The Role of Poverty in California Teenagers’ Fatal Traffic Crash Risk

Teenagers’ high rates of motor vehicle crashes, accounting for 40% of external deaths among 16-19 yearolds, have been ascribed largely to inherent “adolescent risk-taking” and developmental hazards. However, the fact that compared to adults 25 and older, teenagers are twice as likely to live in poverty and low-income areas, risk factors for many types of violent death, has not been assessed. This paper uses Fatality Analysis Reporting System data on 65,173 fatal motor vehicle crashes by drivers in California’s 35 most populous counties for 1994-2007 to analyze fatal crash involvements per 100 million miles driven by driver age, county, poverty status, and 15 other traffic safety-related variables. Fatal crash rates were substantially higher for every driver age group in poorer counties than in richer ones. Multivariate regression found socioeconomic factors, led by the low levels of licensing and high unemployment rates prevalent in low-income areas, were associated with nearly 60% of the variance in motor vehicle crash risks, compared to 3% associated with driver age. The strong association between fatal crash risk and poverty, especially for young drivers who are concentrated in high-poverty brackets and low-income areas, suggests that factors related to poorer environments constitute a major traffic safety risk requiring serious attention. © 2009 Californian Journal of Health Promotion. All rights reserved.


Introduction
Motor vehicle fatalities, comprising 41% of California teenagers' external deaths compared to 24% for adults, represent the widest gap between adolescents' and adults' non-natural mortality (EPICenter, 2008).Researchers typically ascribe teenagers' high rates of traffic casualty to factors allegedly innate to adolescence: developmental immaturity, peer influences, and biologically-impelled risk-taking (Steinberg, 2007;National Research Council, 2006;Hedlund, Shults, & Compton, 2003;Blum, Beuhring, & Rinehart, 2000;Ulmer, Williams, & Preusser, 1997;Chen et al, 2000;Simpson, 2003).At a 2006 conference of leading authorities on the "science of adolescence" sponsored by the National Academy of Sciences, experts cited "the tinderbox in the teenage brain" generating "difficulty in controlling their behavior" as the main cause of adolescents' "high rates of accidents… and reckless behaviors in general" (Dahl, 2006, pp. 7, 15).Another review attributed "adolescents' inclination to engage in risky behavior" to "the temporal gap between puberty, which impels adolescents to thrillseeking, and the slow maturation of the cognitive control system, which regulates these impulses" (Steinberg, 2007, p. 55).
"Teen drivers are different from other drivers," the National Highway Traffic Safety Administration's (2008) discussion."Saving Teenage Lives," declares: On the basis of miles driven, teenagers are involved in three times as many fatal crashes as are all drivers.Why do young drivers have such poor driving performance?Three factors work together to make the teen years so deadly for young drivers: … Inexperience: All young drivers start out with very little knowledge or understanding of the complexities of driving a motor vehicle.Like any other skill, learning to drive well takes a lot of time.Technical ability, good judgment and experience all are needed to properly make the many continuous decisions, small and large, that add up to safe driving… Risk-taking behavior and immaturity: Adolescent impulsiveness is a natural behavior, but it results in poor driving judgment and participation in high-risk behaviors such as speeding, inattention, drinking and driving, and not using a seat belt.Peer pressure also often encourages risk taking.
Greater risk exposure: Teens often drive at night with other teens in the vehicle, factors that increase crash risk.
That NHTSA, like others, fails to assess, or even mention, low socioeconomic status as a risk to teenaged drivers is curious When evaluating the large differences in risks among various racial, ethnic, and regional groups-such as the high rates of homicide among African Americans or motor vehicle deaths among Southernersresearchers typically pursue social and economic explanations (i.e., Fox & Piquero, 2002).However, conclusions about adolescent risktaking and its causes have been reached without first controlling for the differing socioeconomic conditions in which adolescents and adults live (Reyna & Rivers, 2008;Casey, 2008; for critique, see Males, 2009).
That such environmental conditions might be important variables in what is called "adolescent risk-taking" is indicated by the fact that for every race and locale, young people ages 15-19 and 20-24 are two to three times more likely to live in households with incomes below federal poverty thresholds than are adults ages 45-64 (US Census Bureau, 2008, 2008a, 2008b).Agebased income stratification is so pronounced that poverty rates averaging below 10% are found for teenagers in only five of California's 58 counties, versus 32 counties for ages 45-54.Meanwhile average poverty rates of 20% or higher afflict teenagers in 18 counties, versus none for Californians ages 45-54.
The contribution of socioeconomic status to motor vehicle crash mortality, the largest category of external deaths among Americans ages 15 to 34, deserves comprehensive attention, yet only three studies can be located that even peripherally discuss the subject (Aguero-Valverde & Jovanis, 2006;Hasselberg & Laflamme, 2005, 2003) and one, this author's, that provides only a preliminary sketch (Males, 2007).
The present study examines the associations between motor vehicle fatalities, socioeconomic status and related environmental variables among teenagers and adults in California for the purpose of testing the hypothesis that higher levels of poverty more efficiently explains the variance between teenage and adult motor vehicle fatality rates than does innate "adolescent risk-taking."

Data description
This analysis concentrates on characteristics of California resident drivers involved in motor vehicle crashes that caused at least one fatality on public roads, as compiled by the US Department of Transportation's Fatality Analysis Reporting System (FARS, 2008) for the full 1994-2007 period.Ten descriptive variables for drivers and vehicles involved in fatal crashes (single year of age, county of residence, driver's license status, seat belt/restraint use, alcohol or drug intoxication, age of vehicle driven, size of vehicle driven, number of vehicle occupants, and percent of accidents involving single vehicles or vehicle rollover) were entered into the database in conjunction with eight environmental variables by county (poverty rate by age, median percapita personal income, per-capita motor-vehicle registrations, per-capita miles of roadway, percent of the labor force that is unemployed, percent of the population licensed to drive, population density per square mile, and percent of all commuter trips that are made by motor vehicle as opposed to public transportation and other modes) from the California Department of Finance's Demographic Research Unit (2008) and the Census (2008Census ( , 2008aCensus ( , 2008b) (see Table 1).Driver speed is not included because values are missing for 47% of cases.The poverty rate is defined by the Census as the percentage of the population living on incomes below federal poverty guidelines ($13,410 for a family of three in 1999).Where primary data is available only for age groups, values for individual ages were estimated by linear interpolation (Shyrock & Siegel, 1976).
The analysis excluded two sets of outliers.Drivers under age 16 and over age 74 were excluded to eliminate the effect on mortality of age-related physical limitations, and vehicles that were over 60 years old or which had 10 or more occupants also were excluded.To avoid mathematical adjustments necessary to compensate for the effects of small cell size, only the 35 counties with populations of 100,000 or more in the 2000 census are included; these counties account for 96% of California's fatal traffic crashes.With these exclusions, 2,065 county-by-age cells with 65,531 drivers involved in fatal motor vehicle accidents over the 14-year study period remained; the elimination of a few cases with missing values for certain variables for the analyses reduced available cases to 65,173.(Masten & Hagge, 2004;Males, 2007a).
The effect of a law change that disproportionately raised teenage fatal crash involvements relative to older drivers' makes the study hypothesis more difficult to validate.
Estimates of the best measure of risk exposure, the average number of vehicle-miles driven (VMD) per driver, are not available for California.Thus, VMD by driver age and resident county was calculated using a standard, three-step process (McCarthy, 2002) 1) and by county income bracket for all drivers (Table 2).Fatal crash involvement rates by driver age group and poverty bracket for all drivers are shown in Table 3. Tables 4a and 4b provide cross sections of the characteristics of fatal crash involvements for the drivers with the highest rates, those age 16-17, and those with the lowest rates, age 46-47.Table 5 shows the simple bivariate association between each predictor and fatal crash involvement rates.The driver and county variables were then subjected to hierarchical multiple regression (Table 6) to estimate the most important predictors when all variables were controlled.Multicollinearity diagnostics were conducted to assess covariance, especially with regard to the three major demographic variables (driver age, poverty rate, and county per-capita income) and variables that potentially may more efficiently predict the specific criterion of fatal crash involvement risk.
Variables were entered into a stepwise multivariate regression in the order of greatest probability of F as selected by the stepwise program, and the number and order of variables significant at the 0.05 limit shown in Table 6 represents the number of steps and the order of variable entry.

Results
The results of the analysis of the 18 predictors of fatal crash involvement risk are shown in Tables 1-6.As expected, teen drivers ages 16-19 were 2.7 times more at risk of fatal crash involvement than the average for all drivers (Table 1).Interestingly, teens involved in fatal crashes were somewhat more likely than older drivers to have been using seat belts or other restraints and less likely to have been intoxicated.Teens also were nearly twice as likely to occupy highpoverty brackets; were driving vehicles that averaged one year older, were substantially smaller, and contained more occupants (Table 1); and were considerably more likely to live in low-income counties (Table 2).In turn, drivers at every age level who resided in low-income counties were two to three times more likely to suffer fatal crashes, both per capita and per mile driven, than drivers in the wealthiest counties.These risks, which accelerated at poverty rates of 12% and higher and county per-capita incomes of $20,000 and lower, were in part functions of lower use of restraints, generally older vehicles, and higher proportions of unlicensed drivers, vehicle rollovers, and singlevehicle accidents.Low-income counties also were characterized by high levels of poverty and unemployment, lower population density, and less extensive public transit (Tables 2, 3).The average poverty level for teenagers and young adults in the wealthiest quintile was similar to that of middle-aged drivers in the poorest quintiles, and fatal crash risks for older drivers averaged only slightly lower than those of younger ones subjected to similar poverty levels.
Tables 4a and 4b  approximately 1.8 times higher overall and in every quintile than among middle-agers.The 3.9-fold gap between 16-17 year-olds' and 46-47-year-olds' fatal crash rates per mile driven shrinks to 1.5 to 1.8 times higher under reasonably equal poverty levels (e.g., the richest and second richest teen quintiles and versus the fourth richest and poorest quintiles for age 46-47).
Tables 4a and 4b also reveal some unexpected patterns.Teens age 16-17 in poorer counties had higher rates of accidents involving vehicle rollover and generally lower rates of driver licensing and miles driven per driver.However, while crash-involved teens in the poorest income quintile did show considerably higher rates of driving while intoxicated and failure to use restraint, these factors did not vary consistently with county income level.Nor were crashinvolved teens in poorer counties driving smaller or older vehicles or carrying more passengers than teens in more affluent counties.For drivers age 46-47, the sharp increase in fatal crash rates from richer to poorer county income quintiles varied chiefly with a rising percentage of crashes involving vehicle rollover, older vehicle age, and greater vehicle size; other variables were more inconsistent.Teens in fatal crashes were somewhat less likely to have been intoxicated and more likely to have used restraints than their middle-aged counterparts.
The simple bivariate correlation shown in  proved inefficient in explaining variance.Driver age, in particular, explained virtually none of the variance in crash rates once poverty level was controlled and just 12% of the variance once county income was controlled, and the predictor pairs together accounted for less than one-third of total variance in both cases.This suggests that other variables that covary with economic status and age may prove better operational representations of fatal crash risk, a question testable by multivariate regression.
Of the 18 variables entered in the hierarchical multiple regression shown in Table 6, eight remained significant predictors accounting for three-fifths of the total variance in fatal crash risk for all drivers (R=0.786adjusted R 2 =0.616, p=0.000).Poverty and county incomes dropped out as significant predictors and driver age became a minor predictor of fatal crash risk.These were replaced by two major correlates (percent of the population licensed to drive by age and county, and county unemployment rate) and several minor ones (miles driven per year by driver age and county, percent of total trips in the county that involved public transportation, vehicle age by driver age and county, and population per square mile by county).These variables appeared to more efficiently operationalize the general economic measures as driving-related factors.Driver age remained a significant predictor even after these variables were controlled, though it accounted for just 3% of total variance and 5% of explained variance.

Discussion
California drivers of all ages living in poorer areas suffer substantially higher fatal crash rates than those in richer areas.The relationship between poverty and income status and fatal crash risk is strong and consistent both between and within driver age groups.Particularly at young ages, poverty is associated with considerably lower rates of driver's licensing and less driving overall, leading to slower acquisition of driving experience that reduces crash risk (see Dee & Evans, 2001).For example, drivers ages 16-19 in affluent counties such as Marin and San Mateo, despite driving 2,000 to 3,000 more miles each every year, have just one-third the per-person risk and one-fifth the per-mile risk of fatal crash involvement than do teenaged drivers in impoverished counties such as Humboldt and Tulare.Evidence that teenagers who are least at risk tend to be those who are licensed to drive and who drive the most, factors that covary with higher socioeconomic status, challenges the assumption that more teenagers driving more miles necessarily elevates fatality hazards.To assess directly the effect of driver inexperience, a data set that includes not just drivers' ages but years of driving experience (a variable not provided by FARS), would be necessary.It is clear that what aspects of poverty, income, unemployment, rates of driver licensing, population age structure, population density, and public transportation prevalence interact in what ways to contribute to unlicensed driving, driving while intoxicated, driving without restraints, and vehicle rollover incidence that in turn contribute to fatal crash risk are complexities that remain to be explored.
The claim that risk-taking is innate to teenagers has led to advocacy for policies curtailing teenage driving and perhaps even banning it altogether, strategies more likely to add to teenage risks by preventing youths from gaining necessary driving experience.In fact, the most effective measures to combat high rates of motor vehicle fatality may emerge from careful analyses necessary to understand how California's excessive rates of poverty, particularly among young people, and deficient driving conditions in poorer areas interact to substantially elevate fatality risks.

Table 1 . Characteristics of fatal crash rates by driver age group Driver age group → All ages 16-19 20-24 25-34 35-44 45-54 55-64 65-74 Driver characteristics
Since teenaged drivers do not drive as much per person as adults, a third adjustment used estimates of VMD per person by age group from the most recent U.S. Department of Transportation's National Household Travel Survey (Bureau of Transportation Statistics, 2001) to estimate VMD by age group for each California county.The results this estimation technique yielded for VMD by age of California driver are very similar to those of the NHTS, indicating the technique is not biased with regard to estimating the relative proportions of driving by age.Estimates are shown in Table 1.
by age tabulated by the California Department of Motor Vehicles (2007) for each county to produce an estimate of gross VMD by age.

Table 2 . Characteristics of fatal crash rates by county income bracket County income bracket → Richest fifth Second fifth Middle fifth Fourth fifth Poorest fifth County characteristics
AnalysisThe principal outcome measure of interest, fatal crash involvements per 100 million VMD by age of driver and county, was calculated.The summary results are shown in the Appendix.Appendix TableAshows that counties vary substantially in fatal crash risk, as well as in poverty levels, miles driven per resident, personal income, and related variables.Drivers in Central Valley, northern California, and

Table 3 . California driver' fatal crash involvement rates by age and poverty level Poverty quintile, 35 largest California counties, 2000 Driver age: Richest fifth 2nd fifth Middle fifth 4th fifth Poorest fifth Fatal crashes per 100 million VMD by driver age
Each poverty quintile consist of seven counties in the 35 county-ranking by percapita income bracket.See Table2and Appendix A.