Shope JT, Molnar LJ, Elliott MR, Waller PF. Graduated Driver Licensing in MichiganEarly Impact on Motor Vehicle Crashes Among 16-Year-Old Drivers. JAMA. 2001;286(13):1593-1598. doi:10.1001/jama.286.13.1593
Author Affiliations: Transportation Research Institute (Drs Shope and Waller and Ms Molnar) and School of Public Health (Drs Shope and Waller), University of Michigan, Ann Arbor; and Center for Clinical Epidemiology and Biostatistics, University of Pennsylvania, Philadelphia (Dr Elliott).
Context Graduated driver licensing (GDL) programs are being adopted in many
states to address the high rate of motor vehicle fatalities among teens by
requiring teenaged drivers to gain experience and maturity under conditions
of relatively low crash risk before gaining full driving privileges.
Objective To evaluate the early impact of Michigan's GDL program on traffic crashes
among 16-year-old drivers.
Design, Setting, and Subjects Analysis of Michigan motor vehicle crash data from 1996 (before GDL
program implementation) vs 1998 and 1999 (after GDL program implementation)
for 16-year-olds, adjusting for trends among persons 25 years or older.
Intervention Michigan's GDL program, instituted April 1, 1997, for teens younger
than 18 years entering the driver license system, includes 3 licensure levels,
each with driving restrictions and requirements to progress to the next level.
Requirements include extended, supervised practice in the learning level,
2-phase driver education, and night driving restrictions in the intermediate
Main Outcome Measures Rates in 1996 vs 1998 and 1999 for all police-reported crashes; for
fatal injury, nonfatal injury, and fatal/nonfatal injury combined crashes;
for day, evening, and night crashes; for single-vehicle and multivehicle crashes;
and for alcohol-related crashes.
Results Overall, the rate of 16-year-old drivers (per 1000 population) involved
in crashes declined from 154 in 1996 to 111 in 1999 (relative risk [RR], 0.72;
95% confidence interval [CI], 0.71-0.73). After adjusting for populationwide
trends, the overall crash risk for 16-year-olds was significantly reduced
in 1999 from 1996 by 25% (adjusted RR, 0.75; 95% CI, 0.74-0.77). There were
also significant reductions for nonfatal injury and combined fatal and nonfatal
crashes; for day, evening, and night crashes; and for single-vehicle and multivehicle
crashes. Fatal crashes declined from 1996 to 1999, but not significantly (RR,
0.74; 95% CI, 0.49-1.14), and alcohol-related crashes continued at a low rate
(RR, 1.01; 95% CI, 0.80-1.29).
Conclusions Analysis of the first 2 full calendar years following Michigan's GDL
program implementation indicates substantial crash reductions among 16-year-olds.
Future research is necessary to determine if these reductions are maintained
and if other jurisdictions achieve similar results.
Unintentional injury from motor vehicle crashes is the leading cause
of death among US teenagers.1,2
Crash rates among drivers 16 to 19 years of age, per mile driven, are higher
than those for all other age groups,3 and the
crash risk among 16- to 17-year-old drivers is almost 3 times as high as among
18- to 19-year-old drivers.4 Yet, driver licensure
in the United States has been allowed at young ages and after only minimal
classroom education and behind-the-wheel training.
Two factors have likely contributed to the high traffic toll: young
driver inexperience and risk taking. The inexperience of young drivers makes
it difficult for them to recognize and respond to hazards, resulting in unsafe
driving practices, while their immaturity manifests itself in risky driving
practices such as speeding and tailgating.5
Driving is a complex psychomotor skill best acquired with considerable practice
initially conducted in low-risk situations as essential experience is gained.6 The concept of graduated driver licensing (GDL) addresses
youthful risk taking, which may result in traffic violations or crashes, by
limiting access to driving privileges and providing serious consequences,
such as curtailed license privileges, for driving infractions. Graduated driver
licensing thereby ensures that young novice drivers gain experience and maturity
under conditions of low risk before progressing to more risky driving situations.
Several key elements of a comprehensive GDL program for novice drivers
younger than 18 years have been recommended.7,8
First, licensure should be staged to phase drivers into on-the-road driving.
The stages should include a supervised learner's period, an intermediate licensing
stage that permits unsupervised driving only in less risky situations, and
a full-privilege license when the conditions of the first 2 stages have been
met. Second, the learner's stage should be long enough for adequate practice
in increasingly challenging situations and should require fairly extensive
adult supervision of that practice. Third, the intermediate stage, when independent
driving begins, should also be of substantial length and include restrictions
on such risky activities as driving at night and with teenage passengers.9 Fourth, programs should ensure competence before passage
through each stage, using an integrated 2-phase driver education program (1
phase before the first stage and 1 phase before the second stage), written
and road tests, and delays in progress if traffic violations or at-fault crashes
Knowledge about the effect of Michigan's comprehensive law, implemented
April 1, 1997, on young drivers' traffic crashes has been anticipated because
nearly all the recommended GDL components are included. The purpose of our
study was to evaluate the early impact of Michigan's GDL program through examination
of the crash rates of 16-year-old drivers before and after GDL implementation.
Prior to GDL, 16- and 17-year-olds in Michigan could obtain a full driver's
license 30 days after completing a driver's education course and passing a
written test. Under the new GDL program for drivers younger than 18 years
entering the licensing system on or after April 1, 1997, 3 levels of licensure
were instituted. Level 1 is the learner's stage (requiring extensive supervised
practice); level 2 is the intermediate stage (including a night driving restriction);
and level 3 is full licensure (although still under the state's existing 3-year
probationary period for all new licensees). Driver's education under GDL is
conducted in 2 phases, the first phase required before obtaining level 1 licensure
and the second phase required before obtaining level 2 licensure. Teens progress
through the levels only with parental approval, and parents can request that
their teen be retained at any level. Minimum requirements for obtaining licensure
under the previous licensing system and at each level of the new GDL program
are shown in Table 1.
The evaluation design of this administrative policy change, the introduction
of GDL, is a before and after comparison of crashes involving 16-year-old
drivers, using Michigan State Police crash data, which are available for each
calendar year. One full year before the implementation of GDL, 1996, was selected
to be compared with the 2 full calendar years following its implementation,
1998 and 1999. The year 1997 was omitted because of potentially unusual levels
of licensing activity taking place just before and just after the law was
implemented, as well as the time it took for new young drivers to acquire
their level 2 licenses.
In addition to before and after, comparisons were done between crashes
of 16-year-old drivers and crashes of drivers 25 years or older, to control
for the possibility that populationwide changes in crash rates caused observed
changes in crash rates of 16-year-old drivers.
Data on motor vehicle crashes in Michigan in 1996, 1998, and 1999 were
obtained from the Michigan State Police. The police-reported crash files in
Michigan for these years contained records for all crashes reported to local
or state police agencies. Crashes are required to be reported if there is
a personal injury or property damage of $400 or more. These requirements did
not change during the years under study. Injury severity data are coded in
Michigan crash reports using the "KABCO" injury severity scale.10K-level injuries are injuries caused by the crash that
result in death within 90 days of the incident. A-level injuries are incapacitating injuries that prevent injured persons
from walking, driving, or normally continuing the activities they were capable
of performing prior to the injury. B-level injuries
include nonincapacitating injuries that are evident to observers at the scene
of the crash in which the injury occurred. C-level
injuries are possible injuries reported or claimed that are not fatal, incapacitating,
or nonincapacitating evident injuries. O-level refers
to incidents in which no occupants were injured (property damage only). For
our analyses, K-level injuries were used for fatal crashes, and A-level, B-level,
and C-level injuries were used in combination for nonfatal injury crashes.
Outcomes of interest included all crashes, as well as various types:
fatal injury, nonfatal injury, and a combination of those crashes; day (5
AM to 8:59 PM), evening (9 PM to 11:59 PM), and night (midnight to 4:59 AM)
crashes (to capture the period covered by the GDL night restriction as well
as the evening hours that are not covered); single-vehicle and multivehicle
crashes; and alcohol-related crashes ("had been drinking" was on the police
Crash counts were examined, as well as crash rates based on the population
of 16-year-olds in Michigan for each year of interest, obtained from the US
Census Bureau.11 Crash rates per 1000 16-year-olds
in Michigan were used to control for changes in the size of the population.
Crash rates were based on population rather than licensed drivers because
of a difference in recording licenses before and after GDL. Before GDL, learner
permits were not included in the Michigan driver license database, whereas
after GDL, level 1 (learner) licenses were included. The study was approved
by the institutional review board at the University of Michigan. Informed
consent was not necessary because we used archived data that did not contain
private, identifying information.
To estimate the crude change in crash risks between 1996 and 1998 among
Michigan's 16-year-olds, relative risks (RRs) and 95% confidence intervals
(CIs) for each year for each crash type were computed, and the crash rates
after GDL implementation were compared with the crash rates before GDL.
To adjust for the possibility that populationwide changes in crash rates
were the underlying cause of observed changes in the 16-year-old crash rates
between 1996 and 1998, the RR ratios and CIs comparing the RRs between 1998
and 1996 of 16-year-olds and of those 25 years or older were computed, similar
to Mayhew et al12 and Ulmer et al.13 Similar analyses were conducted comparing the RRs
of the 2 groups between 1999 and 1996. This RR ratio is the net impact of
GDL assuming that changes in the crash risk of 16-year-olds in 1996 vs 1998
and 1999 would have exactly paralleled those 25 years or older in the absence
of GDL. Hence a ratio of less than 1 would be indicative of a positive effect
of GDL on the risk of crash for 16-year-olds relative to adults 25 years or
older. This approach is slightly limited by the fact that improved driving
and fewer crashes among 16-year-olds affects crash rates of all drivers to
the extent that 16-year-old drivers are involved in crashes with older drivers.
Nonetheless, it is a reasonable approach as a comparison and has been used
Analyses were performed with SAS version 8 (SAS Institute Inc, Cary,
Based on driver history data from a single point in time each year,
the proportion of the 16-year-old Michigan population who were licensed to
drive independently was 59.7% in 1996 (pre-GDL license), 38.1% in 1998 (6.3%
pre-GDL license plus 31.8% GDL level 2 license), and 37.5% in 1999 (0.2% pre-GDL
license plus 37.3% GDL level 2 license), resulting in a 22.2 percentage point
decrease from 1996 to 1999. It is also of interest to determine the mean age
that 16-year-olds became licensed. Again based on driver history data, the
mean age that licensed 16-year-olds had obtained a full license under the
old system in 1996 was 16.2 years (median, 16.0 years). The mean age of obtaining
a GDL level 2 license in 1998 was 16.3 years (median, 16.1 years), and in
1999 was 16.3 years (median, 16.2 years). The increased age of licensure among
16-year-olds to drive under GDL in 1999 represents about a 1.3-month delay
in licensure to drive independently compared with 1996.
The GDL system was clearly being implemented among new teen drivers.
In January 1998, 70.2% of 16-year-old drivers in Michigan had unrestricted
licenses under pre-GDL regulations, 28.6% were in GDL level 1, and 1.2% were
in GDL level 2. By January 1999, 6.8% of 16-year-old drivers in Michigan had
unrestricted licenses under pre-GDL regulations, 42.8% were in GDL level 1,
and 50.4% were in GDL level 2.
Data on the crash experience of 16-year-old Michigan drivers, before
and after the introduction of GDL, are presented in Table 2. Numbers of crashes, crash rates per 1000 16-year-old population,
and RRs of crash involvement for crashes overall and for selected subtypes
of crashes are included. Sixteen-year-olds in 1998 and 1999 had significantly
lower risks of being involved in a motor vehicle crash than 16-year-olds in
1996 (1998 RR, 0.74; 95% CI, 0.72-0.75; 1999 RR, 0.72; 95% CI, 0.71-0.73).
Reviewing just comparisons of 1999 with 1996, significant reductions in risk
of crash involvement were found for all subtypes of crashes except fatal and
alcohol-related crashes. The greatest reduction in risk occurred for night
crashes (RR, 0.43; 95% CI, 0.38-0.48), the period of time covered by the GDL
level 2 night driving restriction. Similar patterns were found when comparing
1998 with 1996, although reductions were slightly less in that first year
following the introduction of GDL.
Crash risks of 16-year-olds were compared with those of persons 25 years
or older to control for populationwide changes in crash rates that may have
contributed to changes in the 16-year-old crash rates (Table 3). Ratios comparing the RRs of 16-year-olds with persons
25 years or older between 1998 and 1996 and 1999 and 1996 were calculated.
The risk of overall crash involvement for 16-year-olds relative to those 25
years or older declined significantly from 2.17 in 1996 to 1.63 in 1999 (RR
ratio, 0.75; 95% CI, 0.74-0.77). Significant reductions were also found in
the risks of 16-year-olds relative to those 25 years or older for all subtypes
of crashes, except fatal and alcohol-related crashes. Fatal crashes declined,
but the number of fatal crashes was small. In contrast to other crash types
where 16-year-olds had higher rates of each type of crash compared with older
persons, it can be seen that 16-year-olds had and maintained much lower rates
of alcohol-related crashes in 1996, 1998, and 1999 (RR, 0.43, 0.42, 0.44,
respectively). Comparing the RR ratios in Table 3 with the RRs in Table
2 shows that controlling for populationwide trends tempered the
significant post-GDL reductions in risk of each crash type for 16-year-olds
by a factor of 3 to 9 percentage points.
The reductions in crash risk among Michigan's 16-year-olds following
implementation of GDL are impressive, even after the adjustment for populationwide
changes among persons 25 years or older tempered the impact somewhat. Taking
into account these general population trends, the risk of being involved in
a motor vehicle crash in 1999 was 25% lower than in 1996 for 16-year-olds,
a statistically significant difference. Significant reductions in risk were
also found for nonfatal injury crashes (24%) and combined fatal and nonfatal
injury crashes (24%); for day (24%), evening (21%), and night crashes (53%);
and for single-vehicle (29%) and multivehicle crashes (23%). The decrease
seen in the risk for fatal injury crashes (26%) was desirable, although not
statistically significant, likely due to the relatively small numbers of such
crashes. No change was observed in risk of alcohol-related crashes.
The sharp drop in risk of night crash involvement relative to day or
evening crash involvement suggests that the night driving (midnight to 5 AM)
restriction during the intermediate stage of licensure has been very effective.
The drop in evening (9 PM to midnight) crashes is less than the drop for overall
crashes, suggesting that a driving restriction that includes at least some
evening hours might enhance GDL's effect further, as recommended by Foss.9 The expected lack of change in alcohol-related crashes
over the 3-year period and low levels compared with drivers 25 years or older
were very likely due to Michigan's zero tolerance law, implemented in 1994,
whereby any alcohol involvement by teens can result in the loss of their driver
license. The 1999 crash risks were lower than the crash risks in 1998, as
expected. The new cohort of teen drivers subject to the GDL program had to
work its way into the 16-year-old driving population, with 1998 having about
two thirds of the 16-year-olds under the new licensure system, and 1999 having
nearly all the 16-year-olds under the system.
Some reduction in crashes could be explained by teens delaying GDL licensure.14 Indeed, Michigan experienced a 22.2 percentage point
decrease from 1996 to 1999 in the statewide proportion of 16-year-olds licensed
to drive independently. Reductions in crashes, however, exceeded that decrease,
especially in the case of night crashes. The analyses reported herein could
not assess the GDL level of drivers who experienced a crash. When this can
be done, a more precise understanding of the changes in crash risk can be
achieved. Sixteen-year-old drivers who obtained a license did so somewhat
later, an increase of about 1.3 months from 1996 to 1999 in age of licensure
to drive independently. Monitoring licensure levels and crashes over the next
few years will shed light on the contribution that delay in licensure makes
to crash reductions.
These results are among the first in the United States to evaluate a
complete changeover among teen novice drivers into a comprehensive GDL system
that includes a 2-phase driver education requirement and also requires certification
by a responsible adult that a young driver has received extended supervised
practice. Driver education completion is documented by the instructor for
segments 1 and 2 and presented to the secretary of state's driver license
office by the applicant when application is made for levels 1 and 2 licensure.
Parental certification of the supervised practice is required when application
is made for level 2. Michigan does not require the actual number of practice
hours to be recorded, but a survey of 814 parents of level 2 applicants in
July 199815 found that most parents reported
having spent more time (mean = 75.3 hours; 21.4 hours at night) supervising
their child's driving than the required 50 hours, with 10 hours at night.
Many had also kept a written record of their child's driving practice even
though that is not explicitly required by law. Parents were extremely positive
about the supervised practice requirement and the new GDL program in general.
Although no information is available about the hours of teens' driving practice
under the previous licensing system, it is likely to have been considerably
Although a comprehensive approach to GDL for young novice drivers was
proposed as early as the 1970s,16 only recently
has the concept begun to gain widespread acceptance in the United States.
Graduated driver licensing was first implemented in 1987 in New Zealand and
subsequently spread to Australia and Canada. In the United States, legislative
initiatives for GDL generally began in the late 1990s. Thirty-four states
and the District of Columbia in the United States have now adopted a 3-level
GDL system for new drivers, and other states and jurisdictions have some elements
of GDL but not 3-stage systems.17 Comparisons
among program evaluations are difficult because of differences in previous
laws, differences in GDL laws, and differences in the methodology used to
assess changes following GDL. Florida's GDL program, introduced in July 1996,
restricts night driving of 16-year-old intermediate licensees (11 PM to 6
AM) and limits the number of traffic violations that can be accumulated. Evaluation
of the program, using before and after comparisons and a reference group of
older persons unaffected by the program, found an 11% reduction in fatal and
injury crash rates among 16-year-olds between 1995 and 1997.13
A supervised driving requirement was not added until October 2000.17 North Carolina's GDL program, implemented in December
1997, restricts driving between 9 PM and 5 AM during the intermediate stage,
but does not contain a requirement for minimum hours of supervised driving.
Preliminary evaluation of the North Carolina program found a 26% reduction
in crashes among 16-year-olds in 1999 compared with 1997.14
Ohio's GDL law includes a night driving restriction, supervised practice,
and a passenger limitation. Preliminary evaluation found a 23% reduction in
the crash rate.18 Results from evaluations
of GDL programs outside the United States have indicated declines in crashes
among young novice drivers ranging from 7% to 37%.12,19- 21
There are several limitations to this analysis. First, other than for
night crashes, the individual effects of each GDL component could not be separated.
Instead, the evaluation has examined GDL's overall impact. Apparently, the
combined effect of all its components, including requirements for practice,
time in each stage of licensure, 2-phase driver education, clean driving record,
and night driving restriction, has resulted in substantially fewer crashes
among 16-year-olds in Michigan. Second, no driving mileage exposure data are
available for 16-year-olds before and after GDL to determine whether decreased
mileage contributes to the decreased crash risk. Exposure between midnight
and 5 AM is presumed to have been less due to the restriction in the law,
which parents and law officers are responsible to enforce. Third, additional
driving restrictions may have been imposed by some parents on teens, but such
data are unavailable. Indeed, the state's written material for parents encourages
this practice, sometimes in the form of a parent-teen contract. Fourth, and
perhaps the most unfortunate limitation, is that the GDL license level is
not recorded on the crash report, so that analyses of crashes by GDL level
cannot readily be conducted—each crash record must be individually matched
with the driver license file and the data on GDL levels by date. Fifth, it
was not feasible for the study design to include comparison states. But one
can compare Michigan's fatal crash rates with national fatal crash rates over
the study years (fatal crashes are the only types of crashes for which full
national counts are available). Using data for 16-year-old drivers from the
Fatal Analysis Reporting System (FARS)22 with
extrapolated estimates of the 16-year-old population from the US Census Bureau,11 fatal crashes occurred nationally at a rate of 0.34
per 1000 in 1996 and 0.29 per 1000 in 1999, a reduction of 15%. Michigan's
32% reduction rate in the 16-year-old fatal crash rate was twice the reduction
in the national rate.
Future analyses will address crash risk by GDL level, as well as length
of time and age in each GDL level. As more teen drivers move into and progress
through the Michigan GDL system, efforts to assess the longer-term effects
can be conducted. Analyses will extend crash data through 2000, at which time
every 16-year-old and nearly all 17-year-olds will have been subject to GDL.
The crash data on 17-year-olds will be of particular interest because if they
have a level 3 license, they can drive without a night restriction, and their
driving behavior will more completely reflect the full effect of the GDL system.
Crashes occurring during each evening hour can also be examined as a basis
for reviewing the effectiveness of the limited night restriction currently
in place.9 Future analyses will also examine
effects of the sex of the driver, as well as violations, safety belt use,
and the presence of young passengers23 on the
crash experience of teen drivers before and after the introduction of GDL
to determine whether these behaviors were positively, if indirectly, affected
by GDL. There is strong evidence that young male drivers and drivers transporting
passengers in their vehicles are at increased risk for crashes and death.23,24 Safety belt use of young drivers
is also of interest because failure to use safety belts may be part of a constellation
of risk-taking behaviors that can increase crash risk.25- 27
In conclusion, Michigan's comprehensive GDL program was a major change
from the previous teen driver licensing system, and substantial reductions
in motor vehicle crashes among 16-year-old drivers were subsequently observed
that may be attributable to GDL. While these early results in preventing crashes
among young drivers are impressive, evaluation over the longer term is needed
to see if the crash reductions are maintained as new cohorts of teen drivers
enter and progress through the system. In addition, evaluation results from
other jurisdictions are needed to determine if such positive results are achieved
elsewhere. Other jurisdictions' evaluations will also help assess the relative
contributions of the different components to the overall effect of GDL. This
task will be complex, given the variations among former licensing laws, as
well as the considerable differences in the GDL programs being adopted. Nonetheless,
a GDL system for licensing young drivers seems an extremely promising approach
to reducing injury from motor vehicle crashes among teenagers.