«by Johnathon P. Ehsani A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy (Health Behavior ...»
While a sizeable body of evidence suggests that GDL leads to significant declines in 16- and 17-year-old drivers’ crashes (Foss, Feaganes et al. 2001; Shope, Molnar et al. 2001; Williams, Ferguson et al. 2005; Chen, Baker et al. 2006; Trempel 2009; McCartt, Teoh et al. 2010; Williams and Shults 2010; Masten, Foss et al. 2011), the effect of GDL on 18-year-old drivers’ crashes is less clear. A study examining the effect of California's GDL on teen driver fatalities reported a 24% rise in 18-year-old fatal crashes following the introduction of GDL in July 1998 (Males 2007). A recent national study by Masten and colleagues also reported a significant post-GDL increase in fatal crashes of 18-year-old drivers large enough that no net benefit from GDL on overall 16to 19-year-old drivers’ fatalities remained (Masten, Foss et al. 2011). In contrast, McCartt and colleagues found that states with GDL legislation rated by the Insurance Institute for Highway Safety as ‘good’ (i.e., more, stronger components) had significantly lower 18year-old driver fatal crashes than states with ‘fair’ or ‘poor’ GDL (i.e., fewer, weaker components) (McCartt, Teoh et al. 2010; McCartt and Teoh 2011). Using the same taxonomy to rate GDL legislation, Trempel found that ‘good’ GDL programs were associated with significantly fewer insurance collision claims by 18-year-old drivers (Trempel 2009).
Both studies that reported an increase in 18-year-old drivers’ fatalities employed time series analysis, an evaluation method that controls for pre-existing trends, seasonal variation and serial correlation between observations (McCleary and Hay 1982).
Masten’s national evaluation pooled individual states’ time series into a single sample, making this evaluation arguably the most complete study on the effect of GDL that has been conducted to date. However, both studies utilized data from the Fatal Analysis Reporting System (FARS) and, therefore, were limited to only fatal crashes (National Highway Traffic Safety Administration 2010). Neither McCartt (McCartt, Teoh et al. 2010) nor Trempel (Trempel 2009) controlled for pre-existing trends in teen crashes in their analysis, and McCartt was limited to fatal-only crashes. Trempel’s analysis was restricted to collision claims involving new (i.e., 0-3-year-old) motor vehicles (Trempel 2009), a sample that is known to be unrepresentative of teen drivers (Williams, Leaf et al. 2006).
While few studies have examined the effect of GDL on 18-year-old drivers’ crashes and their results are mixed, the studies that reported increased fatal crash rates among 18-year-old drivers raise serious questions about potential unintended consequences of GDL. Two mechanisms have been proposed to explain the observed increase in 18-year-old drivers’ crash rates. The first is an ‘offset effect’, where a proportion of teen drivers, either wishing to avoid the requirements of GDL or for other reasons, do not get licensed to drive until age 18 (Masten and Hagge 2004). As a result, the increased crash risk associated with inexperience for these new drivers is shifted from 16 to 18 years of age. The second mechanism results from a proportion of teens not advancing well through GDL prior to their turning 18 and getting a regular license.
These drivers may still lack independent driving experience when they receive their regular driver license at age 18 (Dee, Grabowski et al. 2005). The result from either mechanism could be an increase in 18-year-old driver crash rates.
To conclusively determine the mechanism responsible for an increase in 18-yearold drivers’ crashes, one would track thousands of young drivers’ using longitudinal state crash and licensure records, linking individual’s crashes with their age at licensure.
However, licensure data reported by the Federal Highway Administration underreport the actual number of licensed teens, and these data are difficult to obtain from individual states (Insurance Institute for Highway Safety 2006). In their absence, we can exploit natural experiments resulting from GDL policy implementation to shed some light on the issue. By comparing teen drivers’ crash rates in states where GDL applies exclusively to 15- to 17-year-olds with rates in states where GDL extends to all novice drivers regardless of age, inferences could be made about the underlying mechanism responsible for an increase in 18-year-old drivers’ crashes.
If an increase in 18-year-old drivers’ crash rates was observed in states where GDL applies to 15- to 17-year-old drivers only, one could infer that the responsible mechanism might be the offset effect. The assumption would be that in these states a sizeable proportion of teens are delaying licensing until age 18, at which point they begin driving without the benefit of experiencing GDL. Meanwhile, in states where GDL applies to novice drivers of all ages, and every new driver experiences GDL, there would not be an increase in crashes at age 18. Alternatively, if an increase in 18-year-old drivers’ crash rates was observed in states where GDL applies only to 15- to 17-year-old drivers and also in states where GDL applies to all novice drivers one could infer that the primary mechanism responsible for the increase in 18-year-old drivers’ crashes is that teens licensed under GDL lack unrestricted driving experience when they get a regular license, and therefore are more likely to crash.
The purpose of this paper is to determine the effect of GDL on 18-year-old drivers’ crash rates, and to shed light on a possible mechanism responsible for some potential increase in crash rates by comparing two states where GDL provisions apply exclusively to 15- to 17-year-old drivers (Florida, Michigan), to a state where GDL applies to novice drivers of all ages (Maryland). The effects of GDL on 16- and 17-yearold drivers will also be examined, to extend the evaluation period relative to previous studies and confirm positive outcomes in this population.
The implementation of a GDL program would result in (1) a significant decline in the crash rates of 16-year-old drivers, and (2) a significant decline in the crash rates of 17-year-old drivers. With regard to an effect of GDL on crashes of 18-year-old-drivers, two hypotheses will be tested: (3) there will be an increase in 18-year-old drivers’ crash rates in states where GDL applies exclusively to 15- to 17-year-old drivers, and no change or a decrease in 18-year-old drivers’ crash rates in states where GDL applies to all novice drivers; and (4) there will be an increase in 18-year-old drivers’ crash rates in states where GDL applies exclusively to 15- to 17-year-old drivers, and an increase in 18-year-old drivers’ crash rates in states where GDL applies to all novice drivers.
To test these hypotheses, state crash record data, which include all reported crashes occurring on public roadways, were obtained for Florida, Michigan and Maryland. These states represent three distinct approaches to the implementation of GDL policy (Table 4.1). Florida is credited with having implemented the first GDL program in the United States in July 1996, when the state required a six-month learner permit period for teen drivers aged 16- or 17-years-old and restricted driving of 16-yearolds between the hours of 11 p.m. and 6 a.m. and of 17-year-olds between the hours of 1 a.m. and 5 a.m. In October 2000, Florida augmented its GDL program by extending the learner permit period to 12 months and requiring 50 hours of supervised practice driving. Complete crash record data for Florida were obtained for the contiguous period 1990 to 2009.
Michigan introduced a three-phase GDL program in April 1997 that included a learner license period lasting a minimum of 6 months and the completion of 50 hours of supervised driving. Teens driving with an intermediate license were restricted from driving between the hours of 12 midnight to 5 a.m. Complete crash record data for Michigan were obtained for the contiguous period 1992 to 2009.
Maryland’s GDL program is unique in the United States, in that it requires novice drivers of all ages to complete the same learner permit phase, but imposes restrictions on passenger and nighttime driving only to novice drivers under 18-years-of-age. Since 1979, Maryland has had a nighttime driving restriction for drivers under 18 years of age, prohibiting their driving between the hours of 1 a.m. and 6 a.m.(Preusser, Ferguson et al. 1998). The hours of the nighttime driving restriction were revised in 1985 to begin at 12 midnight and continue until 5 a.m. From July 1999, novice drivers of all ages in Maryland were required to hold a learner permit for four months and to complete 40 hours of supervised driving. In October 2005, the learner permit period was extended from four to six months, and the required number of supervised driving hours was increased from 40 to 60 hours for novice drivers of all ages. In addition, a passenger driving restriction was added, banning drivers under age 18 from carrying passengers younger than age 18 for the first five months of intermediate licensure. In October 2009, the learner permit period was extended to nine months. Complete crash record data for Maryland were obtained for the contiguous period 1998 to 2009. All state crash record data used in this study were provided by the Center for Management of Information for Safe and Sustainable Transportation of the University of Michigan Transportation Research Institute (University of Michigan Transportation Research Institute 2012).
Measures Novice Driver Crash Rates: Monthly counts of all reported crashes occurring on public roadways (fatal/disabling injury, non-disabling injury, and possible-injury/propertydamage-only (PDO)) involving at least one teen driver (16-, 17- or 18-year-old) in cars, trucks/pickups, and sport utility vehicles were obtained for Florida, Michigan and Maryland. Crashes that involve any injury (fatal or non-fatal) are required to be reported.
However the states differ slightly in their criteria for crashes where only property damage occurs. Florida requires property damage crashes valued at $500 or more to be reported to the police, while in Michigan the property damage should be $1,000 or more. In Maryland, a crash is reported if it involves immobilizing property damage or if an individual involved in the crash requests a report.
The number of licensed teen drivers would provide the most precise estimate of the effect of the GDL programs on crash rates, if it were available. However, licensure data reported by the Federal Highway Administration underreport the actual number of licensed teens, and these data are difficult to obtain from individual states (Insurance Institute for Highway Safety 2006). Miles driven by each teen would also be an ideal indicator of exposure, but are difficult to measure and generally unavailable. Therefore, crash rates were based on the number of 16-, 17-, and 18-year old teens in the overall population of each state. Annual population estimates by state and age were obtained from the United States Census Bureau (Bureau of the Census. U.S. Department of Commerce 1999; Bureau of the Census. U.S. Department of Commerce 2010). Monthly values were interpolated using cubic spline curves, which are the smoothest curves that best fit a set of data points (Bartels, Beatty et al. 1998). Age-group-specific monthly crash involvement rates of 16-, 17-, and 18-year-old drivers per 100,000 population were calculated using monthly crash counts and population estimates. Fifteen-year-old drivers were excluded from the study because driving in this age group typically occurs under supervision, which has the lowest lifetime crash risk (Mayhew, Simpson et al. 2003;
Covariates Adult driver crash rates: Monthly crash rates for drivers age 35 to 54 were estimated using the identical method used to estimate teen driver crash rates. 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 the program. 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 covariate to control for unmeasured factors that affect all drivers enhances the validity of the findings.
Gas prices: An inverse relationship between gas prices and crashes has been identified for drivers of all ages (Sivak and Schoettle 2010); however, research suggests that teen driving behavior may be more sensitive to higher gas prices, relative to older drivers (Morrisey and Grabowski 2010). Monthly national gas prices were obtained from the U.S. Energy Information Administration and included as a covariate (U.S. Energy Information Administration 2011).
GDL effective date: The precise points at which GDL programs were expected to affect crash rates were determined by inspection of the details of each state’s GDL system. This information was obtained from the Insurance Institute for Highway Safety GDL effective date database (Insurance Institute for Highway Safety 2012) and each state’s Department of Motor Vehicles website (Florida Department of Highway Safety and Motor Vehicles 2009; Maryland Motor Vehicle Administration 2012; Michigan Secretary of State 2012). The date at which GDL would apply to each age group in the study sample was calculated by identifying the first month that GDL would impact the age group. For example, Florida’s first GDL program was implemented in July 1996.
From this date onward, GDL applied to all new teen drivers in the state, but existing teen drivers were exempt from the program. Therefore, the point at which 17-year-old drivers were directly affected by Florida’s GDL was July 1997, as the initial cohort of teens affected by GDL advanced through the system. Similarly, the point at which 18-year-old drivers were directly affected by GDL was July 1998. This approach to identifying the date at which GDL affected each year-of-age in the study population was applied to each state in the sample. Because at least two years of post-implementation data points were required to estimate effects, GDL laws introduced after January 2008 were not included in the analyses.