Women’s Issues in Transportation: Summary of the 4th International Conference, Volume 2: Technical Papers (2011)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

1Female Involvement in U.S. Fatal Crashes Under a Three-Level Hierarchical Crash Model Mediating and Moderating Factors Eduardo Romano, Tara Kelley-Baker, and Pedro Torres, Pacific Institute for Research and Evaluation (PIRE) Men have long held the lead in motor-vehicle crashes; how- ever, research indicates that women are closing the gap. The reasons for this relative increase are unclear. To fur- ther investigate this problem, the authors applied a simpli- fied version of the hierarchical levels of driving behavior (HLDB) model to investigate female involvement in fatal crashes in the United States. The HLDB model recognizes that decisions at higher levels affect decisions at lower lev- els. At the top level, the model assumes that the driver’s condition (e.g., inattention, fatigue, impairment) has an effect on the next level (e.g., speeding or other failures to obey traffic laws), which subsequently affects the basic maneuvering skills (i.e., the lowest level). Data for this study were drawn from the Fatality Analysis Reporting System for the years 1982 to 2007. Single-vehicle crashes were used to indicate crash responsibility. Basic descriptive and multilevel analyses were applied to investigate female involvement at each level of the HLDB model. Compared with males, female drivers were less likely to be involved in crashes associated with the highest HLDB level, but more likely to be involved in the lowest level. The relative high prevalence of females in skill-related crashes, however, occurred only when associated with speeding. Variations in this finding due to age and gender were also found. Find- ings from this study should help to develop more efficient (better targeted) traffic safety prevention policies. Despite significant progress in traffic safety during the past decades, motor vehicle crashes (MVCs) remain a major source of injury. U.S. males account for most of the traffic fatalities—three times that of females—and thus, have received most of the resources and focus (Beirness 1989; Cerrelli 1998). Cur- rent data show, however, that the prevalence of women in fatal MVCs is rising. Romano and colleagues (2008) have shown that female involvement in fatal crashes increased in the United States, albeit mostly due to young drivers. Although the National Highway Traffic Safety Administration (NHTSA) reported that the number of male drivers killed in fatal crashes dropped from 45,084 in 1975 to 39,739 in 1994, during the same period, the number of female drivers in fatal crashes increased from 9,356 to 13,430 (NHTSA 1995). Interestingly, the estimated involvement rate in fatal crashes per 100,000 licensed male drivers has continuously declined over the past 30 years (from 62 in 1975 to 42 in 2003), whereas it has remained unchanged for female drivers for about 15 years (NHTSA n.d.). The reason for this increase in crash fatalities is often cited as being related to the different and richer roles modern women are playing in the society. Some expla- nations focus on household roles and their associated stresses, travel patterns, and driving skills. Other expla- nations go beyond exposure and argue that the develop- ment of changes in normative behaviors among women that occur over time translate into new, riskier driving behaviors (Pisarski 1992; Voas et al. 1998). The driving skills of female drivers, particularly vehicle control and maneuvering, have received atten- tion among researchers. McKenna et al. (1991) ana- lyzed a survey of drivers in the United Kingdom and concluded that there is a significant difference between male and female personal assessments of driving skills,

2 wOMEN’S ISSUES IN TRANSPORTATION, VOLUME 2 with males rating themselves higher than females. In an examination of Finnish drivers, Laapotti et al. (2003) also found that female drivers tended to evaluate their own driving skills lower than did their male counter- parts and that, although young female drivers were more safety oriented than male drivers, female drivers had more problems in vehicle handling and mastering traffic situations. Changes in attitudes toward risk may have contrib- uted to those findings. Stamatiadis and Deacon (1995), working with Michigan data, found that young driv- ers as a whole drive less safely now than the same age group did in previous years, with female drivers driv- ing more safely, on average, than male drivers. Data from Michigan, however, show that the rate of aggres- sive, risky driving and speeding-related collisions among female drivers (particularly young women) is increasing (Kostyniuk et al. 1996; waller et al. 2001). In a study on female driving by the Pacific Institute for Research and Evaluation, almost two-thirds (65%) of participants said they drove more than 15 mph over the speed limit occasionally or very often (NHTSA 2001). About half had been stopped for a traffic violation in the past 3 years, primarily for speeding. In 2000, how- ever, an Internet-based, nonscientific survey of 400 clients of an Ohio insurance company reported that women stuck in summer traffic with children in their cars were four times more likely than men to speed, run a red light, or drive on the shoulder (Progressive Insurance 2000). In an application of the hierarchical levels of driv- ing behavior (HLDB) model to archival data, Laapotti and Keskinen (2004) examined changes over time in the MVC patterns of males and females in Finland. They noticed that female drivers had proportionally more accidents connected to vehicle maneuvering and control of traffic situations, such as reversing and loss- of-control accidents when not speeding. Laapotti and Keskinen (1998) also found that female drivers were more prone to lose control of their vehicles when the condition of the road surface was less than optimal (e.g., slippery). Thus, the occurrence of MVCs among female drivers is rising, although MVCs are still much less prevalent among female drivers than among male drivers. why? Unfortunately, very little is known about such appar- ent increase in risk. The main goal of this study was to add to the current literature by applying a simplified version of the HLDB model (Keskinen 1996) to inves- tigate female involvement in fatal crashes in the United States caused by gender differences in three hierarchi- cal levels (from top to bottom): driving context (e.g., impaired driving, fatigue, inattention), the mastering of traffic situations (e.g., red-light running, speeding), and vehicle maneuvering (e.g., loss of vehicle control). Methods Data Crash data for this study were obtained from the Fatality Analysis Reporting System (FARS) for the years 1982 to 2007. FARS is a record system for fatal crashes (defined as a crash on a public roadway causing a death within 30 days of the event) (NHTSA n.d.). FARS provides detailed information about fatally injured drivers’ gen- der, age, level of alcohol consumption, and maneuvering skills. FARS also contains information about the number of vehicles involved in the crash. The data set provides a large representative source of information that allows the making of inferences confidently at the national level and the evaluation of changing trends over time. There were 1,489,277 drivers in the FARS data for 1982 to 2007. To provide a measure of crash responsi- bility, only drivers involved in single-vehicle crashes were included in the data set. Drivers who presented a driver condition signaling a “mentally challenged” state, were involved in a police chase, were driving buses or farm equipment, or were parking vehicles were excluded. Finally, only fatally injured drivers with gender identi- fication were kept in the data set. After these manipula- tions, 384,861 drivers remained in the data set and were used for most analyses. Because FARS began collecting information on race and ethnicity in 1999, only the FARS data for 1999 to 2007 were kept for analyses including this variable (n = 150,400). Measures Driving Exposure As mentioned, driving exposure has been suggested as an explanation for the relative increase in crashes among female drivers. Driving exposure was addressed by apply- ing the estimated vehicle miles traveled (VMT) for each relevant group (e.g., Braver 2001). A common method of normalizing crash fatality data, the VMT-based approach normalizes crash incidence by number of miles driven (thereby adjusting incidence by crash exposure). Esti- mates of the annual mileage driven by each of the four racial–ethnic groups were drawn from the 1995 and the 2001 National Household Travel Surveys (FHwA n.d.). The most recent survey available is for 2001; although the 2008 survey has been completed, it has still not been made available to the public. Because this measure was collected through a national survey that is conducted only once every 5 to 7 years, VMT data had to be estimated for nonsurvey years. In this study, the VMT data from 1995 were applied to the FARS period of 1982 to 1997, and the 2001 VMT data were applied to the FARS period of 1998

3FEMALE INVOLVEMENT IN U.S. FATAL CRASHES to 2007. The analytical limitations that such a generaliza- tion presents are obvious, but the benefits of providing some correction for female drivers’ driving exposure out- weigh the limitations of the measure. HLDB Crash Levels Table 1 shows the FARS variables used and the criteria applied to build the condition variables at each of the three HLDB levels. At the top of the HLDB model is Level 3, which includes the factors of alcohol consump- tion, fatigue, and inattention. It is followed by Level 2, which includes red-light running, failing to obey a traffic signal, speeding, and other aggressive driving, and Level 1, which includes loss of vehicle control associated with improper vehicle maneuvering or bad weather or surface conditions. Alcohol consumption was established when the driver recorded a positive blood alcohol concentration (BAC) result, as determined by the actual BAC as mea- sured and reported in FARS and the multiple imputation of BAC values when the actual BAC values were miss- ing, as is currently done in FARS (Subramanian 2002). The other crash factors were assigned using information from the following FARS variables: the Driver Condition Factor (DR_CF), the Person-Related Factor (P_CF), the Condition Factor (CF), Violation Charge (VIOLCHG), and the Vehicle Condition Factor (VEH_CF) (NHTSA 2008). To inform about multiple crash conditions, FARS allows up to four entries of each of the individual condi- tions (e.g., DR_CF1 to DR_CF4). when multiple crash conditions were present, the presence of a targeted fac- tor in any of these conditions (up to four levels) was sufficient for its identification. Table 1 shows the codes applied to the classification of drivers. Analyses Conceptual Model Drinking-and-driving contexts can be viewed as decision- making situations requiring individuals to choose between riskier and safer courses of action (Labouvie and Pinsky 2001). This study follows a limited version of the HLDB model (Keskinen 1996), which recognizes that decisions at higher levels affect the decisions (and skills) at lower levels. At the top level, Keskinen’s model assumes that the person’s general goals and skills (e.g., the person’s enthu- siasm about cars and driving) affect the decisions at the second level (e.g., drinking and driving), which affect the next level (e.g., speeding or other failures to obey traffic laws), which subsequently affect the basic maneuvering skills (e.g., controlling the vehicle direction). Information TAbLe 1 FARS Variables Used to Create Three Driving Conditions Level and Variable DR_CF P_CF CF VIOLCHG Level 3 Fatigue or inattention 1, 6 3, 10 4 (since 1997) 93–98 (since 1991) Alcohola Level 2 Speeding 44 44 2–3 (until 1996) 46 (since 1998) 21–25 (since 1997) 29 (since 1997) Failure to obey 38, 39 (and 33 31–39 (since 1997) 19<TRA_CONT<22 or 38–41 0<TRA_CONT<7 ) Other aggressive 8 (since 2003) 36 (since 1995) 2–3 (since 1997) 13 (since 2005) 46 4 (until 1996) 46 (until 1994) 8 (since 2004) 27, 36 Level 1 Maneuver 26–36 20 (2000 and 2001) 41–69 (since 1997) 47–48 22 (until 2002) 52 36–37 (since 1994) 58 42–43, 47–49 18 (since 1995) 50 (since 1994) 28, 34–35, 51 59 (since 1995) weather or slippery road conditions 61, 77, 79, 87 60, 73, 75, 83 5 a BAC > .00, tested or imputed (by NHTSA).

4 wOMEN’S ISSUES IN TRANSPORTATION, VOLUME 2 on drivers’ general goals and skills (the top level of the HLDB model) is not available in the FARS data set. Thus, this study applied a simplified model based on the three lower levels of the HLDB model. Analytical Models Basic statistical comparisons were applied to analyze the data. To investigate the gender contribution to each level and crash factor, multinomial (polytomous) logis- tic regression was applied. Two sets of regressions were applied: one modeling the occurrence of a Level 1 crash as a function of Level 2 factors and covariables, and another modeling the occurrence of a Level 2 crash as a function of Level 3 crash factors and covariables. The covariables considered included gender (male, female), age (<21, 21 to 34, 35 to 64, and 65 and older), race– ethnicity (non-Hispanic whites and other), time of day (6 a.m. to 9 p.m. and 10 p.m. to 5 a.m.), and the pres- ence of passengers (none, 1, 2, or more). Although the analysis of main effect models was of interest, the focus was on the interaction between gender and all relevant factors and covariables. For instance, in modeling Level 1 crashes, the focus was on the interactions between gen- der and Level 2 factors. If these interactions were sig- nificant, it would mean that, as compared with males, the contribution of females to the likelihood of a Level 1 crash would partially depend on the occurrence of a Level 2 factor. Several sequential runs for each of the two sets of models described were performed to test the significance and direction of interaction terms involving gender. In each of these series of models, the analysis began with a main-effects-only model and sequentially added the gender interactions to test for the separate and joint contributions of those additions. Results Table 2 shows the number of drivers in the file by gen- der, as well as the corresponding female-to-male (F–M) ratio. The F–M ratios and prevalence estimates (percent) in Table 2 are adjusted by driving exposure. Table 2 shows that overall, for each male driver in the file, there are 0.65 females. Table 2 also shows that the prevalence of Level 3 crashes is significantly larger among males (70.5%) than among females (50.9%) (p < .01). The F–M ratio mimics this finding, showing a much smaller ratio (i.e., relatively less for females) for Level 3 crashes than for non-Level 3 crashes. Almost 90% of the level 3 crashes involved alcohol (78.1% alone, 11.6% in con- junction with fatigue or inattention). Largely, this find- ing is not surprising, for the sample studied included only single-vehicle crashes in an attempt to ensure the drivers’ responsibility in the crash, and single-vehicles crashes (particularly at nighttime) have been used as a proxy for alcohol-related crashes. Not surprisingly, male drivers were more likely to be involved in alcohol-related crashes (81.2% of the Level 3 crashes alone, 11.5% in conjunction with fatigue or inattention) than female drivers (71.4% and 11.8%, respectively). For fatigue TAbLe 2 Number of Drivers, Their Distribution, and Female–Male Ratio by Crash Level Males (%) Females (%) Both (%) Female–Male Ratio Level and Variable (N = 275,329) (N = 109,532) (N = 384,861) 0.65 Level 3 No Level 3 present 29.5 49.1 37.2 1.07 Level 3 present 70.5 50.9 62.8 0.47 Alcohol 81.2 71.4 78.1 0.41 Fatigue or inattention 7.4 16.8 10.4 1.06 Jointa 11.5 11.8 11.6 0.48 Level 2 No Level 2 present 55.5 47.4 50.6 0.75 Level 2 present 44.5 52.6 49.4 0.55 Speeding 66.6 52.2 61.5 0.43 Failure to obey 11.8 28.0 17.6 1.29 Other aggressive 10.8 12.5 11.4 0.64 Jointa 10.8 7.3 9.5 0.37 Level 1 No Level 1 present 34.9 40.9 37.3 0.76 Level 1 present 65.1 59.1 62.7 0.59 Maneuver 94.8 90.9 93.4 0.56 weather or surface conditions 2.1 3.3 2.5 0.94 Jointa 3.2 5.8 4.1 1.06 Note: N = nonweighted number of drivers in file. Female–male ratio is the ratio of female drivers to male drivers in each category. Percentages and female–male ratio are weighted by driving exposure. a Joint denotes more than one factor in the level.

5FEMALE INVOLVEMENT IN U.S. FATAL CRASHES or inattention crashes in which alcohol was absent, the previously observed association reverses; more females (16.8%) than males (7.4%) were involved in this type of crash. The F–M ratios also allow for rapid visualization of this difference between alcohol-related and fatigue– inattention crashes, being much larger among the latter. Female drivers are more prevalent than males among all Level 2 crashes (p < .01). Speeding is the most frequent Level 2 crash factor (61.5%), albeit more predominantly among males (66.6%) than among females (52.2%). Interestingly, female drivers were observed more fre- quently in the crash category of “failure to obey.” Failure- to-obey crashes registered the highest F–M ratio (1.29) in Table 2. This finding is somewhat surprising, as failure- to-obey crashes include crashes associated with red-light running and other failures to obey traffic signals, which female drivers are usually regarded as being less prone to commit than their male counterparts. Male drivers were more prevalent than female drivers among all Level 1 crashes (p < .01). This is not surpris- ing, given how the crash levels were constructed. Most Level 1 crashes involved “maneuver” crashes (93.4% overall), with only 2.5% of Level 1 crashes being due to weather or a slippery road surface. Although the rela- tive involvement of female drivers in crashes involving weather or surface conditions was higher than that of male drivers (3.3% and 2.1%, respectively), almost no gender-related difference occurred for maneuver crashes. Finally, the overall F–M ratio decreased from Level 1 crashes to Level 3 crashes (0.59, 0.55, and 0.47, respec- tively), suggesting that compared with male drivers, female drivers tend to be more represented in Level 1 crashes than in Level 3 crashes. Table 3 reproduces the outcome of the multinomial logistic regression models for the occurrence of the Level 1 crash factors (dependent variable) as a function of gender and age for each of the Level 2 crash factors under consideration. Model 1 in Table 3 reproduces the outcome of the main effects model. This model shows that the likelihood of a maneuver crash increases when the driver is a female, speeds or drives aggressively, is younger than 35 years old, is a non-Hispanic white, and drives alone. For a weather–slippery surface crash, the outcome of the main effects model is similar to that for a maneuver crash, with two exceptions: older driv- ers and minority drivers are more prone to be involved in weather–slippery surface crashes than younger and minority drivers, respectively. Models 2 through 6 in Table 3 show the results of including the dual gender interaction terms. The analysis of these interactions reveals some interesting findings. For maneuver crashes, there is a significant and negative interaction between those age 65 and older and females. This result suggests that females tend to be relatively less involved in maneuver crashes when they are age 65 and older. For crashes involving weather or slippery surface conditions, there is a significant and negative interaction between passengers and female drivers. In other words, having a passenger reduces the likelihood that a female driver will be involved in a weather–slippery surface crash. Perhaps most relevant for this study, for both maneu- ver and weather–slippery surface crashes, there is also a significant and positive interaction between females and speeding. when included in the model, such interaction makes the main effect nonsignificant. In other words, the outcome of Models 2 through 6 suggests that although females are more likely than males to be involved in maneuver or weather–slippery surface crashes, this result is particularly associated with speeding. Table 4 shows the outcome of the multinomial logis- tic regression models for the occurrence of the Level 2 crash factors (dependent variable) as a function of gender and age for each of the Level 3 crash factors under con- sideration. Among the Level 3 factors, alcohol increases the likelihood of a speeding crash, although a driver who shows fatigue or inattention is less likely to speed. Of particular interest is the analysis of the role of females in speeding crashes, for such a role was relevant to the analy- sis of the occurrence of Level 1 crashes (Table 3). Table 4 shows that females consistently show a lower contribu- tion to speeding crashes than males. Models 2, 3, and 4 in Table 4 show a negative interaction between females and alcohol, suggesting that when alcohol is consumed, female drivers are less likely to speed than male drivers. Once the gender and time interaction enters the model (Models 5 and 6), however, the female–alcohol interaction is no longer significant, suggesting that the association between alcohol, gender, and speeding is highly dependent on the time of day in which it occurs. The significance of the female–passenger interaction shows that for female driv- ers, the presence of a passenger reduces the likelihood of a speeding-related crash; that is, the significance of this interaction suggests again that compared with male driv- ers, female drivers tend to be “safer” when a passenger is present than when driving alone. The somewhat surprising result in Table 4 is the posi- tive and significant contribution of female drivers to failure-to-obey crashes. Females are usually viewed as deviating less from norms than males. Less surprising are the significant and negative interactions between female drivers and nighttime driving and between female drivers and the presence of one or more passengers. This finding suggests that female drivers are less likely to fail to obey when driving in the daytime with at least one passenger. discussion of Results The results of the analyses performed in this study show the appropriateness of the HLDB model. The inclusion

T A b L e 3 M ul ti no m ia l R eg re ss io n of L ev el 1 F ac to rs o n L ev el 2 F ac to rs a nd C ov ar ia bl es M od el 1 M od el 2 M od el 3 M od el 4 M od el 5 M od el 6 V ar ia bl e C oe ff . P > z C oe ff . P > z C oe ff . P > z C oe ff . P > z C oe ff . P > z C oe ff . P > z M an eu ve r Fe m al e 0. 05 7 .0 00 0. 02 0 .2 63 0. 04 0 .1 10 0. 03 1 .4 16 0. 03 9 .3 19 0. 03 3 .4 09 Sp ee di ng 0. 04 3 .0 07 2 0. 00 3 .8 82 0. 00 2 .8 99 0. 00 2 .9 04 0. 00 1 .9 48 0. 00 2 .9 33 Fa ilu re t o ob ey 2 2. 81 1 .0 00 2 2. 76 0 .0 00 2 2. 76 9 .0 00 2 2. 76 9 .0 00 2 2. 76 7 .0 00 2 2. 76 7 .0 00 O th er a gg re ss iv e 3. 96 6 .0 00 3. 87 1 .0 00 3. 87 3 .0 00 3. 87 3 .0 00 3. 87 2 .0 00 3. 87 2 .0 00 A ll to ge th er 1. 49 1 .0 00 1. 50 6 .0 00 1. 51 2 .0 00 1. 51 1 .0 00 1. 51 0 .0 00 1. 51 1 .0 00 Fe m al e an d sp ee di ng 0. 13 5 .0 00 0. 12 1 .0 00 0. 12 1 .0 00 0. 12 4 .0 00 0. 12 4 .0 00 Fe m al e an d fa ilu re t o ob ey 2 0. 07 9 .3 54 2 0. 06 3 .4 61 2 0. 06 3 .4 61 2 0. 06 7 .4 34 2 0. 06 8 .4 26 Fe m al e an d ag gr es si ve 0. 26 9 .2 39 0. 26 6 .2 44 0. 26 6 .2 43 0. 26 8 .2 40 0. 26 8 .2 41 Fe m al e an d al l 2 0. 08 3 .4 03 2 0. 09 6 .3 36 2 0. 09 5 .3 37 2 0. 09 2 .3 56 2 0. 09 3 .3 49 <2 1 0. 20 3 .0 00 0. 20 3 .0 00 0. 20 4 .0 00 0. 20 3 .0 00 0. 20 2 .0 00 0. 20 4 .0 00 21 –3 4 0. 12 3 .0 00 0. 12 4 .0 00 0. 12 5 .0 00 0. 12 5 .0 00 0. 12 3 .0 00 0. 12 3 .0 00 65 + 2 0. 39 5 .0 00 2 0. 39 2 .0 00 2 0. 34 7 .0 00 2 0. 34 6 .0 00 2 0. 34 4 .0 00 2 0. 34 3 .0 00 Fe m al e * <2 1 0. 00 1 .9 86 0. 00 1 .9 84 0. 00 4 .9 33 0. 00 1 .9 82 Fe m al e * 21 –3 4 0. 00 1 .9 78 0. 00 2 .9 59 0. 00 7 .8 51 0. 00 5 .8 87 Fe m al e * 65 + 2 0. 09 2 .0 32 2 0. 09 3 .0 31 2 0. 09 9 .0 23 2 0. 09 8 .0 24 N on w hi te s 2 0. 02 4 .1 60 2 0. 02 4 .1 57 2 0. 02 4 .1 54 2 0. 02 8 .1 77 2 0. 02 8 .1 80 2 0. 02 9 .1 71 Fe m al e * no nw hi te 0. 01 2 .7 37 0. 01 1 .7 55 0. 01 3 .7 15 N ig ht ti m e 2 0. 25 8 .0 00 2 0. 25 6 .0 00 2 0. 25 4 .0 00 2 0. 25 4 .0 00 2 0. 24 5 .0 00 2 0. 24 5 .0 00 Fe m al e * ni gh tt im e 2 0. 02 6 .4 27 2 0. 02 6 .4 40 1 Pa ss en ge r 2 0. 06 0 .0 01 2 0. 05 9 .0 02 2 0. 06 0 .0 01 2 0. 06 0 .0 01 2 0. 06 0 .0 01 2 0. 06 7 .0 05 1+ P as se ng er s 2 0. 07 0 .0 05 2 0. 06 9 .0 06 2 0. 07 0 .0 05 2 0. 07 0 .0 05 2 0. 07 1 .0 05 2 0. 07 9 .0 17 Fe m al e * 1 pa ss en ge r 0. 01 7 .6 47 Fe m al e * 1+ p as se ng er s 0. 01 9 .7 10 C on st an t 0. 29 2 .0 00 0. 30 6 .0 00 0. 29 7 .0 00 0. 30 0 .0 00 0. 29 7 .0 00 0. 29 8 .0 00 W ea th er o r Sl ip pe ry S ur fa ce Fe m al e 0. 35 4 .0 00 0. 17 0 .0 39 0. 12 4 .2 28 0. 13 7 .3 59 0. 15 4 .3 12 0. 24 9 .1 10 Sp ee di ng 1. 16 3 .0 00 0. 98 5 .0 00 1. 02 1 .0 00 1. 02 1 .0 00 1. 01 8 .0 00 1. 01 2 .0 00 Fa ilu re t o ob ey 2 1. 52 3 .0 00 2 1. 24 3 .0 00 2 1. 27 3 .0 00 2 1. 27 3 .0 00 2 1. 27 0 .0 00 2 1. 28 4 .0 00 A ll to ge th er 2 0. 01 6 .9 53 2 0. 30 0 .4 21 2 0. 25 7 .4 91 2 0. 25 7 .4 92 2 0. 26 0 .4 86 2 0. 27 2 .4 66 Fe m al e an d sp ee di ng 0. 40 1 .0 00 0. 32 4 .0 04 0. 32 4 .0 04 0. 33 1 .0 03 0. 33 6 .0 03 Fe m al e an d fa ilu re t o ob ey 2 0. 43 6 .1 13 2 0. 38 4 .1 66 2 0. 38 4 .1 66 2 0. 39 2 .1 58 2 0. 36 7 .1 86 Fe m al e an d al l 0. 61 0 .2 64 0. 53 0 .3 32 0. 53 0 .3 33 0. 53 9 .3 25 0. 55 8 .3 07 <2 1 0. 00 3 .9 65 0. 00 4 .9 59 2 0. 15 2 .1 59 2 0. 15 2 .1 59 2 0. 15 5 .1 52 2 0. 18 8 .0 85 21 –3 4 0. 06 1 .3 45 0. 06 5 .3 10 2 0. 02 1 .8 04 2 0. 02 1 .8 11 2 0. 02 6 .7 60 2 0. 04 4 .6 15 65 + 2 0. 33 6 .0 00 2 0. 32 8 .0 00 2 0. 24 0 .0 42 2 0. 24 1 .0 42 2 0. 23 5 .0 48 2 0. 24 1 .0 43 Fe m al e * <2 1 0. 32 8 .0 33 0. 32 8 .0 33 0. 33 5 .0 30 0. 38 7 .0 13 Fe m al e * 21 –3 4 0. 18 9 .1 35 0. 18 8 .1 38 0. 20 0 .1 19 0. 23 1 .0 73 Fe m al e * 65 + 2 0. 16 1 .3 32 2 0. 16 0 .3 37 2 0. 17 2 .3 05 2 0. 17 7 .2 93 N on w hi te s 0. 17 9 .0 06 0. 17 7 .0 07 0. 17 6 .0 07 0. 18 4 .0 32 0. 18 5 .0 32 0. 19 6 .0 23 Fe m al e * no nw hi te 2 0. 01 6 .9 06 2 0. 01 7 .8 95 2 0. 04 9 .7 11 N ig ht ti m e 2 0. 66 7 .0 00 2 0. 65 7 .0 00 2 0. 64 8 .0 00 2 0. 64 8 .0 00 2 0. 62 2 .0 00 2 0. 62 0 .0 00 Fe m al e * ni gh tt im e 2 0. 06 9 .5 78 2 0. 07 6 .5 35

1 pa ss en ge r 0. 05 9 .3 73 0. 06 3 .3 44 0. 06 2 .3 46 0. 06 2 .3 47 0. 06 2 .3 49 0. 19 5 .0 26 1+ p as se ng er s 2 0. 05 2 .5 73 2 0. 04 7 .6 08 2 0. 04 5 .6 30 2 0. 04 5 .6 28 2 0. 04 6 .6 20 0. 12 8 .3 09 Fe m al e * 1 pa ss en ge r 2 0. 29 3 .0 29 Fe m al e * 1+ p as se ng er s 2 0. 35 9 .0 54 C on st an t 2 3. 71 3 .0 00 2 3. 62 4 .0 00 2 3. 60 4 .0 00 2 3. 61 0 .0 00 2 3. 61 8 .0 00 2 3. 65 4 .0 00 M an eu ve r an d W ea th er Fe m al e 0. 44 0 .0 00 0. 38 9 .0 00 0. 35 9 .0 00 0. 14 2 .2 72 0. 17 6 .1 78 0. 17 7 .1 87 Sp ee di ng 0. 89 5 .0 00 0. 80 4 .0 00 0. 84 2 .0 00 0. 83 9 .0 00 0. 83 0 .0 00 0. 83 0 .0 00 Fa ilu re t o ob ey 2 4. 62 2 .0 00 2 3. 84 1 .0 00 2 3. 88 1 .0 00 2 3. 88 1 .0 00 2 3. 87 3 .0 00 2 3. 87 2 .0 00 O th er a gg re ss iv e 2. 96 5 .0 00 3. 08 0 .0 00 3. 09 2 .0 00 3. 08 9 .0 00 3. 08 3 .0 00 3. 08 3 .0 00 A ll to ge th er 0. 71 9 .0 00 0. 98 6 .0 00 1. 03 0 .0 00 1. 02 3 .0 00 1. 01 3 .0 00 1. 01 3 .0 00 Fe m al e an d sp ee di ng 0. 21 5 .0 19 0. 14 1 .1 36 0. 14 6 .1 23 0. 16 4 .0 86 0. 16 3 .0 86 Fe m al e an d fa ilu re t o ob ey 2 1. 50 8 .1 34 2 1. 44 4 .1 52 2 1. 44 5 .1 51 2 1. 46 3 .1 46 2 1. 46 5 .1 46 Fe m al e an d ag gr es si ve 2 0. 19 8 .5 60 2 0. 22 1 .5 15 2 0. 21 4 .5 29 2 0. 20 3 .5 51 2 0. 20 2 .5 53 Fe m al e an d al l 2 0. 99 7 .0 11 2 1. 07 3 .0 06 2 1. 06 5 .0 06 2 1. 04 2 .0 08 2 1. 04 3 .0 08 <2 1 0. 26 4 .0 00 0. 26 4 .0 00 0. 19 9 .0 23 0. 19 8 .0 24 0. 19 0 .0 30 0. 18 9 .0 33 21 –3 4 0. 11 3 .0 42 0. 11 4 .0 40 2 0. 01 8 .8 16 2 0. 02 8 .7 25 2 0. 04 6 .5 62 2 0. 04 6 .5 58 65 + 2 0. 61 8 .0 00 2 0. 61 3 .0 00 2 0. 46 9 .0 00 2 0. 45 8 .0 00 2 0. 44 3 .0 00 2 0. 44 1 .0 00 Fe m al e * <2 1 0. 12 8 .3 04 0. 12 8 .3 03 0. 14 4 .2 46 0. 14 6 .2 46 Fe m al e * 21 –3 4 0. 26 9 .0 14 0. 28 6 .0 09 0. 31 9 .0 04 0. 32 1 .0 04 Fe m al e * 65 + 2 0. 25 9 .0 82 2 0. 28 1 .0 60 2 0. 30 8 .0 39 2 0. 31 3 .0 37 N on w hi te s 0. 29 8 .0 00 0. 29 8 .0 00 0. 29 6 .0 00 0. 18 2 .0 17 0. 18 2 .0 17 0. 18 4 .0 16 Fe m al e * no nw hi te 0. 25 8 .0 28 0. 25 4 .0 30 0. 25 0 .0 34 N ig ht ti m e 0. 00 0 .0 00 0. 00 0 .0 00 0. 00 0 .0 00 0. 00 0 .0 00 2 1. 17 8 .0 00 2 1. 17 8 .0 00 Fe m al e * ni gh tt im e 2 0. 24 4 .0 43 2 0. 24 5 .0 42 1 pa ss en ge r 0. 10 1 .0 74 0. 10 4 .0 67 0. 10 1 .0 75 0. 10 3 .0 71 0. 10 2 .0 73 0. 07 9 .3 25 1+ p as se ng er s 0. 11 9 .1 07 0. 12 2 .0 99 0. 11 8 .1 12 0. 12 1 .1 03 0. 11 8 .1 12 0. 15 2 .1 60 Fe m al e * 1 pa ss en ge r 0. 04 4 .6 96 Fe m al e * 1+ p as se ng er s 2 0. 06 2 .6 76 C on st an t 2 3. 27 6 .0 00 2 3. 25 3 .0 00 2 3. 23 7 .0 00 2 3. 14 4 .0 00 2 3. 16 1 .0 00 2 3. 16 1 .0 00 N o te : R ef er en ce d ep en de nt le ve l i s “N o L ev el 1 c ra sh .” M od el 1 in cl ud es m ai n ef fe ct s on ly . M od el s 2 th ro ug h 6 pr og re ss iv el y in cl ud e du al g en de r in te ra ct io ns .

T A b L e 4 M ul ti no m ia l R eg re ss io n of L ev el 2 F ac to rs o n L ev el 3 F ac to rs a nd C ov ar ia bl es M od el 1 M od el 2 M od el 3 M od el 4 M od el 5 M od el 6 V ar ia bl e C oe ff . P > z C oe ff . P > z C oe ff . P > z C oe ff . P > z C oe ff . P > z C oe ff . P > z Sp ee di ng Fe m al e 2 0. 26 0 .0 00 2 0. 22 5 .0 00 2 0. 16 4 .0 00 2 0. 22 1 .0 00 2 0. 20 7 .0 00 2 0. 12 5 .0 08 A lc oh ol 0. 54 0 .0 00 0. 57 1 .0 00 0. 57 0 .0 00 0. 57 0 .0 00 0. 55 6 .0 00 0. 55 8 .0 00 Fa ti gu e or in at te nt io n 2 0. 41 5 .0 00 2 0. 45 9 .0 00 2 0. 45 9 .0 00 2 0. 45 9 .0 00 2 0. 46 0 .0 00 2 0. 46 2 .0 00 A lc oh ol a nd f at ig ue 2 0. 07 4 .0 20 2 0. 05 5 .1 51 2 0. 05 7 .1 39 2 0. 05 8 .1 29 2 0. 07 1 .0 65 2 0. 07 0 .0 71 Fe m al e * al co ho l -0 .0 83 .0 18 2 0. 08 0 .0 26 2 0. 07 9 .0 28 2 0. 04 2 .2 78 2 0. 04 8 .2 18 Fe m al e * fa ti gu e 0. 09 1 .1 85 0. 09 1 .1 85 0. 09 2 .1 80 0. 09 4 .1 71 0. 10 7 .1 18 Fe m al e * (a lc oh ol a nd f at ig ue ) -0 .0 41 .5 34 2 0. 03 6 .5 82 2 0. 03 3 .6 22 0. 00 4 .9 51 0. 00 4 .9 57 <2 1 0. 76 9 .0 00 0. 77 0 .0 00 0. 78 5 .0 00 0. 78 5 .0 00 0. 78 0 .0 00 0. 76 2 .0 00 21 –3 4 0. 46 2 .0 00 0. 46 1 .0 00 0. 51 2 .0 00 0. 51 0 .0 00 0. 50 3 .0 00 0. 49 4 .0 00 65 + 2 0. 89 5 .0 00 2 0. 89 5 .0 00 2 0. 84 0 .0 00 2 0. 83 7 .0 00 2 0. 83 0 .0 00 2 0. 83 3 .0 00 Fe m al e * <2 1 2 0. 03 5 .4 37 2 0. 03 5 .4 36 2 0. 02 2 .6 34 0. 01 4 .7 66 Fe m al e * 21 –3 4 2 0. 13 9 .0 00 2 0. 13 4 .0 00 2 0. 11 9 .0 02 2 0. 09 7 .0 12 Fe m al e * 65 + 2 0. 13 1 .0 34 2 0. 13 9 .0 25 2 0. 15 5 .0 13 2 0. 16 2 .0 10 N on w hi te s 2 0. 08 2 .0 00 2 0. 08 1 .0 00 2 0. 08 1 .0 00 2 0. 10 6 .0 00 2 0. 10 6 .0 00 2 0. 09 9 .0 00 Fe m al e * no nw hi te 0 .0 72 .0 61 0. 06 9 .0 69 0. 04 3 .2 62 N ig ht ti m e 0. 23 5 .0 00 0 .2 34 .0 00 0. 23 3 .0 00 0. 23 3 .0 00 0. 26 8 .0 00 0. 26 9 .0 00 Fe m al e * ni gh tt im e 2 0. 10 2 .0 07 2 0. 11 0 .0 03 1 pa ss en ge r 0. 19 5 .0 00 0. 19 5 .0 00 0. 19 4 .0 00 0. 19 4 .0 00 0. 19 4 .0 00 0. 28 5 .0 00 1+ p as se ng er s 0. 21 0 .0 00 0 .2 09 .0 00 0. 20 9 .0 00 0. 21 0 .0 00 0. 20 9 .0 00 0. 30 7 .0 00 Fe m al e * 1 pa ss en ge r 2 0. 25 3 .0 00 Fe m al e * 1+ pa ss en ge rs 2 0. 24 8 .0 00 C on st an t 2 1. 00 5 .0 00 2 1. 02 2 .0 00 2 1. 04 6 .0 00 2 1. 02 7 .0 00 2 1. 03 3 .0 00 2 1. 05 8 .0 00 Fa ilu re t o O be y Fe m al e 0. 32 9 .0 00 0 .3 62 .0 00 0. 48 5 .0 00 0. 34 8 .0 00 0. 37 9 .0 00 0. 44 2 .0 00 A lc oh ol 2 0. 44 5 .0 00 2 0. 39 1 .0 00 2 0. 38 4 .0 00 2 0. 38 7 .0 00 2 0. 42 2 .0 00 2 0. 41 8 .0 00 Fa ti gu e or in at te nt io n 2 0. 63 1 .0 00 2 0. 61 6 .0 00 2 0. 61 6 .0 00 2 0. 61 5 .0 00 2 0. 61 6 .0 00 2 0. 62 0 .0 00 A lc oh ol a nd f at ig ue 2 1. 23 6 .0 00 2 1. 20 4 .0 00 2 1. 19 8 .0 00 2 1. 20 3 .0 00 2 1. 23 4 .0 00 2 1. 23 3 .0 00 Fe m al e * al co ho l 2 0. 11 6 .0 37 2 0. 12 6 .0 27 2 0. 12 2 .0 33 2 0. 05 0 .4 00 2 0. 05 6 .3 41 Fe m al e * fa ti gu e 2 0. 02 4 .7 88 2 0. 02 6 .7 67 2 0. 02 6 .7 69 2 0. 02 4 .7 87 2 0. 01 3 .8 82 Fe m al e * (a lc oh ol a nd f at ig ue ) 2 0. 05 9 .6 88 2 0. 07 0 .6 35 2 0. 06 2 .6 71 0. 00 4 .9 81 0. 00 4 .9 79 <2 1 0. 15 1 .0 00 0. 15 2 .0 00 0. 18 3 .0 02 0. 18 2 .0 02 0. 16 5 .0 06 0. 14 7 .0 15 21 –3 4 2 0. 11 8 .0 01 2 0. 11 9 .0 01 0. 00 1 .9 79 2 0. 00 6 .9 13 2 0. 02 9 .5 82 2 0. 03 9 .4 64 65 + 1. 01 9 .0 00 1 .0 19 .0 00 1. 12 3 .0 00 1. 13 2 .0 00 1. 14 8 .0 00 1. 14 9 .0 00 Fe m al e * <2 1 2 0. 04 6 .5 67 2 0. 04 6 .5 68 2 0. 01 5 .8 57 0. 01 1 .8 88 Fe m al e * 21 –3 4 2 0. 23 4 .0 01 2 0. 22 3 .0 02 2 0. 18 5 .0 11 2 0. 16 7 .0 22 Fe m al e * 65 + 2 0. 18 2 .0 01 2 0. 20 0 .0 00 2 0. 22 8 .0 00 2 0. 23 7 .0 00 N on w hi te s 0. 00 5 .8 68 0. 00 6 .8 46 0. 00 6 .8 35 2 0. 07 7 .0 76 2 0. 07 6 .0 84 2 0. 06 5 .1 35 Fe m al e * no nw hi te 0. 17 1 .0 06 0. 16 5 .0 08 0. 14 3 .0 21 N ig ht ti m e 2 0. 77 7 .0 00 2 0. 78 1 .0 00 2 0. 78 0 .0 00 2 0. 78 0 .0 00 2 0. 64 2 .0 00 2 0. 64 0 .0 00 Fe m al e * ni gh tt im e 2 0. 33 8 .0 00 2 0. 34 4 .0 00 1 pa ss en ge r 0. 45 3 .0 00 0. 45 4 .0 00 0. 45 0 .0 00 0. 45 1 .0 00 0. 45 0 .0 00 0. 50 9 .0 00 1+ p as se ng er s 0. 39 1 .0 00 0 .3 91 .0 00 0. 38 9 .0 00 0. 39 1 .0 00 0. 38 9 .0 00 0. 54 3 .0 00 Fe m al e * 1 pa ss en ge r 2 0. 12 1 .0 32 Fe m al e * 1+ p as se ng er s 2 0. 27 7 .0 01 C on st an t 2 1. 91 2 .0 00 2 1. 93 3 .0 00 2 2. 00 2 .0 00 2 1. 93 7 .0 00 2 1. 95 2 .0 00 2 1. 98 1 .0 00

O th er A gg re ss iv e Fe m al e 2 0. 11 8 .0 00 2 0. 10 0 .0 19 2 0. 14 4 .0 11 2 0. 13 0 .0 91 2 0. 12 3 .1 13 2 0. 13 5 .0 94 A lc oh ol 0. 24 0 .0 00 0. 25 5 .0 00 0. 24 6 .0 00 0. 24 6 .0 00 0. 23 9 .0 00 0. 23 9 .0 00 Fa ti gu e or in at te nt io n 2 0. 98 2 .0 00 2 0. 84 0 .0 00 2 0. 83 9 .0 00 2 0. 83 9 .0 00 2 0. 83 9 .0 00 2 0. 83 9 .0 00 A lc oh ol a nd f at ig ue 2 0. 92 1 .0 00 2 0. 95 9 .0 00 2 0. 96 8 .0 00 2 0. 96 7 .0 00 2 0. 97 4 .0 00 2 0. 97 4 .0 00 Fe m al e * al co ho l 2 0. 03 3 .5 89 2 0. 01 1 .8 56 2 0. 01 2 .8 52 0. 00 5 .9 43 0. 00 6 .9 26 Fe m al e * fa ti gu e 2 0. 30 9 .0 42 2 0. 31 0 .0 41 2 0. 31 0 .0 41 2 0. 30 9 .0 42 2 0. 31 2 .0 40 Fe m al e * (a lc oh ol a nd f at ig ue ) 0. 12 7 .4 12 0. 14 5 .3 54 0. 14 4 .3 57 0. 16 0 .3 10 0. 16 0 .3 09 <2 1 0. 24 1 .0 00 0. 24 0 .0 00 0. 17 3 .0 01 0. 17 3 .0 01 0. 17 0 .0 02 0. 17 5 .0 01 21 –3 4 0. 15 0 .0 00 0. 14 9 .0 00 0. 15 6 .0 00 0. 15 7 .0 00 0. 15 3 .0 00 0. 15 5 .0 00 65 + 2 0. 31 4 .0 00 2 0. 31 4 .0 00 2 0. 34 4 .0 00 2 0. 34 4 .0 00 2 0. 34 1 .0 00 2 0. 34 5 .0 00 Fe m al e * <2 1 0. 17 9 .0 37 0. 17 8 .0 37 0. 18 4 .0 32 0. 17 6 .0 42 Fe m al e * 21 –3 4 2 0. 01 8 .7 92 2 0. 02 0 .7 74 2 0. 01 3 .8 54 2 0. 01 9 .7 90 Fe m al e * 65 + 0. 07 0 .4 32 0. 07 3 .4 17 0. 06 6 .4 67 0. 07 6 .4 03 N on w hi te s 2 0. 13 8 .0 00 2 0. 13 8 .0 00 2 0. 13 9 .0 00 2 0. 13 4 .0 01 2 0. 13 4 .0 01 2 0. 13 7 .0 01 Fe m al e * no nw hi te 2 0. 01 9 .7 81 2 0. 02 0 .7 69 2 0. 01 0 .8 87 N ig ht ti m e 0. 00 0 .0 00 0. 00 0 .0 00 0. 00 0 .0 00 0. 00 0 .0 00 0. 13 3 .0 01 0. 13 3 .0 01 Fe m al e * ni gh tt im e 2 0. 04 9 .4 76 2 0. 04 7 .5 00 1 pa ss en ge r 0. 14 5 .0 00 0. 14 5 .0 00 0. 14 7 .0 00 0. 14 7 .0 00 0. 14 6 .0 00 0. 17 9 .0 00 1+ p as se ng er s 0. 23 0 .0 00 0 .2 30 .0 00 0. 23 5 .0 00 0. 23 5 .0 00 0. 23 4 .0 00 0. 17 8 .0 07 Fe m al e * 1 pa ss en ge r 2 0. 07 1 .3 46 Fe m al e * 1+ p as se ng er s 0. 13 8 .1 55 C on st an t 2 2. 15 1 .0 00 2 2. 16 0 .0 00 2 2. 14 3 .0 00 2 2. 14 6 .0 00 2 2. 14 9 .0 00 2 2. 14 8 .0 00 A ll T og et he r Fe m al e 2 0. 53 4 .0 00 2 0. 56 0 .0 00 2 0. 43 2 .0 00 2 0. 66 0 .0 00 2 0. 66 4 .0 00 2 0. 61 0 .0 00 A lc oh ol 0. 57 5 .0 00 0. 57 3 .0 00 0. 56 4 .0 00 0. 56 1 .0 00 0. 56 2 .0 00 0. 56 4 .0 00 Fa ti gu e or in at te nt io n 2 1. 20 3 .0 00 2 1. 32 5 .0 00 2 1. 33 1 .0 00 2 1. 33 1 .0 00 2 1. 33 1 .0 00 2 1. 33 4 .0 00 A lc oh ol a nd f at ig ue 2 0. 91 1 .0 00 2 0. 94 7 .0 00 2 0. 95 9 .0 00 2 0. 96 4 .0 00 2 0. 96 3 .0 00 2 0. 96 3 .0 00 Fe m al e * al co ho l 0. 01 0 .9 02 0. 04 2 .6 06 0. 04 5 .5 78 0. 03 2 .7 18 0. 02 5 .7 77 Fe m al e * fa ti gu e 0. 29 5 .1 93 0. 31 3 .1 67 0. 31 8 .1 60 0. 31 7 .1 61 0. 32 9 .1 46 Fe m al e * (a lc oh ol a nd f at ig ue ) 0. 14 7 .4 88 0. 19 3 .3 64 0. 20 9 .3 26 0. 19 8 .3 57 0. 19 7 .3 59 <2 1 0. 96 1 .0 00 0. 96 1 .0 00 1. 04 9 .0 00 1. 04 7 .0 00 1. 04 7 .0 00 1. 03 6 .0 00 21 –3 4 0. 59 4 .0 00 0. 59 4 .0 00 0. 68 0 .0 00 0. 67 4 .0 00 0. 67 4 .0 00 0. 66 8 .0 00 65 + 2 0. 84 6 .0 00 2 0. 84 5 .0 00 2 1. 04 5 .0 00 2 1. 03 7 .0 00 2 1. 03 7 .0 00 2 1. 03 7 .0 00 Fe m al e * <2 1 2 0. 31 8 .0 02 2 0. 31 9 .0 02 2 0. 32 5 .0 02 2 0. 30 4 .0 04 Fe m al e * 21 –3 4 2 0. 28 0 .0 02 2 0. 26 1 .0 03 2 0. 26 6 .0 03 2 0. 25 1 .0 05 Fe m al e * 65 + 0 .4 40 .0 06 0. 40 7 .0 11 0. 41 3 .0 11 0. 40 1 .0 13 N on w hi te s 2 0. 17 4 .0 00 2 0. 17 3 .0 00 2 0. 17 2 .0 00 2 0. 24 3 .0 00 2 0. 24 2 .0 00 2 0. 23 6 .0 00 Fe m al e * no nw hi te 0. 29 1 .0 01 0. 29 2 .0 01 0. 26 9 .0 03 N ig ht ti m e 0. 29 7 .0 00 0. 29 6 .0 00 0. 29 0 .0 00 0. 29 1 .0 00 0. 29 2 .0 00 0. 29 4 .0 00 Fe m al e * ni gh tt im e 0 .0 24 .7 83 0. 01 6 .8 53 1 pa ss en ge r 0. 41 1 .0 00 0. 41 1 .0 00 0. 41 0 .0 00 0. 41 1 .0 00 0. 41 1 .0 00 0. 44 7 .0 00 1+ p as se ng er s 0. 38 5 .0 00 0 .3 84 .0 00 0. 38 1 .0 00 0. 38 5 .0 00 0. 38 5 .0 00 0. 47 7 .0 00 Fe m al e * 1 pa ss en ge r 2 0. 06 4 .4 78 Fe m al e * 1+ p as se ng er s 2 0. 27 3 .0 25 C on st an t 2 2. 94 5 .0 00 2 2. 93 8 .0 00 2 2. 97 9 .0 00 2 2. 92 6 .0 00 2 2. 92 6 .0 00 2 2. 94 2 .0 00 N o te : R ef er en ce d ep en de nt le ve l i s “N o L ev el 2 c ra sh .” M od el 1 in cl ud es m ai n ef fe ct s on ly . M od el s 2 th ro ug h 6 pr og re ss iv el y in cl ud e du al g en de r in te ra ct io ns .

10 wOMEN’S ISSUES IN TRANSPORTATION, VOLUME 2 of higher-level factor conditions in analytical models explains the occurrence of crashes at lower levels. These results also reveal the implicit complexity of the HLDB model and the analytical difficulties such a model pre- sents. The HLDB model proposed here could be analyti- cally visualized as a multilevel model with multinomial categorical dependent variables, both continuous and categorical independent terms with weights adjusting for driving exposure, and the inclusion of interaction terms. To the authors’ knowledge, none of the statistical pack- ages currently available (e.g., Mplus, STATA-Gllamm, SAS) satisfactorily deals with all of these analytical com- plexities, particularly in a situation like this, in which the need to model interaction terms of meaningful inter- pretation was highly relevant. To address this limitation, the analyses were separated into two parts to look inde- pendently at the contribution of Level 3 to the Level 2 crashes and of the Level 2 crashes to the Level 1 crashes. Although this approach did not allow for a full integra- tion of the three-level factors, it did provide some mean- ingful results. Focusing on the gender-related interaction terms and going from Level 1 to Level 3 showed that, compared with males, female drivers were overrepresented in fatal maneuver and weather–slippery surface (i.e., Level 1 fac- tor) crashes, but mostly when they were speeding. That is, although female drivers are not more prone to skill- related crashes than males, they seem to be more vulner- able to these crashes than males when speeding (a Level 2 crash factor) is involved. This overall finding is moder- ated by the presence of passengers and the time of day in which the crash occurs, for female drivers are less likely to speed than males when carrying a passenger during the daytime. Interestingly, although speeding seems to induce more “loss of control” crashes (i.e., Level 1 crashes) among female drivers than among males, the study shows that females are less likely to speed than males. Moreover, even when alcohol (a Level 3 crash fac- tor that induces speeding) is present, female drivers are less likely to speed than their male counterparts. Although female drivers are less likely to be involved in alcohol-related crashes, they are overrepresented in the other Level 3 crash factor under consideration: fatigue or inattention. The overrepresentation of females in this Level 3 factor did not, however, affect the involvement of female drivers in either Level 2 or Level 1 crashes. Perhaps the most surprising finding of this study is the overrepre- sentation of female drivers in failure-to-obey crashes, for which this paper does not offer an explanation. In summary, this study presents new evidence against the popular belief that female drivers are more prone to be involved in skill-related crashes than males. At least such evidence is strong under “normal” traffic condi- tions. It might not be the case, however, when speeding is involved. Nevertheless, the finding that females are less likely to speed than males tends to compensate for such negative outcome, at least at the aggregated level. Despite its interesting results, this study is not free of shortcomings. Data for this study include only fatal crashes. Further, to maximize the likelihood that driv- ers included in this study were responsible for the crash, only single-vehicle crashes were included in the data set. It is very likely that the application of the HLDB model to nonfatal crashes might have yielded very different results. Another shortcoming of this study involves the construction of the levels and factors applied. Such a con- struction unavoidably implies some degree of arbitrari- ness; however, to minimize such arbitrariness in building these variables, the FARS instructions were followed to the degree possible. Another weakness of the study involves the data used to adjust for driving exposure. As mentioned, estimates of VMT are currently available only for the years 1995 and 2001. Extrapolating those estimates to other years may have led to sudden or recent changes, or both, in driving patterns being ignored in the analyses. Finally, another limitation to this study (perhaps the most relevant) involves the possibility of bias on the part of the officers who coded the crashes. Police officers may have been more prone to assign different codes to females than males (e.g., by assigning a maneuver code to females more frequently than to males). Nevertheless, the study only found an association between female drivers and maneuver crashes in situations in which speeding was involved. If there was gender-related bias in the way that these types of crashes were registered, then such an association between speed- ing and maneuver crashes should be revisited. AcknowledgMent This research was supported by a grant from the National Institute of Child Health and Human Develop- ment (NICHD). RefeRences Beirness, D. J. 1989. Female Drivers in Canada: Trends in Acci- dent Involvement. In Women, Alcohol, Drugs and Traffic: Proceedings of the International Workshop, Almquist and wiksell, Stockholm, Sweden. Braver, E. R. 2001. Race, Hispanic Origin, and Socioeconomic Status in Relation to Motor Vehicle Occupant Death Rates and Risk Factors Among Adults. Insurance Institute for Highway Safety, Arlington, VA. Cerrelli, E. C. 1998. Crash Data and Rates for Age-Sex Groups of Drivers, 1998. Research Note. National Highway Traf- fic Safety Administration, U.S. Department of Transporta- tion, washington, D.C.

11FEMALE INVOLVEMENT IN U.S. FATAL CRASHES Federal Highway Administration. n.d. National Household Travel Survey. U.S. Department of Transportation, wash- ington, D.C. http://nhts.ornl.gov/download.shtml. Keskinen, E. 1996. why Do Young Drivers Have More Acci- dents? In Berichte der Bundesanstalt für Strassenwesen, B. S. Mensch & Sicherheit, Bergisch Gladbach, Germany. Kostyniuk, L. P., L. J. Molnar, and D. w. Eby. 2000. Are women Taking More Risks while Driving?: A Look at Michigan Drivers. In Women’s Travel Issues: Proceedings from the Second National Conference, October 1996, FHwA, U.S. Department of Transportation, washington, D.C., pp. 502–516. http://www.fhwa.dot.gov/ohim/wom ens/chap26.pdf. Laapotti, S., and E. Keskinen. 1998. Differences in Fatal Loss- of-Control Accidents Between Young Male and Female Drivers. Accident Analysis and Prevention, Vol. 30, No. 4, pp. 435–442. Laapotti, S., and E. Keskinen. 2004. Has the Difference in Acci- dent Patterns Between Male and Female Drivers Changed Between 1984 and 2000? Accident Analysis and Preven- tion, Vol. 36, No. 4, pp. 577–584. Laapotti, S., E. Keskinen, and S. Rajalin. 2003. Comparison of Young Male and Female Drivers’ Attitude and Self- Reported Traffic Behaviour in Finland in 1978 and 2001. Journal of Safety Research, Vol. 34, No. 5, pp. 579–587. Labouvie, E. w., and I. Pinsky. 2001. Substance Use and Driv- ing: The Coexistence of Risky and Safe Behaviors. Addic- tion, Vol. 96, No. 3, pp. 473–484. McKenna, F. P., R. A. Stanier, and C. Lewis. 1991. Factors Underlying Illusory Self-Assessment of Driving Skill in Males and Females. Accident Analysis and Prevention, Vol. 23, No. 1, pp. 45–52. National Highway Traffic Safety Administration. n.d. Traffic Safety Facts 2005: A Compilation of Vehicle Crash Data from the Fatality Analysis Reporting System and the Gen- eral Estimates System. DOT HS 810 631. National Center for Statistics and Analysis, U.S. Department of Transpor- tation, washington, D.C. http://www-nrd.nhtsa.dot.gov/ pubs/tsf2005.pdf. National Highway Traffic Safety Administration. 1995. Fatal- ity Analysis Reporting System (FARS) 1995. DOT HS 809 726. National Center for Statistics and Analysis, U.S. Department of Transportation, washington, D.C. National Highway Traffic Safety Administration. 2001. Tar- geting Young Female Drivers with Impaired Driving Mes- sages. Traffic Tech Technology Series Number 240. U.S. Department of Transportation, washington, D.C. National Highway Traffic Safety Administration. 2008. FARS Analytic Reference Guide 1975 to 2007. DOT HS 810 937. U.S. Department of Transportation, washing- ton, D.C. http://www-nrd.nhtsa.dot.gov/Pubs/810937 .PDF. Pisarski, A. E. 1992. Travel Behavior Issues in the 90’s. 1990 Nationwide Personal Transportation Survey (NPTS) Pub- lication Series. Federal Highway Administration, U.S. Department of Transportation, washington, D.C. Progressive Insurance. 2000. Fatigue, Alcohol and Child Safety Issues Expected to Rise Dramatically This Summer (Sur- vey): The Auto Channel. Romano, E., T. Kelley-Baker, and R. B. Voas. 2008. Female Involvement in Fatal Crashes: Increasingly Riskier or Increasingly Exposed? Accident Analysis and Prevention, Vol. 40, No. 5, pp. 1781–1788. Stamatiadis, N., and J. A. Deacon. 1995. Trends in Highway Safety: Effects of an Aging Population on Accident Pro- pensity. Accident Analysis and Prevention, Vol. 27, No. 4, pp. 443–459. Subramanian, R. 2002. Transitioning to Multiple Imputa- tion: A New Method to Estimate Missing Blood Alcohol Concentration (BAC) Values in FARS. DOT HS 809 403. Mathematical Analysis Division, National Center for Statistics and Analysis, National Highway Traffic Safety Administration, U.S. Department of Transportation, washington, D.C. Voas, R. B., J. wells, D. Lestina, A. williams, and M. Greene. 1998. Drinking and Driving in the United States: The 1996 National Roadside Survey. Accident Analysis and Prevention, Vol. 30, No. 2, pp. 267–275. waller, P., M. R. Elliott, J. T. Shope, T. E. Raghunathan, and R. J. A. Little. 2001. Changes in Young Adult Offense and Crash Patterns over Time. Accident Analysis and Preven- tion, Vol. 33, No. 1, pp. 117–128.