Predicting Young Drivers’ Time-to-Licensure from Sociodemographic Characteristics and Quality of Adult-Supervised Practice

Abstract

Many jurisdictions now require adult-supervised practice for learner drivers. For many younger drivers this entails supervision from parents. There is not consensus on how to best optimize this supervision or which factors influence how quickly young drivers progress through the learner period to obtain an independent license. Additionally, young male drivers are over-represented in fatal and serious crashes and it is unknown if this risk has a basis in their learner driver experiences. Using data from the control arm of the Drivingly Trial (554 parent-teen dyads) we examined how sociodemographic factors and parent-teen driving practice behaviours contributed to how quickly teens were licensed and determined if there were sex differences in how practice was experienced by teens. Greater practice variety was associated with faster licensure. Parent engagement with practice supervision increased over the learner period and teens became more supportive of their parents’ supervision as their anticipated license date neared. Male and female teens did not differ with respect to their experience of supervised practice, pre-permit driving or time-to-licensure. White teens and teens from non-urban areas were licensed faster than other teens in our sample, all of whom began the study with a learner’s permit and intention to get licensed. Licensing confers risks and opportunities for young people. Ensuring young learner drivers practice in a range of different driving environments is important. More research is needed to determine how to reduce structural and social barriers to licensure without conferring an increased crash risk.

1.0. Introduction

Motor vehicle crashes (MVC) are a leading cause of death and disability to adolescents in the United States and around the world.^1,2 The learner period is the safest period for young drivers; crashes among learners are rare compared with other license stages.^3–7 MVC rates among newly licensed teenagers rapidly increase at the point of independent licensure, then decrease thereafter as experience is obtained.^4,6 In addition to inexperience, risk for a MVC following licensure can be affected by sex, age, and a variety of other factors that can moderate crash risk.^6,8–11 Common to all new drivers, however, is the practical issue that one cannot “practice” driving independently until the supervised period is over. To address the issue of young driver inexperience prior to licensure, various jurisdictions have passed laws and regulations mandating the number of adult-supervised practice hours needed, the length of time required to be spent as a learner or provisional driver, and if any formal educational or on-road training by a qualified driver educator is required prior to taking the behind-the-wheel license test. It is not a requirement that such supervision be undertaken by a parent, but it is common especially in jurisdictions with lower licensing ages like The United States (US).¹²

Self-report studies using convenience samples tend to show positive associations with the implementation of mandatory supervised practice requirements and hours of supervised practice reported.^12,13 Higher amounts of self-reported practice hours have been linked with fewer driver evaluator assessed risky driving maneuvers and driver errors during the permit period.^14,15 To our knowledge the only experimental study on this topic found that increasing the average amount of practice per week and the variety of practice overall (i.e., the number of different driving environments practiced in) reduced young drivers’ risk of making safety critical driving errors at the end of the permit period.¹⁶

Linkages between pre-license practice and post-license real-world outcomes are less clear. A naturalistic study of learner drivers found that greater consistency of practice had a small protective association with future MVC involvement as a licensed driver.¹⁷ In the US, Gilpin (2019) used administrative data sets and found that the implementation of Graduated Driver Licensing (GDL) supervised practice requirements increased fatal MVCs by 5.9–6.3%,¹⁸ while another US study of administrative data found no association between GDL requirements for practice supervision and fatal MVCs.¹⁹ Gilpin hypothesized that perhaps parents were not ready for the magnitude of the supervision task.¹⁸ In Australia, Kettlewell and Siminski (2020) analyzed the effect of GDL policy changes implemented in New South Wales on the MVC involvement of newly licensed young drivers and found that instituting a 50 hour practice requirement (from no requirement) reduced the probability of MVC involvement by 21% in the first year of licensed driving, but no effect was found when the policy changed to 120 required hours from the 50 hour requirement.²⁰

It is important to consider how aspects of the supervisory climate and parent-teen relationship affect practice driving and the licensure process, but there are very few studies on this topic. What we do know is that there is some evidence that aspects of the parent-teen relational context, such as support and trust, have been linked with more engaged and conscientious supervised practice.^21,22 Parents’ driving-related self-efficacy has also been associated with their provision of higher quality supervised practice for their learner teens.²³ Observational, cross-sectional self-report studies, and qualitative studies with parents and teens all indicate that parent supervised practice tends to be low in practical driving challenges and primarily consists of drives of convenience consistent with the family’s routine (e.g., travel to school, extra-curriculars, and errands).^12,24 Studies have documented lack of awareness of supervised practice requirements, which would naturally translate into difficulty with compliance.²⁵ Indeed, a naturalistic study found that most learner drivers did not actually meet the required number of supervised practice hours.²⁶ There is also some evidence that supervised practice interventions might be more effective for female teens than male teens in that an intervention designed to improve parental supervision during the learner period led to higher practice variety for female teens compared with male teens.²⁷ This finding could be due to parental perceptions and practices related to being more protective of their daughters (i.e., being less willing to expose them to safety risks like driving) compared to their sons, thus resulting in more capacity for daughters’ practice variety to increase with parent-directed intervention. This interpretation would be consistent with prior research that illustrates that parents socialize boys’ and girls’ safety behaviours differently.²⁸

With respect to obtaining a driver’s license, there is emerging evidence that both greater practice variety (i.e., the range of environments where practice occurs) is associated with faster licensure and that parents’ holding stronger perceptions of their adolescent’s driving skill relative to their peers is also associated with faster licensure.²⁷ These associations are sensible in that they likely reflect interest and aptitude on the part of the learner teen and support from their parent. Contextual factors such as living in a more urban environment or areas higher in social deprivation are both associated with comparatively slower time to licensure.^29–31 Compared to White teens, Latino and Black teens have slower licensing trajectories.²⁹ These contextual and sociodemographic factors are often interwoven with one another,³² but to our knowledge research has not yet been conducted that measured person-level urbanicity, race, and license status, along with practice and supervisory variables, in the same study. Teens with slower licensing trajectories cite practical reasons for not getting licensed including lack of a car, not needing a car, and the cost of vehicle ownership.³³ These licensing studies all pre-date the COVID-19 pandemic, so it is unclear how the pandemic might have affected adolescents’ plans to obtain a license. One study from Alabama (US) indicated a decrease in driving among licensed adolescents during the pandemic.³⁴

Finally, driving a vehicle prior to obtaining a permit has been shown to be indicative of future crash involvement in New Zealand.³⁵ In New South Wales, pre-permit driving has been linked with a variety of deleterious behaviours such as alcohol and other drug use while driving and higher engagement in risky driving behaviours.³⁶ These findings suggest that parents allowing, or being unaware of, pre-permit driving is a significant safety issue.

Thus, all available evidence suggests that it is important for young people and their practice supervisors to fully engage with the process of supervised driving, but that a variety of factors may influence the effectiveness of these “learner period” experiences for mitigating future crash risk for any given young person. The purpose of the current study was to use data from the control group of the Drivingly Randomized Controlled Trial to shed light on how families enter, experience, and exit the learner permit period. This information is important because these initial interactions and experiences can set the stage for parents’ and teens’ expectations for what their practice will be like together, potentially influencing their engagement and willingness to participate and invest throughout the learning-to-drive process, and consequently their preparation for independent licensure. To our knowledge no studies on this topic have been conducted with families since restrictions associated with the COVID-19 pandemic were lifted.

1.1. Current Study

Using data from the Drivingly Trial (NCT03639753),³⁷ we first describe the sample characteristics of parents and young learner drivers in the control group of the larger study and then determine if young learner drivers’ initial driving behaviours including variety of practice, practice driving-related social support, and parental engagement as a practice supervisor vary by teen sex. We then investigate how these factors, along with race and contextual variables (e.g., urbanicity), might contribute to the acquisition of a driver’s license. Analyses were conducted in an exploratory framework, so for example while we were interested in exploring sex differences in how practice supervision might be experienced by male and female teens owing to males being over-represented in fatal crashes,¹ we had no specific hypotheses relating to how their practice experiences might differ.

2.0. Methods

Drivingly is a comprehensive behavioral intervention for learner teen drivers and their parents. The purpose of the trial is to evaluate the effect of the Drivingly Program on young drivers’ risk of being in a MVC in the first 12 months of licensure in the Commonwealth of Pennsylvania, US. In Pennsylvania, learner drivers must be at least 16 years of age when they apply for the supervised learner’s permit, hold this permit for at least 6 months, and their parent/guardian must certify the learner completed at least 65 hours of supervised behind-the-wheel practice driving, and 10 of these hours must be at night and 5 need to be in inclement weather. The permit is valid for 1 year, after which it can be renewed.

Eligible adolescents were between 16–17.33 years age and held a learner’s permit (Pennsylvania) at the point of enrolment, which they intended have for 6 months or more; and they practiced no more than 10 hours, and had access to the internet, a practice vehicle, and at least one parent/caregiver willing to supervise them. Adolescents’ enrolled parent had to be 18 years or older, have internet access, and be a licensed driver. Adolescents with visual, medical, or physical impairments that would require a handicap placard or assistive device to drive, or adolescents with pervasive developmental delays were ineligible to participate. Similarly, non-fluency in written or spoken English (either parents or adolescents), having a sibling already enrolled in the study, being enrolled in other driving studies, and being unable to access intervention components (e.g., no internet and no practice vehicle) also conferred ineligibility. All parent participants provided informed consent and parental permission and adolescents provided informed consent. Inquiries about accessing the study data can be directed to the corresponding author.

Drivingly enrollment began on 8/18/2021 and concluded on 12/15/2023 with 1,108 parent-dyads enrolled. Parents and teens completed a set of surveys at enrollment, three months later, and again at licensure. The baseline survey window opened at randomization and closed 4 weeks later. The 3-month “learner survey” window opened 91 days after randomization and closed 4 weeks later. The post-license survey opened at the point the study team learned that the teen had obtained their driver’s license. Families were contacted at regular intervals close to their projected license date. When the team learned the teen was licensed, they validated the license date visually by inspecting a digital image of the license and sent the teen and parent their post-license surveys. They each had 4 weeks to complete the post-license survey from this point. Every effort was made to open this survey completion window as close to the real license date as possible. More detail about the trial protocol can be found in Hafetz et al. 2023.³⁷

To maintain integrity of the intention-to-treat analysis plan for the primary outcome of interest, outside of the sociodemographic characteristics we report only results from the control group (554 dyads), which was a usual care group, in the current analysis. Through October 31, 2024, of the 554 enrolled dyads, 548 (99%) teens and 551 (99%) parents completed the baseline survey (546 complete dyads), 540 (97%) teens and 543(98%) parents completed the 3-month learner survey (535 complete dyads), and 467 (84%) teens and 468 (84%) parents completed the licensure survey (458 complete dyads).

2.1. Measures

Variety of parent supervised practice was assessed using items adapted from a prior measure of Practice Variety.¹⁶ Respondents indicated yes/no if supervised practice occurred in 8 specific environments then selected the number of hours practiced using a closed-ended response format (less than 1 hour; 1–2 hours; 3–5 hours; 6–10 hours; more than 10 hours). Participants reported retrospectively over the past 3 months. Practice time cut-offs to indicate a reasonable amount of practice occurred in that environment were determined as follows: minimum of 1 hour in parking lots and on country roads and a minimum of 3 hours on residential roads, one-lane roads, two-lane roads, commercial roads, at night, and in bad weather. For each environment, participants were given a 0 if they did not meet the cut-point and 1 if they did. We then created a composite 8-level practice variety variable (Practice Variety) that ranged from 0 to 8 based on meeting these cut-points for each environment. This survey was completed by parents at all three time points and teens at the first two time points. The three-month survey was used for modeling as it was the last point that had both parent and teen report and the license survey may have captured some post-license driving. This approach is intentionally agnostic to exactly where families have practiced and for how long and is instead concerned with the range of environments where practice occurred.

Parent practice engagement was measured by 10 items on the Parental Engagement In Practice Supervision Scale (PEPSS¹⁶) which uses a five-point scale ranging from “never” (1) to “always” (5) (e.g., I decided what driving skills I wanted my teen to practice; I determined the driving route my teen used for each practice drive; I scheduled dates and times to practice driving with my teen; I gave my teen feedback about his/her driving skills). It produces an average total score from 1–5. This measure was completed by parents at all three time points. The average (across all three time points) α=for this sample was 0.73.

Parents and teens reported on perceived support received during the practice driving process using the Practice Driving Support Scale (PDSS)²¹, which has 4 items (e.g., Overall, I felt my parent/teen supported me in the practice driving process; My parent/teen helps to decide what to practice; I talked to my parent/teen about my fears and worries about driving; and I could easily communicate with my teen/parent about driving issues) using 5-point scale Never (1) to Always 5). It produces an average total score 1–5. This was a minor change from its original scoring, which was on a 7-point scale. Teens reported on parents and parents reported on teens at all three time points. The average (across all three time points) α=for this sample of parents was 0.70 for parents and α=for teens was 0.79.

At baseline, teen respondents also answered the following 4 questions: 1. Did you ever drive a car or truck, not including farm equipment, before getting your permit? Response choices are yes (if yes, how many hours?) or no; 2. Have you ever had a lesson with a professional driving instructor? Response choices are yes, (if yes, how many?), No, and Not Sure; 3. How many practice drives have you taken with a practice supervisor since you got a permit? Responses are collected in a write-in open text box; and 4. How much time have you spent on practice driving since you got a permit? Responses are collected in a write-in open text box.

Sociodemographic data (e.g., age, sex, race, ethnicity and zip code) were collected at enrollment. Zip code was used to identify participants who lived in Philadelphia County. Those living in Philadelphia were classified as urban and other participants as living in a non-urban location. The City of Philadelphia is one of the most populated cities in the United States.

2.3. Analysis Plan

Teen and parent characteristics and supervised practice driving patterns were summarized using appropriate statistical methods including median (interquartile range (IQR)) and means (standard deviation (SD)) for continuous variables, and frequency (percentage) for categorical variables. Comparisons by teen sociodemographic data were performed using Wilcoxon rank sum tests (for continuous variables) or Pearson chi-square tests (for categorical variables).

To create the practice variety score, the sum of the number of environments with practice exceeding specific thresholds (parking lots and country roads: at least 1 hour; all other environments: at least 3 hours) were calculated separately from teen and parent reports. T-tests were used to compare support, engagement, and practice driving variety scores by sex. Repeated measures mixed effect models with an autoregressive correlation structure were used to look at the difference in social support and parent engagement scores over supervisory period. For time-to-licensure analysis, time was calculated from permit date and initial analysis consisted of log-rank tests to examine differences in time-to-licensure by sociodemographic, practice variety scores at 3-months, and perceived support characteristics. In a model containing all predictors, multivariable Cox proportional hazards regression models were fitted separately for teen and parents to examine the combined effect of sociodemographic, practice variety score, and perceived support variables on time-to-licensure. Statistical significance level was set to 0.05 for all tests. Statistical analyses were performed using SAS version 9.4, and R version 4.2.3.³⁸

3.0. Results

Sociodemographic data for the sample are presented in Table 1. Control group teens were 51% female, 71% of the sample was White, 14% was Black or African American, and 7% described themselves as multiple races or another race, and 6% as Hispanic or Latino. Most parents in the control group were mothers (88%) with similar racial profiles as the teens. At enrolment, teens had a mean age of 16.33 years (SD, 0.31 years) and parents had a mean age of 48.62 years (SD, 5.93 years).

Table 1:

Drivingly Demographics for Control Participants

	Teens	Parents
Number of control participants enrolled	554	554
Permit time at enrollment (months), mean(SD)	1.71 (1.93)
Age at enrollment (years), mean(SD)	16.33 (0.31)	48.62 (5.93)
Sex
Male	269 (49%)	64 (12%)
Female	284 (51%)	490 (88%)
Unknown/not reported	1 (<1%)	0 (0%)
Race
White	396 (71%)	423 (76%)
Black or African American	80 (14%)	67 (12%)
More than one race	38 (7%)	13 (2%)
Other/unknown	40 (7%)	51 (9%)
Ethnicity
Hispanic or Latino	35 (6%)	26 (5%)
Not Hispanic or Latino	515 (93%)	525 (95%)
Unknown/not reported	4 (1%)	3 (1%)
Geographical Classification
Non-Urban	443 (82%)	443 (82%)
Urban	95 (18%)	95 (18%)
Relationship to teen
Mother		487 (88%)
Father		62 (11%)
Aunt/uncle (LG)		1 (<1%)
Brother/sister (LG)		0 (0%)
Grandparents (LG)		3 (1%)
Other (LG)		1 (<1%)
Missing		0 (0%)

Teens held their permit on average 1.71 months (SD, 1.93 months) at the point of study entry. At enrollment, 76 (14%) male teens reported they engaged in pre-permit driving compared with 60 (11%) female teens, p=0.06. Of the teens who drove prior to having a permit, male teens reported a median of 1 hour of pre-permit driving, with interquartile range (IQR) of (1, 3) compared with female teens who reported a median of 1 hour (IQR: 1, 2), p=0.15. Only 19 (3%) male teens and 24 (4%) female teens had any instruction with a driving instructor at study entry and they reported practicing a median of 3 total practice hours (IQR: 1, 5.5) with no differences by teen sex, p=0.75.

There were no meaningful or consistent differences by teen sex with respect to social support given or received as it related to supervised driving or parents’ engagement as a practice supervisors (Table 2). Parent self-reported engagement as a practice supervisor increased over the supervisory period (p<0.001). Teens reported receiving less practice driving social support from their parents over the supervisory period (p <0.001), while parents reported receiving more social support from their teens over the supervisory period (p = 0.04). At three-months post-baseline (roughly half-way through the permit period), teens reported a median practice variety score of 3, with an IQR of (1,5). Parents reported a median practice variety score of 3.5, with an IQR of (1,5) at three months. Practice variety could range from 0–8.

Table 2.

Parent Supervised Practice: Teen and Parent Report, Overall and By Sex

	Timepoint	Overall		Male		Female
Parent Engagement		n	Mean(SD)	n	Mean(SD)	n	Mean(SD)	p-value*
	Baseline	548	3.2 (0.7)	63	3.2 (0.7)	485	3.2 (0.7)	0.39
	3 months post baseline	541	3.4 (0.7)	60	3.3 (0.5)	481	3.4 (0.7)	0.30
	Post-licensure	467	3.7 (0.5)	56	3.6 (0.5)	411	3.7 (0.5)	0.05
Social Support
Teen	Baseline	547	4.4 (0.7)	267	4.3 (0.7)	280	4.5 (0.6)	0.03
	3 months post baseline	539	4.2 (0.7)	259	4.2 (0.8)	280	4.2 (0.7)	0.76
	Post-licensure	466	4.3 (0.8)	224	4.3 (0.8)	242	4.3 (0.8)	0.81
Parent	Baseline	548	3.9 (0.8)	64	3.8 (0.8)	484	3.9 (0.8)	0.30
	3 months post baseline	540	3.9 (0.7)	60	3.9 (0.5)	480	3.9 (0.7)	0.23
	Post-licensure	466	4.0 (0.7)	56	4 (0.5)	410	4.0 (0.7)	0.29
Practice Variety
Teen	Baseline	504	1.1 (1.1)	244	1.1 (1.2)	260	1.1 (1)	0.10
	3 months post baseline	520	3.4 (2.1)	246	3.4 (2.2)	274	3.3 (2.1)	0.65
Parent	Baseline	495	1.0 (1)	60	1 (0.9)	435	1.1 (1.1)	0.73
	3 months post baseline	520	3.5 (2.2)	58	3.4 (2.1)	462	3.5 (2.3)	0.89
	Post-licensure	468	5.6 (2.1)	56	5.8 (2.0)	412	5.6 (2.1)	0.65

Note: One teen did not report sex at birth and thus was excluded from analysis.

^*P-value from t-test comparing scores across sex. Parent Engagement = parent practice supervision scale, Social Support = parents reported on support received from teens and teens reported on support received from parents.

3.1. Licensure

Teens were licensed a median time of 7.87 months after obtaining their permit, with IQR of (6.33, 11.5) months. In initial unadjusted time-to-licensure analyses, we found that teens from urban areas had longer times until licensure (p<0.0001) but observed no differences by teen sex (p=0.77). We also observed different licensure timing patterns by race, with teens identifying as White obtaining licenses faster than other racial groups (p<0.0001). Individually, we did not see associations between licensure timing and tertiles of social support, reported by teens (p=0.21), whereas there was a significant association with the parent-reported measure of social support (i.e., how much the parent felt the teen was supporting the parent and the learning-to-drive process) (p=0.0075). Additionally, tertiles of parental engagement as practice supervisor were associated with licensure timing, with higher engagement scores corresponding to shorter time-to-licensure (p<0.0001). Practice variety scores at 3-months were associated with faster time-to-licensure, both the scores reported by teens (p<0.0001) and the variety scores reported by parents (p<0.0001).

To enable the joint evaluation of these factors with licensure timing, two models were evaluated: the first model evaluated the impact of parent reported practice-driving social and behavioral factors, along with demographic variables, on time-to-licensure. These variables included practice variety, social support related to supervised practice driving, and parental engagement as a supervisor. The second model evaluated the same variables but was based on teen reported data with the exception of parental engagement as a supervisor, which was not reported by teens.

Cox regression models using teen covariates indicated that when considered jointly, teen race, urbanicity, and practice variety all significantly contributed to time-to-licensure, with non-white race, urban residence, and lower practice variety scores all associated with longer time-to-licensure. For the parent covariate model, statistically significant parent covariates included parent sex, race, urbanicity, and practice variety. Specifically, longer time-to-licensure was associated with male parent supervisor/primary parent participant (father), urban residence, non-white race, and lower practice variety. Interestingly, after adjustment for demographic characteristics, parent engagement was no longer a significant predictor of time-to-licensure. Please see Table 3 for the hazard ratios and p-values associated with these models.

Table 3:

Predicting licensure based on parent and teen report: Results from the multivariable cox models

Characteristics	Teen-reported Cox Model			Parent-reported Cox Model
	HR	95% CI	P-value	HR	95% CI	P-value
Sex
Male	—	—		—	—
Female	1.02	0.85, 1.23	0.8	1.47	1.09, 1.99	0.011
Race
White	—	—		—	—
Black or African American	0.68	0.50, 0.93	0.016	0.56	0.40, 0.77	<0.001
More than 1 race	0.77	0.54, 1.12	0.2	0.69	0.37, 1.27	0.2
Other / Unknown	0.55	0.38, 0.80	0.002	0.60	0.42, 0.83	0.003
Geographical classification
Non-Urban	—	—		—	—
Urban	0.61	0.46, 0.81	<0.001	0.70	0.53, 0.92	0.011
Practice variety score at 3 months	1.20	1.15, 1.26	<0.001	1.23	1.17, 1.28	<0.001
Social Support tertile
Tertile 1	—	—		—	—
Tertile 2	1.16	0.94, 1.43	0.2	0.94	0.74, 1.18	0.6
Tertile 3	1.17	0.92, 1.48	0.2	1.05	0.83, 1.33	0.7
Parent Engagement tertile1
Tertile 1	—	—	—	—	—
Tertile 2	—	—	—	0.96	0.75, 1.22	0.7
Tertile 3	—	—	—	1.29	1.00, 1.67	0.048

Note: Parent Engagement = parent practice supervision scale, Social Support = parents reported on support received from teens and teens reported on support received from parents. This table illustrates the results from two statistical models, the first column presents the results from the teen model that used the teen reported data and the second from the parent model that used the parent reported data. Sex refers to male teens / fathers or female teens mothers depending on the column.

4.0. Discussion

4.1. Family relationships and driver licensing

Families reported a generally positive experience with respect to engagement, social support, and supervision during the learner period. Families seemed to grow into the supervisory experience with parent engagement increasing over the learner period along with parents’ perceptions of teen support, while teen perception of parent support decreased. This “exchange of support” likely reflects a developmentally appropriate reversal of the teen taking on more responsibility and leading on practice and license activities as they gain experience and the practical test gets nearer. Average levels of parent and teen-reported support and parent engagement as practice supervisors were over the mid-point of the 5-point scales on which these constructs were assessed. We did not assess whether the teen in our study was the first teen in the family that learned to drive, so some parents might have more experience and comfort with the process than others – although that might also be contingent on how well the first experience went.

4.2. Disparities in driver licensing

Consistent with research conducted prior to the pandemic and with population-level data, Black or African American teens and teens living in urban environments had slower licensing trajectories than White teens and teens from non-urban environments in our study sample. In the composite models when race and urbanicity were considered together, each variable retained its unique predictive power associated with slower licensing. This is an important finding because every teen in our study had a permit when they enrolled, indicative of their intention to become licensed. Other studies that have reported similar findings often relied on administrative data that did not assess if the teen was ever permitted,³⁰ or have permit information but not person-level information about race.³⁹ Thus, these results are unique in that we have person-level information on both characteristics, along with urbanicity, and analyzed these characters together in conjunction with social and behavioral factors (e.g., parent engagement). These results suggest that more research is required to understand barriers and facilitators to licensure, along with goals of licensure for these populations.

4.3. Parent supervision and engagement

Parents who reported being more engaged supervisors had teens who progressed to the license faster than those parents who reported less engagement as a practice supervisor, but this effect disappeared when the contribution of the covariates were considered. So, while these parents may be more goal-directed in general, as well as have more time available to support their teen obtain a license, this mattered less when other social and contextual factors were accounted for. Parent perceptions concerning how much support their teen provided to the supervisory dynamic was also tenuously associated with faster licensure, but this association was not as strong or reliable. We found an association with being supervised by a mother and faster licensure. It may be that in our sample of parents, father-involvement (i.e., the father was the parent who consented and enrolled as the primary supervisor in the study) is a proxy for a two-parent household. Prior research has found teens from two-parent households practice more hours than teens from single-parent households,¹² which potentially may lead families to push back licensing dates. Importantly, we had very few fathers in the study overall so drawing strong inferences from this association is not recommended and additional research is needed to explore this association further.

US states vary considerably with respect to how much say parents have over their teen’s licensure journey. For example, the Special Minor’s Restricted License in Iowa allows teenagers at least 14.5 years of age who have completed a driver education course and have held their learner’s permit for 6 months with no citations to drive to school and school activities, to work and to complete farm work unsupervised if their parent signs a form that allows it.⁴⁰ Iowa is a rural and sparsely populated state that contrasts considerably with New Jersey a densely populated state that has the oldest age requirement, 17 years, to enter into an unsupervised driving license stage during which teens must drive with a decal on their vehicle.⁴¹ Our study focused on families in the Philadelphia metro area in the Commonwealth of Pennsylvania, in a recruitment area that was largely suburban and urban as opposed to rural. The extent to which parents would be supportive of more or less independence regarding when and how their teens are licensed is likely associated with very practical factors tied to where and how they live.

4.4. Policy implications

These results have implications for jurisdictions considering requiring affidavits or driving logs (paper-based or technology-based) to demonstrate teens have complied with practice hour requirements in order for the teen to take their practical road test. Logbooks are not often completed with fidelity but typically this is owing to real life practical issues.⁴² Teens and parents have consistently said that they find app- and vehicle-based telematic devices intrusive^42,43 so this type of technology may not be a viable option that can be mandated easily. Telematics apps can be activated when the teen is a passenger or not in the vehicle at all, so it should not be assumed that a telematics-based practice log is inherently a more accurate practice record than a paper-based practice log. These types of apps and tracking devices have also been used in the context of coercive control and intimate partner violence and care will need to be taken to ensure that attempts to mitigate crash risk via mandatory use of enhanced tracking and monitoring applications does not increase risk for coercive control and stalking behavior.⁴⁴ Some demographic groups may also be very uncomfortable with government-mandated electronic tracking of their or their children’s whereabouts in any capacity owing to safety, legal, and privacy concerns. Most importantly, there is not consistent or clear evidence that there is a specific number of practice hours that matters. Rather, the range of environments practiced in seems to be more predictive of time-to-licensure; which appears to be related to interest in driving, means to drive, and ability to pass the practical driving test.

4.5. Practice variety

We chose to focus on practice variety as opposed to practice quantity. Linear assessment of practice hours with post-license outcomes in cross-sectional and longitudinal studies may mask non-linear and more complex associations. The relationship between the number of hours or the length of time the permit is held, and future MVC outcomes, may be confounded such that longer numbers of hours practiced, or time spent in the learner period, can both plausibly be associated with qualitatively different profiles such as: a) an anxious teen with slowly developing skills who is very reticent about licensure, b) a competent teen with normally developing skills who is conscientious and/or has conscientious parents seeking to delay licensure until they feel they driving is “mastered”, or c) a risk prone teen with engaged parents seeking to delay licensure until the teen’s behaviour improves. Additionally, there is growing evidence that practice variety, or the range of environments practiced in, is a much better indicator of the quality of supervised practice. The reasons for this pattern could have to do with unmeasured error variance as explained above, or just that quantity of practice really does not matter as much as other facets of supervised practice.

4.6. Behavioral interventions and young driver training

The current analysis focused on parent-supervised practice and not on the effects of other types of driver training or behavioral interventions; however, the broader Drivingly trial was designed to evaluate the combined effect of a variety of interventions for learner drivers and these results are not yet available. A recent systematic review of behavioral interventions and driver training programs directed towards young drivers of any license stage (i.e., learners or recently licensed teens) subjected to randomized control trials (RCT) found no evidence that individual-level behavioral interventions can reduce post-license MVC risk.⁴⁵ Almost all the RCTs were under-powered to detect main effects on MVCs and thus this lack of evidence is due in large part to the lack of appropriately powered trials in conjunction with the complexity of driving environments (i.e., drivers can crash for a panoply of reasons, not just those targeted by the particular intervention that was being trialed). Another systematic review focused specifically on the efficacy of driving simulators for young drivers found inconclusive evidence for simulation as an effective approach, and this was mainly due to too few studies that met the inclusion criteria, heterogeneity of outcome variables, and small sample sizes.⁴⁶ There is evidence that driver training programs and behavioral interventions can positively affect more proximal variables. For example, there is evidence that hazard training programs can improve eye-glance behavior and young drivers’ behavior as evaluated in virtual driving environments.⁴⁷ Parent-teen agreements can codify house rules and parenting programs can improve parent-adolescent communication and supervisory behaviors.⁴⁸ However, lack of statistical power in experimental studies continue to be a challenge in this field, especially when MVCs are the outcome of interest, which could potentially be overcome through more collaboration among researchers and innovative study designs.⁴⁵

4.7. Strengths and limitations

This study of over 500 families benefited from strong survey measures, multiple informants (parents and teens), excellent retention for survey completion, and direct observation of the driver’s license ensuring validity of the license data. Twenty-nine percent of the sample was non-White, which is generally reflective of the geographic areas where recruitment was conducted. Despite these strengths there are limitations that should be noted that affect generalizability of the study findings. There were very few teens from rural areas in our study. Our study did not focus on young people known to have additional support needs. For example, autistic young people face unique barriers accessing driver education and training and may have more anxiety in relation to driving^49,50 but also engage in more driver training, when it is available.⁵¹ Parents of young people with attention deficit hyperactivity disorder (ADHD), who are at higher risk for a number of negative health and social outcomes,^52,53 including MVCs,⁵⁴ have to negotiate a more difficult learning to drive experience for their young person who may be more distractable, hyperactive, and/or impulsive than a young person without ADHD. Young drivers with ADHD make more driving errors than young learner drivers not diagnosed with ADHD during the learner period.⁵⁵ Collectively, the presence of these or other conditions can translate into a more challenging supervisory experience for parents and young people.

Conclusion

Driver licensing conveys increased social mobility (e.g., access to education, employment, and healthcare) that can be beneficial for young people, but also has well-documented safety risks. Pathways for safely reducing systemic barriers to driver licensing should be explored to ensure that those young people who can be licensed safely are able to do so. Top-down supervised practice mandates and affidavits might exacerbate existing problems (e.g., driving without a permit) and further disenfranchise families from participating in licensing schemes. Alternatively, they may provide licensing agencies confidence that supervised practice has occurred as required. Pragmatic approaches to safe licensing are crucial along with additional research with families to understand how their behavior would be influenced by such policies.

Highlights.

• Male and female teens had similar parent-supervised practice experiences

• Greater practice variety was associated with faster licensure

• White teens and teens from non-urban areas were licensed faster than other teens