Research report no.12 2005

In the driver's seat:
Understanding young adults' driving behaviour

A report on the ATP Young Drivers Study from the collaborative partnership between the Australian Institute of Family Studies, the Royal Automobile Club of Victoria and the Transport Accident Commission of Victoria.

by Diana Smart and Suzanne Vassallo
with Ann Sanson, Samantha Cockfield, Anne Harris, Warren Harrison and Allison McIntyre

4. Precursors and correlates of risky driving, crash involvement and speeding violations.

Having examined young adults' learner driver experiences and current driving behaviour, the precursors of three different types of problematic driving outcomes were next investigated - risky driving behaviour, crash involvement, and speeding violations. The results of these analyses are presented in this section, beginning with separate descriptions of the longitudinal precursors and concurrent correlates of risky driving behaviour, crash involvement, and speeding violations. This is followed by a detailed discussion of the similarities and differences in the profile of precursors and correlates identified for each of the different driving outcomes (Section 5). Finally, the degree of overlap in the membership of the risky driving, crash involvement and speed violation groups is examined (Section 6). The theoretical, practical and policy implications of these findings are discussed in detail later in the report (Section 7).

As the amount of driving undertaken by an individual is associated with his/her chance of becoming involved in a crash (Diamantopoulou et al. 1996; Harrison and Christie 2003) or being apprehended by police for a speeding violation (Diamantopoulou et al. 1997), the effects of driving exposure (the average number of hours young adults reported driving each week) were controlled20 when examining the correlates and precursors of crash involvement and speeding violations. However, as risky driving behaviour was measured over a predetermined period (participants' last ten trips), driving exposure was not controlled when examining this outcome.

Precursors and correlates of risky driving behaviour

To assist with the identification of precursors and correlates of risky driving, participants were grouped on the basis of their recent self-reported driving behaviour. Their level of risky driving was assessed by the frequency of engagement in the following risky driving behaviours during their previous ten trips: drove up to 10 km/h over the limit; drove between 10 and 25 km/h over the limit; drove more than 25km/h over the limit; did not wear a seat belt (or helmet) at all; forgot seatbelt (or helmet) for part of the trip; drove when very tired; drove when probably affected by alcohol; and drove when probably affected by an illegal drug.

Formation of groups

Cluster analysis was used to identify groups with differing levels of risky driving within the ATP sample. A three-cluster solution was seen as providing the best representation of the patterns of risky driving. 21 The three clusters identified were:

The profile of the three risky driving clusters on each of the eight risky driving measures is displayed in Figure 15, which shows that the high risky driving group reported the highest frequencies of all risky driving behaviours; the low risky driving group reported the lowest levels, and the moderate risky driving groups reported intermediate levels on most risky driving behaviours examined (the exception being driving without a seat-belt (or helmet) for the duration of the trip, on which the low and moderate groups were similar). 22 Differences between the groups were most noticeable on levels of speeding behaviour.

Frequency of risky driving

Group characteristics

Gender

The gender composition of the three groups (low, moderate and high risky driving groups) is shown in Table 8. From this Table it can be seen that that there were significant differences in the gender composition of the low and high risky driving groups. 23 While there was an equal proportion of young men and women in the moderate risky driving group, there were significantly more young men in the high risky driving group (77 per cent of this group were young men, 23 per cent were young women), and significantly fewer young men in the low risky driving group (39 per cent of this group were young men and 61 per cent were young women).

Gender of risky driving groups
Crash involvement and speeding violations

In order to assess whether the clusters accurately discriminated between types of risky drivers, the three groups were compared on two outcomes commonly associated with risky driving behaviour: crash involvement and speeding violations. The results of these comparisons are presented in Table 9.

As this Table shows, the groups differed significantly on these outcome measures in the manner expected, 24 with the high risky driving group reporting the highest crash involvement rate and highest number of speeding violations, and the low risky driving group reporting the lowest. Thus it seemed that the groups identified by cluster analysis were substantially different in terms of their driving behaviours. It should also be noted that there was considerable variability in the number of crashes and speeding violations reported by the young adults in each of the groups (as indicated by the large standard deviations). This within-group variability was particularly noticeable in the high risky driving group, which was much smaller than the other two cluster groups.

Crashes and speeding

Group differences by stage of development

Having examined young adults' learner driving experiences and current driving behaviour (Section 3), the factors (past and present) that were associated with a risky driving style were next investigated. Separate Multivariate Analyses of Variance (MANOVAs) were performed for each developmental stage 25 (infancy, toddlerhood, early childhood, mid childhood, late childhood, early adolescence, mid/late adolescence, and early adulthood) and source of report (parent, teacher, and self report). The strategy of examining each developmental stage separately enables the identification of the stage of development at which significant group differences emerge. Furthermore, it maximises the number of participants included in the analyses, which is of importance to maintain power.

Due to the large number of variables being examined during adolescence and early adulthood, a number of MANOVAs were performed for each source of report from early adolescence on. For example, during adolescence, two MANOVAs were performed at each timepoint for each source of report (one focusing on individual attributes, the other examining environmental factors), while in early adulthood, three MANOVAs were performed for each source of report (examining individual attributes, environmental factors, and current life circumstances).

As a large number of analyses were performed, the Bonferroni adjustment procedure was used to adjust the significance level to reduce the possibility of Type 1 error (the likelihood of showing group differences of statistical significance when they are not actually there). Results are reported for differences at the Bonferroni adjusted level of 0.0023 or lower. However, when results were significant at a less stringent level of p<.05, this is noted as a trend. Scheffe post-hoc contrasts were used to identify significant differences between particular groups.

Effect sizes26 were used to assess the strength of group differences across various domains of functioning. As a guide to interpreting these statistics, larger effect sizes indicate more powerful group differences, and hence, more powerful precursors.

The results of these analyses are summarised separately for each stage of development in Tables 10 through 13, as well as being described in the text below. When examining the Tables, it should be noted that variables displayed in CAPITALS showed significant group differences (p<.0023), while the other variables listed showed only trends (p<.05). The discussion that follows will generally focus on significant group differences.

Infancy, toddlerhood and early childhood (0-4 years)
No significant differences were found between the three risky driving groups (low, moderate and high) during the earliest years of life.

Mid childhood (5-8 years)
During mid childhood, significant differences emerged between the high risky driving group and the other two groups (moderate and low risky driving) in the areas of temperament style, behaviour problems and school adjustment and achievement (see Table 10). Teacher reports indicated that children in the high risky driving group were significantly more aggressive and hyperactive than other children, less task-oriented (less able to maintain their attention on tasks or activities), and had greater difficulty adjusting to the routines and demands of school life. These differences were all in the medium effect size range.

There was also a trend for multivariate group differences in parent reports. However, the strength of these differences did not reach the adjusted significance level (p<.0023).

Late childhood (9-12 years)
Significant differences were once again found between the high risky driving group and the other two groups during late childhood (see Table 10). Interestingly, as in mid childhood, they were only apparent from teacher reports. As before, teachers rated the high risky driving group as less task-oriented than the moderate and low risky driving groups. At this time, teachers also noted group differences in social skills, with the high risky driving group being rated as less cooperative than the other two groups. These differences were in the medium effect size range.

There was a trend for multivariate group differences in mid childhood by parent report. In late childhood, no significant multivariate group differences were apparent by parent or self report.

Teacher-reported differences

Early adolescence (12-14 years)
By early adolescence, the high risky driving group was displaying significantly more difficulties than both the moderate and low risky driving groups over a range of domains according to parent and adolescent self reports (see Table 11). As before, many of these group differences centred on personal attributes (temperament style, behaviour problems and social skills); however, group differences were also evident in the areas of school adjustment and achievement, peer relationships and the parent-adolescent relationship.

Parent-reported and self-reported differences in early adolescence

In terms of individual attributes, the high risky driving group was rated by parents as less persistent (less able to remain focused on tasks or activities), as engaging more frequently in antisocial acts by their own report (such as violence, property offences, or the use or sale of illegal drugs), and as less cooperative (by parent and self report), less responsible (by parent report) and less empathic (by self report) than the other two groups. In relation to environmental factors, parent and adolescent reports suggested that the high risky driving group was experiencing more school adjustment difficulties and poorer quality friendships (by self report) than the moderate and low risky driving groups, and associated more often with antisocial peers (by self-report) than those in the low risky driving group.

The first signs of differences among the moderate and low risky driving groups emerged at this age. Thus, while the majority of group differences at this time were between the high risky driving group and the other two groups, self reports suggested that both the high and moderate risky driving groups were more aggressive, less self-controlled and had a more difficult relationship with parents than the low risky driving group. Furthermore, while not as problematic as the high risky driving group, the moderate risky driving group was also significantly more antisocial and less cooperative than the low risky driving group. Almost all group differences were within the medium effect size range.

Mid/late adolescence (15-18 years)
There were fewer significant group differences at mid/late adolescence, with only adolescent-reported differences reaching significance (see Table 12). While there was an overall significant multivariate group difference on individual attributes by parent report, none of the variables contributing to this difference reached the adjusted significance level. On parent-reported environmental factors there was a trend for multivariate group differences, but these differences did not reach the adjusted significance level.

Parent-reported and self-reported differences in mid/late adolescence

Group differences were found on levels of behaviour problems, use of coping strategies and interpersonal relationships, and in contrast to previous developmental stages, many of these differences were between both the high and moderate risky driving groups and the low risky driving groups.For example, while the self reports of the high risky driving group revealed them to be more hyperactive, more antisocial and to experience less warmth in their relationships with their parents than other drivers, both the moderate and high risky driving groups were more aggressive, more reliant on acting out, less adaptive coping strategies (such as using drugs or reacting explosively), and reported socialising with antisocial peers more frequently than low risky drivers. Furthermore, the moderate risky driving group was more hyperactive and antisocial than the low risky driving group (although they were less problematic on these characteristics than the high risky driving group).

Most group differences were in the small to medium effect size range, with the exception of self-reported antisocial behaviour, which had a large effect size.

Early adulthood (19-20 years)
By early adulthood, group differences were evident on a wide range of domains (refer to Table 13). Both parents and young adults rated the high risky driving group as displaying more externalising behaviour problems (higher parent-reported aggression and self-reported antisocial behaviour), lower social skills (lower empathy and responsibility by parent and self report), more problematic peer relationships (greater association with antisocial peers by self report), and as having had more police contact for driving offences (by self report) than the other two groups. In addition, the high-risky driving group was less involved in community activities (lower civic engagement) than the low risky driving group, by their own report.

Parent-reported and self-reported differences in early adulthood

The moderate risky driving group continued to be increasingly differentiated from the other two groups. For example, there were a number of aspects on which the moderate risky driving group was more problematic than the low risky driving group, but less problematic than the high risky driving group, (self-reported antisocial behaviour, responsibility, association with antisocial peers and police contact for driving offences) and others on which both the moderate and high risk driving groups were higher than the low risky driving group (multi-substance use, time spent driving per week). In addition, young adults in the moderate risky driving group were rated by parents as less anxious and more assertive than both the other groups.

Interestingly, the three groups did not significantly differ in their family characteristics (parental marital or employment status; educational level; level of financial strain by parent report); or their learner driver experiences and their own current life circumstances (employment and educational status, living arrangements, highest year of completed schooling etc) by self report.

More than two-thirds of the group differences were in the medium to large effect size range. The most powerful differences were found for antisocial behaviour, association with antisocial peers and unsurprisingly, police contact for driving offences.

Summary

The three groups (low, moderate, and high risky driving) could not be distinguished during infancy, toddlerhood and early childhood. The first significant group differences were reported by teachers in mid childhood, and centred on temperament style (task orientation), behaviour problems (aggression and hyperactivity), and school adjustment difficulties. During late childhood, teachers noted similar differences and also reported significant group differences in levels of social competence (cooperation). All significant group differences were between the high risky driving group and both the moderate and low risky driving groups, with the former group exhibiting more difficulties than the other two groups. Group differences during mid and late childhood were in the medium effect size range.

By early adolescence, the first parent- and self-reported differences emerged. Parent and adolescent reports suggested that the high risky driving group was displaying significantly more difficulties than the moderate and low risky driving groups over a range of domains. As before, many of these group differences centred on personal attributes (temperament style, behaviour problems, social competence). However, group differences were also evident in the areas of school adjustment and achievement, peer relationships and the quality of the parent-adolescent relationship. While the majority of group differences were between the high risky driving group and the other two groups, some differentiation between the low and moderate risky driving groups became evident.

This differentiation continued into mid/late adolescence, with the moderate risky driving group significantly differing from one or both of the other groups (but particularly the low risky driving group) on a number of aspects. Group differences in coping styles (greater use of less adaptive, acting out strategies by the high and moderate groups) were also found at this developmental stage. Group differences during adolescence ranged in strength from small to large effect sizes. The most powerful group difference was found for self-reported antisocial behaviour during mid/late adolescence.

By early adulthood, group differences were evident on a wide range of domains. In addition to those problematic characteristics described previously, the high risky driving group also reported lower levels of civic engagement than the low risky driving group and more police contact for driving offences, while both the high and moderate risky driving groups reported engaging in more multi-substance use, and spending more time driving than the low risky driving group. At this time, the moderate risky driving group continued to become more distinct from the other two groups, being rated as more problematic than the low risky driving group, but less problematic than the high risky driving group, on many variables. Furthermore, the moderate risky driving group was rated as more assertive and less anxious than both other groups during early adulthood. The majority of group differences were in the moderate to large effect size range, with the most powerful differences evident on antisocial behaviour, association with antisocial peers and police contact for driving offences.

Precursors and correlates of crash involvement

In order to identify predictors of crash involvement, young adults were grouped according to number of crashes they had been involved in as a driver or rider since commencing driving a car or riding a motorcycle.

Formation of groups

Table 14 displays the distribution of crashes among the ATP sample. As the Table shows, while the majority (57 per cent) had not been involved in a crash, approximately one-third had been in one crash, and about one-in-ten had been involved in two crashes. Very few (less than 3 per cent) reported that they had been involved in more than two crashes as a driver.

Individuals were classified into three groups:

Distribution of crashes

Group characteristics

Gender composition

The gender composition of the three groups (no, single, and multiple crash groups) is shown in Table 15. From this Table it can be seen that that there was a weak but significant difference in the gender composition.27 For example, there was a trend for more young men than women to be members of the multiple crash group (54 per cent of this group were male). Hence, while young men and women did not significantly differ in their overall rates of crash involvement, when they were grouped according to the number of crashes they had been involved in, a weak but significant gender difference emerged.

Gender and crash groups
Crash characteristics

The single and multiple crash groups were compared to determine whether the two groups differed in the circumstances surrounding their crashes (whether they were alone or with passengers) and/or the consequences of these crashes (whether the crash resulted in property damage, or injury or death). These comparisons revealed that a higher proportion of the multiple crash group had been involved in crashes resulting in property damage (regardless of whether they were alone28 or had passengers when the crash occurred29) than the single crash group (see Table 16).

Crash characteristics
Group differences by stage of development
The precursors and correlates of crash involvement were next examined. Using a

similar strategy to that employed previously, separate analyses were performed for each developmental stage and source of report. 30 However, unlike the previous analyses of risky driving, participants' level of driving exposure (number of hours driving per week) was controlled in this next set of analyses.

The results of these analyses are summarised separately for each stage of development in Tables 17 and 18, as well as being described in the text below. As described previously, variables displayed in CAPITALS showed significant group differences (p<.0023), while the other variables listed showed only trends (p<.05). The subsequent discussion will focus on significant group differences.

Infancy, toddlerhood, and early childhood (0-4 years)
The no, single and multiple crash groups did not significantly differ during infancy, toddlerhood and early childhood.

Mid childhood (5-8 years)
There was a trend for multivariate group differences in parent reports at mid childhood. However, the strength of these differences was not sufficient to reach the adjusted significance level of p<.0023. 31 No significant group differences were evident by teacher reports at this age.

Late childhood (9-12 years)
Parent-, teacher- and self-reports indicated no significant differences between the groups during late childhood.

Early adolescence (12-14 years)
There were trends for multivariate group differences on parent- and self-reported individual attributes and parent-reported environmental characteristics during early adolescence. However, these differences did not reach the adjusted significance level of p<.0023. No significant differences were evident on self-reported environmental factors at this age.

Mid/late adolescence (15-18 years)
Significant group differences first emerged during mid/late adolescence. These differences were apparent on a wide range of domains, namely, temperament style, behaviour problems, social skills, coping styles, and peer relationships (see Table 17). Only on parent-reported environmental factors were no significant multivariate group differences found, although there was a trend for such differences (p<.05).

The majority of group differences were between the two crash groups (single and multiple crashes) and the no crash group. For example, the two crash groups were rated as more aggressive (by parents and adolescents), more antisocial (by adolescents), and as more reliant on less adaptive acting out coping strategies such as reacting explosively or using drugs (by adolescents), than the no crash group. In addition, the multiple crash group (but not the single crash group) was rated as less persistent (by parents), and as using more licit and illicit substances (by adolescents) than the no crash group.

There were also some aspects reflecting interpersonal skills on which the single crash group was rated more positively than either the multiple and no crash groups (self-reported level of communication with peers) or the no crash group alone (parent-reported assertiveness).

Group differences were generally quite modest at this stage, typically falling in the small effect size range. The most powerful group differences were found for selfreported antisocial behaviour and use of less adaptive, acting out coping strategies, which both were in the medium effect size range.

Early adulthood (19-20 years)
The majority of group differences during early adulthood emerged from self reports, with only one parent-reported difference (conflict in parent-young adult relationship) evident at this stage (see Table 18). There was a trend for multivariate group differences on parent-reported individual attributes, but this did not reach the adjusted significance level.

Similar to the previous stage, many of the group differences were between the two crash groups (single and multiple crashes) and the no crash group, with those who had been involved in a crash displaying a more 'difficult' temperament style (lower persistence and higher reactivity, by self report) engaging in more antisocial behaviour and multi-substance use (as reported by young adults); having more police contact for driving offences; and associating more frequently with antisocial peers than those in the no crash group.

However, as in mid/late adolescence, there were also some aspects on which the multiple crash group alone was reported to experience more difficulties than the no crash group (parent-reported conflict in their relationship with their ATP young adult, self-reported stress experienced during driving practice), and others on which the multiple crash group reported more difficulties than the single crash group (self-reported antisocial behaviour, police contact for driving offences, and association with antisocial peers).

Parent-reported and self-reported significant differences between the no, single and multiple crash groups over mid/late adolescence

The three groups did not significantly differ in their family characteristics according to parent reports (parental marital or employment status, educational level, level of financial strain) or the young adults' current life circumstances (employment and educational status, living arrangements, highest year of completed schooling) by self reports.

As before, most group differences were in the small effect size range. However, medium effect sizes were observed for the degree of conflict between parents and young adults (parent-report), antisocial behaviour (self-report), and police contact for driving offences (self-report).

Parent-reported and self-reported significant differences between the no, single and multiple crash groups in early adulthood

Summary

The three crash groups (no, single, and multiple) did not significantly differ during infancy, toddlerhood, childhood and early adolescence. The first group differences emerged in mid/late adolescence. Group differences were apparent on a wide range of domains at this time (temperament style, behaviour problems, social skills, coping styles, and peer relationships) and were generally between the two groups that had experienced crashes and the no crash group. Parent and self reports suggested that those who had been involved in a crash displayed more problematic characteristics than those who had not. Group differences were generally in the small effect size range, with the most powerful group differences found for antisocial behaviour and a less adaptive, acting out coping style, which were both in the medium effect size range.

While there were few parent-reported differences in early adulthood, self reports indicated differences between the groups across a range of domains (temperament style, behaviour problems, peer relationships, learner driver experiences and contact with the criminal justice system). Similar to mid to late-adolescence, many group differences emerged between the two crash groups and the no crash group, with the two crash groups being rated as more problematic than the no crash group in these areas of functioning. There were also some aspects (antisocial behaviour, police contact for driving offences, association with antisocial peers) on which the two crash groups were clearly differentiated from each other, with single crash group being rated as more problematic than the no crash group, but less problematic than the multiple crash group. Once again, most group differences were in the small effect size range. The most powerful group differences were found for antisocial behaviour, police contact for driving offences and the amount of conflict in the relationship between parents and young adults.

Precursors and correlates of speeding violations

To assist with identifying predictors of speeding violations, young adults were grouped according to the number of times they had been apprehended by police for exceeding the speed limit, according to self report.

Formation of groups

Table 19 displays the distribution of speeding violations among the ATP sample.

Distribution of speeding violations among the ATP sample

As this Table shows, the majority of participants (69 per cent) had never been detected speeding. About one-in-five had been caught speeding once, and just over one-in-ten reported multiple speeding violations. Very few participants reported having been caught speeding more than twice (less than 5 per cent). Consequently, participants were divided into three groups:

Group characteristics

Gender composition

The gender composition of the three groups (no, single, and multiple speeding violations) is shown in Table 20, which shows that that there was a significant difference in the gender composition of the no and multiple speeding violation groups. 32 While there was a similar proportion of young men and women in the single speeding violation group, there was a higher proportion of males in the multiple speeding violation group (68 per cent of this group were young men and 32 per cent were young women), and significantly more females in the no speeding violation group (62 per cent of this group were young women and 38 per cent were young men).

Gender composition of the no, single and multiple speeding violation groups
Self-reported speeding behaviour

In order to determine whether the three speeding violation groups differed in their levels of speeding behaviour (and not just their propensity to attract police attention), the frequency of engagement in different levels of speeding behaviour across the three groups was examined. The results of these comparisons are displayed in Table 21.

As this Table shows, the three groups significantly differed in their levels of self-reported speeding behaviour.33 As expected, the multiple speeding violation group reported speeding more often (at all levels) than the no speeding violation group, while the single speeding violation group fell between these two groups, reporting more low to moderate speeding behaviour (up to 25km/h over the limit) than the no speeding violation group, but less high-level speeding (in excess of 25km/h above the limit) than the multiple speeding violation group.

Self-reported speeding behaviour of the no, single and multiple speeding violation groups during their past 10 trips
Group differences by stage of development

The correlates and precursors of speeding violations are examined in this next section. Using the same strategy as employed in the previous section, separate Multivariate Analyses of Covariance (MANCOVAs) were performed for each developmental stage and source of report, with driving exposure included as a covariate to control for its influence on speeding behaviour.34

Tables 22 through 25 summarise the results of these analyses. As a guide to interpreting these tables, variables displayed in CAPITALS indicate significant group differences (p<.0023),35 while the other variables listed indicate group trends (p<.05). The following discussion will focus only on significant differences.

Infancy, toddlerhood and early childhood (0-4 years)
The no, single, and multiple speeding violation groups did not significantly differ from each other during infancy, toddlerhood, and early childhood.

Mid childhood (5-8 years)
There were trends for multivariate group differences in parent and teacher reports during mid childhood. However, these differences were not strong enough to reach the adjusted significance level (p<.0023).

Late childhood (9-12 years)
The first significant group differences emerged during late childhood (see Table 22). These differences were between the multiple speeding violation group and the other two groups (single and no speeding violations) and all but one were teacher-reported. Both teachers and parents rated the multiple speeding violation group as more hyperactive than the single and no speeding violation groups, while teachers also rated the multiple speeding violation group as more aggressive, less cooperative, less self-controlled and less task-oriented than other children. Most group differences were within the medium effect size range.

There was a trend for multivariate group differences in self reports during latechildhood. However, these differences were not sufficiently strong to reach statistical significance (at the adjusted significance level of p<.0023).

Parent-reported and teacher-reported significant differences between the no, single and multiple speeding violation groups in late childhood

Early adolescence (12-14 years)
By early adolescence, numerous differences were apparent between the speeding violation groups (Table 23). As before, the majority of group differences were between the multiple speeding violation group and the other two groups, with the multiple speeding violation group appearing more problematic than the single and no speeding violation groups across a wide range of domains. For instance, the multiple speeding violation group exhibited a less persistent temperament style (as reported by parents), higher levels of hyperactivity (as reported by parents), lower social skills (lower parentreported cooperation and self-reported empathy) and more school adjustment difficulties (as reported by parents and adolescents) than the single and no speeding violation groups. There were also some characteristics on which only the multiple and no speeding violation groups differed (parent- and self-reported association with antisocial peers, self-reported aggression), to the disadvantage of the former group.

The first sign of differences between the single speeding violation group and the no speeding violation group was found at this age, with participants in both the multiple and single speeding violation groups rating themselves as more antisocial than those in the no speeding violation group.

Group differences ranged in strength from small to medium effect sizes. The most powerful group differences were found on persistence (parent report), hyperactivity (parent report), antisocial behaviour (adolescent report), empathy (adolescent report), and school adjustment (parent report).

Parent-reported and self-reported significant differences between the no, single and multiple speeding violation groups in early adolescence

Mid/late adolescence (15-18 years)
Few significant group differences were found during mid/late adolescence (Table 24). The only significant differences that emerged at this time were in the areas of temperament style and behaviour problems.For example, the multiple speeding violation group was found to be more aggressive (according to parent reports) and more antisocial (by self reports) than the other two groups.

The single violation group showed further signs of differentiation from the no speeding violation group at this age. For example, like the multiple violation group, parents rated the single violation group as more sociable than the no violation group, and self reports indicated that they more often engaged in antisocial behaviour than the no violation group (although the single speeding violation group was not rated as highly on this aspect as the multiple violation group).

Group differences were generally in the small effect size range at this developmental stage. However, a medium effect size was found for differences on self-reported antisocial behaviour. There were trends for significant multivariate group differences on environmental factors by both parent and self reports (p<.05), but they did not reach the adjusted significance level.

Parent-reported and self-reported significant differences between the no, single and multiple speeding violation groups over mid/late adolescence

Early adulthood (19-20 years)
The domains in which group differences appeared became more varied by early adulthood (Table 25). Group differences emerged in some areas of functioning in which differences had been found in earlier developmental stages (behaviour problems, peer relationships), as well as domains that had not been examined before (current life circumstances, contact with the criminal justice system).

Parent-reported and self-reported significant differences between the no, single and multiple speeding violation groups in early adulthood

While many group differences were between the multiple speeding violation group and the other two groups, there were also some characteristics on which both the single and multiple speeding violation group differed from the no speeding violation group; the multiple speeding violation group alone differed from the no speeding violation group; or all three groups differed from each other.

In terms of behavioural and emotional difficulties, self reports indicated that multiple speeding violation group engaged in more multi-substance use than other young adults, and both the single and multiple speeding violation groups were more antisocial (by self report) and had experienced more contact with the criminal justice system (by parent report) than the no speeding violation group. Furthermore, both speeding violation groups also reported experiencing more police contact for driving offences than the no violation group, although the single speeding violation group was not as problematic as the multiple speeding violation group on this aspect. While the two speeding violation groups exhibited higher levels of behaviour problems, the no speeding violation group reported more emotional adjustment difficulties, with self reports indicating that the no speeding violation group was more depressed than the multiple speeding violation group, and more anxious than the other two groups.

In relation to current life circumstances, the multiple speeding violation group was less likely to have completed school than the other two groups, and less likely to be currently undertaking a course of study (according to their own reports). Perhaps as a result, the multiple speeding violation group was more likely to be in paid employment than the other two groups.

The multiple speeding violation group also differed from the other groups in terms of their peer relationships, with young people in the multiple speeding violation group associating more frequently with antisocial peers (by self report).

No significant differences were found between the groups in their parent-reported family characteristics (parental marital, educational or occupational status; financial strain) or their self-reported learner driver experiences, although there was a trend for group differences on the amount of driving practice young adults received when learning to drive.

Most group differences were in the small to medium effect size range with the exception of police contact for driving offences, for which a large effect size was found.

Summary

The three speeding violation groups (no, single and multiple violations) could not be distinguished from each other during the earliest years of life. While there were trends for multivariate group differences in mid childhood, the first significant group differences did not emerge until late childhood. At this time, significant group differences were found between the multiple speeding violation group and both the single and no speeding violation groups on temperament style (task orientation), behaviour problems (hyperactivity and aggression) and social skills (cooperation and self-control), with the former group exhibiting more difficulties than the other two groups. Most group differences were teacherreported, and were within the medium effect size range.

By early adolescence, numerous differences were evident between the speeding violation groups. Consistent with earlier findings, the majority of group differences were between the multiple speeding violation group and the other two groups, with the multiple speeding violation group exhibiting more difficulties than the single and no speeding violation groups across a wide range of domains (exhibiting a more difficult temperament style, more behaviour problems, lower social skills, and more school adjustment difficulties). Group differences ranged in strength from small to medium effect sizes, with the most powerful group differences being found for persistence, hyperactivity, antisocial behaviour, empathy and school adjustment.

Interestingly, there were few significant group differences between the speeding violation groups during mid/late adolescence. The only significant group differences emerged in the areas of temperament style and behaviour problems. Furthermore, the pattern of group differences was not as consistent as at previous time points. While the multiple speeding violation group was still more problematic than both the other groups on some characteristics (aggression and antisocial behaviour), the single speeding violation group showed some differentiation from the no speeding violation group at this age. For example, like the multiple speeding violation group, the single speeding violation group was found to be more sociable than the no speeding violation group and to engage in higher levels of antisocial behaviour (although not as high as in the multiple speeding violation group). Group differences were relatively small at this developmental stage. However, a medium effect size was found for differences on self-reported antisocial behaviour.

The differentiation of the single speeding violation group from the no speeding violation group continued in early adulthood. While the multiple speeding violation group was still generally the most problematic group, the single speeding violation group also exhibited higher levels of antisocial behaviour and higher rates of contact with the criminal justice system than the no speeding violation group. Interestingly, the no speeding violation group displayed more emotional adjustment difficulties (anxiety and depression) than one or both of the other groups at this time.

The domains in which group differences appeared also became more varied at this time. Group differences emerged in some areas of functioning in which differences had been found in earlier developmental stages (behaviour problems, peer relationships), as well as domains that had not been examined before (current life circumstances, contact with the criminal justice system). Most group differences were in the small to medium effect size range, with the exception of police contact for driving offences, which, not surprisingly, was large.

Gender differences in the precursors and correlates of risky driving, crash involvement and speeding violations

As noted earlier, the risky driving, crash involvement and speeding violation groups differed significantly in gender composition. More specifically, young men tended to be over-represented in the most problematic driving groups (high risky driving, multiple crash and multiple speeding violations), while a higher proportion of young women tended to be members of the low problem driving groups (particularly the low risky driving and no speeding violation groups). This gender difference was particularly notable in the high risky driving and multiple speeding violation groups, with more than three-quarters of high risky drivers, and two-thirds of the multiple speeding violation group being young men. These gender differences may have contributed to the group differences found.

While it would have been interesting to examine the precursors and correlates of risky driving, crash involvement and speeding violations separately for young men and women, the small number of women in these groups, particularly in the high risky driving (n=17) and multiple speeding violation groups (n=36), would have meant that the statistical power of such analyses would have been severely limited, and hence, meaningful results would have been difficult to obtain. Consequently, it was decided not perform gender-specific analyses.

An alternative approach would have been to statistically control for the influence of gender on risky driving, crash involvement and speeding violations. However, given the clear interconnectedness between gender and the three problematic driving outcomes (particularly risky driving and speeding violations), controlling for the effects of gender was not considered sensible, since this would have effectively cancelled out the very real contribution of gender, and may have resulted in findings of limited relevance to real world settings (Tabachnick and Fidell 1996), where young males are in fact more likely to engage in risky driving behaviour and speeding (Beirness and Simpson 1988; Shope, Lang and Waller 1997). Hence, the effects of gender on risky driving, crash involvement and speeding violations were not controlled in the ATP Young Drivers Study.


  1. Analyses of Covariance (ANCOVAs) were performed, in which the effects of the covariate (driving exposure) were statistically controlled before group comparisons were undertaken.
  2. A two-step clustering procedure was undertaken of young adults' responses to the eight risky driving items. The first step involved identifying the appropriate number of clusters in the data. Random samples of approximately 200 cases were hierarchically clustered using Ward's method, with squared Euclidean distance used to measure the inter-object similarity. Examination of dendograms showed that a three-cluster solution produced good differentiation between groups. However, solutions between two to five factors were also possible. Therefore, in the second step, K-means two to five cluster solutions were imposed on the whole sample, using SPSS QUICK CLUSTER, a procedure that groups cases into clusters once the number of clusters is provided. These different solutions were inspected to find the most meaningful one. The three-cluster solution using running means was selected as the most appropriate representation of the data (Hair et al. 1998; SPSS Inc. 2002).
  3. The three risky driving groups significantly differed in the frequency with which they exceeded the speed limit by up to 10km/h (F (2,1052)=1678.17, p<.001); exceeded the speed limit by between 10 and 25km/h (F (2, 1052) = 1834.61, p<.001); exceeded the speed limit by over 25 km/h (F (2, 1052) = 660.75, p<.001), drove when probably affected by alcohol (F (2, 1052) = 41.38, p<.001); forgot to wear their seat-belt (or helmet ) for part of the trip (F (2, 1052) = 24.50, p<.001); drove when very tired (F(2, 1052) = 71.62, p<.001); or drove when probably affected by an illegal drug ( F(2, 1052) = 23.80, p <.001). The high risky driving group significantly differed from the other two groups in the frequency with which they forgot to wear a seat-belt (or helmet) for the entire duration of a trip (F (2, 1052) = 27.75, p<.001).
  4. A chi-square test of independence revealed a significant difference in the proportion of the males and females in the risky driving groups, x2 (2) = 44.71, p<.001. An examination of the standardised residuals revealed a significantly higher proportion of males (and a lower number of females) in the high risky driving group than would be expected by chance; as well as a significantly higher proportion of females in the low risky driving group, and a lower number of males in this group, than would be expected by chance.
  5. One-way Analyses of Variance (ANOVAs) revealed significant differences between the risky driving groups on rates of crash involvement (F (2, 1032)=5.70, p<.01) and speeding violations (F (2, 1024)=32.04, p<.001). Scheffe tests revealed that the high risky driving group was involved in significantly more crashes than those in the low risky driving group, while all three groups significantly differed in their number of speeding violations, with the high risky driving group, on average, reporting the highest number of speeding violations and the low risky driving group, the least.
  6. As described in Section 2.
  7. Cohen's (1988) effect size criteria were used to assess the strength of group differences across the various domains. For Analyses of Variance an effect size of .10 represents a small effect, .25 a medium effect, and .40 a large effect (Cohen 1988).
  8. A chi-square test of independence revealed a significant difference in the proportion of the males and females in the crash involvement groups, x2 the standardised residuals revealed a trend for a higher number of males (and a smaller number of females) in the multiple crash group than would be expected by chance.
  9. x2 (1) = 26.56, p<.001.
  10. x2(1) = 12.75, p<.001.
  11. As before, several analyses (in this case, Multivariate Analyses of Covariance - MANCOVAs) were performed for each source of report from early adolescence on, due to the large number of variables under investigation during adolescence and early adulthood.
  12. As a large number of analyses were performed, the Bonferroni adjustment procedure was used to adjust the significance level to reduce the possibility of Type 1 error (the likelihood of showing group differences of statistical significance when they are not actually there). Results are reported for differences at the Bonferroni adjusted level of 0.0023 or lower.
  13. A chi-square test of independence revealed a significant difference in the proportion of the males and females in the speeding violation groups, x2 (2) = 40.97, p<.001. An examination of the standardised residuals revealed a significantly higher proportion of males (and a lower proportion of females) in the multiple speeding violations group than would be expected by chance; as well as a significantly higher proportion of females in the no violations group (and a lower number of males in this group) than would be expected by chance.
  14. One-way Analyses of Variance revealed significant differences between the no, single and multiple speeding violation groups in the frequency with which they drove up to 10km/h above the speed limit, F (2,1028)=31.2, p<.001; drove between 10-25 km/h over the speed limit, F(2, 1026)=30.5, p<.001; and drove more than 25km/h over the speed limit, F(2,1027)=41.2, p<.001.
  15. As before, multiple MANCOVAs were performed for each informant (parent, adolescent/young adult) during adolescence and early adulthood, due to the large number of variables under investigation during this period.
  16. Due to the large number of analyses performed, the Bonferroni adjustment procedure was used to adjust the significance level to reduce the possibility of Type 1 error (the likelihood of showing group differences of statistical significance when they are not actually there). Results are reported for differences at the Bonferroni adjusted level of 0.0023 or lower.

Research Report 12: Contents | Next | Previous