The effect of age, gender and driver status on pedestrians’ intentions to cross the road in risky situations

Abstract

The theory of planned behaviour (TPB) has been used successfully in the past to account for pedestrians’ intentions to cross the road in risky situations. However, accident statistics show age and gender differences in the likelihood of adult pedestrian accidents. This study extends earlier work by examining the relative importance of the model components as predictors of intention to cross for four different adult age groups, men, women, drivers and nondrivers. The groups did not differ in the extent to which they differentiated between two situations of varying perceived risk. The model fit was good, but accounted for less of the variance in intention for the youngest group (17–24) than for other age groups. Differences between the age groups in intention to cross seemed to be due to differences in perceived value of crossing rather than differences in perceived risk. Women were less likely to intend to cross than men and perceived more risk, and there were important age, gender and driver status differences in the importance of the TPB variables as predictors of intention. A key implication of these findings is that road safety interventions need to be designed differently for different groups.

Introduction

Despite an excellent international record for road safety of drivers and passengers, the UK is ranked 8th in the pre-2004 EU for pedestrian safety (Department for Transport [DfT], 2004a), with pedestrians 19 times more at risk of being killed per distance travelled than drivers (Department of the Environment, Transport and the Regions [DETR], 1999). Within adults, the risk of being involved in a pedestrian accident varies with age and gender, with older adults at greatest risk of being seriously injured or killed per distance walked (DfT, 2004b) and men at all ages being at greater risk of serious injury than women (DfT, 2004a).

Giving the lowest risk group (those aged 25–59 for each gender group) a baseline index of 1, it can be illustrated that gender interacts with age, based on accident statistics (DfT, 2004a). For men, the 16–24 years group is at the most risk (risk index of 2.13), more than twice the risk of the 25–75 years group, after which the risk index rises again (to 1.73). For women, the over 75s are at the greatest risk (risk index of 3.23, with risk index for young women aged 17–25 being 2.35). The increase for women is greater and occurs earlier than that for men, with an increase of 21% between 25–59 years and 60–74 years (no corresponding increase for men).

The important question is what differences in pedestrian characteristics may be contributing to these differences in risk? The contributions of changes in cognitive and sensory skills with increasing age to pedestrian safety have been reviewed elsewhere (Dunbar et al., 2004), as have characteristics and attitudes of younger adult road users that may lead to them taking greater risks on the roads or failing to perceive hazards (e.g. Parker et al., 1992a). One key difference between the age groups and genders which may contribute to the risk differences is the extent to which people are drivers. More men than women drive at all ages (e.g. see DTLR, 2001a) and therefore may be expected to be pedestrians less and/or to have differing traffic perception skills. The most obvious explanation of increased accident risk for older women is that people who drive less, probably walk more, and are thus more exposed to risky situations as pedestrians. However, examination of figures given by the DTLR (2001a), in combination with accident data (DTLR, 2001b), suggests that although accident rate decreases with increase in percentage of licence holders for both genders, the relationship with distance walked is not straightforward. Women are less likely to have a driving licence at all ages, but walk more than men only during the middle adulthood years. As they get older and licences become less common, their level of walking reduces to below that of men of the same age. This suggests something other than the amount one walks that may make nondrivers at more risk as pedestrians than drivers. There is some evidence that skill differences between drivers and nondrivers may account for differences in accident rates (Carthy et al., 1995), though this study found no age or gender effects. However, the risk differences seen in adulthood may also be the result of differences in beliefs regarding crossing in less than ideal situations and behaviour in such situations. There is substantial evidence that differences in attitudes and beliefs towards risky driving behaviour (such as speeding) do indeed mirror age and gender differences in accident data (e.g., Conner et al., 2003, Parker et al., 1992b) and predict actual behaviour (e.g. Elliott et al., 2003a, Elliott et al., 2003b). This study therefore investigates pedestrian behaviour in order to examine differences in attitudes and beliefs predicting intention to cross in risky situations in relation to the demographic differences in accident data outlined. Further, differences in demographics, basic skill, cognitive or visual function in terms of relation with accident data, may not be as amenable to change or intervention as beliefs that have been shown to predict risky behaviour. Knowledge of any demographic differences in such beliefs is therefore vital for designing interventions.

One way of investigating such beliefs is through the use of a model such as Ajzen's (1991) theory of planned behaviour (TPB). In this model, intention is given a key role in the prediction of actual behaviour. Intention is presumed to reflect the motivation a person has to perform the behaviour, and their plans and expectations that they will perform it in the near future. Ajzen highlights the issue of control in the success of intention actually predicting behaviour in that people can only do what they intend if their behaviour is under their own volitional control (Ajzen, 1991). Thus accurate perception of control over performing a behaviour (perceived behavioural control, PBC) is also a key predictor of both behaviour and intention in the TPB model. This perception of the ease or difficulty of performing a behaviour may be expected to vary as a function of the situation as perceived by the person, e.g. in terms of difficulty or risk. PBC can account for situations where behaviour of choice may not be carried out due to lack of volitional control, an important factor in a study involving older people whose road crossing decisions may be affected by mobility constraints. The TPB also puts forward attitude towards the behaviour, and subjective norms, as predictors of intention. Attitude is concerned with a person's evaluation of the behaviour, e.g., would it produce a positive or negative outcome? Subjective norm is concerned with the perceived social pressure to perform the behaviour or not. The relative importance of these three direct predictors of intention is presumed to vary across situations and behaviours (Ajzen, 1991), but the general principles which guide the manner in which they vary have not been fully explored, particularly with reference to crossing at risky locations.

In the TPB, the direct measures of PBC, attitude and subjective norm are each underpinned by indirect measures. Underpinning PBC are beliefs about the extent to which a person believes that they are in control of their behaviour or risk in this particular situation, or the extent to which such behaviour is believed to affect the risk. More specific than general locus of control, PBC is the product of the control beliefs related to the behaviour, e.g., “factors that increase or reduce the perceived difficulty of performing the behaviour in question” (Ajzen, 1991, p. 196) and the perceived power of the control belief to actually control the outcome (perform the behaviour or not). Attitudes are presumed to be predicted by behavioural beliefs. Ajzen (1991) proposes that attitudes towards behaviours are formed by beliefs that people hold about them in terms of expectancy of the outcome of that behaviour, e.g., the cost of performing the behaviour. Thus people have positive or negative values of a behaviour which automatically form an attitude towards that behaviour. However, the subjective value of the perceived outcome also contributes to the attitude in direct proportion to the strength of the outcome belief. Thus, attitudes are predicted by the product of behavioural beliefs (belief about the outcome of the behaviour) and evaluation of that outcome, that is, the TPB is an “expectancy-value” model. Finally, subjective norms are presumed to be predicted by a person's normative beliefs about what significant other groups would expect them to do, combined multiplicatively with their motivation to comply with the opinions of others, i.e., the value component. Studies vary as to whether they adopt the direct measures to predict intention, or the indirect, combination measures, with the correlation between the two being variable and sometimes disappointing (Ajzen, 1991). However, in a road safety context, Elliott et al. (2005) showed that the combined measures predict the direct measures and intentions to a similar degree, and underline the added predictive power given by the multiplicative combination over and above the contribution made by each component separately. For this reason, and in line with previous studies of pedestrian behaviour (e.g. Evans and Norman, 1998, Evans and Norman, 2003), this study uses the indirect multiplicative measures of attitudes and subjective norm to predict intention.

A crucial issue for studies of safety related behaviour, where it is difficult to measure actual behaviour, is the relationship between intention, which we can measure, and actual behaviour, which we cannot safely measure. Meta-analyses have found the average amount of variance in actual behaviour predicted by intention in the TPB model to be 27% (Armitage and Conner, 2001), with an upper limit of 38% and a further 2–12% predicted when PBC is used as an additional direct predictor of behaviour, separate to its prediction of intention (Rivis and Sheeran, 2003). More specifically, studies of actual behaviour in traffic have found the TPB to be a useful model of prediction (e.g. Quine et al., 1998, Elliott et al., 2003a, Elliott et al., 2003b). For example, Elliott et al. (2003a) found that intention was a significant independent predictor of observed speeding behaviour in a validated driving simulator. Moreover, Holland and Conner (1996) found that changes in intention to speed taken from randomly selected drivers along a road that was the subject of a police anti-speeding campaign mirrored the actual changes in speeding measured along the road. Thus it is proposed that using such models and intention as at least a partial proxy for actual behaviour is scientifically appropriate and useful in situations where experimentally assessing actual behaviour may present risk to the participant and serious ethical concerns.

Two significant studies that have used TPB to analyse adults’ road crossing decisions are those by Evans and Norman (1998) and Diaz (2002), with further evidence from Evans and Norman (2003) on adolescents’ road crossing decisions. Evans and Norman (1998) found that subjective norm, PBC and attitudes accounted for 39–52% of the variance in intention to cross the roads in three risky situations, with PBC emerging as the most important predictor variable. Although Evans and Norman (1998) did not examine any of the above variables separately for different age groups, they did examine age as a predictor variable for intention to cross, and found that older people were generally less likely to say that they would cross the roads in the situations presented. No gender differences were found, although driving studies have generally found that men, especially young men, display more risk-taking behaviour and attitudes than women (e.g., Parker et al., 1995, Laapotti et al., 2001).

Diaz (2002) contrasted two adult age groups, under and over 26 years, using a questionnaire examining prediction of intention to perform illegal mid-block crossing. As with Evans and Norman (1998), no effects of gender were found, and younger people were significantly more likely to intend to cross in the situation given. Diaz (2002) also found that younger people had a more positive attitude towards crossing in such risky situations than the older group, perceived more acceptance from significant others, and had a lower PBC. However, the amount of variance in intention that the TPB variables account for together is not given in this study, and although the results report a significant relation between intention and actual behaviour, the measurement of actual behaviour is not reported.

The aim of this study is to extend the above work by examining effects of age, gender and driver status, as they relate to the components of the model, examining both differences in beliefs between demographic groups, and differences between groups in the contribution of each belief to the variance in intention to cross in less than ideal circumstances.

Previous studies addressing crossing intentions have not compared the TPB model fits for different age, gender and driver status groups. However, group differences have been found in related areas. For instance, Conner et al. (2003) found that normative beliefs were stronger predictors of intention to speed for men than for women. Moreover, Parker et al. (1992b) found differences between age groups in outcome evaluations with regard to driving violations, behavioural, normative and control beliefs in a study comparing attitudes of high and low risk young male drivers. They also found that subjective norm was a stronger predictor of intention than attitudes. Quine et al. (1998) found that subjective norm and PBC were particularly influential in younger adults; PBC also emerged as the most important predictor of road crossing intentions in the adolescents studied by Evans and Norman (2003), again underlining the central role of control. Conner et al. (2003) found that normative beliefs were particularly influential for men. While previous road safety research examining TPB variables specifically with older adults was not found, there is research in health education fields such as exercise (Conn et al., 2003) and breast cancer screening (Grunfeld et al., 2003) which found that beliefs about outcome were particularly influential for older women, again supporting the hypothesis that components of the TPB may be differentially important for different demographic groups.

These group differences are in line with Ajzen's (1991) suggestion that the relative importance of the TPB variables in the prediction of intention would be expected to vary across situations, behaviours (and populations). Demographic differences are investigated here by examining interactions between age, gender and driver status for the different TPB variables. Age and gender differences are well documented in various risk-taking areas (e.g. DeJoy, 1992) and in actual road crossing behaviour specifically (Hamed, 2001), with the latter study finding that women, older pedestrians and drivers took fewer risks.

The fact that the TPB does not fully predict intention implies that the TPB is open to the incorporation of further variables, particularly if, as Ajzen (1991) points out, it can be shown that they predict a significant proportion of the variance in intention or behaviour further to that predicted by the TPB variables (see also Conner and Armitage, 1998). With regard to road crossing decisions, moral norms and the anticipated emotion of taking the risk have been included, with the latter being shown to contribute a significant proportion of variance to the prediction of intention in a pedestrian study with adolescents, beyond the impact of the TPB components (Evans and Norman, 2003). This concept can be defined as the “enjoyment” factor of risk taking (or the positive implications of avoiding frustration that waiting longer or walking further may imply) and can be seen as an important determinant of intention in this situation. Anticipated affective reaction has been shown to be a significant predictor of intention (behavioural expectancy) after taking into account attitudes, subjective norms and PBC in a variety of risk related areas (Richard et al., 1996), and, specifically, in other road safety contexts, e.g., intentions to commit motorway violations (Parker et al., 1995). While it may be argued that emotion towards the behaviour is an attitude, these specific road safety studies support its inclusion as a separate predictor. Indeed, Ajzen (1991) advocates distinguishing between evaluative and affective outcomes and refers to two studies that provided strong evidence for the discriminant validity of separate evaluative and affective predictors (Ajzen and Driver, 1991, Breckler and Wiggins, 1989). Importantly, in an intervention aimed at changing intention to break speed limits, Parker et al. (1996) found that increasing the salience of anticipated affect was more effective than interventions directed at influencing the basic TPB components. The implication is that affect does need to be considered separately from evaluative attitudes, as a direct predictor of intention. However, one can distinguish between anticipated affect (factors such as regret after the behaviour) and affective attitudes (emotions towards the behaviour) and it is the latter which are examined here (see Parker et al., 1998). The examination here of interactions between demographic groups will help determine whether the role of affect is specific to particular age, gender or driver status groups.

A further variable of potential importance here is that of perceived risk. Although it is known that differences in perceived risk exist between younger and older male drivers (Deery, 1999), no studies to date have looked at such differences among pedestrians and the subsequent impact on intention to cross. Previous work with drivers in a related area (but including pedestrian situations) has shown that the amount of self-bias in perceived accident likelihood reduced with increasing age, i.e., older drivers were less likely to perceive themselves to be less at risk than their peers (generally, people perceive themselves to be less at risk than their peers), but self-bias increased with greater driving experience and with greater perceived control (Holland, 1993). In a general review of risk perception, Van der Plight (1996) concluded that although perceived risk is an important determinant of preventive behaviour, this relationship is dependent on the measure of perceived risk being conditional (upon performing the behaviour) rather than more general, and that in some circumstances its effect is mediated by subjective norms and attitudes. Given the evidence above, this study will incorporate measures of affective attitudes and risk perception in addition to the traditional TPB variables and also examine the differences between situations of varying perceived risk. In particular, interactions between age and gender are of interest to determine if younger men do indeed perceive less risk or intend to cross more than other groups, and possible sources of such differences will be examined in terms of the beliefs that predict intention. Interactions of the demographic variables with level of risk of situation are also crucial to determine whether the groups differ in their response to differing situations. Such interactions will also enable us to examine older at risk groups, for example, older women and nondrivers.

A key aim of the study is to examine whether the impact of the components of this extended TPB model varies between groups or situations per se, but also as a function of their contribution to the prediction of behavioural intention. The purpose is to inform both our efforts to understand demographic differences in accident risk, and also education and future interventions. For example, road safety campaigns rarely focus on adult pedestrians, but those few that are aimed at older pedestrians tend to focus on passive changes, such as wearing brighter colours, or having up-to-date spectacle prescriptions, as opposed to the decisions people make about where to cross (see Dunbar et al., 2004, for a review).

In summary, in an attempt to explain the demographic differences in pedestrian accident statistics and thereby inform road safety education, this study uses the TPB to identify group differences in intention to cross roads in two situations of varying perceived risk. Specifically, it aims to investigate whether:

• Older people, women and drivers are less likely to intend to cross than other groups, particularly in the more risky of the situations depicted.

• Different age, gender and driving status groups differ in their perception of risk in two road crossing situations and whether such differences will affect their intention to cross, in addition to the effect of TPB components.

• There are age, gender, driving status and risk level (situational) differences in the TPB predictor variables and affective attitudes that provide indications of the source of pedestrians’ intention formation process.

• There are group differences in the relative importance of TPB variables and additional variable of affective attitudes that predict intention to cross.