet, no studies have found good evidence of an injury reducing effect in countries that have introduced bicycle helmet legislation. Two of the most promising explanations for why helmet laws do not work as intended are risk compensation and shifts in the cycle population as a response to the law.

The present article investigates whether the lack of effect of helmet wearing laws is due to risk compensation mechanisms or population shifts (i.e. discouraging cyclists with the lowest accident risk, and thereby increasing the overall average risk per cyclist). A random sample of 1504 bicycle owners in Norway responded to a questionnaire on among other things helmet use, bicycle equipment use, accident involvement, cycling behaviour and risk perception. Data were analysed by using structural equation model (SEM). The results show that the cyclist population in Norway can be divided into two sub-populations: one speed-happy group that cycle fast and have lots of cycle equipment including helmets, and one traditional kind of cyclist without much equipment, cycling slowly. With all the limitations that have to be placed on a cross sectional study such as this, the results indicate that at least part of the reason why helmet laws do not appear to be beneficial is that they disproportionately discourage the safest cyclists.

Introduction

Falling of a bicycle can be painful, but it can also be dangerous. In the EU (19 member states) road accidents kill approximately 2000 cyclists each year (ERSO, 2010). A device that has the potential of reducing these numbers drastically is the bicycle helmet. As yet, only a few countries have legislated mandatory bicycle helmet use as a measure, but several countries are considering introducing helmet laws. What would be the effect of such a law? Would fatality rates be reduced?

Most case-control studies show injury reducing effects of bicycle helmets, as has been summarised in several reviews (Attewell et al., 2001, Thompson et al., 2000). These results mainly come from cross sectional studies of helmet users vs. non-helmet users. However, the evidence from countries that have introduced helmet laws is mixed. Some studies report that head injuries among cyclists have been reduced following the helmet use law (Carr et al., 1997, Hendrie et al., 1999). Other population studies show that these reductions are not larger than for other road user groups (i.e., other accident reducing mechanisms than the helmet are at work) and that the reductions over time in other injuries are of similar magnitude to the reductions in head injuries (Rissel, 2012, Robinson, 2006). This has been interpreted as an indication that the main reason for the reductions is reduced cycling and not an effect of the helmet. Furthermore the case-control findings are often criticised for not having sufficient control for other factors, i.e., that there are many other factors that differ between cases and controls in these studies, and that the effects are related to these factors and not to helmet wearing (Elvik, 2011). Robinson (2007) shows that lack of effects from helmet laws seem to be the rule rather than the exception.

The explanations given in population studies for why helmet laws do not work as intended are most often risk compensation; i.e., that cyclist wearing a helmet encourages cyclists to ride faster and take more risks (Robinson, 2006). The reasoning is that people perceive risk to be lower when wearing a helmet (than not wearing a helmet), and compensate for this perceived decrease in risk and increase in safety by cycling faster and more aggressively. The issue of risk compensation connected with helmet use has been the focus of a quite heated debate within the research community (e.g. Adams and Hillman, 2001a, Curnow, 2005, Elvik, 2011, Robinson, 2006, Robinson, 2007, Thompson et al., 2000).

Another explanation why helmet laws do not seem to give the results hoped for could be that helmet laws reduce the number of cyclists and thus put the remaining cyclists more at risk. The more cyclists on the road, the more will car drivers be aware of them. Such a “safety in numbers” effect has been documented by Jacobsen, 2003, Turner et al., 2011 and others. Furthermore, and an issue we will address in particular in the present paper is that helmet wearing laws may generate shifts in the cyclist population as a response to the law. Helmet laws generally reduce the number of cyclists (Gidske et al., 2007, Robinson, 2006), and if the cyclists remaining after the introduction of a law are the ones behaving most risky it is not surprising that the law does not give the expected result. In particular one may expect a decrease in traditional cyclists, who do not have many accidents anyways. Indeed, it could also be that the helmet law is introduced as a response to increased cycle accidents – which again could be related to changes in the cycle population towards a more training oriented type of cyclists with fast cycles and special equipment, including helmets. If so, a helmet wearing law would only boost such a trend.

Perceived risk is normally not studied in relation to the risk compensation theory, even if it is quite clearly an integral part of the risk compensation mechanism. An underlying assumption of the risk compensation hypothesis is that potential changes in risk perception following the introduction of a safety device are more or less “cancelled out” due to behavioural changes. In other words, when someone starts to use a helmet, they perceive risk to be reduced, and thus allow themselves to cycle faster off-setting some of the safety effect. Given such risk compensation, one would expect the perceived risk to return the previous level after a while. According to Wilde’s risk homeostasis model (1994) it would return to the exact same level as before.

By contrast, the population shift hypothesis implies that after a mandatory helmet wearing law is introduced the most risky cyclists remain in the population whilst others are discouraged, so the population average risk-taking increases. The hypothesis also implicates that in a situation where a helmet wearing law is not yet introduced, helmet users perceive risk as being greater than non-users. This is mainly due to the fact that helmet wearing is part of an equipment “package” suitable for training and fast cycling. However, there might also be another subgroup of cyclists that voluntarily wear helmets because they are particularly safety oriented, and not because the helmet is part of a larger equipment package.

Fig. 1 is a simplified illustration of the potential implications of using vs. not using a helmet according to the risk compensation theory (left panel) and population shift theory (right panel). The figure is an attempt at operationalising the two theories to fit our cross-sectional research design. We also believe that such a concrete representation can provide a testable conceptual basis for future empirical studies and can contribute to further advancing the scientific debate concerning helmet use among bicyclists. As the figure illustrates, both models indicate that helmet users cycle faster than non-users. However the models differ in their prediction about risk perception. According to risk compensation theory helmet users do not perceive the risk of an accident as higher than non-users, and most likely they would perceive it as lower. According to a population shift explanation, helmet users perceive the risk as higher than non-users.

To our knowledge, no large scale population studies have investigated and tested the risk compensation hypothesis or the population shift hypothesis with respect to bicycle helmets. The risk compensation hypothesis have been mentioned by several researchers as a possible explanation for the lack of effect for helmet wearing laws (Adams and Hillman, 2001b, Robinson, 2007), but generally not confronted with empirical data. One exception is Walker (2007) who reported a tendency for car drivers to overtake cyclists with less safety margins when the cyclist used a helmet - thus implying risk compensation, not among cyclists but among fellow road users. Also, significant risk compensation was observed when children ran an obstacle course wearing a helmet and wrist guards; tripping, falling and bumping into things increased by 51% compared to running the course without protective equipment (Morrongiello, Walpole, & Lasenby, 2007). A recent field experiment (Phillips, Fyhri, & Sagberg, 2011) showed that when not wearing a helmet, routinized helmet users reported higher experienced risk (explicit measure) and cycled more slowly. No such differences was found when cyclists unaccustomed to helmets were asked to use them.

An important question in the discussion about risk compensation of safety devices is: “why are some devices or measures compensated and others not?” A traditional assumption is that the intervention in question has to be either intrusive or conspicuous in order to be compensated. However, some researchers also claim that there is a distinction between injury reducing and accident reducing interventions, and that normally only the latter are compensated (Bjørnskau, 1995, Graham, 1982, Lund and O’Neill, 1986, OECD, 1990, Sagberg et al., 1997). The bicycle helmet is not an accident reducing device and hence should not “fall victim” of risk compensation. However, it can be argued that the accident/injury distinction makes less sense for the bicycle helmet than it does for a typical safety device for cars, such as seat belts. As a cyclist the perceived difference between being in an accident and having an injury is rather small, whereas for a car driver an accident may not necessarily imply being injured due to the protection inherently offered by the mass of the car. It has been argued that if the ratio of personal injury to non-personal injury is large also injury reducing measures will be victim of risk compensation (Bjørnskau, 1995, Fridstrøm, 1999). Thus it might well be that the helmet is potentially the subject of a risk compensation mechanism.

To our knowledge there are no studies that have systematically investigated risk perception as such among different groups of cyclists, and especially not helmet users vs. others. The closest we get to this is a study by McGuire and Smith (2000) who looked at the correlation between helmet use and use of other safety equipment, and a study by Lajunen and Rasanen (2004) who studied correlations between helmet use and positive health behaviour. These studies suggest that helmet users are more safety conscious than other cyclists. However, the results cannot really differentiate between what one would expect according to the risk compensation hypothesis and the population shift hypothesis.

The purpose of the present article is to investigate whether the lack of effect of helmet wearing laws is due to risk compensation mechanisms or population shifts, by looking at risk perception, cycling behaviour, accident involvement, and use of various cycling equipment.

The population shift hypothesis that will be focused in the following is the shift that normally follows from the introduction of helmet wearing laws, namely that a large number of cyclists abandon cycling. This self selection changes the cyclist population and possibly in such a fashion that the remaining cyclist population on average is a more equipped and training oriented type of cyclist. An important premise for this hypothesis is that these high equipment users cycle faster and more aggressively and subsequently have more accidents than other cyclists.

The risk compensation hypothesis states that the individual cyclist changes his behaviour as a response to wearing a helmet whereas the population shift hypothesis states that it is the group characteristics of helmeted and non-helmeted cyclists – and the changes in the ratio between these groups – that is the central mechanism behind the lacking effects of helmet wearing laws.

In Norway helmet use is not mandatory. The use of bicycle helmets has been annually registered through behaviour observations on counting stations since 1999. In 2008 39% (35% females and 41% males) of passing bicyclists above 17 years of age in eastern Norway used a helmet (Muskaug, Nygaard, Rosland, Johansen, & Sjøvold, 2009).