- Email: [email protected]
The effect of the learner license Graduated Driver Licensing components on teen drivers’ crashes
Accident Analysis and Prevention 59 (2013) 327–336
Contents lists available at ScienceDirect
Accident Analysis and Prevention journal homepage: www.elsevier.com/locate/aap
The effect of the learner license Graduated Driver Licensing components on teen drivers’ crashes Johnathon Pouya Ehsani a,b,∗ , C. Raymond Bingham a,b , Jean T. Shope a,b a b
University of Michigan School of Public Health, Ann Arbor, MI, United States University of Michigan Transportation Research Institute, Ann Arbor, MI, United States
a r t i c l e
i n f o
Article history: Received 8 February 2013 Received in revised form 7 May 2013 Accepted 1 June 2013 Keywords: Graduated Driver Licensing Learner permit Teen driving Motor vehicle crash
a b s t r a c t Background: Most studies evaluating the effectiveness of Graduated Driver Licensing (GDL) have focused on the overall system. Studies examining individual components have rarely accounted for the confounding of multiple, simultaneously implemented components. The purpose of this paper is to quantify the effects of a required learner license duration and required hours of supervised driving on teen driver fatal crashes. Methods: States that introduced a single GDL component independent of any other during the period 1990–2009 were identified. Monthly and quarterly fatal crash rates per 100,000 population of 16- and 17-year-old drivers were analyzed using single-state time series analysis, adjusting for adult crash rates and gasoline prices. Using the parameter estimates from each state’s time series model, the pooled effect of each GDL component on 16- and 17-year-old drivers’ fatal crashes was estimated using a random effects meta-analytic model to combine findings across states. Results: In three states, a six-month minimum learner license duration was associated with a significant decline in combined 16- and 17-year-old drivers’ fatal crash rates. The pooled effect of the minimum learner license duration across all states in the sample was associated with a significant change in combined 16- and 17-year-old driver fatal crash rates of −.07 (95% Confidence Interval [CI] −.11, −.03). Following the introduction of 30 h of required supervised driving in one state, novice drivers’ fatal crash rates increased 35%. The pooled effect across all states in the study sample of having a supervised driving hour requirement was not significantly different from zero (.04, 95% CI −.15, .22). Conclusion: These findings suggest that a learner license duration of at least six-months may be necessary to achieve a significant decline in teen drivers’ fatal crash rates. Evidence of the effect of required hours of supervised driving on teen drivers’ fatal crash rates was mixed. © 2013 Elsevier Ltd. All rights reserved.
1. Introduction Motor vehicle crashes are the leading cause of death and a major contributor to nonfatal injury of teens in the United States (National Highway Traffic Safety Administration, 2010b). To address this public health threat, Graduated Driver Licensing (GDL) laws have been enacted by all fifty states and the District of Columbia (Williams and Shults, 2010). GDL was introduced in the United States beginning in the mid-1990s, replacing laws that allowed quick and easy access to full-privilege licenses. GDL laws vary in their requirements, but commonly include two levels that impose restrictions on teens’ driving (Foss and Goodwin, 2003). The first is a learner license that allows teens to gain driving experience under the supervision of a
∗ Corresponding author at: University of Michigan School of Public Health, Ann Arbor, MI, United States. Tel.: +1 734 763 9938. E-mail address: [email protected] (J.P. Ehsani). 0001-4575/$ – see front matter © 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.aap.2013.06.001
fully-licensed driver (typically a parent or parent designate over age 21). The second is an intermediate license, which allows teens who have gained some initial experience driving with a learner license to drive independently but with restrictions that limit their exposure to the highest risk driving conditions: driving at night (Williams and Shabanova, 2003) and driving with young passengers (Chen et al., 2000). There is little question that GDL reduces 16- and 17-year-old driver crashes (Masten et al., 2011; Shope, 2007); however, the elements responsible for the greatest reductions in crashes and the mechanism by which these reductions are achieved are not well understood. Evaluations of individual components of GDL have rarely accounted for the confounding effect of multiple GDL components being implemented simultaneously, and have assumed independent implementation of each component, which does not reflect the reality of how the majority of these laws were introduced. Further, most evaluations used a pre- and post-GDL study design that is unable to distinguish whether a decline in crashes was
328
J.P. Ehsani et al. / Accident Analysis and Prevention 59 (2013) 327–336
directly attributable to GDL, or the result of a preexisting downward trend (Elliot and Shope, 2003; Sivak and Schoettle, 2010). Currently in the U.S., learner license requirements are the most widely implemented of all GDL components, existing individually or side-by-side with other components in all fifty states and the District of Columbia. However, evidence of the effectiveness of the details of the learner license requirements is not well established. For example, the optimal number of months a learner license should be held has not been determined. Studies in Kentucky, Connecticut, and Nova Scotia indicated substantial crash reductions for 16-year-old drivers when a learner license period was newly mandated or an existing period was extended (Agent et al., 2001; Mayhew et al., 2003; Ulmer et al., 2001). While these findings indicated an extension of the learner license period reduced crashes, none of these studies used licensure data, so it is unclear whether a delay in the age of independent driving, or improvements in driving ability due to an extension in supervised driving beyond the required minimum were responsible for the crash reductions. Furthermore, little is known about the optimal number of months a learner license should be held for the best safety benefit. In all three states above, the learner license was mandated to be six months; however, there is no evidence to suggest whether or not a sixmonth period of supervised driving is adequate (Foss, 2007). For example, for the states discussed above, it is unknown whether a doubling of the learner license period (to twelve months) would have resulted in the same or a larger crash reduction. Similarly, the safety effect of a required number of supervised driving hours on teen drivers is also poorly understood. The small body of research examining the subject is inconclusive. A study of Swedish teens found an average of 120 h of supervised driving was associated with a significant reduction in crash involvement during independent licensure, compared to those who had approximately 50 h of supervised driving practice (Gregersen et al., 2000, 2003; Sagberg and Gregersen, 2005). Teen drivers in the northeastern U.S. who completed a period of supervised driving, however, were no different in their time-to-first-crash from those who did not have supervised driving experience (McCartt et al., 2003). Similarly, French teens who received professional driving instruction with an extensive period of supervised driving (equivalent to approximately 3000 miles) had the same crash likelihood as teens who only received professional driving instruction (Page, 2004). Due to the small number of studies, it is not possible to determine whether or not 120 h is the optimal number of supervised driving hours. The purpose of this paper is to quantify the effect of two required GDL components of the learner license on 16- and 17-year-old drivers’ fatal crash rates: the length of the learner license (months) and the number of supervised driving hours. 1.1. Research hypotheses This study tested the following hypotheses: 1. The introduction of a learner license required minimum holding period as part of GDL will be followed by a reduction in 16- and 17-year-old drivers’ fatal crash rates. 2. The introduction of a minimum number of required supervised driving hours as part of GDL will be followed by a decline in 16and 17-year-old drivers’ fatal crash rates. 2. Method 2.1. Inclusion criteria To test these hypotheses, states that introduced a learner license holding period, or required hours of supervised driving independent
of other GDL components during the period 1990–2007, were identified (Tables 1 and 2). The evaluation period spanned 1990–2009, however, because at least two years of data post-implementation were required to estimate the effect of a component, the sample was limited to states introducing GDL components prior to December 31st 2007. States were also excluded from the sample if they introduced multiple GDL components simultaneously with the component of interest, or had a learner license age below 15. 2.2. Data and measures Monthly counts of fatal crashes involving at least one teen driver (aged 16 or 17 years) in cars, trucks/pickups, vans/minivans, and sport utility vehicles were obtained for the contiguous period 1990–2009 from the Fatality Analysis Reporting System (FARS) for the states being analyzed (National Highway Traffic Safety Administration, 2010a). Ideally, data from all injury crashes (not just fatal crashes) occurring in each candidate state would also be included; however, only a limited number of states make their injury crash data available to researchers (National Highway Traffic Safety Administration, 2011), so such an approach could not be taken for this study. Furthermore, injuries are not recorded consistently across states or sometimes even across police agencies within a state. Fatal crash rates would ideally be based on the number of licensed teen drivers, however, licensure data reported by the Federal Highway Administration underreport the actual number of licensed teens, and licensure data are difficult to obtain from individual states (Insurance Institute for Highway Safety, 2006). Miles driven by each teen would also be ideal, but are difficult to measure and are unavailable. Therefore, crash rates were based on the number of teens in the overall population. Annual population estimates by state and year of age were obtained from the U.S. Census Bureau (Bureau of the Census US Department of Commerce, 1999, 2010). Monthly values were interpolated using cubic spline curves, which are the smoothest curve that exactly fits a set of data points (Bartels et al., 1998). The combined monthly fatal crash involvement rates of 16- and 17-year-old drivers per 100,000 population were calculated using monthly fatal crash counts and monthly population estimates. Data for drivers younger than 16 years were excluded because only a few states allow unsupervised driving by 15 year olds (Insurance Institute for Highway Safety, 2012). Several states in the sample had relatively small populations, increasing the probability of a floor effect, where crash rates cannot take on a value lower than zero. To compensate for this effect, states with a 16- to 17-year-old population below 85,000 (Maine, Nebraska, New Hampshire, Rhode Island, and Utah) were modeled using quarterly data. Quarterly fatal crash involvement rates were calculated using the monthly crash counts and population estimates. 2.3. Covariates 2.3.1. Comparison population The monthly fatal crash rate for drivers age 25–54 was used as a covariate, representing crashes for the typical adult driving population. Applying the identical method used to estimate 16- and 17-year-old fatal crash rates, monthly fatal crash rates of 25- to 54-year-old drivers per 100,000 population were calculated using monthly fatal crash counts and monthly population estimates. The purpose of the comparison population was to adjust for variability in the teen driver crash rates due to extraneous factors affecting drivers of all ages and to test the effect of GDL against a comparison population of persons unaffected by GDL. Although time series analyses control for pre-existing secular trends in crash rates, the inclusion of the crash rates of another age group as a historical
J.P. Ehsani et al. / Accident Analysis and Prevention 59 (2013) 327–336
329
Table 1 Percentage change in crash rate by state, following introduction of learner license.* State
Connecticut Hawaii Kentucky Minnesota South Carolina Tennessee Virginia Utah a b c *
Effective date
January July October February July January July August
Learner license duration
1997 1997 1996 1997 1998 1996 1996 2006
6 months 3 months 6 months 6 months 3 months 3 months 6 months 6 months
Fatal crash rate per 100,000 population Pre-introductiona
Post-introductionb
2.17 1.85 4.61 2.76 3.76 4.95 2.94 2.58
1.65 1.36 4.10 2.59 3.75 4.86 2.62 1.39
Percentage change in fatal crash ratesc
−16.6 −10.6 −8.9 −18.9 −2.8 0.8 −5.5 10.1
Prior to the introduction of the learner license. Until the introduction of new GDL component(s). Values based on ARIMA. Statistically significant percentages highlighted in bold.
Table 2 Percentage change in crash rate by state, following introduction of required supervised practice driving hours.* State
Effective date
Arizona Kentucky Maine Minnesota New Hampshire Rhode Island
January October January January September July
a b c *
Required supervised practice driving hours
2000 2006 1998 1999 1999 2003
25 h 60 h 35 h 40 h 20 h 50 h
Fatal crash rate per 100,000 population Pre-Introductiona
Post-Introductionb
3.27 4.31 3.07 2.59 2.55 1.56
2.54 3.25 3.33 2.46 2.27 1.46
Percentage change in fatal crash ratesc
−17.7 −4.2 −13.7 34.1 5.6 10.6
Prior to the introduction of the required supervised practice driving hours. Until the introduction of new GDL component(s). Values based on ARIMA. Statistically significant percentages highlighted in bold.
covariate to control for unmeasured factors that affect all drivers enhances the validity of the findings. 2.3.2. Gasoline prices An inverse relationship between gasoline prices and fatal crashes has been identified for drivers of all ages (Sivak and Schoettle, 2010); however, research suggests teen driving behavior may be more sensitive to higher gasoline prices, relative to older drivers (Morrisey and Grabowski, 2010). Monthly national average gasoline prices, obtained from the U.S. Energy Information Administration (U.S.Energy Information Administration, 2011), were used as a covariate in the analyses to adjust for their effect on the amount of driving exposure and resulting crash risk level. 2.3.3. GDL effective date The points at which GDL components were expected to affect crash rates were determined from information about each state’s GDL system, obtained from the Insurance Institute for Highway Safety’s GDL effective date database (Insurance Institute for Highway Safety, 2012) and each state’s website (Florida Department of Highway Safety and Motor Vehicles, 2009; Maryland Motor Vehicle Administration, 2012; Michigan Secretary of State, 2012). 2.4. Analytical method Monthly crash rates per 100,000 population of 16- and 17year-old drivers were analyzed using Auto-Regressive Integrated Moving Average (ARIMA) interrupted time series analysis for each state. The AR component represents the lingering effects of previous observations, while the MA component represents lingering effects of previous random shocks (or error) (McCleary and Hay, 1982). This approach accounts for trend, seasonality and
time-related autocorrelation in the data and eliminates the need to exclude the data points prior to and immediately after the implementation of GDL. It is superior to other before-after evaluation methods that do not take these factors into account, and exclude data points immediately before and after the introduction of GDL. Within ARIMA, a transfer function relates the enactment of GDL laws to crash rates using two parameters. The first parameter, ω, is the magnitude of the change (rise or fall) in crash rates. If ω is statistically significant, the size of the change is directly interpretable as the percentage change in the post-intervention series relative to the pre-intervention series using the formula 100 × [eω − 1] (McDowall et al., 1980; Tabachnick and Fidell, 2007). The second parameter, ı, reflects the onset of the change. For these analyses, ı was fixed at 0, meaning the anticipated change in crash rates would be abrupt and lasting. All models were estimated using the natural logarithm of the crash rates. Complete details of the model parameters can be found in Appendices A and B.
2.5. Analytical strategy The analyses were conducted in three stages. First, a linear regression model was estimated for each state for combined 16and 17-year-old driver crash rates and the covariates: adult crash rates, gasoline prices, and GDL laws. Second, the model for each state was statistically adjusted for trends and seasonal variation. Autoregressive and moving average orders were identified using auto-correlation and partial-auto-correlation functions of the series residuals. The original regression model was re-estimated with the inclusion of the autoregressive or moving average orders identified in the second stage. Outliers were also detected and controlled in each model. Analyses were conducted using the SCA
330
J.P. Ehsani et al. / Accident Analysis and Prevention 59 (2013) 327–336
Fig. 1. Pooled effect of the learner license duration on 16- and 17-year-old fatal driver crash rates. (1) Square area around the point estimates corresponds to the weighting of the state in the overall model. (2) Diamond represents the pooled point estimate with 95% Confidence Interval.
Time Series and Forecasting System, a time-series analysis software package (Scientific, 2011). The effects of each GDL component on combined 16- and 17year-old drivers’ crashes were pooled across states using a random effects meta-analytic model. This approach assumes differences in the implementation and effectiveness of GDL in each state, and takes these differences into account as additional sources of variation (Deeks et al., 2008). These analyses were conducted using Comprehensive Meta-Analysis software (Borenstein et al., 2005). 3. Results 3.1. Learner license duration The results of the interrupted time series analysis of the effect of the learner license duration requirement partially confirm the first hypothesis. In three of the five states that introduced a six-month learner license period, there was a significant reduction in fatal teen driver crash rates. Specifically, there were significant reductions in fatal teen driver crash rates following the introduction of a sixmonth learner license in Connecticut, Minnesota, and Virginia of 16.6%, 18.9% and 5.5%, respectively (Table 1). There were no significant changes in the fatal crash rates of 16- and 17-year-old drivers in states that introduced a minimum learner license duration of three months. Adult fatal crashes accounted for some of the variability in teen fatal crashes for each state in the sample, with the exception of Hawaii and Connecticut. When the effect of a minimum learner license duration was pooled across all states in the sample, there was a statistically significant change in combined 16- and 17-year-old driver fatal crash rates of −.07 (95% Confidence Interval [CI] −.11, −.03). Fig. 1 depicts the pooled effect of the learner license duration on 16- and 17-year-old drivers’ fatal crash rates. 3.2. Supervised driving hours The introduction of required hours of supervised driving was not followed by a significant reduction in teen drivers’ fatal crashes in any state included in the sample (Table 2), and therefore the second hypothesis were rejected. In Minnesota, the introduction of 30 h of required supervised practice driving corresponded with a significant 34.5% increase in combined 16- and 17-year-old drivers’ fatal
Fig. 2. Pooled effect of required supervised driving hours on 16- and 17-year-old driver fatal crash rates. (1) Square area around the point estimates corresponds to the weighting of the state in the overall model. (2) Diamond represents the pooled point estimate with 95% Confidence Interval.
crash rates. Adult fatal crashes predicted some of the variability in teens’ fatal crashes in each state except New Hampshire. The pooled effect of supervised driving hours across all states in the study sample was not significantly different from zero (.04, 95% CI −.15, .22) in the random effects meta-analytic model (Fig. 2). 4. Discussion The purpose of this study was to examine the independent effects of a required learner license duration and required hours of supervised driving on teen drivers’ fatal crashes. Based on the pooled estimates of the meta-analytic models, the introduction of the learner license minimum holding period resulted in a significant decline in teen drivers’ fatal crash rates, while the introduction of a minimum number of required supervised driving hours had no effect. It could be argued that one primary mechanism for crash reductions in teen drivers’ fatal crash rates may be a delay in the age of independent driving. This explanation is consistent with the findings that the six-month learner license resulted in a significant decline in 16- and 17-year-old drivers’ fatal crashes in three states, and the three-month learner license may have reduced crash rates, but less than the six-month learner license (did not reach statistical significance). It is also consistent with the finding that supervised driving hour requirements had no effect on fatal crash rates. However, this interpretation is complicated by the relationship between the licensing age and the learner license duration that was introduced in each state (Williams, 2007). For example, in states where the minimum age for both the learner license and the intermediate license are identical (e.g., 16 years), adding a learner license holding period guarantees a delay past the minimum age for an intermediate license. A single state in the study sample (Connecticut) fit these criteria. Whereas in Virginia and Minnesota, where a delay was not guaranteed, significant declines in 16- and 17-year-old driver fatal crashes were also observed. These observations suggest that delayed licensure may contribute, but is not entirely responsible for the crash reductions associated with the learner license components of GDL. The sample for this study was limited to states that introduced a single GDL component independently of other components. In some of the states, components were introduced within an existing GDL system. Consequently, the additive effect of multiple co-existing GDL components could not be disentangled. Significant changes in fatal crash rates were observed in large states,
J.P. Ehsani et al. / Accident Analysis and Prevention 59 (2013) 327–336
suggesting that the absence of an effect in smaller states could be due to insufficient power. However, estimating the models for small states using quarterly data did not alter the results that were based on monthly data. Crash data for 16- and 17year-old drivers were combined because using each year of age would have resulted in data that were too sparse to permit meaningful interpretation (Kochanek et al., 2011). This also provided a consistent outcome measure across all states in the study. It should also be noted that crash data for this sample were limited to fatal crashes. Some have argued that fatal crashes represent a small and atypical subset of all crashes, and the etiology of fatal crashes may differ from that of less serious crashes (Lam, 2003). This study investigated the effect of two required GDL components of the learner license on 16- and 17-year-old drivers’ fatal crash rates. This age group, however, includes teenage drivers who are at different stages of the licensing process. The rationale for selecting this age group was to evaluate the effect of the learner license components on early driving, particularly independent driving. While the licensure status of the 16- and 17- year-olds involved in fatal crashes could not be determined, epidemiological evidence indicates that driving under supervision during the learner license stage entails a very low crash risk (Mayhew, 2003). Therefore, it is likely that the majority of fatal crashes included in
331
these analyses occurred while teens where driving using an intermediate license, which allows teens to drive independently but with restrictions. In addition, compliance and enforcement of the GDL components under investigation were not accounted for in this study. Future research examining the effect of GDL components on teen drivers should extend the analysis to include all crash types: property-damage-only crashes and injury crashes, as well as fatal crashes. Such data would allow an examination of the effects of GDL on crash types of differing severity. In addition, studies should use state level licensure data to take into account the proportion of teens licensed at each year of age, and the GDL stage associated with each individual driver’s crash record. With this more detailed information, time-to-event analysis could be used to compare first-time crash incidence at each level of licensure and determine whether the primary mechanism for crash reductions is delayed licensure or safer driving. This study adds to the literature by presenting new information because GDL evaluations to date have largely measured the effect of several requirements imposed at the same time, rather than isolating the effects of specific components. Using this analytical approach, the learner license duration was found to be a critical component of GDL, and a six-month learner license duration seems to be the minimum period necessary to be associated with a significant decline in teen drivers’ fatal crash rates.
332
Appendix A. Parameters of ARIMA models estimating the effect of learner license duration on 16- and 17-year olds’ fatal crash rates per capita, 1990–2009 Model component
Parameter (Lag)
Estimate
Connecticut
6 month learner license Effective January 1997 No passengers for first three months (with exceptions) Effective October 2003 20 h supervised driving 12 a.m.–5 a.m. driving restriction No passengers for first six months (with exceptions) Effective October 2005 40 h supervised driving No passengers for first 12 months (with exceptions) Effective August 2008 Control series (25–54 year-olds) Gasoline price Noise (AR or MA component) Constant
ω
−.1820
.04
ω
−.0067
.97
ω
−.1128
.56
ω
−.3240
.05
ˇ ˇ None
.3730 −.0004
.09 .97
.7112
<.01
ω
−.1122
.35
ω
−.0310
.90
ˇ ˇ None
.1957 −.0012
.33 .39
.6307
.01
ω
−.0942
.15
ω
−.0433
.81
ω
−.0166
.93
ˇ ˇ None
.4850 −.0014
.01 .02
1.2077
<.01
ω
−.2092
.03
ω
.2931
<.01
ω
−.2698
.02
ˇ ˇ None
.5796 −.0014
<.01 <.01
.8546
<.01
−.0283
.75
Hawaii
Kentucky
Minnesota
South Carolina
3 month learner license Effective July 1997 6 month learner license 11 p.m.–5 a.m. driving restriction One passenger younger than 18 Effective September 2003 Control series (25–54 year-olds) Gasoline price Noise (AR or MA component) Constant 6 month learner license Effective October 1996 60 h supervised driving Effective October 2006 12 a.m.–6 a.m. driving restriction One passenger younger than 20 Effective April 2007 Control series (25–54 year-olds) Gasoline price Noise (AR or MA component) Constant 6 month learner license Effective February 1997 30 h supervised driving Effective January 1999 12 a.m.–5 a.m. driving restriction One passenger younger than 20 for first 6 months Three passengers younger than 20 for second 6 months Effective August 2008 Control series (25–54 year-olds) Gasoline price Noise (AR or MA component) Constant 3 month learner license Effective July 1998
ω
p
J.P. Ehsani et al. / Accident Analysis and Prevention 59 (2013) 327–336
State
6 month learner license 40 h supervised driving 6 p.m.–6 a.m. ESTa driving restriction 8 p.m.–6 a.m. EDTb driving restriction Two passengers younger than 21 Effective September 2003 Control series (25–54 yr-olds) Gasoline price Noise (AR or MA component) Constant Tennessee
Utah
6 month learner license Effective July 1996 Three passengers younger than 16 Effective July 1998 12 a.m.–4 a.m. driving restriction 9 month learner license 40 h supervised driving One passenger younger than 18 until age 17 Three passengers younger than 18 thereafter Effective July 2001 One passenger younger than 18 for first 12 months Three passengers younger than 18 thereafter Effective July 2003 45 h supervised driving Effective July 2008 Control series (25–54 yr-olds) Gasoline price Noise (AR or MA component) Noise (AR or MA component) Noise (AR or MA component) Constant 12 a.m.–6 a.m. driving restriction 30 h supervised driving Effective July 1999 No passengers for first 6 months Effective July 2001 30 h supervised driving Effective July 2003 40 h supervised driving Effective July 2004 6 month learner license Effective August 2006 Control series (25–54 year-olds) Gasoline price Noise (AR or MA component) Constant
−.0512
.63
ˇ ˇ MA (12)
.4586 −.0016 −.1293 1.0368
.01 .01 .05 <.01
ω
.0083
.90
ω
−.2063
.01
ˇ ˇ None
.8871 −.0010
<.01 .06
.6486
.01
ω
−.0562
.04
ω
.0390
.15
ω
−.0044
.86
ω
−.0154
.62
ω
−.0257
.43
ˇ ˇ MA (1) MA (6) AR (1) None
.5128 −.0002 .7223 .2422 .6592
<.01 .25 <.01 <.01 <.01
ω
.0295
.83
ω
−.0881
.60
ω
.1449
.49
ω
.0890
.70
ω
.0967
.58
.6120 −.0030
<.01 .03
.8262
<.01
ˇ ˇ None
J.P. Ehsani et al. / Accident Analysis and Prevention 59 (2013) 327–336
Virginia
3 month learner license Effective January 1996 6 month learner license 50 h supervised driving 11 p.m.–6 a.m. driving restriction Single passenger restriction Effective July 2001 Control series (25–54 year-olds) Gasoline price Noise (AR or MA component) Constant
ω
a
333
Eastern standard time. b Eastern daylight time.
334
Appendix B. Parameters of ARIMA models estimating the effect of supervised driving hours on 16- and 17-year-olds’ fatal crash rates per capita, 1990–2009 Model component
Parameter (Lag)
Estimate
p
Arizona
25 h supervised driving Effective January 2000 5 month learner license Effective July 2000 6 month learner license 30 h supervised driving 12 a.m.–5 a.m. driving restriction One passenger younger than 18 for first 6 months Effective July 2008 Control series (25–54 yr-olds) Gasoline price Noise (AR or MA component) Constant
ω
−.1955
.27
ω
.0597
.74
ω
−.2520
.05
ˇ ˇ None
.6422 −.0006
<.01 .29
.6250
.01
ω
−.0942
.15
ω
−.0433
.81
ω
−.0166
.93
ˇ ˇ None
.4850 −.0014
.01 .02
1.2077
<.01
ω
−.1492
.43
ω
.0173
.94
ω
.2022
.39
.4254 −.0026
.04 .07
.9811
<.01
Kentucky
Maine
6 month learner license Effective October 1996 60 h supervised driving Effective October 2006 12 a.m.–6 a.m. driving restriction One passenger younger than 20 Effective April 2007 Control series (25–54 yr-olds) Gasoline price Noise (AR or MA component) Constant 35 h supervised driving Effective January 1998 No passengers for first 3 months Effective August 2000 6 month learner license 12 a.m.–5 a.m. driving restriction No passengers for first 6 months Effective September 2003 Control series (25–54 yr-olds) Gasoline price Noise (AR or MA component) Constant
ˇ ˇ None
J.P. Ehsani et al. / Accident Analysis and Prevention 59 (2013) 327–336
State
Minnesota
Rhode Island
3 month learner license 1 a.m.–5 a.m. driving restriction Effective January 1998 20 h supervised driving Effective September 1999 One passenger for first 6 months Effective January 2003 40 h supervised driving Effective June 2009 Control series (25–54 yr-olds) Gasoline price Noise (AR or MA component) Constant 6 month learner license 1 a.m.–5 a.m. driving restriction Effective January 1999 50 h supervised driving Effective July 2003 One passenger younger than 21 for the first 12 months Effective July 2005 Control series (25–54 yr-olds) Gasoline price Noise (AR or MA component) Constant
ω
−.2092
.03
ω
.2931
<.01
ω
−.2698
.02
ˇ ˇ None
.5796 −.0014
<.01 <.01
.8546
<.01
ω
−.0820
.70
ω
.0542
.82
ω
−.0268
.90
ω
.0052
.99
.2238 −.0011
.26 .39
.8102
<.01
ω
−.0639
.64
ω
.1006
.63
ω
−.0956
.73
ˇ ˇ None
.3132 −.0009
.02 .95
.3022
.17
ˇ ˇ None
J.P. Ehsani et al. / Accident Analysis and Prevention 59 (2013) 327–336
New Hampshire
6 month learner license Effective February 1997 30 h supervised driving Effective January 1999 12 a.m.–5 a.m. driving restriction One passenger younger than 20 for first 6 months Three passengers younger than 20 for second 6 months Effective August 2008 Control series (25–54 yearr-olds) Gasoline price Noise (AR or MA component) Constant
335
336
J.P. Ehsani et al. / Accident Analysis and Prevention 59 (2013) 327–336
References Agent, K., Steenbergen, L., Pigman, J., Stinson Kidd, P., McCoy, C., Pollack, S., 2001. Impact of partial graduated driver’s license program on teen motor vehicle crashes in Kentucky. Transportation Research Record: Journal of the Transportation Research Board 1779, 54–60. Bartels, R.H., Beatty, J.C., Barsky, B.A., 1998. An Introduction to Splines for Use in Computer Graphics and Geometric Modelling. Morgan Kaufmann, San Francisco, CA. Borenstein, M., Hedges, L., Higgins, J., Rothstein, H., 2005. Comprehensive Metaanalysis Version 2. Biostat, Englewood, NJ. Bureau of the Census, US Department of Commerce, 1999. 1990–999 Annual Time Series of State Population Estimates by Age and Sex: 1990–1999. Bureau of the Census, US Department of Commerce, 2010. 2000–2009 State Single Year of Age and Sex Population Estimates. Chen, L., Baker, S.P., Braver, E.R., Li, G., 2000. Carrying passengers as a risk factor for crashes fatal to 16- and 17-year old drivers. JAMA: The Journal of the American Medical Association 283 (12), 1578–1582. Deeks, J.J., Altman, D.G., Bradburn, M.J., 2008. Statistical methods for examining heterogeneity and combining results from several studies in meta-analysis. BMJ Publishing Group. Elliott, M.R., Shope, J.T., 2003. Use of a Bayesian changepoint model to estimate effects of a graduated driver’s licensing program. Journal of Data Science 1, 43–63. Florida Department of Highway Safety and Motor Vehicles, 2009. License Requirements for Teens – Graduated Driver Licensing. 2012. Foss, R.D., 2007. Improving graduated driver licensing systems: a conceptual approach and its implications. Journal of Safety Research 38 (2), 185–192. Foss, R., Goodwin, A., 2003. Enhancing the effectiveness of graduated driver licensing legislation. Journal of Safety Research 34 (1), 79–84. Gregersen, N.P., Berg, H.Y., Engstrom, I., Nolen, S., Nyberg, A., Rimmo, P.A., 2000. Sixteen years age limit for learner drivers in Sweden: an evaluation of safety effects. Accident Analysis and Prevention 32 (5), 25–35. Gregersen, N.P., Nyberg, A., Berg, H.Y., 2003. Accident involvement among learner drivers – an analysis of the consequences of supervised practice. Accident Analysis and Prevention 35 (5), 725–730. Insurance Institute for Highway Safety, 2006. Motor vehicle registration and licensed driver information. Insurance Institute for Highway Safety, 2012. Effective dates of graduated driver licensing components. 2012. ˜ A.M., Kung, H., 2011. Deaths: final data Kochanek, K.D., Zu, J., Murphy, S.L., Minino, for 2009. National Center for Health Statistics 60 (3). Lam, L.T., 2003. Factors associated with young drivers car crash injury: comparisons among learner, provisional, and full licensees. Accident Analysis and Prevention 35 (6), 913–920. Masten, S.V., Foss, R.D., Marshall, S.W., 2011. Graduated driver licensing and fatal crashes involving 16- to 19-year-old drivers. JAMA: The Journal of the American Medical Association 306 (10), 1098–1103.
Maryland Motor Vehicle Administration. Rookie Driver Program. 2012 [cited 2012 3rd March]; Available from: http://www.mva.maryland.gov/ Driver-Services/RookieDriver/default.htm. Mayhew, D.R., 2003. The learner’s permit. Journal of Safety Research 34 (1), 35–43. Mayhew, D.R., Simpson, H.M., Desmond, K., Williams, A.F., 2003. Specific and long-term effects of Nova Scotia’s graduated licensing program. Traffic Injury Prevention 4 (2), 91–97. McCartt, A.T., Shabanova, V.I., Leaf, W.A., 2003. Driving experience, crashes and traffic citations of teenage beginning drivers. Accident Analysis and Prevention 35 (3), 311–320. McCleary, R., Hay, R.A., 1982. Applied Time Series Analysis for the Social Sciences. Sage Publications, Beverly Hills, CA. McDowall, D., McCleary, R., Meiringer, E.E., Hay, A.H., 1980. Interrupted Time Series Analysis. Sage Publications, Thousand Oaks, CA. Michigan Secretary of State, 2012. Graduated Driver Licensing. 2012. Morrisey, M.A., Grabowski, D.C., 2010. Gas prices, beer taxes and GDL programmes: effects on auto fatalities among young adults in the US. Applied Economics 43 (25), 3645–3654. National Highway Traffic Safety Administration, 2010a. Fatality Analysis Reporting System. National Highway Traffic Safety Administration, 2010b. Traffic Safety Facts 2009: A Compilation of Motor Vehicle Crash Data from the Fatality Analysis Reporting System and the General Estimates System. U.S. Department of Transportation. National Highway Traffic Safety Administration, 2011. SDS Overview. 2011. Page, Y., 2004. An evaluation of the effectiveness of the supervised driver-training system in France. Annual Proceedings at Association for the Advancement of Automotive Medicine 48, 131–145. Sagberg, F., Gregersen, N.P., 2005. Effects of Lowering the Age Limit for Driver Training. Elsevier Science, Amsterdam. Scientific, C.A., 2011. Time-series analysis and forecasting software. WorkBench Version 6.2.0. Shope, J.T., 2007. Graduated driver licensing: review of evaluation results since 2002. Journal of Safety Research 38 (2), 165–175. Sivak, M., Schoettle, B., 2010. Toward understanding the recent large reductions in U.S. road fatalities. Traffic Injury Prevention 11, 561–566. Tabachnick, B.G., Fidell, L.S., 2007. Using Multivariate Statistics. Allyn and Bacon, Boston. U.S.Energy Information Administration, 2011. Energy Prices. 2011. Ulmer, R.G., Ferguson, S.A., Williams, A.F., Preusser, D.F., 2001. Teenage crash reduction associated with delayed licensure in Connecticut. Journal of Safety Research 32 (1), 31–41. Williams, A.F., Shabanova, V.I., 2003. Responsibility of drivers, by age and gender, for motor-vehicle crash deaths. Journal of Safety Research 34 (5), 527–531. Williams, A.F., Shults, R.A., 2010. Graduated driver licensing research, 2007 – present: a review and commentary. Journal of Safety Research 41 (2), 77–84.
Contents lists available at ScienceDirect
Accident Analysis and Prevention journal homepage: www.elsevier.com/locate/aap
The effect of the learner license Graduated Driver Licensing components on teen drivers’ crashes Johnathon Pouya Ehsani a,b,∗ , C. Raymond Bingham a,b , Jean T. Shope a,b a b
University of Michigan School of Public Health, Ann Arbor, MI, United States University of Michigan Transportation Research Institute, Ann Arbor, MI, United States
a r t i c l e
i n f o
Article history: Received 8 February 2013 Received in revised form 7 May 2013 Accepted 1 June 2013 Keywords: Graduated Driver Licensing Learner permit Teen driving Motor vehicle crash
a b s t r a c t Background: Most studies evaluating the effectiveness of Graduated Driver Licensing (GDL) have focused on the overall system. Studies examining individual components have rarely accounted for the confounding of multiple, simultaneously implemented components. The purpose of this paper is to quantify the effects of a required learner license duration and required hours of supervised driving on teen driver fatal crashes. Methods: States that introduced a single GDL component independent of any other during the period 1990–2009 were identified. Monthly and quarterly fatal crash rates per 100,000 population of 16- and 17-year-old drivers were analyzed using single-state time series analysis, adjusting for adult crash rates and gasoline prices. Using the parameter estimates from each state’s time series model, the pooled effect of each GDL component on 16- and 17-year-old drivers’ fatal crashes was estimated using a random effects meta-analytic model to combine findings across states. Results: In three states, a six-month minimum learner license duration was associated with a significant decline in combined 16- and 17-year-old drivers’ fatal crash rates. The pooled effect of the minimum learner license duration across all states in the sample was associated with a significant change in combined 16- and 17-year-old driver fatal crash rates of −.07 (95% Confidence Interval [CI] −.11, −.03). Following the introduction of 30 h of required supervised driving in one state, novice drivers’ fatal crash rates increased 35%. The pooled effect across all states in the study sample of having a supervised driving hour requirement was not significantly different from zero (.04, 95% CI −.15, .22). Conclusion: These findings suggest that a learner license duration of at least six-months may be necessary to achieve a significant decline in teen drivers’ fatal crash rates. Evidence of the effect of required hours of supervised driving on teen drivers’ fatal crash rates was mixed. © 2013 Elsevier Ltd. All rights reserved.
1. Introduction Motor vehicle crashes are the leading cause of death and a major contributor to nonfatal injury of teens in the United States (National Highway Traffic Safety Administration, 2010b). To address this public health threat, Graduated Driver Licensing (GDL) laws have been enacted by all fifty states and the District of Columbia (Williams and Shults, 2010). GDL was introduced in the United States beginning in the mid-1990s, replacing laws that allowed quick and easy access to full-privilege licenses. GDL laws vary in their requirements, but commonly include two levels that impose restrictions on teens’ driving (Foss and Goodwin, 2003). The first is a learner license that allows teens to gain driving experience under the supervision of a
∗ Corresponding author at: University of Michigan School of Public Health, Ann Arbor, MI, United States. Tel.: +1 734 763 9938. E-mail address: [email protected] (J.P. Ehsani). 0001-4575/$ – see front matter © 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.aap.2013.06.001
fully-licensed driver (typically a parent or parent designate over age 21). The second is an intermediate license, which allows teens who have gained some initial experience driving with a learner license to drive independently but with restrictions that limit their exposure to the highest risk driving conditions: driving at night (Williams and Shabanova, 2003) and driving with young passengers (Chen et al., 2000). There is little question that GDL reduces 16- and 17-year-old driver crashes (Masten et al., 2011; Shope, 2007); however, the elements responsible for the greatest reductions in crashes and the mechanism by which these reductions are achieved are not well understood. Evaluations of individual components of GDL have rarely accounted for the confounding effect of multiple GDL components being implemented simultaneously, and have assumed independent implementation of each component, which does not reflect the reality of how the majority of these laws were introduced. Further, most evaluations used a pre- and post-GDL study design that is unable to distinguish whether a decline in crashes was
328
J.P. Ehsani et al. / Accident Analysis and Prevention 59 (2013) 327–336
directly attributable to GDL, or the result of a preexisting downward trend (Elliot and Shope, 2003; Sivak and Schoettle, 2010). Currently in the U.S., learner license requirements are the most widely implemented of all GDL components, existing individually or side-by-side with other components in all fifty states and the District of Columbia. However, evidence of the effectiveness of the details of the learner license requirements is not well established. For example, the optimal number of months a learner license should be held has not been determined. Studies in Kentucky, Connecticut, and Nova Scotia indicated substantial crash reductions for 16-year-old drivers when a learner license period was newly mandated or an existing period was extended (Agent et al., 2001; Mayhew et al., 2003; Ulmer et al., 2001). While these findings indicated an extension of the learner license period reduced crashes, none of these studies used licensure data, so it is unclear whether a delay in the age of independent driving, or improvements in driving ability due to an extension in supervised driving beyond the required minimum were responsible for the crash reductions. Furthermore, little is known about the optimal number of months a learner license should be held for the best safety benefit. In all three states above, the learner license was mandated to be six months; however, there is no evidence to suggest whether or not a sixmonth period of supervised driving is adequate (Foss, 2007). For example, for the states discussed above, it is unknown whether a doubling of the learner license period (to twelve months) would have resulted in the same or a larger crash reduction. Similarly, the safety effect of a required number of supervised driving hours on teen drivers is also poorly understood. The small body of research examining the subject is inconclusive. A study of Swedish teens found an average of 120 h of supervised driving was associated with a significant reduction in crash involvement during independent licensure, compared to those who had approximately 50 h of supervised driving practice (Gregersen et al., 2000, 2003; Sagberg and Gregersen, 2005). Teen drivers in the northeastern U.S. who completed a period of supervised driving, however, were no different in their time-to-first-crash from those who did not have supervised driving experience (McCartt et al., 2003). Similarly, French teens who received professional driving instruction with an extensive period of supervised driving (equivalent to approximately 3000 miles) had the same crash likelihood as teens who only received professional driving instruction (Page, 2004). Due to the small number of studies, it is not possible to determine whether or not 120 h is the optimal number of supervised driving hours. The purpose of this paper is to quantify the effect of two required GDL components of the learner license on 16- and 17-year-old drivers’ fatal crash rates: the length of the learner license (months) and the number of supervised driving hours. 1.1. Research hypotheses This study tested the following hypotheses: 1. The introduction of a learner license required minimum holding period as part of GDL will be followed by a reduction in 16- and 17-year-old drivers’ fatal crash rates. 2. The introduction of a minimum number of required supervised driving hours as part of GDL will be followed by a decline in 16and 17-year-old drivers’ fatal crash rates. 2. Method 2.1. Inclusion criteria To test these hypotheses, states that introduced a learner license holding period, or required hours of supervised driving independent
of other GDL components during the period 1990–2007, were identified (Tables 1 and 2). The evaluation period spanned 1990–2009, however, because at least two years of data post-implementation were required to estimate the effect of a component, the sample was limited to states introducing GDL components prior to December 31st 2007. States were also excluded from the sample if they introduced multiple GDL components simultaneously with the component of interest, or had a learner license age below 15. 2.2. Data and measures Monthly counts of fatal crashes involving at least one teen driver (aged 16 or 17 years) in cars, trucks/pickups, vans/minivans, and sport utility vehicles were obtained for the contiguous period 1990–2009 from the Fatality Analysis Reporting System (FARS) for the states being analyzed (National Highway Traffic Safety Administration, 2010a). Ideally, data from all injury crashes (not just fatal crashes) occurring in each candidate state would also be included; however, only a limited number of states make their injury crash data available to researchers (National Highway Traffic Safety Administration, 2011), so such an approach could not be taken for this study. Furthermore, injuries are not recorded consistently across states or sometimes even across police agencies within a state. Fatal crash rates would ideally be based on the number of licensed teen drivers, however, licensure data reported by the Federal Highway Administration underreport the actual number of licensed teens, and licensure data are difficult to obtain from individual states (Insurance Institute for Highway Safety, 2006). Miles driven by each teen would also be ideal, but are difficult to measure and are unavailable. Therefore, crash rates were based on the number of teens in the overall population. Annual population estimates by state and year of age were obtained from the U.S. Census Bureau (Bureau of the Census US Department of Commerce, 1999, 2010). Monthly values were interpolated using cubic spline curves, which are the smoothest curve that exactly fits a set of data points (Bartels et al., 1998). The combined monthly fatal crash involvement rates of 16- and 17-year-old drivers per 100,000 population were calculated using monthly fatal crash counts and monthly population estimates. Data for drivers younger than 16 years were excluded because only a few states allow unsupervised driving by 15 year olds (Insurance Institute for Highway Safety, 2012). Several states in the sample had relatively small populations, increasing the probability of a floor effect, where crash rates cannot take on a value lower than zero. To compensate for this effect, states with a 16- to 17-year-old population below 85,000 (Maine, Nebraska, New Hampshire, Rhode Island, and Utah) were modeled using quarterly data. Quarterly fatal crash involvement rates were calculated using the monthly crash counts and population estimates. 2.3. Covariates 2.3.1. Comparison population The monthly fatal crash rate for drivers age 25–54 was used as a covariate, representing crashes for the typical adult driving population. Applying the identical method used to estimate 16- and 17-year-old fatal crash rates, monthly fatal crash rates of 25- to 54-year-old drivers per 100,000 population were calculated using monthly fatal crash counts and monthly population estimates. The purpose of the comparison population was to adjust for variability in the teen driver crash rates due to extraneous factors affecting drivers of all ages and to test the effect of GDL against a comparison population of persons unaffected by GDL. Although time series analyses control for pre-existing secular trends in crash rates, the inclusion of the crash rates of another age group as a historical
J.P. Ehsani et al. / Accident Analysis and Prevention 59 (2013) 327–336
329
Table 1 Percentage change in crash rate by state, following introduction of learner license.* State
Connecticut Hawaii Kentucky Minnesota South Carolina Tennessee Virginia Utah a b c *
Effective date
January July October February July January July August
Learner license duration
1997 1997 1996 1997 1998 1996 1996 2006
6 months 3 months 6 months 6 months 3 months 3 months 6 months 6 months
Fatal crash rate per 100,000 population Pre-introductiona
Post-introductionb
2.17 1.85 4.61 2.76 3.76 4.95 2.94 2.58
1.65 1.36 4.10 2.59 3.75 4.86 2.62 1.39
Percentage change in fatal crash ratesc
−16.6 −10.6 −8.9 −18.9 −2.8 0.8 −5.5 10.1
Prior to the introduction of the learner license. Until the introduction of new GDL component(s). Values based on ARIMA. Statistically significant percentages highlighted in bold.
Table 2 Percentage change in crash rate by state, following introduction of required supervised practice driving hours.* State
Effective date
Arizona Kentucky Maine Minnesota New Hampshire Rhode Island
January October January January September July
a b c *
Required supervised practice driving hours
2000 2006 1998 1999 1999 2003
25 h 60 h 35 h 40 h 20 h 50 h
Fatal crash rate per 100,000 population Pre-Introductiona
Post-Introductionb
3.27 4.31 3.07 2.59 2.55 1.56
2.54 3.25 3.33 2.46 2.27 1.46
Percentage change in fatal crash ratesc
−17.7 −4.2 −13.7 34.1 5.6 10.6
Prior to the introduction of the required supervised practice driving hours. Until the introduction of new GDL component(s). Values based on ARIMA. Statistically significant percentages highlighted in bold.
covariate to control for unmeasured factors that affect all drivers enhances the validity of the findings. 2.3.2. Gasoline prices An inverse relationship between gasoline prices and fatal crashes has been identified for drivers of all ages (Sivak and Schoettle, 2010); however, research suggests teen driving behavior may be more sensitive to higher gasoline prices, relative to older drivers (Morrisey and Grabowski, 2010). Monthly national average gasoline prices, obtained from the U.S. Energy Information Administration (U.S.Energy Information Administration, 2011), were used as a covariate in the analyses to adjust for their effect on the amount of driving exposure and resulting crash risk level. 2.3.3. GDL effective date The points at which GDL components were expected to affect crash rates were determined from information about each state’s GDL system, obtained from the Insurance Institute for Highway Safety’s GDL effective date database (Insurance Institute for Highway Safety, 2012) and each state’s website (Florida Department of Highway Safety and Motor Vehicles, 2009; Maryland Motor Vehicle Administration, 2012; Michigan Secretary of State, 2012). 2.4. Analytical method Monthly crash rates per 100,000 population of 16- and 17year-old drivers were analyzed using Auto-Regressive Integrated Moving Average (ARIMA) interrupted time series analysis for each state. The AR component represents the lingering effects of previous observations, while the MA component represents lingering effects of previous random shocks (or error) (McCleary and Hay, 1982). This approach accounts for trend, seasonality and
time-related autocorrelation in the data and eliminates the need to exclude the data points prior to and immediately after the implementation of GDL. It is superior to other before-after evaluation methods that do not take these factors into account, and exclude data points immediately before and after the introduction of GDL. Within ARIMA, a transfer function relates the enactment of GDL laws to crash rates using two parameters. The first parameter, ω, is the magnitude of the change (rise or fall) in crash rates. If ω is statistically significant, the size of the change is directly interpretable as the percentage change in the post-intervention series relative to the pre-intervention series using the formula 100 × [eω − 1] (McDowall et al., 1980; Tabachnick and Fidell, 2007). The second parameter, ı, reflects the onset of the change. For these analyses, ı was fixed at 0, meaning the anticipated change in crash rates would be abrupt and lasting. All models were estimated using the natural logarithm of the crash rates. Complete details of the model parameters can be found in Appendices A and B.
2.5. Analytical strategy The analyses were conducted in three stages. First, a linear regression model was estimated for each state for combined 16and 17-year-old driver crash rates and the covariates: adult crash rates, gasoline prices, and GDL laws. Second, the model for each state was statistically adjusted for trends and seasonal variation. Autoregressive and moving average orders were identified using auto-correlation and partial-auto-correlation functions of the series residuals. The original regression model was re-estimated with the inclusion of the autoregressive or moving average orders identified in the second stage. Outliers were also detected and controlled in each model. Analyses were conducted using the SCA
330
J.P. Ehsani et al. / Accident Analysis and Prevention 59 (2013) 327–336
Fig. 1. Pooled effect of the learner license duration on 16- and 17-year-old fatal driver crash rates. (1) Square area around the point estimates corresponds to the weighting of the state in the overall model. (2) Diamond represents the pooled point estimate with 95% Confidence Interval.
Time Series and Forecasting System, a time-series analysis software package (Scientific, 2011). The effects of each GDL component on combined 16- and 17year-old drivers’ crashes were pooled across states using a random effects meta-analytic model. This approach assumes differences in the implementation and effectiveness of GDL in each state, and takes these differences into account as additional sources of variation (Deeks et al., 2008). These analyses were conducted using Comprehensive Meta-Analysis software (Borenstein et al., 2005). 3. Results 3.1. Learner license duration The results of the interrupted time series analysis of the effect of the learner license duration requirement partially confirm the first hypothesis. In three of the five states that introduced a six-month learner license period, there was a significant reduction in fatal teen driver crash rates. Specifically, there were significant reductions in fatal teen driver crash rates following the introduction of a sixmonth learner license in Connecticut, Minnesota, and Virginia of 16.6%, 18.9% and 5.5%, respectively (Table 1). There were no significant changes in the fatal crash rates of 16- and 17-year-old drivers in states that introduced a minimum learner license duration of three months. Adult fatal crashes accounted for some of the variability in teen fatal crashes for each state in the sample, with the exception of Hawaii and Connecticut. When the effect of a minimum learner license duration was pooled across all states in the sample, there was a statistically significant change in combined 16- and 17-year-old driver fatal crash rates of −.07 (95% Confidence Interval [CI] −.11, −.03). Fig. 1 depicts the pooled effect of the learner license duration on 16- and 17-year-old drivers’ fatal crash rates. 3.2. Supervised driving hours The introduction of required hours of supervised driving was not followed by a significant reduction in teen drivers’ fatal crashes in any state included in the sample (Table 2), and therefore the second hypothesis were rejected. In Minnesota, the introduction of 30 h of required supervised practice driving corresponded with a significant 34.5% increase in combined 16- and 17-year-old drivers’ fatal
Fig. 2. Pooled effect of required supervised driving hours on 16- and 17-year-old driver fatal crash rates. (1) Square area around the point estimates corresponds to the weighting of the state in the overall model. (2) Diamond represents the pooled point estimate with 95% Confidence Interval.
crash rates. Adult fatal crashes predicted some of the variability in teens’ fatal crashes in each state except New Hampshire. The pooled effect of supervised driving hours across all states in the study sample was not significantly different from zero (.04, 95% CI −.15, .22) in the random effects meta-analytic model (Fig. 2). 4. Discussion The purpose of this study was to examine the independent effects of a required learner license duration and required hours of supervised driving on teen drivers’ fatal crashes. Based on the pooled estimates of the meta-analytic models, the introduction of the learner license minimum holding period resulted in a significant decline in teen drivers’ fatal crash rates, while the introduction of a minimum number of required supervised driving hours had no effect. It could be argued that one primary mechanism for crash reductions in teen drivers’ fatal crash rates may be a delay in the age of independent driving. This explanation is consistent with the findings that the six-month learner license resulted in a significant decline in 16- and 17-year-old drivers’ fatal crashes in three states, and the three-month learner license may have reduced crash rates, but less than the six-month learner license (did not reach statistical significance). It is also consistent with the finding that supervised driving hour requirements had no effect on fatal crash rates. However, this interpretation is complicated by the relationship between the licensing age and the learner license duration that was introduced in each state (Williams, 2007). For example, in states where the minimum age for both the learner license and the intermediate license are identical (e.g., 16 years), adding a learner license holding period guarantees a delay past the minimum age for an intermediate license. A single state in the study sample (Connecticut) fit these criteria. Whereas in Virginia and Minnesota, where a delay was not guaranteed, significant declines in 16- and 17-year-old driver fatal crashes were also observed. These observations suggest that delayed licensure may contribute, but is not entirely responsible for the crash reductions associated with the learner license components of GDL. The sample for this study was limited to states that introduced a single GDL component independently of other components. In some of the states, components were introduced within an existing GDL system. Consequently, the additive effect of multiple co-existing GDL components could not be disentangled. Significant changes in fatal crash rates were observed in large states,
J.P. Ehsani et al. / Accident Analysis and Prevention 59 (2013) 327–336
suggesting that the absence of an effect in smaller states could be due to insufficient power. However, estimating the models for small states using quarterly data did not alter the results that were based on monthly data. Crash data for 16- and 17year-old drivers were combined because using each year of age would have resulted in data that were too sparse to permit meaningful interpretation (Kochanek et al., 2011). This also provided a consistent outcome measure across all states in the study. It should also be noted that crash data for this sample were limited to fatal crashes. Some have argued that fatal crashes represent a small and atypical subset of all crashes, and the etiology of fatal crashes may differ from that of less serious crashes (Lam, 2003). This study investigated the effect of two required GDL components of the learner license on 16- and 17-year-old drivers’ fatal crash rates. This age group, however, includes teenage drivers who are at different stages of the licensing process. The rationale for selecting this age group was to evaluate the effect of the learner license components on early driving, particularly independent driving. While the licensure status of the 16- and 17- year-olds involved in fatal crashes could not be determined, epidemiological evidence indicates that driving under supervision during the learner license stage entails a very low crash risk (Mayhew, 2003). Therefore, it is likely that the majority of fatal crashes included in
331
these analyses occurred while teens where driving using an intermediate license, which allows teens to drive independently but with restrictions. In addition, compliance and enforcement of the GDL components under investigation were not accounted for in this study. Future research examining the effect of GDL components on teen drivers should extend the analysis to include all crash types: property-damage-only crashes and injury crashes, as well as fatal crashes. Such data would allow an examination of the effects of GDL on crash types of differing severity. In addition, studies should use state level licensure data to take into account the proportion of teens licensed at each year of age, and the GDL stage associated with each individual driver’s crash record. With this more detailed information, time-to-event analysis could be used to compare first-time crash incidence at each level of licensure and determine whether the primary mechanism for crash reductions is delayed licensure or safer driving. This study adds to the literature by presenting new information because GDL evaluations to date have largely measured the effect of several requirements imposed at the same time, rather than isolating the effects of specific components. Using this analytical approach, the learner license duration was found to be a critical component of GDL, and a six-month learner license duration seems to be the minimum period necessary to be associated with a significant decline in teen drivers’ fatal crash rates.
332
Appendix A. Parameters of ARIMA models estimating the effect of learner license duration on 16- and 17-year olds’ fatal crash rates per capita, 1990–2009 Model component
Parameter (Lag)
Estimate
Connecticut
6 month learner license Effective January 1997 No passengers for first three months (with exceptions) Effective October 2003 20 h supervised driving 12 a.m.–5 a.m. driving restriction No passengers for first six months (with exceptions) Effective October 2005 40 h supervised driving No passengers for first 12 months (with exceptions) Effective August 2008 Control series (25–54 year-olds) Gasoline price Noise (AR or MA component) Constant
ω
−.1820
.04
ω
−.0067
.97
ω
−.1128
.56
ω
−.3240
.05
ˇ ˇ None
.3730 −.0004
.09 .97
.7112
<.01
ω
−.1122
.35
ω
−.0310
.90
ˇ ˇ None
.1957 −.0012
.33 .39
.6307
.01
ω
−.0942
.15
ω
−.0433
.81
ω
−.0166
.93
ˇ ˇ None
.4850 −.0014
.01 .02
1.2077
<.01
ω
−.2092
.03
ω
.2931
<.01
ω
−.2698
.02
ˇ ˇ None
.5796 −.0014
<.01 <.01
.8546
<.01
−.0283
.75
Hawaii
Kentucky
Minnesota
South Carolina
3 month learner license Effective July 1997 6 month learner license 11 p.m.–5 a.m. driving restriction One passenger younger than 18 Effective September 2003 Control series (25–54 year-olds) Gasoline price Noise (AR or MA component) Constant 6 month learner license Effective October 1996 60 h supervised driving Effective October 2006 12 a.m.–6 a.m. driving restriction One passenger younger than 20 Effective April 2007 Control series (25–54 year-olds) Gasoline price Noise (AR or MA component) Constant 6 month learner license Effective February 1997 30 h supervised driving Effective January 1999 12 a.m.–5 a.m. driving restriction One passenger younger than 20 for first 6 months Three passengers younger than 20 for second 6 months Effective August 2008 Control series (25–54 year-olds) Gasoline price Noise (AR or MA component) Constant 3 month learner license Effective July 1998
ω
p
J.P. Ehsani et al. / Accident Analysis and Prevention 59 (2013) 327–336
State
6 month learner license 40 h supervised driving 6 p.m.–6 a.m. ESTa driving restriction 8 p.m.–6 a.m. EDTb driving restriction Two passengers younger than 21 Effective September 2003 Control series (25–54 yr-olds) Gasoline price Noise (AR or MA component) Constant Tennessee
Utah
6 month learner license Effective July 1996 Three passengers younger than 16 Effective July 1998 12 a.m.–4 a.m. driving restriction 9 month learner license 40 h supervised driving One passenger younger than 18 until age 17 Three passengers younger than 18 thereafter Effective July 2001 One passenger younger than 18 for first 12 months Three passengers younger than 18 thereafter Effective July 2003 45 h supervised driving Effective July 2008 Control series (25–54 yr-olds) Gasoline price Noise (AR or MA component) Noise (AR or MA component) Noise (AR or MA component) Constant 12 a.m.–6 a.m. driving restriction 30 h supervised driving Effective July 1999 No passengers for first 6 months Effective July 2001 30 h supervised driving Effective July 2003 40 h supervised driving Effective July 2004 6 month learner license Effective August 2006 Control series (25–54 year-olds) Gasoline price Noise (AR or MA component) Constant
−.0512
.63
ˇ ˇ MA (12)
.4586 −.0016 −.1293 1.0368
.01 .01 .05 <.01
ω
.0083
.90
ω
−.2063
.01
ˇ ˇ None
.8871 −.0010
<.01 .06
.6486
.01
ω
−.0562
.04
ω
.0390
.15
ω
−.0044
.86
ω
−.0154
.62
ω
−.0257
.43
ˇ ˇ MA (1) MA (6) AR (1) None
.5128 −.0002 .7223 .2422 .6592
<.01 .25 <.01 <.01 <.01
ω
.0295
.83
ω
−.0881
.60
ω
.1449
.49
ω
.0890
.70
ω
.0967
.58
.6120 −.0030
<.01 .03
.8262
<.01
ˇ ˇ None
J.P. Ehsani et al. / Accident Analysis and Prevention 59 (2013) 327–336
Virginia
3 month learner license Effective January 1996 6 month learner license 50 h supervised driving 11 p.m.–6 a.m. driving restriction Single passenger restriction Effective July 2001 Control series (25–54 year-olds) Gasoline price Noise (AR or MA component) Constant
ω
a
333
Eastern standard time. b Eastern daylight time.
334
Appendix B. Parameters of ARIMA models estimating the effect of supervised driving hours on 16- and 17-year-olds’ fatal crash rates per capita, 1990–2009 Model component
Parameter (Lag)
Estimate
p
Arizona
25 h supervised driving Effective January 2000 5 month learner license Effective July 2000 6 month learner license 30 h supervised driving 12 a.m.–5 a.m. driving restriction One passenger younger than 18 for first 6 months Effective July 2008 Control series (25–54 yr-olds) Gasoline price Noise (AR or MA component) Constant
ω
−.1955
.27
ω
.0597
.74
ω
−.2520
.05
ˇ ˇ None
.6422 −.0006
<.01 .29
.6250
.01
ω
−.0942
.15
ω
−.0433
.81
ω
−.0166
.93
ˇ ˇ None
.4850 −.0014
.01 .02
1.2077
<.01
ω
−.1492
.43
ω
.0173
.94
ω
.2022
.39
.4254 −.0026
.04 .07
.9811
<.01
Kentucky
Maine
6 month learner license Effective October 1996 60 h supervised driving Effective October 2006 12 a.m.–6 a.m. driving restriction One passenger younger than 20 Effective April 2007 Control series (25–54 yr-olds) Gasoline price Noise (AR or MA component) Constant 35 h supervised driving Effective January 1998 No passengers for first 3 months Effective August 2000 6 month learner license 12 a.m.–5 a.m. driving restriction No passengers for first 6 months Effective September 2003 Control series (25–54 yr-olds) Gasoline price Noise (AR or MA component) Constant
ˇ ˇ None
J.P. Ehsani et al. / Accident Analysis and Prevention 59 (2013) 327–336
State
Minnesota
Rhode Island
3 month learner license 1 a.m.–5 a.m. driving restriction Effective January 1998 20 h supervised driving Effective September 1999 One passenger for first 6 months Effective January 2003 40 h supervised driving Effective June 2009 Control series (25–54 yr-olds) Gasoline price Noise (AR or MA component) Constant 6 month learner license 1 a.m.–5 a.m. driving restriction Effective January 1999 50 h supervised driving Effective July 2003 One passenger younger than 21 for the first 12 months Effective July 2005 Control series (25–54 yr-olds) Gasoline price Noise (AR or MA component) Constant
ω
−.2092
.03
ω
.2931
<.01
ω
−.2698
.02
ˇ ˇ None
.5796 −.0014
<.01 <.01
.8546
<.01
ω
−.0820
.70
ω
.0542
.82
ω
−.0268
.90
ω
.0052
.99
.2238 −.0011
.26 .39
.8102
<.01
ω
−.0639
.64
ω
.1006
.63
ω
−.0956
.73
ˇ ˇ None
.3132 −.0009
.02 .95
.3022
.17
ˇ ˇ None
J.P. Ehsani et al. / Accident Analysis and Prevention 59 (2013) 327–336
New Hampshire
6 month learner license Effective February 1997 30 h supervised driving Effective January 1999 12 a.m.–5 a.m. driving restriction One passenger younger than 20 for first 6 months Three passengers younger than 20 for second 6 months Effective August 2008 Control series (25–54 yearr-olds) Gasoline price Noise (AR or MA component) Constant
335
336
J.P. Ehsani et al. / Accident Analysis and Prevention 59 (2013) 327–336
References Agent, K., Steenbergen, L., Pigman, J., Stinson Kidd, P., McCoy, C., Pollack, S., 2001. Impact of partial graduated driver’s license program on teen motor vehicle crashes in Kentucky. Transportation Research Record: Journal of the Transportation Research Board 1779, 54–60. Bartels, R.H., Beatty, J.C., Barsky, B.A., 1998. An Introduction to Splines for Use in Computer Graphics and Geometric Modelling. Morgan Kaufmann, San Francisco, CA. Borenstein, M., Hedges, L., Higgins, J., Rothstein, H., 2005. Comprehensive Metaanalysis Version 2. Biostat, Englewood, NJ. Bureau of the Census, US Department of Commerce, 1999. 1990–999 Annual Time Series of State Population Estimates by Age and Sex: 1990–1999. Bureau of the Census, US Department of Commerce, 2010. 2000–2009 State Single Year of Age and Sex Population Estimates. Chen, L., Baker, S.P., Braver, E.R., Li, G., 2000. Carrying passengers as a risk factor for crashes fatal to 16- and 17-year old drivers. JAMA: The Journal of the American Medical Association 283 (12), 1578–1582. Deeks, J.J., Altman, D.G., Bradburn, M.J., 2008. Statistical methods for examining heterogeneity and combining results from several studies in meta-analysis. BMJ Publishing Group. Elliott, M.R., Shope, J.T., 2003. Use of a Bayesian changepoint model to estimate effects of a graduated driver’s licensing program. Journal of Data Science 1, 43–63. Florida Department of Highway Safety and Motor Vehicles, 2009. License Requirements for Teens – Graduated Driver Licensing. 2012. Foss, R.D., 2007. Improving graduated driver licensing systems: a conceptual approach and its implications. Journal of Safety Research 38 (2), 185–192. Foss, R., Goodwin, A., 2003. Enhancing the effectiveness of graduated driver licensing legislation. Journal of Safety Research 34 (1), 79–84. Gregersen, N.P., Berg, H.Y., Engstrom, I., Nolen, S., Nyberg, A., Rimmo, P.A., 2000. Sixteen years age limit for learner drivers in Sweden: an evaluation of safety effects. Accident Analysis and Prevention 32 (5), 25–35. Gregersen, N.P., Nyberg, A., Berg, H.Y., 2003. Accident involvement among learner drivers – an analysis of the consequences of supervised practice. Accident Analysis and Prevention 35 (5), 725–730. Insurance Institute for Highway Safety, 2006. Motor vehicle registration and licensed driver information. Insurance Institute for Highway Safety, 2012. Effective dates of graduated driver licensing components. 2012. ˜ A.M., Kung, H., 2011. Deaths: final data Kochanek, K.D., Zu, J., Murphy, S.L., Minino, for 2009. National Center for Health Statistics 60 (3). Lam, L.T., 2003. Factors associated with young drivers car crash injury: comparisons among learner, provisional, and full licensees. Accident Analysis and Prevention 35 (6), 913–920. Masten, S.V., Foss, R.D., Marshall, S.W., 2011. Graduated driver licensing and fatal crashes involving 16- to 19-year-old drivers. JAMA: The Journal of the American Medical Association 306 (10), 1098–1103.
Maryland Motor Vehicle Administration. Rookie Driver Program. 2012 [cited 2012 3rd March]; Available from: http://www.mva.maryland.gov/ Driver-Services/RookieDriver/default.htm. Mayhew, D.R., 2003. The learner’s permit. Journal of Safety Research 34 (1), 35–43. Mayhew, D.R., Simpson, H.M., Desmond, K., Williams, A.F., 2003. Specific and long-term effects of Nova Scotia’s graduated licensing program. Traffic Injury Prevention 4 (2), 91–97. McCartt, A.T., Shabanova, V.I., Leaf, W.A., 2003. Driving experience, crashes and traffic citations of teenage beginning drivers. Accident Analysis and Prevention 35 (3), 311–320. McCleary, R., Hay, R.A., 1982. Applied Time Series Analysis for the Social Sciences. Sage Publications, Beverly Hills, CA. McDowall, D., McCleary, R., Meiringer, E.E., Hay, A.H., 1980. Interrupted Time Series Analysis. Sage Publications, Thousand Oaks, CA. Michigan Secretary of State, 2012. Graduated Driver Licensing. 2012. Morrisey, M.A., Grabowski, D.C., 2010. Gas prices, beer taxes and GDL programmes: effects on auto fatalities among young adults in the US. Applied Economics 43 (25), 3645–3654. National Highway Traffic Safety Administration, 2010a. Fatality Analysis Reporting System. National Highway Traffic Safety Administration, 2010b. Traffic Safety Facts 2009: A Compilation of Motor Vehicle Crash Data from the Fatality Analysis Reporting System and the General Estimates System. U.S. Department of Transportation. National Highway Traffic Safety Administration, 2011. SDS Overview. 2011. Page, Y., 2004. An evaluation of the effectiveness of the supervised driver-training system in France. Annual Proceedings at Association for the Advancement of Automotive Medicine 48, 131–145. Sagberg, F., Gregersen, N.P., 2005. Effects of Lowering the Age Limit for Driver Training. Elsevier Science, Amsterdam. Scientific, C.A., 2011. Time-series analysis and forecasting software. WorkBench Version 6.2.0. Shope, J.T., 2007. Graduated driver licensing: review of evaluation results since 2002. Journal of Safety Research 38 (2), 165–175. Sivak, M., Schoettle, B., 2010. Toward understanding the recent large reductions in U.S. road fatalities. Traffic Injury Prevention 11, 561–566. Tabachnick, B.G., Fidell, L.S., 2007. Using Multivariate Statistics. Allyn and Bacon, Boston. U.S.Energy Information Administration, 2011. Energy Prices. 2011. Ulmer, R.G., Ferguson, S.A., Williams, A.F., Preusser, D.F., 2001. Teenage crash reduction associated with delayed licensure in Connecticut. Journal of Safety Research 32 (1), 31–41. Williams, A.F., Shabanova, V.I., 2003. Responsibility of drivers, by age and gender, for motor-vehicle crash deaths. Journal of Safety Research 34 (5), 527–531. Williams, A.F., Shults, R.A., 2010. Graduated driver licensing research, 2007 – present: a review and commentary. Journal of Safety Research 41 (2), 77–84.