California ’ s Graduated Driver License Law : Effects on Older Teenagers Mike Males Center on Juvenile and Criminal Justice

Many traffic safety researchers believe Graduated Driver Licensing (GDL) laws save lives by imposing restrictions, stronger licensing requirements, and delayed licensure status on drivers under age 18. To determine the effects of California’s GDL law on traffic fatalities among older (age 18-19) and younger (age 16-17) teens by age, sex, accident characteristics, and license status, mortality data from California’s Center for Health Statistics, driver and accident data from the Fatality Analysis Reporting System, and population data from the California Department of Finance were analyzed for the 1995-2004 period. Compared to California who began driving before the GDL law took effect and to corresponding trends among Californians ages 20 through 44, 16and 17-year-olds subject the GDL law experienced net decreases of 13% in motor vehicle fatality rates and 14% in driver involvements in fatal accidents. However, 18and 19-year-olds subjected to GDL programs experienced net increases of 11% in traffic fatalities and 10% in involvements of drivers in fatal accidents, more than offsetting the declines among younger teenagers. These results support reassessment of the effects of the GDL law, including its specific requirements, on older teenagers. © 2006 Californian Journal of Health Promotion. All rights reserved.


Introduction
High rates of motor vehicle crashes and fatalities involving teenage drivers have been attributed to inexperience with driving and risk-taking due to immature age (Chen et al., 2000;Ulmer, Williams, & Preusser, 1997).To allow new drivers under age 18 to obtain experience while minimizing risks, states have replaced simple, single-stage drivers' licensing procedures that required brief training and testing with graduated drivers' licensing (GDL) laws, which provide lengthy, multistaged, supervised training and probationary periods.
California's Graduated Driver Licensing (GDL) law, effective July 1, 1998, is considered among the most restrictive of any state's (Masten & Hagge, 2003;Smith, Pierce, Ray & Murrin, 2001).As summarized in Appendix A, the law requires new teenage drivers to successfully complete a year-long, three-stage process to obtain a full privilege license (California Department of Motor Vehicles, 2006).
Preliminary studies report that GDL laws reduce traffic crashes and fatalities among 16-year-olds and, in some cases, 17-year-olds (Chen, Baker & Li, 2006;Foss, Feaganes & Rodgman, 2001;Morrisey, Grabowski, Dee & Campbell, 2006;Shope, Molnar, Elliott & Waller, 2001).Several studies, none citing data more recent than 2001, find that California's GDL law reduced fatalities and certain types of fatal crashes among 16year-olds (Cooper, Gillen & Atkins, 2004;Leaf, 2002;Rice, Peek-Asa & Kraus, 2004;Williams, Nelson & Automobile Club of Southern California, 2000, 2001), though one initial study warns that GDL laws may have unexpected consequences for older teenagers (Masten & Hagge, 2003).Unfortunately, few longer-term analyses exist (Dee, Grabowski, Morrisey, 2005;Hedlund Shults & Compton, 2003;Simpson, 2003).This study examines the effects of California's GDL law on fatalities involving teenagers and drivers ages 16-19 licensed before and after the GDL law took effect.If high rates of crashes among 16-17 year-old drivers are due mainly to immaturity and risk-taking inherent in young age, as most researchers argue, GDL laws would be expected to save teenage lives by delaying the age at which full licensing privileges are allowed (see Appendix A).

Data and Analysis
Data for this study include mortality vital statistics, driver and passenger involvements in fatal crashes, and population counts.California's Center for Health Statistics (1995Statistics ( -2004)) The outcome of interest is motor vehicle fatality rates involving California residents during the specified study period (1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004).Motor vehicle fatality victims listed by the CHS as having resided in California's 58 counties, and drivers and passengers involved in fatal accidents listed by FARS as residing in zip codes 90000 through 96162, are California residents.Drivers are cited here as licensed California drivers if they are coded by FARS as holding a valid California driver's license.CHS vital statistics, FARS fatal crash involvement, and Department of Finance population data are used to calculate and compare California teenagers' motor vehicle fatality rates per 100,000 population for ages 16 through 19, categorized by whether or not they were subject to the requirements of California's GDL law.

Mortality Statistics
Motor vehicle deaths and fatal traffic crash involvements for these teenaged groups are examined in two ways.The "birth cohort" analysis examines the 2,021 California residents ages 16 through 19 who were born from January 1, 1980, through December 31, 1984, who died in motor vehicle accidents during the 1996-2004 period.This cohort is divided into symmetrical subcohorts: (a) Prelaw (born January 1, 1980-June 30, 1982) who turned age 16 in the 30month period before the GDL law took effect, and (b) Postlaw (born July 1, 1982-December 31, 1984) who turned 16 in the first 30 months after the law took effect and were subjected to its licensing requirements.
The preliminary comparison of the motor vehicle fatality rates of the postlaw versus prelaw subcohorts by sex is presented in Appendix B and Appendix C. A change of less than one indicates a decline in postlaw fatality rates compared to prelaw rates.
The comparisons in Appendix B and Appendix C are described as preliminary because they do not control for larger, population-level trends caused by changes in seat belt, drunken driving, and other traffic safety laws, economic cycles, weather, and other factors that affect traffic death rates for all age groups.To control for these larger factors, motor vehicle fatality and driver involvement rates for each teenage group in the prelaw and postlaw cohorts are expressed as Incident rate ratios (IRRs) to the corresponding traffic death and driver involvement rates for California residents 20 through 44 (see Foss, Feaganes & Rodgman, 2001;Shope, Molnar, Elliott & Waller, 2001).The necessity of using IRRs expressing teenaged fatality rates as ratios to those of older residents, as opposed to just using simple rates, is evident in Appendix B: there has been a general increase in traffic fatality rates among all age groups over the 1995-2004 period that, if not factored out, would make it appear that the GDL program is associated with a major increase in teenage deaths.To accomplish this, a comparison set of traffic death rates among 20-44-year-olds is calculated for each teen age being examined to control for these larger factors.Note that for each 30-month subcohort, deaths among 16 year-olds will occur at the beginning of each subcohort's death period (from January 1, 1996, through June 30, 1999, for the prelaw subcohort, and from July 1, 1998, through January 31, 2002, for the postlaw subcohort), while those of 19-year-olds will occur at the end (from January 1, 1999, through June 30, 2002, for the prelaw subcohort, and from July 1, 2001, through December 31, 2004, for the postlaw subcohort), with those of 17-and 18-year-olds occupying time periods in between.Thus, the time periods of traffic deaths and populations for 20-44-yearolds used to calculate IRRs must match those of the teen age to which they are compared.The motor vehicle death counts and populations for each teen age and for the 20-44-year-olds used as the comparison group for each teen age are shown in Appendix B.
In Tables 1, Appendix D, and Appendix E, an IRR of less than one indicates the teen age has a lower risk of fatality than corresponding 20-44year-olds.The IRRs for postlaw versus prelaw subcohorts are compared by Adjusted IRRs, which express the difference in IRRs in the postlaw versus the prelaw period.An Adjusted IRR of less than one means that a teenage traffic fatality rate relative to the corresponding rate among 20-44 year-olds is lower in the postlaw period than in the prelaw period, indicating the GDL law is associated with a decline in teenage fatalities or fatal driver involvements.An Adjusted IRR is statistically significant if its Confidence Interval (CI, also shown) does not contain 1.00.

Driver involvement statistics
The second analysis uses FARS data to compare the IRRs of postlaw versus prelaw cohorts of the 4,806 California residents ages 16-19 involved as drivers in fatal motor vehicle crashes during 1995-2004.Because FARS does not provide birth dates of traffic crash victims that would allow true cohorts to be delineated as for mortality vital statistics, the prelaw cohort in this analysis consists of drivers involved in fatal accidents from 1995 through the year before the date their age first was subjected to GDL requirements, and the postlaw cohort consists of drivers in fatal crashes in the years after the date their age first was subjected to the GDL law through 2004.Crashes occurring in the 12 months before, and the 12 months after, the date on which each age is first subjected to the GDL law are not included for that age, since this "transition period" includes both prelaw and postlaw drivers and evidences temporary fluctuations in the licensing of new drivers (Foss, Feaganes & Rodgman, 2001;Shope, Molnar, Elliott & Waller, 2001).The GDL law's restrictions and requirements first applied to 16 year-olds on July 1, 1998; the first 16 year-old drivers licensed under its regulations would have been on the road beginning on January 1, 1999 (along with older 16-year-olds licensed before the law took effect); and all 16 year-olds driving after July 31, 1999, would have been subjected to the GDL law.Further, there is evidence of a surge in licensing of 16 year-olds prior to the GDL law taking effect.Thus, the prelaw period for 16 year-olds is January 1, 1995, through June 30, 1997, the transition period is July 1, 1997, through June 30, 1999, and the postlaw period is July 1, 1999, through December 31, 2004.Rates of fatal crash involvements are calculated for each age and compared for each cohort (Table 2).As for the birth cohort, the teenaged driver cohorts' postlaw versus prelaw IRRs, expressed as ratios to corresponding fatal crash involvement rates of California residents ages 20-44, are compared to produce Adjusted IRRs.These Adjusted IRRs form the basis for evaluating the changes in fatal accident involvements among teenaged drivers associated with the GDL law.
Appendix D shows the IRRs and Adjusted IRRs by gender and driver's license status for California teenaged residents involved as drivers in fatal crashes.Appendix E examines the effects on teenage fatal crash experience of three specific restrictions imposed by GDL lawsprohibitions on driving by novice drivers under age 18 (unless supervised by adults over age 25) alone, late at night (midnight to 5 a.m.), or with teenaged passengers.Also presented are fatal crash IRRs for teenaged drivers listed as driving improperly (which means those who tested positively for drug or alcohol intoxication, were mentioned on police reports for improper driving, or were formally cited for driving improperly) compared to rates of drivers ages 20-44 similarly noted for driving improperly during the same periods.Improper driving notations and citations often involve subjective judgment by officers and should be regarded as such.
The driver cohort analysis includes all deaths for each age in each year, 1995 through 2004, prior to and following the date of initial exposure to the law.The birth cohort analysis includes all deaths for each age level in adjacent, symmetrical populations differing only in their exposure versus non-exposure to the GDL requirements.The mortality and driver involvement cohorts examined in this study reflect the maximum, equivalent periods of prelaw and postlaw accident exposure that can be examined using the most recent data (2004).

Results
Tables 1, Appendix D, and Appendix E show the key results.After the GDL law was adopted, younger teenagers and female teenagers showed declines in motor vehicle deaths and fatal accident involvements while older teens and males showed increases.Because older teens and males suffer fatal traffic mishaps at considerably higher rates than younger teens and females, the overall result is that California's GDL law is associated with a small, nonsignificant increase in teenaged traffic deaths.
The change in teenage motor vehicle death rates after the GDL law took effect varied considerably by gender as well as by age.Under the GDL program, traffic death rates did not change significantly among teenaged girls but rose among boys (Adjusted IRR, 1.09; 95% CI 1.04-1.15).Among both males and females, a shift in fatalities from younger to older ages is evident.The changes are significant for 16-yearold females (down 28%), and males (down 20%), and for males ages 18 (up 34%).
Sixteen-year-old drivers show significant declines in crash involvements for both sexes.Nineteen-year-old drivers suffered significant increases for both males and females, as did 18year-old males.For both sexes, the decline among 16-year-olds was sufficient to produce significantly lower postlaw IRRs among 16-17year-olds (down 14%), while the increase among 18-year-olds and males produced a significant rise for 18-19-year-olds (up 10%).
The GDL law was also followed by significant increases in involvements in fatal crashes by unlicensed teenaged drivers.Before the GDL law took effect, 19.4% of drivers under age 18 and 22.5% ages 18-19 involved in fatal crashes were unlicensed; after the law, 28.9%, and 25.8%, respectively.Among drivers 20-44, the proportion of fatal crashes involving unlicensed drivers remained stable at around 24% during the period.However, licensed teenaged drivers did not consistently fare better than unlicensed ones after the GDL law.Postlaw 16-17 year-olds show a significant decrease in fatal crash involvement rates by licensed drivers and a significant increase in such accidents among unlicensed drivers.Teenagers ages 18-19 show significant postlaw increases in both licensed and unlicensed-driver crash IRRs.Appendix E evaluates three common restrictions associated with GDL laws by license status.The pattern is unexpected.Fatal crashes involving single drivers (that is, ones unsupervised by adults) were higher among California-licensed teenaged drivers of all ages after the GDL took effect (significantly so for 18-year-olds and for unlicensed older teens), but not for unlicensed 16-17-year-olds.Late-night fatal crash IRRs rose sharply among 16-17-year-olds legally licensed under GDL laws, though not among 16-17-year-olds driving without licenses or among 18-19-year-olds.Crashes involving teenaged drivers with teenaged passengers declined significantly for 16-17-year-olds after the GDL program took effect; no change occurred among older teens, and there was no difference in trend between licensed and unlicensed drivers.Finally, postlaw crash IRRs involving improper driving were significantly lower among younger teens, both licensed and unlicensed, but were significantly higher among 18-19-year-olds, including those with valid California licenses.

Fatality changes associated with California's GDL law
Applying the significant IRRs and confidence intervals for traffic fatality trends, California's GDL law is associated with an annual average of 20 fewer traffic deaths (95% CI, 9-29) and 23 fewer driver fatal crash involvements (95% CI, 16-31) among 16-17-year-olds, offset by 24 more traffic deaths (95% CI, 11-38) and 28 more driver involvements in fatal crashes (95% CI, 17-39) among 18-19-year-olds.These fatality numbers are not additive due to the different time periods involved in their generation.The most reasonable conclusion from these data is that California's GDL program is associated with no effect on teenage traffic fatality or fatal crash risks.

Limitations
California's youth population, though large, is composed of many diverse elements.Larger numbers and more specific study of the detailed demographics and individual case histories of 16-19-year-old drivers involved in fatal crashes are needed to pinpoint which aspects of GDL laws may produce greater risks to which populations of teenagers, particularly older ones.Further, the ratio variables used as the main statistic in this study, while efficient and used by many authors, present some methodological problems (error terms in both the numerator and denominator) that make their comparison somewhat more problematic than for simple rates.Basic means-test and regression analyses produced results similar to this analysis.

Discussion
The results of this study are based on more than 3,700 fatal crash involvements and 2,000 traffic fatalities among California teenagers over the most recent period available as of this writing (1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004).Regardless of type of measure (fatalities or driver involvements), time period (initial years or longer-term), or index (singledriver, late-night, improper-driving involvements, etc.) employed, the results are consistent: California's Graduated Driver Licensing (GDL) law is associated with shifts in traffic death risks, but not with improved traffic safety, among teenagers.If anything, there may be a slight net increase in teenage fatalities following the GDL law.

Effects by age
This "seesaw" effect consists primarily of a shift of 25-30% toward more fatal crash involvements and traffic deaths among older (age 18-19) teenagers relative to younger (age 16-17) ones, and, secondarily, from a 15-20% increase in traffic deaths among male versus female teens.While fatal crashes involving unlicensed teenage drivers increased by 34% relative to those involving licensed teenage drivers after the GDL law took effect, the impact of this trend on overall teenage fatality and fatal crash involvements is inconsistent.In several categories, teenaged drivers licensed under the GDL program experienced increases in traffic death and fatal crash risks comparable to or worse than those of teenagers driving without valid licenses.
The sizeable, consistent increase in fatalities and fatal accident involvements among 18-year-olds (and, in several comparisons, 19-year-olds) who were subjected to GDL requirements as younger teens is the chief, unintended consequence associated with California's GDL law.This seesaw effect, in which laws delaying hazardous adult behaviors are found to reduce risks for younger teenagers at the expense of raising them for older teenagers and young adults, has also been found following implementation of state laws raised the drinking age to 21 (Asch & Levy, 1987;Dee & Evans, 2001).
These findings suggest two alternative hypotheses.First, driving at age 18-19 entails lower risks per mile driven than driving at age 16-17 not because of the greater maturity of older ages, but mainly because of the experience gained driving at younger ages.The increases in traffic fatality outcomes among older teenaged drivers subject to the GDL law's supervisions and restrictions, especially those attributed to improper driving and driving alone, suggest older postlaw teens may lack the realistic experience with riskier situations that was once gained at ages 15, 16, or 17.In fact, age 16 may be a safer age than age 18 (or older) to initiate full-privilege driving, perhaps because of greater family controls and more rapid learning curves at younger ages.This point is suggested by the fact that the prelaw-to-postlaw shift from younger to older drivers in fatal driver involvements (25%-30%) was greater than the shift in younger to older driver's licensing numbers (20%) alone would have predicted.Under this interpretation, prohibiting persons under age 18, or 21, from driving holds the potential of larger increases in risks to young adults licensed without previous teenage driving experience.This hypothesis is supported by the unusually large increase in fatal accident involvements among 18-year-olds (and, to a lesser extent, 19-year-olds) first allowed full driving privileges under GDL laws.

Effects by gender
The increases in fatalities involving males and male drivers (but not females) at age 18, and of greater fatality increases among female 19 yearolds than males, indicate a differing response to the GDL law by gender.Whether this is due to varying responses of parents, authorities, or girls themselves to curtail younger females' driving privileges more than young males' under increased legal restriction is not clear.

Effects by accident type
The alternative hypothesis that young-driver inexperience, rather than young age per se, is paramount is both supported and complicated by the unexpected findings regarding the specific restrictions imposed by California's GDL program.Increased fatal accident IRRs among postlaw 16-17 year-old drivers involving singledriver and late-night driving are not easily explained, given specific GDL restrictions aimed at preventing these very outcomes.It may be that the GDL law's restrictions on first-year drivers' (usually 16-year-olds') driving at night prevents them from gaining experience necessary to drive safely during late-night hours and encourages more hazardous, clandestine driving.The only significant decrease in fatal crashes relating to a GDL mandate -the reduction involving 16-17-year-old drivers transporting teenaged passengers -may be a questionable benefit.Banning younger teens from transporting peers means these same teenagers must either drive alone or be transported by older drivers (generating more crashes involving 20-44-year-olds with teen passengers).The near-significant postlaw increase in lone-driver fatal crashes among 16-17-year-olds (Adjusted IRR, 1.09; 95% CI, 0.98-1.21) is a plausible outcome of preventing younger teenagers from transporting each other, resulting in more teenagers driving alone rather than carpooling.

Effects by license status
The increase in fatal accidents involving unlicensed teenaged drivers of all ages suggests a main effect of California's GDL law has been to deter some of the postlaw population from obtaining driver's licenses at all.The increase in accidents by unlicensed teenage drivers after the law took effect does not reflect a larger trend, since the proportion of accidents involving unlicensed drivers 20-44 did not increase.However, 18-19-year-old drivers lawfully licensed under the GDL program also show increased fatal crash risks (Adjusted IRR, 1.07; 95% CI, 1.03-1.12).

Conclusion
The second alternative hypothesis is that GDL programs are an inefficient means of reducing young-driver risks because they target all teenagers rather than the specific high-risk fraction responsible for most serious accidents, and because their major element of teenage driver training (supervision by parents or other nearby adults) may add to rather than reduce risks in that problematic fraction.Despite recent suggestions in the press of extreme teenage risks, traffic fatalities caused by teenaged drivers are rare events, averaging approximately one fatality per four million trips (Chen et al, 2000) by 16-17 year-old drivers, or per 15 million miles driven by teens (calculated from Table A-17, U.S. Department of Transportation, 2001).Given that the best predictor of a teenagers' driving record is their parents' driving record (Bianchi & Summala, 2004;Taubman-Ben-Ari, Mikulincer & Gillath, 2005), GDL laws deputizing any parent or nearby adult over age 25 with a license as a driving instructor for novice teenagers may work to perpetuate intergenerational bad driving habits.
The results of this study add to previously expressed concerns that GDL laws, especially restrictive ones, have negative effects on older teenaged drivers, particularly males.This potential merits reassessment of the advisability and structure of policies designed to delay adult privileges and greater caution in recommending such policies until longer-term effects on both younger and older teenagers and young adults can be quantified.

Table 1
Incident rate ratios (IRRs), prelaw and postlaw teen motor vehicle death rates vs. age 20-44 by sex, and change in (Adjusted) IRRs from prelaw to postlaw period

Table 2
Drivers involved in fatal motor vehicle crashes by age, prelaw and postlaw periods, populations, and crash rates per 100,000 population, with comparative figures for age20-44 (N=3,772)