Deaths and injuries arising from road crashes are a major problem in Australia, as throughout the world. It is well recognised that young people represent a high-risk group within the road system (International Transport Forum-OECD, 2019). In consequence many jurisdictions, including New South Wales (NSW), Australia, have introduced a graduated driver licensing system (GDLS). Young novice drivers undergo a specified number of hours (120 in NSW) learning to drive under supervision, before passing into one or two provisional driving phases (two in NSW, P1, P2) prior to full licensure. During these provisional phases, young novice drivers are unsupervised but are subject to several restrictions affecting, amongst others, limits in the carriage of young passengers at night-time in the first year, all phone use and alcohol consumption (Senserrick & Williams, 2015).
Freed from the restraints of supervision after obtaining a provisional driving permit, there is a noticeable increase in the frequency with which young drivers are involved in road crashes that gradually decline over the following six months (Ma et al., 2013; McCartt et al., 2003; Senserrick et al., 2021). Over the months of initial independent driving, young novice drivers gain a broad range of driving experiences and, due to age related development, their capacity for self-regulation improves, enabling better decision-making (Johnson & Jones, 2011; Watson-Brown et al., 2021). As such, deliberate risky driving behaviours are known to reduce with age, particularly for young male drivers (Rolison & Moutari, 2020). Moreover, research has shown that for each yearly increase in age, crash risk declines (McCartt et al., 2009).
Gender is a known contributing factor to crashes with males of all ages, but particularly for young and newly licensed drivers who are at greater risk. Young novice male drivers have a tendency for a lower risk perception compared to females, consequently engaging in more risky driving behaviours and being over-represented in road crashes (Cordellieri et al., 2016; Møller & Janstrup, 2021). Young drivers with high offending rates, particularly speeding, running a red light and seat belt infringements, are more likely to be involved in a fatal or serious injury crash (Meyer et al., 2021). Recency, seriousness of an offence, and frequency of offences have also been found to increase the likelihood of a crash (Factor, 2014; Walter & Studdert, 2015). Furthermore, young drivers in rural locations tend to perceive speeding, amongst other risky behaviours, as the norm and, coincidently, there are more crashes, particularly for males, in rural areas compared to metropolitan locations (Cullen et al., 2021; Knight et al., 2012).
There is a considerable body of research on young driver crashes and contributing factors in the early months and years of driving. Example results from the New Zealand Drive Study suggest that time spent in the restricted phase of the DDLS, alcohol use and recency of migration to New Zealand are risk factors for crashing and may also affect timing of crash (Begg et al., 2017; Boufous et al., 2023; Lewis-Evans, 2010). There seems however to be little specific research into whether young drivers involved in a traffic crash during the first six months differ from those who crash later or have no recorded crashes in the first few years. Using data obtained from the NSW transport authority, this question has been explored in a cohort of young novice drivers, all under 25 years of age at first learner licence, who obtained the learner licence in the financial year, 2007-2008.
Personal and demographic data and information on traffic offences and crashes were supplied in three files by the NSW transport authority for all persons who were in the driver licensing system, that is, persons who were in at least one of the learner or the two provisional phases, between 1 July 2007 and 31 January 2014, inclusive.
Data included in each file were:
personal, licensing and demographic data including: gender, date of birth, dates of transition to the various phases and postcode at first licencing.
offence information, including date and offence code
police notified crashes with their dates, severity (fatal, serious (hospitalisation required) or other), at-fault determination and crash type according to Road User Movement (RUM) codes (Transport for NSW, 2019).
Data were linked using a unique identifying number, common to all three files, specially generated by the providers for this study. A combined file of all learners first licensed in the year from 1 July 2007 to 30 June 2008 was created from the three files. Only one offence, the most serious according to a hierarchy to be described below, was selected per offence date, up to a maximum of six dates. Only 2.2 percent of novices had more than six offence dates. All crashes were included in the final combined file, four being the maximum recorded. Scrutiny of the file revealed no missing information apart from a small number of missing postcodes.
Data inclusions and exclusions
Only novices under age 25 years when first licensed and who had transitioned to unsupervised driving were included in the analysis. Three crash categories were created:
early crashers, novice drivers who had a recorded crash within the first six months of entering the P1 phase
later crashers, novice drivers who had a recorded crash only after six months since transitioning to P1
non-crashers, novice drivers who did not have a recorded crash in the observation period. Non-crashers were included for completeness and to provide context
Data for learners aged under 16 years and records with missing postcodes (< 0.01%) or postcodes outside of mainland NSW (0.04%) were excluded. Also excluded were data for individuals with atypical progressions through the system such as omitting one or more phases or progressing too fast through a phase. The latter, comprising 4.8 percent of the cohort, were taken to be a system artefact associated with persons who had reasons for special consideration (such as prior driving experience in another state) or due to transition problems in the first year of the changes to the GDLS. A more detailed description of the final file and its preparation is available in Siskind, Faulks and Sheehan (2019).
Offences were categorised by seriousness into eight hierarchical classes: dangerous, reckless or negligent driving; drink or drug driving; driving while suspended or disqualified; speeding; failure to obey traffic signs and signals; usage offences (driving offences other than those specifically categorised); failure to comply with GDLS rules; and administrative offences (Corbett, 2010; Siskind et al., 2019). These were assigned by police officers attending the crash scene. Dangerous/negligent driving, drink or drug driving and speeding (three of the first four categories) were grouped into a further category of serious offences for the purposes of this analysis.
Statistical analysis was by contingency table analysis for discrete variables with the Mantel-Hansen test for ordered variables. The Kruskal-Wallis non-parametric analysis of variance was employed for continuous or quantitative variables. Apart from the analysis of proportions of males and females, all comparisons were sex-specific. Comparisons were firstly between the early and later crashers and secondly between crashers and non-crashers. Because of the large sample sizes, differences of modest statistical significance are to be treated with caution.
Variables considered for analysis were: sex; age at first learner’s licence (age at L) and at commencing unsupervised driving (age at P1); time spent under supervision (time in L); numbers of crashes and of offences post P1 in various categories, and category of residence as defined by postcode at first licencing. Postcodes were grouped into ‘major cities’ or ‘rest of state’, as determined by the Accessibility/Remoteness Index of Australia classification (Australian Institute of Health and Welfare, 2004) and by quintile of the Socio-Economic Index of Relative Disadvantage as derived from the 2006 Census (Australian Bureau of Statistics, 2008).
To study the evolution of crash types over the period following entry into the P1 phase, the distribution of individual RUM codes in the first and second three-month period and after six months of unsupervised driving was derived. The proportions of single-vehicle crashes in these intervals were also calculated. The relative proportion of crashes occurring on the same dates as a traffic offence, which in most cases can be presumed to refer to the same event, was also determined.
Table 1 lists the results for all variables other than traffic offence and crash data by early, later and non-crashers. There were statistically significant differences among the subgroups on all variables. The largest differences among these variables were between novice drivers who experienced a crash and those who did not, with smaller differences between early and later crashers, of which few were statistically significant.
Non-crashers were proportionately more female and, among both sex groups, were less likely to live in a major city at learner licensure, older at learner licensure and at transition to P1 and to have spent more time in the learner phase. These novice drivers were also somewhat less likely to have resided in a postcode classified as of greater socioeconomic disadvantage at entry into the licensing system.
Between early and later crashers there was little difference in the proportion of females, age at learner licensing, major city or rest of state residence at learner licensing and socioeconomic classification of postcode residence at learner licensing. There was also little difference in age at first attaining P1 among females at only 5 percent significance, and a small if statistically significant difference among males (p < 0.001). Overall, in this sample, females were 2.6 months older than males at L and six months older at P1, differences which are statistically highly significant due in part to the very large numbers involved. Early crashers spent more time in the learner phase.
Frequency of traffic offences as a driver in the post-learner phase are summarised by the three subgroups in Table 2, overall and by the most serious. Mean numbers with standard errors (s.e.) of all and serious offences are presented in Table 3 for early crashers, later crashers, these two combined and non-crashers. Whether a crash occurred on the same day as a traffic offence is also reported in Table 4.
Non-crashers had far fewer reported traffic offences, including serious offences, than early or later crashers after the learner phase. Later crashers had noticeably fewer both traffic offences and total and at-fault crashes than early crashers (Tables 2, 3, 4) in this phase. However, excluding offences associated with a crash reduced the difference between early and later crashes in number of both all and serious offences (Table 3). The difference between crashers and non-crashers is also reduced but remains highly significant.
Multivariate analysis by logistic regression of early versus later crashers and of all crashers versus non-crashers confirmed all these results. There was little difference between early and later crashers in numbers of serious crashes, involving injury or death, while there were only a handful of fatal crashes in both groups (Table 4). All differences noted here were statistically significant to a high degree partly as a consequence of the large numbers involved.
As remarked previously, it is plausible to assume that the majority of crashes and traffic offences that occurred on the same date referred to the same event. Early crashers (63.5%) had a significantly higher proportion of such crashes compared to later crashers (52.7%). According to the police in attendance, the young driver was at-fault in almost all crashes (over 99%). The principal offences associated with these crashes were: dangerous/negligent driving (86%); failure to obey signs or signals (10%); and, drink/drug driving (3%). As noted above, only the most serious offence on a given date was included in the data file, so for example, speeding would not be reported here if dangerous or negligent driving was also cited.
Given the apparent relationship between time of crash and whether the crash was associated with an offence, it might be expected that the most common traffic offence reported among early crashers in the first six months after commencing unsupervised driving is dangerous or negligent driving (64% of total offences). This offence category was relatively uncommon in late crashers and non-crashers at about 8 percent each. Bearing in mind the caveat above about multiple offences, failure to give way (11.8%) and speeding (11.1%) were noted to be the next most common offences in the early crasher group. Among the other two groups the most common category of offence within six months of P1was speeding at over 40 percent of total offences. The GDLS offences in these two groups formed at least 20 percent of total offences in this period although this is likely to be an under-estimate. In the early crasher group, if the dangerous/negligent driving category is excluded, speeding offences and offences involving failure to observe traffic signs and signals then make up 30 percent of traffic offences in the first six months post P1 and GDLS offences at least 20 percent.
Beyond six months after transition to the P1 phase the distribution of offence types is different, with speeding by far the largest category at close to half of early crashers and non-crashers. In late crashers the category, dangerous/negligent driving, comprised 27.5 percent of offences; if this category is excluded, speeding offences also make up almost half of the total.
In all three groups drink/drug and improperly licensed driving are reported relatively uncommonly in each period, each making up 2-3 percent of total offences. Examination of a sample of drivers in this cohort suggests that the proportion under-reported may be of the order of 5 percent. From the same sample it appears that the proportion of speeding offences under-reported is even less. Serious offences are less subject to under-reporting.
In terms of the types of first post-P1 crash experienced by these young drivers, there is very little difference in the distribution of the most common RUM codes between the novice drivers who crashed within three months of reaching P1, crashed in the next three months or had their first crash later (Table 5). All other RUM codes had frequencies of 2 percent or less. Nor did the proportion of single vehicle crashes differ much in these three intervals (≤ 3 months, 27.0%; 4 - 6 months, 26.0%, ≥ 6 months, 27.1%).
Young drivers who experienced crashes in the first few months after their period of supervised driving differed in several respects from young drivers who did not crash. The young drivers who crashed are somewhat younger both at first licensing and on reaching the provisional phase, more often male and spent less time driving under supervision. In addition, the young drivers who crashed tended to be from major cities and postcodes classified as more socio-economically disadvantaged. On the other hand, there were fewer differences in these respects between the young novice drivers who crashed within the first six months of unsupervised driving and novices who crashed later. Two characteristics of note were later crashers spent less time driving under supervision and were slightly younger.
In respect of traffic offence histories, here too the largest differences are between young drivers who crashed compared to young drivers who did not crash with smaller but statistically significant differences between early and later crashers. The former had on average more total offences and also more serious offences reported than the latter who in turn had many more than non-crashers, which was in large part due to the greater number of offences occurring on the same day as crashes.
The results for serious offences are consistent with a recent study using crash data from Victoria where repeat serious offenders were significantly more likely to be involved in a crash within the following three years (Meyer et al., 2021). In the first six month period after reaching P1, early crashers incurred many more citations for dangerous or negligent driving than later crashes or non-crashers, whose principal reported offence was speeding. After six months, speeding was the main traffic offence reported in the current datafile for all three groups, although in this period dangerous/negligent driving comprised over a quarter of offences in later crashers. Due to its low position in the offence hierarchy GDLS offences may be under-reported.
There were also notable differences between early and later crashers in their total, severity and at-fault crash experience, with the former having significantly more of all crash classes. Some 64 percent of the crashes of early crashes occurred on the same date as a traffic offence compared with 54 percent of later crashes. Almost all such crashes, police reported the young driver was at-fault.
The combination of factors that contribute to young driver crashes during the first six months of independent driving implicate age and, particularly, driving inexperience (McCartt et al., 2009). Research has found that the combination of road conditions, speed choice, inexperience, and loss of control account for a large proportion of variance in young drivers’ single vehicle crashes (Rolison & Moutari, 2020). The first six months reflects the most dramatic decline in crashes with single vehicle, night-time and weekend crashes, and crashes when carrying two or more passengers having the greatest decreases for young drivers (Mayhew et al., 2003). Furthermore, in the first six months of independent driving the crash risk of youths who drive more frequently decreases, with one study finding the pivotal amount of driving at 73 hours (Li et al., 2017). Given the average amount of time a young driver typically drives during their first six months of independent licensure, 73 hours is approximately aligned with five to six months (Li et al., 2017; Mayhew et al., 2003). These findings will become more relevant when there is information on the personal information and on-road exposure of young drivers who crash while in the GDL system from studies suggested below.
There is an extensive literature on reasons for and circumstances of young driver crashes including studies of cognitive and psychological factors. One study of interest used an American database containing detailed analyses of a representative sample of serious crashes, classified the crashes of teenage drivers (under 19 years of age). The authors found that errors of judgement or inadequate attention led to the great majority of instances where the teenaged drivers was judged at fault, with aggressive driving being uncommon at about 3 percent (Curry et al., 2011). However, the timing of crashes relative to the commencement of unsupervised driving was not considered. Cooper and his colleagues (1995, p. 104) suggested that “those new drivers whose attitudes, access to vehicles and reasons for driving expose them to an excessive level of risk” may crash sooner rather than later but they do not further define attitudes. McCartt and colleagues (2009, p. 209) stated that “the weight of evidence suggests a steep learning curve among drivers of all ages”. One would therefore expect there to be some change over the first months of unsupervised driving in the types of crashes experienced by these novices although this is not observed in this study.
Only crashes reported to police are included in the database and there may be some degree of under-reporting. Since reporting of property damage only crashes requires that the vehicle(s) be towed away, under-reporting is unlikely to be common enough to influence the results and is not likely to occur more often in one crash group than another. Similarly crashes which occur in another state, which would not appear in these data, are likely to be equally common in both groups.
Routinely collected data such as these have limited explanatory value concerning the social and cognitive factors underlying the observed differences, in particular those between early and late crashers. Also unavailable is information on on-road exposure, a common deficiency in data collected for administrative purposes. It has been assumed that drivers in the three crash classes drive on average similar amounts, although it would be of interest to know to what extent this is true.
The principal aim of this study was to determine if there were ways in which young novice drivers who crash within six months of entering the provisional, unsupervised phase of the GDL system differed from those who crashed after six months. The early crashers spent about two months longer in the supervised phase had somewhat more traffic offences and crashes (both total and serious crashes) with crashes judged by attending police to be more likely at fault. In addition, the crashes of early crashers were more often associated with a traffic offence.
The study also explored the differences between young novice drivers who do not crash after commencing to drive unsupervised, at least during the first few years and the novices who crashed. The groups were found to differ in several respects. The novices who did not crash within the observation period were more likely to be female, older at first licensing and at starting to drive unsupervised, to have spent longer in the supervised phase and to be residents of major cities and suburbs of higher socio-economic status at first licensure. They also had far fewer total traffic offences and serious offences.
In order for these results to be of more practical value there needs to be a study of the attitudes and personality traits of young drivers who crash early or later as defined in the current study. This could readily be carried out in a large database, if such exists, containing information on attitudes, personality and exposure collected by questionnaire or interview at or near the start of the learner phase, plus driver histories over at least four years.
The data used in this study were provided by Transport for New South Wales, to whom I am grateful. Dr Ian Faulks was instrumental in securing the data.
Victor Siskind planned and executed the analysis, and also compiled and edited the manuscript.
No funding was provided for this research.
Human Research Ethics Review
The Queensland University of Technology (QUT) Human Research Ethics Committee assessed this research as meeting the conditions for exemption from HREC review in November 2013 and approval in accordance with section 5.1.22 of the National Statement on Ethical Conduct in Human Research (2007) (Exemption number: 1400001000; Exemption number: 1400001002; Exemption number: 1600000204).
Data availability statement
Data were provided by the New South Wales Government, Transport for New South Wales on the understanding that they not be shared with third parties.
Conflicts of interest
The author declares that there are no conflicts of interest.