Driving Beliefs and Behaviors of Novice Teen Drivers and their Parents: Implications for Teen Driver Crash Risk

Abstract

This project was a cross sectional survey of 739 matched family pairs of recently licensed teens and their parents. It was designed to assess the relationship of driving behaviors between parents and teens and to investigate predictors of teen crashes. One third of teens reported being involved in a crash during an average of 14 months of driving while 19.9% reported received a moving violation. Teens reported engaging in most risky driving behaviors more frequently than their parents. Teen and parental driving behaviors were associated, but the level of association was low, suggesting that other factors may outweigh parental influence. In multivariate analysis, only the teen’s belief about their crash risk and whether the teen had received a moving violation were associated with reporting a crash.

In 1999, teenagers accounted for about seven percent of licensed drivers in the United States but were involved in 15 percent of all fatal crashes [NHTSA, 1999]. Of crashes involving teens, the adolescent driver is at fault more than 80 percent of the time and the crash rate for teen drivers is up to 6 times that of older drivers [Ryan, Legge, and Rosman, 1998]. Sixteen-year-old drivers are particularly at risk [MMWR, 1996], and most likely to be the driver at fault [Ulmer, Williams, and Presser, 1997]. The high crash risk for teens is believed to be the result of a combination of immaturity and inexperience. Immaturity is age related and associated with decreased perception of risk, increased willingness to attempt behaviors identified as risky, and higher suggestibility by peers. Inexperience is skills based and associated with less ability to accurately perceive specific road situations, poor estimation of safe turning speeds, traffic gaps, and stopping distances, as well as slowed or inappropriate responses to traffic situations [Deery, 1999].

As part of an evaluation of Pennsylvania’s recently enhanced graduated driver licensing system, a survey of matched pairs of teens and parents was conducted. This provided an opportunity to study the etiology of teens’ risky driving behaviors. With rare exceptions [Hemenway and Solnick, 1993], previous research has tended to focus on personality traits or general attitudes toward risk (immaturity characteristics) rather than on risks specific to driving [Deery and Fildes, 1999; Arnett, Offer and Fine, 1997; Jonah, 1997]. In fact, very little research has evaluated the specific etiology for risky driving behaviors in teens. One recent paper suggests that having a parent with a driving history of three or more reported crashes increases the teenage child’s risk of having a reportable crash. [Ferguson, Williams, Chapline, et al., 2001] The aim of this project was to evaluate the effect of parental beliefs and driving behaviors on teen driving behaviors and crash risk. Finally, the goal was to create a model including both teen and parent behaviors and beliefs that would predict increased crash risk for the teen.

METHODS

A statewide, random-sample mail survey of teen drivers and their parents was conducted in Pennsylvania. To be eligible, the teen must have been 16 or 17 years old when they obtained an initial learner’s permit and that permit must have been issued between 1 January and 1 July 2000. This ensured that the teens were licensed under the most recent graduated driver licensing system in Pennsylvania.

SAMPLE SELECTION

48,372 teens met inclusion criteria. Assuming a response rate between 20 and 40 percent, surveying 2500 families was expected to result in enough matched family pairs to allow for multivariate regression to be performed using both parent and teen driving behaviors as predictors. Because there was no previous data on responses from matched pairs, no formal power analysis was performed. A random sample of 2500 teens (and their parents) were selected and packets were mailed to the addresses available in the Pennsylvania driver license database. A total of 739 matched pairs responded and are included in this analysis.

SURVEY DESIGN

The survey was created using previously performed telephone surveys [PennDOT Junior License Survey; California New Driver Survey] and published results as a starting point [MMWR, 1994; Williams, Ferguson, Leaf, et al., 1998; Ferguson and Williams, 1996; Begg, Langley, Reeder, et al., 1995]. After consensus regarding the questions was reached among the research team, the questions were tested in eight focus groups held in Western PA during October, 2000. These groups included both teens and parents. Questions that were unclear or difficult to respond to or those that created discomfort for the teens or parents (such as questions surrounding alcohol) were re-written or removed from the survey.

Data on length of supervised driving during the learner phase were collected as these were thought to correlate with teen crash risk. In addition to the number of crashes and violations throughout the driving period, data on current driving behavior, including frequency of cell phone or radio use, speed, and six specific driving behaviors were requested from each driver.

Previous work on aggressive driving or willingness to engage in risky driving behaviors has focused on generalized personality traits, psychiatric illness, alcohol use, and age related willingness to take risks [Deery and Fildes, 1999; Donovan, Marlatt and Salzberg, 1983, Jonah, 1986; Stevenson, Palamara, Morrison, et al., 2001]. This work suggests more deviant personality types, hyperactive teens [Nada-Raja, 1997], and generally increased aggressiveness are associated with an increased risk of teen crashes. Rather than focus on risky behaviors in general, this study concentrated on driving behaviors.

Of these behaviors, two were identified as potentially protective; one was thought to identify driver aggressiveness; and three are risky behaviors. The two protective behaviors were personally wearing a seat belt and ensuring that passengers were wearing theirs. Shouting, gesturing, or blowing the horn were combined and used as a category of potentially aggressive behaviors. Crossing into an oncoming lane of traffic to pass, speeding up to run through a yellow light, and tailgating another driver may place the driver at increased risk of a crash [Retting, Williams, Preusser, et al., 1995; Retting, Ulmer and Williams, 1999]. Finally, subjects were asked to evaluate teen crash risk as a whole and to compare their personal crash risk to others of the same age.

Because this project involved minors, implied consent was obtained from the parents by addressing the entire survey package “To the Parents of – “. Parental consent and assent of the minor was assumed if surveys were returned. Two return envelopes were provided to allow teens to respond without the oversight of their parents. An incentive was used, consisting of an entry for each returned survey into a lottery to win one of four $100 gas cards. Expedited approval was received from the Institutional Review Board of Allegheny General Hospital prior to beginning this study.

DATA ANALYSIS

Responses between the parent and teen groups were initially compared using χ² or McNemar where appropriate. In addition, the kappa statistic and/or correlation methods were applied to compare the responses of the matched pairs of teens and their parents for each behavior. Then, using whether or not the teen had reported having had a crash as the outcome variable, the behavior reports of both groups were tested for univariate association. Finally, those behaviors or beliefs that were univariately associated with the outcome were analyzed using multivariate logistic regression.

Analysis was performed using SPSS (v.11, 2001) and Primer of Biostatistics [Glantz, 1996]. Significance was set with α = 0.05.

RESULTS

DEMOGRAPHICS

A total of 739 matched pairs of teens and parents responded to the survey. The teen survey respondent group did not significantly differ from the population or the sampled group on age and gender. The majority of respondents (56.3%) were 17, 42.2% were 18 and 1.5% were 19 years old at the time they responded to the survey. 53.3% of the teen respondents were female. All but 14 of the teens had advanced to junior licensure, and 47.8% had achieved a regular license.

The majority (74.4%) of the parent respondents were female. 95.8% of the respondents were white and over 47.9% of those answering the income question had household incomes above $60,000. These numbers are not consistent with the general population of Pennsylvania. To evaluate this further, geocoding was performed comparing the sample group and the respondent group to census tracts by car ownership. With greater than 75% matching, there was excellent correlation for both the sample and the respondents to geographic areas of higher vehicle ownership. There was no significant geographic difference between the sample and the respondents. This implies but does not prove that the bias exists in the eligible population rather than the respondent group.

Although the small proportion of non-white respondents did not allow subgroup analysis by race, such analysis was performed by income. This resulted in significant differences only on two questions: speeding, where both teens and parents of higher income report speeding more frequently; and cell phone use, where both groups were more likely to report using a cell phone while driving if they were of higher income (p<0.02 for both).

CRASH RISK ESTIMATES

An overwhelming majority of both parents (99.7%) and teens (98.2%) reported being at least as safe as other drivers their age. (See Table 1.) Among the teens, boys were more likely (63.0%) than girls (52.2%, p=0.001) to report being safer drivers than their age matched counterparts, but there was no such gender gap among the parents.

Table 1.

Driver Estimates Of Personal Crash Risk, Relative To Other Drivers Of The Same Age

	Parents		Teens
	N	(%)	N	(%)
Less likely to crash	389	(52.1)	426	(57.8)
Equally likely to crash	341	(46.6)	298	(40.4)
More likely to crash	3	(0.3)	13	(1.8)
Total	732		737

kappa = 0.161, p<0.001; r = 0.176, p<0.001; χ² p<0.001

IN TRAFFIC DRIVING TEST

There is no available standard for the failure rate on the driving test in Pennsylvania. 73.2% of the 724 teen subjects who had taken Pennsylvania’s in-traffic driving test passed on their first try (range, 1 – 7 attempts; mean 1.29 attempts). Teens reported driving for a mean of 14.1 months since passing the test (N= 729, range 1–36, stnd dev 4.89).

SPEEDING

For reasons potentially related to the formatted appearance of the teen survey, 43.1% of the teen subjects did not respond to this question, even when they responded to all of the others. 76.7% of the parents reported driving no more than 5 m/h over the speed limit. Of those teens that answered this question, 67.7% reported driving 6–10 m/h above the speed limit. (See Table 2.) Teens reported driving significantly faster than parents reported (p<0.001), but there was no gender gap in speeding reported by the teens. (Table 2.)

Table 2.

Speed

	Parents		Teens
	N	(%)	N	(%)
At or below speed limit	105	(14.2)	66	(15.7)
1–5 m/h above	462	(62.5)	62	(14.3)
6–10 m/h above	146	(19.8)	285	(67.7)
11–15 m/h above	14	(1.9)	1	(0.2)
> 16 m/h above	2	(0.3)	8	(1.9)
Total Responding	729		421

r = 0.251, p< 0.001; χ² p < 0.001

DISTRACTIONS

Two distractions were evaluated: willingness to use a cell phone without pulling over and the number of times the driver manipulates the radio in an average 30-minute drive.

Cell phones

49.7% of parents and 54.1% of teens who responded to the question reported using a cell phone without pulling over. Of those who used cell phones, teens were more likely (p<0.001) to report using a cell phone “routinely” while driving. (Table 3.)

Table 3.

Reasons To Use A Cell Phone

	Parents		Teens
	N	(%)	N	(%)
911 call ONLY	13	(3.5)	16	(4.0)
911 or Urgent call	231	(62.8)	191	(48.0)
Routine call	124	(33.7)	191	(48.0)
Total Responding	368		400

kappa = 0.195, p=0.001 for N = 260 cases where both used a phone; χ² p<0.001

There was a significant gender difference in teen cell phone use; 57.4% of girls compared with 40.9% of boys reporting using a cell phone without pulling over (p<0.001), but of those using a cell phone, boys and girls were equally likely to report using it “routinely”. There was no significant gender difference in cell phone use among parents.

Radio Use

Parents and teens were asked how many times they adjusted the radio or CD player in the car during an average 30 minutes of driving. Only 6.8% of parents but 46.9% of teens reported adjusting the audio equipment more than 5 times in a thirty minute drive (p<0.001). (Table 4.) There was no significant relationship within the matched pairs for use of the radio, and there were no income or gender differences in either group.

Table 4.

Number Of Times Adjusted the Radio In 30 Minutes Of Driving

	Parents		Teens
	N	(%)	N	(%)
None	238	(32.8)	54	(7.4)
1–5 times	438	(60.3)	334	(45.6)
6–10 times	32	(4.4)	194	(26.6)
>10 times	28	(2.4)	149	(20.3)
Total Responding	726		731

kappa = 0.33 p=0.66; r=0.089 p=0.17, χ² p<0.001

SAFE, DANGEROUS, AND AGGRESSIVE DRIVING BEHAVIORS

Teens were more likely than their parents to report “always” wearing their own seat belt (71.7% versus 68.2%, p=0.016), but less likely to ensure that passengers were wearing theirs (66.2% “nearly always” or “always”, compared with 82.6% for parents, p<0.001). (Table 5.)

Table 5.

Safe and Aggressive Behaviors

	Wear Your Seatbelt*				Ensure Passengers Wear Seat Belts**
	Parents		Teens		Parents		Teens
	N	(%)	N	(%)	N	(%)	N	(%)
Always	501	(68.2)	528	(71.7)	435	(59.2)	294	(40.1)
Nearly Always	134	(18.2)	107	(14.5)	172	(23.4)	192	(26.1)
Sometimes	54	(7.3)	66	(9.0)	79	(10.7)	142	(19.3)
Seldom	34	(4.6)	21	(2.9)	35	(4.8)	64	(8.7)
Never	12	(1.6)	14	(1.9)	14	(1.9)	43	(5.9)
Total	735		736		735		735

^*Parent-Teen r = 0.323, p<0.001; χ² p=0.016

^**Parent-Teen r = 0.290, p<0.001; χ² p<0.001

Among parents, there was a significant gender gap in seatbelt use, with women being more likely to report “always” wearing their belt (71.7%) compared with men (58.8%, p = 0.018). The same was true for the teen group; 79.2% of girls reported “always” wearing a belt compared with 63.5% of boys (p<0.001). Mothers and girls were also more likely to report “always” or “nearly always” ensuring their passengers were belted compared with fathers or boys (p<0.001 for both groups).

Teens were more likely to act aggressively toward other drivers, with 38.6% of them shouting, gesturing or blowing the horn at least “sometimes”, compared with 21.7% of parents (p<0.001). (Table 6.) There was no significant gender gap nor income effect for the parents or teens in regard to their self-report of these behaviors.

Table 6.

Shouting, Gesturing, or Blowing the Horn

	Parents		Teens
	N	(%)	N	(%)
Always	4	(0.5)	23	(3.1)
Nearly Always	12	(1.6)	35	(4.8)
Sometimes	143	(19.6)	225	(30.7)
Seldom	386	(52.8)	269	(36.6)
Never	186	(25.4)	182	(24.8)
Total Responding	731		734

r = 0.178, p<0.001; χ² p<0.001

On two lane undivided highways or routes, drivers wanting to pass must cross into the oncoming traffic lane to do so, with the risk of a head-on collision. Teens reported performing this maneuver less often than parents (p<0.001), with 48.0% of teens “never” engaging in this maneuver, compared with 36.6% of parents. (Table 7.) There were no gender differences for either group in their self-reports of this maneuver.

Table 7.

Potentially Dangerous Maneuvers

	Cross Into Oncoming Traffic To Pass*				Speed Up To Run Through Yellow Light**				Tailgate Another Driver ▴
	Parents		Teens		Parents		Teens		Parents		Teens
	N	(%)	N	(%)	N	(%)	N	(%)	N	(%)	N	(%)
Always	6	(0.8)	9	(1.2)	1	(0.1)	19	(2.6)	3	(0.4)	4	(0.5)
Nearly Always	25	(4.3)	21	(2.9)	17	(2.3)	55	(7.5)	5	(0.7)	9	(1.2)
Sometimes	152	(21.0)	149	(20.3)	245	(33.2)	305	(41.5)	50	(6.8)	102	(13.9)
Seldom	276	(38.1)	202	(27.6)	366	(49.7)	273	(37.1)	239	(32.3)	239	(32.5)
Never	265	(36.6)	352	(48.0)	107	(14.5)	83	(11.3)	440	(59.5)	382	(51.8)
Total	724		733	736	735	737		735

^*r = 0.263, p<0.001; χ² p<0.001

^**r = 0.187, p <0.001; χ² p<0.001

^▴r = 0.180, p<0.001; χ² p=0.001

In this study, 51.6% of teens reported at least “sometimes” speeding up to run through a yellow light compared with 35.6% of parents (p <0.001). There were no gender differences for either group in their self-report of speeding up to run through lights.

Teenagers in Pennsylvania have previously been shown to be more likely to be involved in crashes involving tailgating, careless lane changing, and careless passing [Staplin and Lyles, 1991]. Although the majority of parents and teens reported “never” tailgating another driver, teens were still more willing to tailgate than their parents (p = 0.001). For teens, there was no significant gender gap, but mothers were less likely to tailgate, with 62.0% of them reporting “never” tailgating compared with 48.6% of fathers (p=0.009).

MOVING VIOLATIONS

To limit recall bias, parents were asked to report the number of tickets they had received in the last 5 years, while teens were asked to report the number they had ever received (average of 14 months of driving unsupervised). 548 (74.2%) of the 738 parents responding to this question reported no moving violations. The remaining 190 individuals accounted for 252 tickets. The number of violations reported per parent ranged from 0 to 9 with 17.4% of those ticketed having received more than one ticket in the last 5 years. For teens, 589 (79.8%) of the 738 respondents reported no moving violations. The remaining 149 teens had received a total of 161 violations (range 0 to 8) and of these 32 (21.5%) had more than one violation. The number of violations was not correlated between matched pairs of parent and teen, however, the dichotomized outcome of ticket/no ticket was significantly associated (p=0.015). (Table 8.)

Table 8.

Matched Family Pairs and Their Tickets

		Teens (N= 737)
		No Tickets	At Least One Ticket
Parents (N= 737)	No Tickets	438	110
	At least one Ticket	150	39

McNemar p=0.015

There was a relationship between whether or not a teen had received any tickets and their gender, with 24.6% of boys receiving at least one ticket compared with 16.6% of girls (p<0.001). The same was true for the parents, with 30.3% of fathers and 23.9% of mothers having had a ticket (p<0.001).

CRASHES

Parents were asked to report the number of crashes they had had over the last five years of driving, while teens were asked to report the total number of crashes. For this question, crashes were defined as “accidents…(in which) you (were) the driver…(even a fender bender), no matter who was at fault.”

Of the 737 parents responding, 230 (31.2%) reported a total of 308 crashes over the previous five years. 55 parents (23.9% of the crashers) reported more than one crash (range 2 to 8).

Of the 739 teens responding to this question, 490 (66.3%) reported never being the driver in a crash. The remaining 249 (33.6%) teen subjects reported a total of 352 crashes. The number of crashes reported by teens ranged from 1 to 8 with 27% of crashing teens having crashed more than one time in an average of 14 months of unsupervised driving exposure.

There were no gender differences in reported crashes for either the parents or teens.

In our results, using self-reported crashes, there was no relationship between the parent reporting a crash within the last 5 years and the teen reporting a crash in their first period of driving (average of 14 months). However, using the same 3 crash criteria for the parents found recently by Ferguson, et al. [2001], there was a statistically significant association between parents reporting 3 or more crashes in the previous 5 years and their teen driver reporting having had a crash in their first few months of driving (p<0.001). (Table 9.)

Table 9.

Parents With 3 Or More Crashes And Teen Crash Risk

		Teens (N= 739)
		No Crashes	At Least One Crash
Parents(N= 739)	0–2 Crashes	486	239
	3+ Crashes	6	8

McNemar p<0.001

PREDICTORS OF TEEN CRASH RATE

Using a dichotomized dependent variable of crash/no crash for the teen and responses to belief and behavior questions as independent variables, analysis of univariate association was performed using chi squared or Fisher’s exact test as appropriate. Potential independent predictors included all of the variables evaluated above and also included family income level, teen’s report of the amount of time spent driving during the learner’s permit phase, and beliefs about the teen crash risk in general. The cut off for inclusion in further analysis was p≤0.10. Subgroup analysis by teen gender showed no significant differences.

Teen and parent predictors were then modeled using logistic regression using a forward modeling, likelihood ratio method for determining the inclusion or exclusion of each variable in the model. Only two variables remained after this procedure, “Teen’s belief that they are less likely to crash than others” and whether or not the teen had received a moving violation. (Table 10.)

Table 10.

Odds Ratios for Predictors of Teen Crashes

Variable	Odds Ratio	95% CI	p
Teen believes they are less likely to have a crash than others their age	0.66	0.56 – 0.77	0.01
Teen has received a moving violation	1.77	1.34 – 2.34	0.01

The hypothesis was that self-report of driving behaviors by teens and their parents would allow us to identify risk factors for crashes in novice teen drivers and several univariate risk factors were successfully identified. However, the main predictor of a teen reporting a crash was that teen’s assessment of their personal crash risk. In the multivariate analysis, parental driving behaviors had no significant effect on teen crash risk.

LIMITATIONS

There were several limitations to this work. The first issue is the survey methodology; this study has all of the biases of self-reporting that other surveys suffer. Attempts were made to limit recall and observation bias, and prior work does suggest that self-reporting of serious traffic crashes has high agreement with traffic records [Begg, Langley and Williams, 1999].

A second limitation is the response rate of just over 30%. No further analysis of the non-respondents could be performed. The package method of survey delivery aimed at the parent was successful in producing a large number of responses coming from matched family pairs, but there was no assessment of the teen’s exposure to the responding parent’s driving behaviors. Of note, the majority of the parental respondents were female. The effect of this finding was limited by performing subgroup analysis where relevant. There was no assurance that the teen identified on the package was in fact the child who returned the survey; and older or younger sibling may have filled it out. There may have been a Hawthorne effect of family members even though they were instructed to answer independently.

In addition, this survey asked about all crashes rather than serious, injury, or police-reported crashes. No further details were obtained on the crashes such as whether or not the driver was at fault. During the survey design process, it became clear that such a question could not have been reliably answered by the teen; even when there was only one driver (themselves) they often would not have labeled themselves as “at fault”. Other factors important in predicting serious or injury crashes may have been omitted from this analysis as a result.

Finally, the vast majority of the respondents were white and of relatively high income. This does not parallel census information from the state of Pennsylvania where 85.4% of the population is white, 10.0% African American, and the rest of another minority race [PA Census 2000]. The median income in PA was $43,742 in 2000 with a standard error of $1,297 [US Census Bureau]. No racial or income information is available from the Pennsylvania Bureau of Driver Licensing. However, the subjects of interest were families having a young teen driver who obtained a learner’s permit nearly as early as possible under the new licensing system. It is reasonable to suppose that these families are more likely to own or have easy access to a car and to rely on a personal motor vehicle for activities of daily living, such as travel to work, school, or social activities. It is possible that the issues that push teenagers to want a driver’s license as early as possible select for a socioeconomic group that, in Pennsylvania, is more likely to be Caucasian.

DISCUSSION

The hypothesis that parental driving behaviors influence teens’ driving is supported by the fact that there was a statistically significant association between matched family pairs on almost every behavior tested. However, the level of association was low, with r-values below 0.3 and kappa statistics below 0.6. This suggests that while the responding parent had some influence on the teen’s beliefs and behavior, other effects are also present, and may outweigh the parent. If all of the influential parents or guardians could have been captured as data points for this study, this association might be appear stronger; but this may indicate that influences outside the home play a large role in teen driving behaviors.

While several of the parent’s beliefs and driving behaviors were univariately associated with whether or not the teen had crashed, all of the parental variables fell out of the final model, indicating the absence of any direct effects of parental driving beliefs and behaviors on their teen’s risk of crashing

In this study, the two characteristics most associated with whether the teen reported a crash were their perception of their crash risk and whether or not they had received a moving violation. Crash estimates held an interesting position in this analysis. Previous work has shown that teen drivers, and teenage boys in particular, tend to overestimate their driving ability [Gregersen, 1996] and underestimate their crash risk [Finn and Bragg, 1986]. 57.8% of the teens in our sample believed they were less likely to crash than others their age.

Internal belief of decreased crash risk was univariately associated with the outcome and remained in the final model as the only predictor associated with a decreased risk of crashing. Of course, internal beliefs about crash risk may be the result of having crashed rather than a predictor of crashing, and this result could simply indicate confounding. However, in other age groups, driver self-confidence has been demonstrated to be remarkably stable and uninfluenced by adverse events or direct evidence of poor driving skills [Marottoli and Richardson, 1998]. The topic requires further, prospective investigation.

That the strongest predictor positively associated with reporting a crash was the teen having received a moving violation suggests a real opportunity for intervention. Unfortunately, this study cannot conclude that the driver received a moving violation before having a crash. Provided that is the case, the teen driver with a moving violation may be “marked” as at high risk of a crash and there may be an opportunity for risk reduction strategies to be employed. Currently, teens must acquire 6 points or be clocked at 26 or more miles above the speed limit before incurring a license suspension. Further work may support more stringent sanctions or requiring retraining or retesting before the suspension is removed.

The absence of persistent gender differences in driving exposure and behaviors in these study results is contrary to most previous work. The population and subject group of teen drivers were evenly divided between male and female. Previous work from New Zealand in 1995 [Harre, Field and Kirkwood, 1996] suggested that teenage boys push to obtain a driver’s license at a younger age than girls, and that they report engaging in more unsafe driving behaviors than girls. In addition, in 2000 there were slightly more boys (173,068, 51.6%) than girls (162,086, 48.4%) aged 16 or 17 in Pennsylvania. [2000 US Census]. This would lead us to expect a preponderance of boys in the youngest driver group, which is not supported by this data. These results may be the result of changing social gender expectations, due to an increased female interest in obtaining a license as early as possible because of the new licensing system, or be specific to the social and political environment in Pennsylvania.

Significant gender differences (males having a higher crash rate than females) in reportable crash rates have been documented previously, often with the greatest gender gap in the youngest age groups [Massie, Campbell and Williams, 1995; McGwin and Brown, 1999; Li, Baker, Langlois, et al., 1998]. The gender gap for fatal crashes declines when corrected for miles driven, but remains present with males aged 16–19 at higher risk than their female counterparts [Ulmer, Williams and Preusser, 1997]. However, injury crashes overall may be more likely to involve female drivers when corrected for miles driven [McGwin and Brown, 1999].

Between boys and girls, this study found no difference in the number of months driving unsupervised, radio use, reported speed, willingness to cross into traffic to pass, shouting/gesturing/horn blowing, or frequency of speeding up to run yellow light. Girls reported using a cell phone more often, but of those using a cell phone, boys and girls were equally likely to report using it “routinely”. Girls were more likely to wear their seatbelt “always” and to ensure their passengers were belted. Boys more often reported that believing they are safer drivers than others their own age. Finally, although boys reported more tickets than girls, there was no difference between boys and girls or mothers and fathers in regard to their reports of crashing.

The lack of a gender difference in crashes could be an effect of self-reporting; girls may be more willing to report or have better recall of their crashes. Alternately, it may be the result of asking for “all” accidents, including fender benders. Girls may have as many but less severe crashes. However, the lack of a gender gap in the responses to the driving behavior questions parallels the reported crash rate and the absence of gender differences in the modeling procedure. This suggests a new phenomenon, girls driving more like boys, could be occurring; this raises interesting issues for further research.

CONCLUSIONS

This study used cross section survey methodology to examine the relationships between matched pairs of novice teen drivers and their parents for driving beliefs and behaviors and assesses predictors of teen crash risk. Three major findings are revealed: 1) Parental beliefs and driving behaviors have a real but small correlation with their teen’s driving beliefs and behaviors but do not predict teen crashes, 2) while there are no strong predictors of teen crashes, a teen’s perceptions of personal crash risk and having moving violations appear to be associated with their risk of crashing, and 3) gender had little effect on driving style and no effect on crash rate for either the parents or their children.