Accident Analysis and Prevention 72 (2014) 161–168

Contents lists available at ScienceDirect

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

Stability of risky driving from late adolescence to early adulthood Suzanne Vassallo a,∗ , Diana Smart a , Melinda Spiteri b , Samantha Cockfield c , Anne Harris b , Warren Harrison d a

Australian Institute of Family Studies, Level 20/485 La Trobe Street, Melbourne, Vic. 3000 Australia Royal Automobile Club of Victoria, 550 Princes Highway, Noble Park North, Vic. 3174, Australia Transport Accident Commission, GPO Box 742, Geelong, Vic. 3220, Australia d Eastern Professional Services, 3 Birdwood Avenue, Ferntree Gully, Vic. 3156, Australia b c

a r t i c l e

i n f o

Article history: Received 14 January 2014 Received in revised form 15 June 2014 Accepted 3 July 2014 Keywords: Risky driving Young adults Longitudinal Drink-driving Speeding

a b s t r a c t This study examined the stability of risky driving behaviour from late adolescence to early adulthood among 823 young Australian drivers participating in an ongoing longitudinal study. This issue was explored by examining the stability of risky driving between the ages of 19–20 and 23–24 years (1) across the cohort and (2) among individuals. Focusing on cohort-wide trends, a modest reduction in the occurrence of speeding was observed across the sample between 19–20 and 23–24 years. However, drink-driving increased markedly over this period, and driving without a seatbelt or helmet for part of a trip also rose. Rates of other risky driving behaviours remained relatively unchanged. With regard to trends among individuals, while a decrease was evident in the risky driving propensities of many who had been classified as moderate or high risky drivers at age 19–20, 48% of the former group, and 77% of the latter group, still exhibited risky driving tendencies at 23–24 years. Together, these findings suggest a fair degree of stability in risky driving from late adolescence to early adulthood among this sample of Australian youth, highlighting the continuing need for road safety initiatives targeting young drivers beyond their first years of licensure. Crown Copyright © 2014 Published by Elsevier Ltd. All rights reserved.

1. Introduction Considerable research has sought to clarify the factors that underlie risky driving among young novice drivers. This focus is clearly warranted, as young people are consistently overrepresented in serious crashes resulting in injury or death (OECD, 2006; Engström et al., 2003), and are at their highest lifetime risk of being in a crash during their early driving years (Lee et al., 2011; VicRoads, 2005). However, much less attention has been given to young drivers past the first years of their driving careers, to determine whether driving behaviours change as they mature and gain on-the-road experience. Do young people maintain the same tendencies to drive safely or riskily over time, or do their driving styles change with increasing age and driving exposure? The scant longitudinal research available suggests that risky driving generally decreases as young people move from adolescence to adulthood (Bingham et al., 2008; Jessor et al., 1997). Jessor et al. (1997) noted a decline in levels of risky driving among 1659

∗ Corresponding author. Tel.: +61 3 9214 7845; fax: +61 3 9214 7839. E-mail addresses: [email protected] (S. Vassallo), [email protected] (D. Smart), Melinda [email protected] (M. Spiteri), samantha cockfi[email protected] (S. Cockfield), [email protected] (A. Harris). http://dx.doi.org/10.1016/j.aap.2014.07.001 0001-4575/Crown Copyright © 2014 Published by Elsevier Ltd. All rights reserved.

young American drivers between the ages of 18 and 25 years. Similarly, Begg and Langley (2001) reported a decrease in the prevalence of almost all types of risky driving among 476 New Zealand male drivers between the ages of 21 and 26 (this decrease was not evident for females who tended to engage in less risky driving overall). One explanation for this trend points to the positive effects that greater experience may have on young people’s driving (Bingham et al., 2008; Jessor et al., 1997). Crash statistics suggest that after an initial period of high risk, young drivers’ crash risk declines sharply over the first few years of licensure (Lee et al., 2011; Engström et al., 2003). An improvement in young people’s driving, resulting from greater experience on the road, is seen as a major contributor to this trend (Bingham et al., 2008; Engström et al., 2003). New drivers tend to lack many of the cognitive and perceptual skills needed to make them safe and effective drivers (Cavallo and Triggs, 1996; Triggs and Smith, 1996). Furthermore, research suggests that inexperienced drivers tend to underestimate the level of risk associated with certain types of driving behaviour or situations, and overestimate their own level of driving ability or capacity to deal with such situations (Cavallo and Triggs, 1996; Triggs and Smith, 1996). Hence, some forms of risky driving, such as unsafe lane changes or following too closely, may not be intentional, but instead reflect inexperience (Cavallo and Triggs, 1996; Williams, 1998). Consequently, it is possible that young people drive in a less risky manner

162

S. Vassallo et al. / Accident Analysis and Prevention 72 (2014) 161–168

as they grow older because their driving improves with experience, as does their appreciation of the dangers of risky driving (Bingham et al., 2008). Developmental changes that occur in early adulthood may also contribute to a decrease in risky driving behaviour. The early adult years are generally a time of considerable change in multiple aspects of life (Arnett, 2000). For many, this period witnesses the transition from school to employment and/or further education, a move out of the parental home, as well as the development of intimate relationships, marriage and parenthood. The assumption of adult occupational, marital and parental roles has been linked to decreased involvement in problem behaviours (Jessor et al., 1997). For example, Bingham et al. (2008) found that both the degree and frequency of risky driving diminished with increasing levels of psychosocial maturity, as reflected by the adoption of adult roles and the completion of developmental tasks associated with becoming an adult. It has also been suggested that a decrease in characteristics that underlie risk-taking, such as sensation-seeking, may contribute to the decline in problem behaviour often seen during early adulthood (Arnett, 1991). Brain development continues through adolescence and early adulthood, with capacities associated with impulse control, decision-making, planning and judgement being the last to evolve (Giedd, 2004; Sowell et al., 1999). Hence, it is likely that cognitive and emotional maturation also contribute to the decline in risk-taking observed over this period. The finding that risky driving among young people tends to decrease with greater driving experience and increasing age is a general, population-level trend. Nevertheless, risky driving remains a frequent occurrence among this age group. Thus, within the broad population trend, sub-groups exhibiting divergent across-time trajectories are probable. For example, research shows that a small number of young people persist in highly unsafe driving practices beyond their first years of driving (Begg and Langley, 2004). It is also plausible that some initially safe young drivers may engage in risky driving as they gain confidence on the road, although research is lacking on this point. Longitudinal research is needed to investigate the possibility of differing across-time trajectories among sub-groups, as well as the factors that may influence progression along differing trajectories. To the authors’ knowledge, few studies have investigated these issues. The Australian Temperament Project (ATP), a longitudinal study of a large community sample followed from infancy into early adulthood, offers a rare opportunity to investigate these issues. At the time the study reported here was undertaken, the ATP had completed 14 waves of data collection over the first 24 years of life, with information on driving history and behaviours collected in the last two waves at 19–20 and 23–24 years, thus enabling examination of across-time trends in risky driving proclivities. The current study has two aims: (1) to examine the stability of a range of risky driving behaviours (e.g. speeding, drink-driving) in a large cohort of Australian youth from 19–20 to 23–24 years, and (2) to explore the stability of risky driving among individuals over this time span. Three patterns of risky driving behaviour (low, moderate and high) have previously been identified in this cohort, when study members were aged 19–20 years (see Vassallo et al., 2007). This paper investigates whether individuals exhibit the same risky driving propensities four years later, at 23–24 years, thereby enabling examination of stability and change in risky driving tendencies over time. Based on the available theoretical and empirical literature, which suggests that engagement in risk-taking behaviours lessens as young people mature and gain driving experience (Bingham et al., 2008; Jessor et al., 1997), it is anticipated that rates of risky driving will be lower among the cohort at 23–24 years, than at 19–20 years. However, it is possible that this decline will be less

marked than in previous studies, because outcomes are assessed at an earlier age (23–24 as opposed to 25–26), hence, developmental changes associated with reduced risk-taking are likely to still be ongoing. Given the lack of research on the stability of risky driving at an individual level among young people, we hypothesise that there will be a reduction in risky driving among individuals who exhibited high or moderate risky driving tendencies at 19–20 years in line with the above normative trends, but make no predictions about the stability of safe driving tendencies. 2. Method 2.1. Participants Participants were members of the Australian Temperament Project (ATP). The ATP is a longitudinal community study following the psychosocial development of a large cohort of children born in the State of Victoria, Australia, between September 1982 and January 1983 (for more details, see Prior et al., 2000; Vassallo and Sanson, 2013). The initial sample comprised 2443 infants (aged 4–8 months) and their parents, who were recruited through Maternal and Child Health Centres during a two-week period in 1983. Participants were recruited from urban (1604 children) and rural (839 children) locations, selected on the advice of the Australian Bureau of Statistics to provide a representative sample of the State population. Comparison of the demographic profile of the obtained sample to Census data confirmed that the sample was representative (Prior et al., 2000). Fourteen waves of data were completed up to 2006 when the young people were aged 23–24 years. Parents, maternal and child health nurses, primary school teachers, and from the age of 11 years, the children themselves, have acted as informants. A fifteenth wave of data was collected in 2010–2011 at 27–28 years, but is not reported here.1 The findings presented in the paper are taken from the 13th and 14th data collection waves, when young people were 19–20 years (in 2002), and 23–24 years (in 2006), respectively. Sixty-five percent of the cohort remained members of the study at 19–20 years and 61% at 23–24 years. The attrition rate at both waves was slightly higher for families experiencing socioeconomic disadvantage or from a non-Australian ethnic background. Attrition was also slightly higher at 23–24 years for participants showing a more difficult temperament style in infancy. However, there were no significant differences between those retained and those lost to attrition on behavioural problems in infancy. In total, 1157 young adults participated in Wave 13 (73% of the retained sample, 56% female) and 1000 in Wave 14 (66% of the retained sample, 61% female). In Victoria, where all participants were born, young people are able to commence learning to drive under supervision at 16 years of age. The minimum licensing age is 18 years, with novice drivers spending the first four years of licensure on a probationary licence2 before graduating to a full licence (prior to 2007, the probationary period was three years). The data collection periods thus cover the very early driving years of this sample (19–20 years, cohort mean time probationary licence held = 21 months, 1.75 years), and their driving behaviours after several years of licensure (23–24 years; cohort mean time licence held = 71 months, almost 6 years). While almost all respondents held a driving licence, a small minority did not. This paper focuses

1

A manuscript which reports on the Wave 15 data is in preparation. Victoria currently has a two-stage probationary licence system. A P1 licence is issued for the first 12 months and a P2 licence for the next 3 years. If a driver is 21 years or older when they obtain their probationary licence, they are issued a P2 licence. One condition of these probationary licences is that drivers maintain a zero blood alcohol concentration (BAC) at all times. 2

S. Vassallo et al. / Accident Analysis and Prevention 72 (2014) 161–168

163

Table 1 Characteristics of the ATP sample at ages 19–20 and 23–24 years. Characteristic

N Gender composition Held car licence Held motorcycle licence Held learner’s permit (for car/motorcycle) Mean length of licensure Mean time spent driving (per week) Ever been detected speeding Ever been involved in a crash when driving

Age 19–20 years

23–24 years

1157 56.0% female 86.2% 2.2% 7.2% 20.8 months (SD = 7.9) 11.9 h (SD = 10.1) 31.1% 42.9%

1000 61.0% female 96.9% 6.4% 2.6% 71.2 months (SD = 15.5) 10.5 h (SD = 8.91) 57.8% 59.7%

on all participants with a licence or learner’s permit, giving a total of 1068 available for inclusion in analyses at 19–20 years, 974 at 23–24 years, and 823 for comparison of trends across waves. Table 1 provides a summary of the characteristics of the sample at both timepoints. 2.2. Measures 2.2.1. Engagement in risky driving This was assessed at 19–20 years and 23–24 years by a set of items which asked participants to report the number of trips in their past 10 in which they had engaged in a range of risky driving acts. The length of the driving trips was not restricted or defined. This measurement approach was adopted as it used a time frame within most participants’ recent recall, minimising the likelihood of recall bias. Respondents were asked: “Please think back over the last 10 times you drove a car (or rode a motorcycle), and circle the number that shows on how many occasions you: drove up to 10 km/h over the limit; drove between 11 and 25 km/h over the limit; drove more than 25 km/h over the limit; did not wear your seat belt/helmet for part of the trip; did not wear your seat belt/helmet for all of the trip; drove when very tired; drove when affected by alcohol”. At 19–20 years, respondents were asked whether they had driven when affected by an illegal drug, while at 23–24 years they were asked three items on this issue: whether they had driven when affected by marijuana/cannabis/THC; ecstasy; or amphetamines (speed, uppers, fast etc.). Other additional items included at 23–24 years were: “nearly fell asleep or fell asleep when driving”; “talked on a hands-free mobile phone when driving (i.e., not holding the phone at all)”; “talked on a handheld mobile phone when driving”; and “used a mobile phone function when driving (e.g., read or sent an SMS message)”. 2.2.2. Driving exposure at 23–24 years Respondents were asked: “In a normal week (think about the last week as a guide), how many hours would you spend driving a car (or riding a motorbike) on: Monday to Friday – daylight hours; Monday to Friday – night time hours; Saturday and Sunday – daylight hours; Saturday and Sunday – night time hours” (i.e., four responses were sought). The four scores were summed to provide an estimate of the total number of hours usually spent driving per week. 2.2.3. Detection for speeding at 23–24 years This was assessed by a single item, which asked respondents to indicate the number of times they had been detected speeding. 2.2.4. Crash involvement at 23–24 years Respondents were asked: “Since starting to drive (or ride a motorbike), have you crashed or had an accident when you were

the driver (this includes any crash or accident, whether or not you were at fault, and even if not serious)?” Participants who gave an affirmative response to this item, were then asked to report the number of crashes or accidents they had been involved in. 2.2.5. Licencing Respondents’ licensing status was measured by a series of items that asked about their licencing history (if they had ever held a car or motorcycle licence, how long they had done so, whether their licence had ever been suspended/cancelled) as well as the type of licence/s they currently held (no licence, learner’s permit, probationary licence or full licence). 2.3. Procedure The study used a mail survey methodology in which questionnaires were mailed to all participants along with reply-paid envelopes to facilitate their return. A second mail-out was later conducted to those who did not respond to the initial mailing, followed by telephone reminders. 2.4. Statistical analysis strategy Groups exhibiting differing levels of risky driving at 19–20 years were previously identified using cluster analysis (Vassallo et al., 2007). In brief, three groups were found: (1) a low risky driving group (n = 675, 64% sample, 39% male); (2) a high risky driving group (n = 74, 7% of the sample, 77% male), and (3) a group with rates of risky driving that were intermediate to the low and high groups – the moderate risky driving group (n = 306, 29% sample, 50% male). On average, the low risky driving group engaged in 1.9 (of a possible 8) different types of risky driving, the moderate group, 3.5 types, and the high group, 4.8 types. Group comparisons indicated that the high group had been involved in significantly more crashes than the low group, and had been detected speeding significantly more often than both the moderate and low groups. These results were seen as supporting the validity of the cluster solutions (Vassallo et al., 2007). To determine whether the cluster group members identified at 19–20 years would continue to be classified in a similar way at 23–24 years, cluster analysis was once again used to identify groups with differing risky driving profiles at 23–24 years. A wider set of risky driving behaviours was assessed at 23–24 years (14 items) than at 19–20 years (8 items). A decision was taken to include the full set of items measured at 23–24 years, as this would provide a more complete and valid assessment of young people’s driving propensities. As in Vassallo et al. (2007), a two-step clustering procedure was undertaken of responses to the 14 risky driving items, using the SPSS statistical package. The first step involved identifying the appropriate number of clusters. Random samples of approximately

164

S. Vassallo et al. / Accident Analysis and Prevention 72 (2014) 161–168

Table 2 Percentage who had engaged in differing risky driving behaviours and mean number of trips behaviours had occurred, by age. Risky driving behaviours (in past ten trips)

Age 19–20 years

Driven up to 10 km/h above the limit Driven between 10 and 25 km/h over the limit Driven more than 25 km/h over the limit Not worn a seatbelt (or helmet) for part of the trip Not worn a seatbelt (or helmet) at all Driven when very tired Driven when affected by alcohol

23–24 years

%

M

SD

%

M

SD

t (823)

84.4 48.0 17.7 10.2 7.4 64.0 13.4

4.04 1.45 0.44 0.27 0.19 1.53 0.21

3.36 2.28 1.32 1.08 0.90 1.82 0.67

84.2 46.0 14.0 13.9 5.1 63.2 22.5

3.70 1.33 0.31 0.44 0.19 1.50 0.36

3.13 2.15 1.07 1.54 1.11 1.78 0.84

2.74** 1.60 2.62** −3.12** 0.08 0.35 −4.70***

* p < .05, ** p < .01, *** p < .001.

200 cases were hierarchically clustered using Ward’s method (1963), with squared Euclidian distance used to measure interobject similarity. The dendogram and agglomeration schedules were examined to identify the most appropriate cluster solution. In the second step, a series of non-hierarchical cluster solutions were imposed on the whole sample, to confirm the appropriateness of the solution identified in the first step. The viability of the cluster solution was assessed in two ways. Firstly, multiple discriminant function analysis was conducted to determine the stability of group membership, using the same 14 risky driving behaviour items used in the cluster analysis. Secondly, cluster groups were compared on two indices commonly associated with risky driving: crash involvement and detection for speeding. ANCOVA were undertaken with driving exposure included as a covariate, to control for its effect. Following this, the degree of commonality in the cluster groups identified at ages 19–20 and 23–24 years was examined to determine whether the same individuals who engaged in low-, moderate- and high risky driving at 19–20 years continued to do so at 23–24 years. Due to the small number of participants within some of the resulting groups, analyses examining the characteristics of groups exhibiting different across-time patterns were not undertaken. 3. Results 3.1. Rates of risky driving at 19–20 and 23–24 years Table 2 shows the percentage of young people who reported engaging in each unsafe driving behaviour at least once during their past 10 trips at each age (19–20 and 23–24 years). Results are presented for the whole sample and for those behaviours that were assessed at both timepoints. As Table 2 shows, the proportion of study members who engaged in high-level speeding (driving at more than 25 km/h) decreased between the ages of 19–20 and 23–24 years. Rates of most other risky driving behaviours remained fairly constant over the two time periods. However, the percentage who drove when affected by alcohol markedly increased, and the proportion who drove without a seatbelt for part of a trip also rose. As well as showing the proportion who engaged in each behaviour at each timepoint, Table 2 also shows the average number of trips (out of the past ten) in which each behaviour had occurred. In general, young people tended to speed less often at age 23–24 than they did at 19–20. For instance, the average number of trips during which young people engaged in low-level speeding significantly decreased from 4.04 (or 40%) of their past ten trips at 19–20 to 3.70 (or 37%) at 23–24 (t(820) = 2.74, p = .006); and high-level speeding decreased from 0.44 (4%) of trips to 0.31% (3%) of trips (t(818) = 2.62, p = .009). On the other hand, the frequency with which participants engaged in drink-driving significantly increased from 2% of trips at age 19–20, to 3.6% of

trips at 23–24 (t(822) = −4.70, p < .001) and driving without a seatbelt or helmet also occurred more frequently at age 23–24 (t(820) = −3.12, p = .002). 3.2. Identification of risky driving behaviour sub-groups at 23–24 years As noted earlier, cluster analysis was used to identify groups with differing profiles of risky driving at 23–24 years. While two-, three-, four-, and five-cluster solutions were viable, the threecluster solution was considered the most appropriate for use in future analyses. The small size of some of the clusters identified in the four- and five-cluster solutions, (ranging from 9 to 19), precluded their use in further analyses, while the two-cluster solution did not provide sufficient differentiation between groups. Thus, the three-cluster solution was retained. A small number of participants (n = 26) could not be classified and were therefore excluded at this stage, resulting in a final sample at 23–24 years of 948 participants. The groups formed exhibited similar profiles to those identified at 19–20 years, and thus were labelled identically. The three groups were: a low risky driving group (n = 630, 66% of the sample, 34% male), a moderate risky driving group (n = 261, 28% of the sample, 50% male) and a high risky driving group (n = 57, 6% of the sample, 58% male). The three groups significantly differed in the frequency in which they engaged in a range of risky driving behaviours at 23–24 years (see Fig. 1). On average the low risky driving group had engaged in 3.4 (of a possible 14) different types of risky driving at 23–24 years, the moderate-risky driving group had engaged in 5.6 different types and the high risky driving group, 7.7 types. The three risky driving groups also significantly differed in gender composition (2 (2) = 29.24, p < .001) with more males than females in the high risky driving group at 23–24 years and fewer males than females in the low risky driving group. Multiple Discriminant Function analysis supported the reliability of the three-cluster solution, with 97.3% cases being correctly classified. All of the cases in the low group were correctly classified, as were 91.6% of the moderate group and 93.0% of the high group. ANCOVA added further support for the viability of this solution. Significant group differences were found on rates of crash involvement (F(2, 890) = 11.68, p < .001) and detection for speeding F(2, 912) = 38.50, p < .001), even after amount of driving exposure was controlled. Post-hoc comparisons revealed that the high group had been detected speeding significantly more often in their driving careers than the low and moderate groups (low: M = 1.65, moderate: M = 1.96, high: M = 3.88). The moderate group also reported significantly more speeding violations than the low group. Regarding crash involvement, the high and moderate level groups reported almost identical rates, which were higher than the low group (low: M = 0.87, moderate: M = 1.26, high: M = 1.23), although only the moderate group significantly differed from the low group on this measure (the small size of the high level group,

S. Vassallo et al. / Accident Analysis and Prevention 72 (2014) 161–168

165

Fig. 1. Frequency of risky driving behaviours at 23–24 years by cluster groups (identified at 23–24 years). Abbreviations: a, p < .001; b, p < .01, c, p < .05.

Table 3 Level of risky driving behaviour among the age 19–20 risky driving clusters, four years later. Risky driving cluster Low at 19–20 years Remained stable at 23–24 years Increased to moderate at 23–24 years Increased to high at 23–24 years Moderate at 19–20 years Remained stable at 23–24 years Decreased to low at 23–24 years Increased to high at 23–24 years High at 19–20 years Remained stable at 23–24 years Decreased to moderate at 23–24 years Decreased to low at 23–24 years

n

% of group

% of sample (n = 808)

395 115 10

76.0 22.1 1.9

48.9 14.2 1.2

90 125 26

37.3 51.9 10.8

11.1 15.5 3.2

13 23 11

27.7 48.9 23.4

1.6 2.8 1.4

n = 57, may have limited the statistical power available to detect group differences). 3.3. Commonality in cluster group membership at 19–20 and 23–24 years The stability of cluster ranking from 19–20 to 23–24 years was examined. Eight hundred and eight participants had cluster rankings for both timepoints, and were consequently included in these comparisons. While the rate of attrition from the high cluster identified at 19–20 years (37%), was higher than that from the moderate and low clusters identified at the same age (24% for both), those high risky drivers who participated at 23–24 did not significantly differ from those who did not on their risky driving behaviours at 19–20. Table 3 reveals a relatively high degree of stability in the low group, with 76% of those classified as low risky drivers at 19–20 being classified similarly at 23–24 years. Of the remaining quarter whose risky driving behaviour had increased over this time period, almost all were classified as moderate risky drivers at 23–24 years (22%) with very few (2%) becoming high risky drivers. The moderate- and high groups were less stable, with the majority of individuals being classified as less problematic at 23–24 than at 19–20 years. Fifty-two percent of the moderate group and 72% of the high group showed improvement. However, some had become

more risky. About 11% of those identified as moderate risky drivers at 19–20 years showed an increase in their risky driving behaviour between 19–20 and 23–24 years. While it was extremely uncommon for a low risky driver at age 19–20 to become a high risky driver at 23–24, a change in the opposite direction was not as unusual, with almost a quarter of those classified as high risky drivers at 19–20 years being classified as low at 23–24 years. Looking next at trends across the entire sample, cluster group position remained stable for the majority, with 62% classified as displaying the same pattern of risky driving (low-, moderate- or high) at 23–24 as at 19–20 years. Almost one-in-five (19%) showed an increase in risky driving over this time span, while a similar proportion (20%) showed a decrease in risky driving behaviour. 4. Discussion This paper examined the stability of driving behaviour from late adolescence to early adulthood among a sample of young Australian drivers participating in an ongoing longitudinal study. Two issues were investigated: the stability of risky driving between the ages of 19–20 and 23–24 across the entire cohort, and the stability of risky driving tendencies among individuals. 4.1. Cohort trends from 19–20 to 23–24 years A modest reduction in the occurrence of both low- and highlevel speeding was found between the ages of 19–20 and 23–24 years. However, there was a sizeable increase in drink-driving, and driving without a seatbelt for part of a trip rose. Rates of other risky driving behaviours remained virtually unchanged. These findings suggest that risky driving is as serious an issue (when present) in the mid-20s as in the late teens, and points to the importance of sustaining road safety measures targeted beyond the first years of driving. The increase in the prevalence of drink-driving is notable and is consistent with the higher rates of crashes associated with drinkdriving found at this age in local accident statistics (Senserrick, 2003; Transport Accident Commission, 2013). It is likely that the increase in drink-driving reflects differences in licence restrictions at the two ages. At 19–20 years, all of the participants in this study held probationary licences or learner’s permits that required them to have a zero blood alcohol content (BAC) level when driving.

166

S. Vassallo et al. / Accident Analysis and Prevention 72 (2014) 161–168

However, at 23–24, when rates of drink-driving were noticeably higher, most (96%) held full licences and were permitted to have a low level of alcohol in their system when driving (0.05 g/100 ml). Hence, it is likely that the more stringent restrictions at 19–20 years had a stronger deterrent effect than those placed on fully licenced drivers. Indeed, research by Senserrick (2003) suggests that the transition from a zero to .05 BAC limit may be difficult for some young drivers. She found that some of the most common strategies used by 21–26 year-old drivers to avoid drink-driving (i.e., counting or spacing drinks, getting someone else to drive – a designated driver) were also the most likely to be misused (i.e., the young adult underestimated their own BAC level; the designated driver was over the legal BAC limit). Senserrick (2003) suggested that the removal of licencing restrictions may occur at a time when drivers are not completely cognizant of their own tolerance for alcohol, or how many drinks they can safely consume without exceeding the legal alcohol limit. As mentioned earlier, in 2007, new licencing requirements were introduced in the state of Victoria, Australia, extending the overall probationary period from three to four years, thus extending the zero BAC requirement of predominantly young drivers for a further year. Interestingly, while there was a trend for lower rates of nonseatbelt use over an entire trip at 23–24 years than at age 19–20, rates of non-use for part of trip significantly rose over the study period. Research suggests that drivers generally fall into three categories in terms of their seatbelt use – consistent users, consistent non-users, and inconsistent users – and that different factors motivate the behaviours of these groups (Oxley et al., 2009; Harrison et al., 2000). For instance, a driver who never wears a seatbelt may do so as an act of defiance against authority or the law or because they have a propensity for risk-taking; while a driver who does not wear a seatbelt for part of a trip, may do so out of habit (i.e., they are not cued to put on their seat belt until they reach a major road), or because they assess the likelihood of being involved in a crash in certain situations as being very low. The increase in inconsistent seatbelt use observed in this sample may indicate that a small proportion of the sample became more lackadaisical in their driving habits with increasing age, experience on the road, and graduation to an unrestricted licence.3 Driving when fatigued was also common among study members and had not decreased in prevalence between the two ages (almost two thirds of participants at both ages reported that this had occurred in their last 10 trips). This is a concerning trend as fatigue is a well-established contributor to crash involvement (Clarke et al., 2002; Dobbie, 2002). However, a small decrease in high-level speeding was observed between the ages of 19–20 and 23–24 years, which is encouraging given the strong links between excessive speeding and crash involvement (Clarke et al., 2002; Engström et al., 2003). Taken together these findings provide little support for our hypothesis that rates of risky driving would be lower among the cohort at 23–24 years than at 19–20 years. Our findings are somewhat at odds with past research, which suggests that risky driving generally decreases over early adulthood (Bingham et al., 2008; Jessor et al., 1997). A number of explanations are possible. Firstly, the indicators of risky driving used here (speeding, drink-driving, driving when affected by illegal drugs, driving while fatigued and non-use of seatbelts) differed from those used in several other studies (see Begg and Langley, 2001 and Jessor et al., 1997 for examples), which have focused on behaviours such as tailgating, unsafe passing and running red lights or stop signs. These measurement

3 Although not statistically significant, we noted a trend for a higher proportion of fully licenced drivers to have driven without a seatbelt for part of a trip at 23–24 years, than among those on restricted licences (probationary or learner drivers).

differences might have contributed to the divergence in results. Secondly, countries vary in licensing age and licence conditions, hence it is possible that ATP study members may have been at a different stage in their driving career to those in other international studies. Finally, studies that have observed a decline in risky driving over early adulthood have generally utilised later assessment points than the current study (i.e., Jessor et al. (1997) noted a decline by age 25; Begg and Langley (2001) observed a decline by age 26). Hence, it is possible that had we assessed risky driving behaviour at a later age than 23–24 years, the expected decline may have been evident. As the ATP follows study members into later life, we will be able to test this supposition in future research. The relative stability found in cohort-wide rates of risky driving emphasises the need for road safety measures to continue or for new initiatives to be developed that target young people beyond the first years of their driving careers. Given the widespread prevalence of risky driving found among these young Australian drivers in their mid-twenties, broad-based initiatives targeting this age group (e.g. community campaigns, police and enforcement efforts) would appear beneficial. In particular, initiatives that are designed to discourage drink driving, driving while fatigued, speeding and mobile phone use while driving would be helpful given the relatively high prevalence of these behaviours exhibited by this group. 4.2. Stability of risky driving tendencies among individuals Turning now to the stability of an individual’s propensity to engage in risky driving, an important finding emerging from this research was that the majority of those who did not engage in risky driving as probationary licence holders continued to refrain from unsafe driving practices after moving to an unrestricted licence. These findings suggest that avoidance of risky driving during the early years of a person’s driving career is likely to persist and be linked to safe driving practices at a later age. Another encouraging finding was that high- or moderate-level risky driving tendencies had diminished for many. These findings suggest that young problem drivers are not destined to continue posing a high road safety risk as they grow older. Rather, it seems that improvement is not only possible, but also relatively common. Nevertheless, improvement does not mean cessation of risky driving, as only one-quarter of high risky drivers had become low risky drivers four years on, as had approximately half the moderate risky drivers. Also of concern, close to a quarter of low risky drivers, and about one-in-ten moderate risky drivers became riskier drivers over the study period. While high risky drivers would appear to present the highest and most immediate risk, moderate risky drivers should not be overlooked, as they also take risks on the road, and are a numerically larger group than high risky drivers. Moderate risky drivers may also be more amenable to change than high risky drivers. Hence, achieving a reduction in the risky driving behaviour of this group may result in greater road safety gains. Hence, interventions and road safety policy aimed at both groups would appear valuable. It is possible that a combination of approaches – targeted approaches aimed at the small group of frequent risky drivers and broad-based interventions aimed at the larger group of moderate risky drivers – may be most effective. Clearly, further research is needed to compare the efficacy of any new or existing intervention and policy approaches on these two different groups. Together, these findings provide some support for our hypothesis that there would be a reduction in risky driving among individuals who exhibited high- or moderate-risky driving tendencies at age 19–20. However, they also suggest that a small proportion of young drivers become less safe with increasing experience on the road. Consequently, while many individuals exhibited changes in their risky driving propensities over time, the net effect

S. Vassallo et al. / Accident Analysis and Prevention 72 (2014) 161–168

of these changes was that the overall risk profile for the sample remained fairly stable at both timepoints. These findings underscore the importance of examining stability at an individual, as well as group, level. 4.3. Strengths and limitations One strength of this study was that it utilised data from a largescale longitudinal study to examine across-time trends in risky driving behaviour, at both an individual and cohort-wide level. Furthermore, rather than focusing on a specific type of risky driving (e.g., speeding, or drink-driving), this study assessed a wide range of risky driving behaviours including speeding, failure to wear a seatbelt, fatigued driving, and drink-driving. Nevertheless, the study also had several limitations. For instance, as previously noted, selective attrition between the two survey waves meant that there were proportionately fewer high risky drivers available for analyses of across-time trends than moderate- or low risky drivers. Hence, the rates of risky driving reported in this study are likely to be an underestimate of rates of risky driving amongst Australian drivers in their mid-20s. We also acknowledge that the inclusion of a larger set of risky driving items in our cluster analysis at 23–24 years than at 19–20 years might have introduced a small amount of measurement error, but deemed it more important to use the most robust measure possible. Reliability analysis showed the broader 14-item scale to have higher internal consistency than the 7-item scale4 (Cronbach alpha of .73 compared with .71). Furthermore, comparison of the cluster solutions resulting from the use of 7 versus 14 items generally revealed a high degree of consistency in the manner in which participants were classified.5 Also, we had hoped to compare those individuals who had reduced their engagement in risky driving, with those who had remained stable or increased their risk profile to explore whether particular characteristics or experiences may have been instrumental in facilitating change. However the small size of some of these across-time stability groups (i.e. increasing low to high: n = 10; decreasing high to low: n = 11; stable high: n = 13) precluded such comparisons. Additionally, while there were suggestions of gender differences, with a higher proportion of males being classified as ‘high risky drivers’ at both 19–20 and 23–24, and fewer males than females classified as ‘low risky drivers’ at both timepoints, small cell sizes prevented us from examining whether the across-time groups differed on gender composition. Hence, further research, employing larger samples would be desirable to confirm the existence of distinct groups with different across-time patterns of risky driving, and identify factors that distinguish groups with differing risky driving trajectories.

167

failure to wear a seatbelt for part of a trip also rose, with minimal change observed on the remaining indicators. Furthermore, while an improvement was evident in the subsequent driving careers of many individuals who had been moderate or high risky drivers at 19–20 years, risky driving tendencies were still evident among the majority of initially high young risky drivers and about half of the initially moderate risky drivers at 23–24 years. Overall, these findings suggest a fair degree of stability in risky driving from late adolescence to early adulthood among this sample of Australian youth, suggesting a continuing need for road safety measures targeting drivers beyond their first years of driving. Disclaimer The views expressed in this paper are those of the authors and may not reflect those of the collaborating organisations involved in this research – the Australian Institute of Family Studies, the Royal Automobile Club of Victoria, the Transport Accident Commission of Victoria and the Australian and Victorian Governments. Acknowledgements The role of the Royal Automobile Club of Victoria and the Transport Accident Commission of Victoria in funding this research is gratefully acknowledged. We would also like to especially thank Thanuja Gunatillake for the important role that she played in this research. This paper is dedicated to our co-author, Warren Harrison, who sadly passed away before this paper was published. With over 20 years experience in road safety research, Warren was an intelligent and passionate researcher, whose insightful comments helped shape this research, and our understanding of the study findings. We would also like to acknowledge other ATP Principal Investigators – Professors Ann Sanson, John Toumbourou, Margot Prior, Frank Oberklaid and Associate Professor Craig Olsson – for the contributions they have made to the ATP, and Dr Ben Edwards and Dr Daryl Higgins for the helpful comments they provided on a previous draft. Finally we would like to sincerely thank the young people and their parents who have participated in the ATP, without whom this study would not have been possible. The ATP is a collaborative project between the Australian Institute of Family Studies, the University of Melbourne, Deakin University and the Melbourne Royal Children’s Hospital. Further information about the study is available from the ATP website (http://www.aifs.gov.au/atp).

5. Conclusion

References

In conclusion, cohort-wide statistics suggested that a small decline in speeding occurred among this sample of young Australian drivers between the ages of 19–20 and 23–24 years. However, there was a sizeable increase in drink-driving and

Arnett, J.J., 1991. Still crazy after all these years: reckless behaviour among young adults aged 23–27. Pers. Indiv. Differ. 12, 1305–1313, http://dx.doi.org/10.1016/0191-8869(91)90205-P. Arnett, J.J., 2000. Emerging adulthood. A theory of development from the late teens through the twenties. Am. Psychol. 55, 469–480, http://dx.doi.org/10.1037/0003-066X.55.5.469. Begg, D., Langley, J., 2001. Changes in risky driving behavior from age 21 to 26 years. J. Saf. Res. 32, 491–499, http://dx.doi.org/10.1016/S0022-4375(01)00059-7. Begg, D.J., Langley, J.D., 2004. Identifying predictors of persistent non-alcohol or drug-related risky driving behaviours among a cohort of young adults. Accid. Anal. Prev. 36, 1067–1071, http://dx.doi.org/10.1016/j.aap.2004.03.001. Bingham, C.R., Shope, J.T., Zakrajsek, J., Raghunathan, T.E., 2008. Problem driving behaviour and psychosocial maturation in young adulthood. Accid. Anal. Prev. 40, 1758–1764, http://dx.doi.org/10.1016/j.aap.2008.06.014. Cavallo, A., Triggs, T.J., 1996. Directions for Improving Young Driver Safety Within Victoria: A Discussion Paper. Monash University Accident Research Centre, Melbourne. Clarke, D.D., Ward, P., Truman, W., 2002. In-depth accident causation study of young drivers (TRL Report TRL542). Transport Research Laboratory, Berkshire, UK.

4 The item “driven when affected by an illegal drug” (assessed at 19–20 years) was replaced by three items at 23–24 years (“driven when affected by marijuana”, “driven when affected by ecstasy”, “driven when affected by amphetamines”). Hence, there were only 7-items that were common at both timepoints. 5 99% of those subsequently labelled ‘low risky drivers’ in the 14-item analysis, and 88% labelled ‘moderate risky drivers’ were classified identically in the 7-item analysis. A higher number of participants were allocated to the ‘high risky driving’ cluster in the 14-item analysis (n = 57) than the 7-item analysis (n = 21). It is likely that the broader scope of the 14-item scale made it more sensitive to detecting high risky driving behaviour.

168

S. Vassallo et al. / Accident Analysis and Prevention 72 (2014) 161–168

Dobbie, K., 2002. An analysis of fatigue-related crashes on Australian roads using an operational definition of fatigue (Report OR23). ACT: Australian Transport Safety Bureau. Engström, I., Gregersen, N.P., Hernetkoski, K., Keskinen, E., Nyberg, A., 2003. Young Novice Drivers, Driver Education and Training: Literature Review. Swedish National Road and Transport Research Institute, Linköping. Giedd, J.N., 2004. Structural magnetic resonance imaging of the adolescent brain. Ann. N.Y. Acad. Sci. 1021, 77–85, http://dx.doi.org/10.1196/annals.1308.009. Harrison, W.A., Senserrick, T.M., Tingvill, C., 2000. Development and trial of a method to investigate the acceptability of seat belt reminder systems (Report 170), Retrieved from the Monash University Accident Research Centre website: http://www.monash.edu.au/miri/research/reports/muarc170.html Jessor, R., Turbin, M.S., Costa, F.M., 1997. Predicting developmental change in risky driving: the transition to young adulthood. Appl. Dev. Sci. 1, 4–16, http://dx.doi.org/10.1207/s1532480xads0101 2. Lee, S.E., Simons-Morton, B.G., Klauer, S.E., Ouimet, M.C., Dingus, T.A., 2011. Naturalistic assessment of novice teenage crash experience. Accid. Anal. Prev. 43, 1472–1479, http://dx.doi.org/10.1016/j.aap.2011.02.026. Organisation for Economic Co-operation and Development (OECD), 2006. Young Drivers: The Road to Safety. Paris, France, OECD Publications. Oxley, J., Langford, J., Palamara, P., Muir, C., Koppel, S., Bohensky, M., Williamson, A., 2009. Non-wearing of Adult Seat Belts in Australia: Where to Next? (APR346/09). Austroads Pty Ltd, Sydney, Australia. Prior, M., Sanson, A., Smart, D., Oberklaid, F., 2000. Pathways from Infancy to Adolescence (Research Report No. 4). Australian Institute of Family Studies, Melbourne.

Senserrick, T.M., 2003. Graduation from a zero to.05 BAC restriction in an Australian Graduated Licencing System: a difficult transition for young drivers? Annu. Proc. Assoc. Adv. Automot. Med. 47, 215–231. Sowell, E.R., Thompson, P.M., Holmes, C.J., Jernigan, T.L., Toga, A.W., 1999. In vivo evidence for post-adolescent brain maturation in frontal and striatal regions. Nat. Neurosci. 2, 859–861, http://dx.doi.org/10.1038/13154. Transport Accident Commission, 2013. TAC Road Safety Statistical SumRetrived from mary, http://www.tac.vic.gov.au/ data/assets/pdf file/ 0012/51051/RSSS Jul13.pdf Triggs, T.J., Smith, K.B., 1996. Young Driver Research Program: Digest of Reports and Principal Findings of Research. Australian Government Publishing Service, Canberra. Vassallo, S., Smart, D., Sanson, A., Harrison, W., Harris, A., Cockfield, S., McIntyre, A., 2007. Risky Driving Among Young Australian drivers: Trends, Precursors and Correlates. Accid. Anal. Prev. 39, 444–458, http://dx.doi.org/10. 1016/j.aap.2006.04.011. Vassallo, S., Sanson, A. (Eds.), 2013. The Australian Temperament Project: The First 30 years, Australian Institute of Family Studies, Melbourne. VicRoads, Discussion paper 2005. Young Driver Safety and Graduated Licensing. Melbourne, VicRoads. Ward, J.H., 1963. Hierarchical grouping to optimize an objective function. J. Am. Stat. Assoc. 58, 236–244, http://dx.doi.org/10.1080/01621459.1963.10500845. Williams, A.F., 1998. Risky driving behaviour among adolescents. In: Jessor, R. (Ed.), New Perspectives on Adolescent Risk Behaviour. Cambridge University Press, New York, pp. 221–237.