National Academies Press: OpenBook

Guidelines for Slope Traversability (2019)

Chapter: Chapter 3. Crash Data Analysis

« Previous: Chapter 2. Literature Review
Page 30
Suggested Citation:"Chapter 3. Crash Data Analysis." National Academies of Sciences, Engineering, and Medicine. 2019. Guidelines for Slope Traversability. Washington, DC: The National Academies Press. doi: 10.17226/25415.
×
Page 30
Page 31
Suggested Citation:"Chapter 3. Crash Data Analysis." National Academies of Sciences, Engineering, and Medicine. 2019. Guidelines for Slope Traversability. Washington, DC: The National Academies Press. doi: 10.17226/25415.
×
Page 31
Page 32
Suggested Citation:"Chapter 3. Crash Data Analysis." National Academies of Sciences, Engineering, and Medicine. 2019. Guidelines for Slope Traversability. Washington, DC: The National Academies Press. doi: 10.17226/25415.
×
Page 32
Page 33
Suggested Citation:"Chapter 3. Crash Data Analysis." National Academies of Sciences, Engineering, and Medicine. 2019. Guidelines for Slope Traversability. Washington, DC: The National Academies Press. doi: 10.17226/25415.
×
Page 33
Page 34
Suggested Citation:"Chapter 3. Crash Data Analysis." National Academies of Sciences, Engineering, and Medicine. 2019. Guidelines for Slope Traversability. Washington, DC: The National Academies Press. doi: 10.17226/25415.
×
Page 34
Page 35
Suggested Citation:"Chapter 3. Crash Data Analysis." National Academies of Sciences, Engineering, and Medicine. 2019. Guidelines for Slope Traversability. Washington, DC: The National Academies Press. doi: 10.17226/25415.
×
Page 35
Page 36
Suggested Citation:"Chapter 3. Crash Data Analysis." National Academies of Sciences, Engineering, and Medicine. 2019. Guidelines for Slope Traversability. Washington, DC: The National Academies Press. doi: 10.17226/25415.
×
Page 36
Page 37
Suggested Citation:"Chapter 3. Crash Data Analysis." National Academies of Sciences, Engineering, and Medicine. 2019. Guidelines for Slope Traversability. Washington, DC: The National Academies Press. doi: 10.17226/25415.
×
Page 37
Page 38
Suggested Citation:"Chapter 3. Crash Data Analysis." National Academies of Sciences, Engineering, and Medicine. 2019. Guidelines for Slope Traversability. Washington, DC: The National Academies Press. doi: 10.17226/25415.
×
Page 38
Page 39
Suggested Citation:"Chapter 3. Crash Data Analysis." National Academies of Sciences, Engineering, and Medicine. 2019. Guidelines for Slope Traversability. Washington, DC: The National Academies Press. doi: 10.17226/25415.
×
Page 39
Page 40
Suggested Citation:"Chapter 3. Crash Data Analysis." National Academies of Sciences, Engineering, and Medicine. 2019. Guidelines for Slope Traversability. Washington, DC: The National Academies Press. doi: 10.17226/25415.
×
Page 40
Page 41
Suggested Citation:"Chapter 3. Crash Data Analysis." National Academies of Sciences, Engineering, and Medicine. 2019. Guidelines for Slope Traversability. Washington, DC: The National Academies Press. doi: 10.17226/25415.
×
Page 41
Page 42
Suggested Citation:"Chapter 3. Crash Data Analysis." National Academies of Sciences, Engineering, and Medicine. 2019. Guidelines for Slope Traversability. Washington, DC: The National Academies Press. doi: 10.17226/25415.
×
Page 42
Page 43
Suggested Citation:"Chapter 3. Crash Data Analysis." National Academies of Sciences, Engineering, and Medicine. 2019. Guidelines for Slope Traversability. Washington, DC: The National Academies Press. doi: 10.17226/25415.
×
Page 43
Page 44
Suggested Citation:"Chapter 3. Crash Data Analysis." National Academies of Sciences, Engineering, and Medicine. 2019. Guidelines for Slope Traversability. Washington, DC: The National Academies Press. doi: 10.17226/25415.
×
Page 44
Page 45
Suggested Citation:"Chapter 3. Crash Data Analysis." National Academies of Sciences, Engineering, and Medicine. 2019. Guidelines for Slope Traversability. Washington, DC: The National Academies Press. doi: 10.17226/25415.
×
Page 45
Page 46
Suggested Citation:"Chapter 3. Crash Data Analysis." National Academies of Sciences, Engineering, and Medicine. 2019. Guidelines for Slope Traversability. Washington, DC: The National Academies Press. doi: 10.17226/25415.
×
Page 46
Page 47
Suggested Citation:"Chapter 3. Crash Data Analysis." National Academies of Sciences, Engineering, and Medicine. 2019. Guidelines for Slope Traversability. Washington, DC: The National Academies Press. doi: 10.17226/25415.
×
Page 47
Page 48
Suggested Citation:"Chapter 3. Crash Data Analysis." National Academies of Sciences, Engineering, and Medicine. 2019. Guidelines for Slope Traversability. Washington, DC: The National Academies Press. doi: 10.17226/25415.
×
Page 48
Page 49
Suggested Citation:"Chapter 3. Crash Data Analysis." National Academies of Sciences, Engineering, and Medicine. 2019. Guidelines for Slope Traversability. Washington, DC: The National Academies Press. doi: 10.17226/25415.
×
Page 49
Page 50
Suggested Citation:"Chapter 3. Crash Data Analysis." National Academies of Sciences, Engineering, and Medicine. 2019. Guidelines for Slope Traversability. Washington, DC: The National Academies Press. doi: 10.17226/25415.
×
Page 50
Page 51
Suggested Citation:"Chapter 3. Crash Data Analysis." National Academies of Sciences, Engineering, and Medicine. 2019. Guidelines for Slope Traversability. Washington, DC: The National Academies Press. doi: 10.17226/25415.
×
Page 51
Page 52
Suggested Citation:"Chapter 3. Crash Data Analysis." National Academies of Sciences, Engineering, and Medicine. 2019. Guidelines for Slope Traversability. Washington, DC: The National Academies Press. doi: 10.17226/25415.
×
Page 52
Page 53
Suggested Citation:"Chapter 3. Crash Data Analysis." National Academies of Sciences, Engineering, and Medicine. 2019. Guidelines for Slope Traversability. Washington, DC: The National Academies Press. doi: 10.17226/25415.
×
Page 53
Page 54
Suggested Citation:"Chapter 3. Crash Data Analysis." National Academies of Sciences, Engineering, and Medicine. 2019. Guidelines for Slope Traversability. Washington, DC: The National Academies Press. doi: 10.17226/25415.
×
Page 54
Page 55
Suggested Citation:"Chapter 3. Crash Data Analysis." National Academies of Sciences, Engineering, and Medicine. 2019. Guidelines for Slope Traversability. Washington, DC: The National Academies Press. doi: 10.17226/25415.
×
Page 55
Page 56
Suggested Citation:"Chapter 3. Crash Data Analysis." National Academies of Sciences, Engineering, and Medicine. 2019. Guidelines for Slope Traversability. Washington, DC: The National Academies Press. doi: 10.17226/25415.
×
Page 56
Page 57
Suggested Citation:"Chapter 3. Crash Data Analysis." National Academies of Sciences, Engineering, and Medicine. 2019. Guidelines for Slope Traversability. Washington, DC: The National Academies Press. doi: 10.17226/25415.
×
Page 57

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.

23 CHAPTER 3. CRASH DATA ANALYSIS 3.1 INTRODUCTION This chapter presents a study conducted for this project to identify the types of vehicles that are more likely to rollover on slopes through an analysis of existing crash databases. The presentation includes: (1) a review of the existing national crash databases in terms of their capability to provide the necessary data to accomplish the objective of the study, (2) a discussion on possible selections of vehicle classification schemes to define vehicle types for the analysis based on available data and the choice made by this study in consideration of available project resources, (3) an introduction of the risk measures and statistical methods used in the study to quantify relative vehicle rollover risks on slopes, (4) a summary of the data analysis procedures, results, and key findings, and (5) a table listing vehicles most likely to rollover of slopes. Based on the type of roadside design issues of interest to this project, the data analysis focused on crashes that involved passenger vehicles and occurred on roadways with a posted speed limit between 45 mph and 75 mph (72 km/h and 121 km/h) inclusive. In addition, to ensure that the injury to occupants and damage to vehicles did not occur before the vehicle encroached onto the roadside areas, the analysis was further restricted to single-vehicle ran-off- road (SVROR) crashes for which the location of their first harmful event (FHE) took place outside the shoulder and on either roadside or median areas. Note that crashes with FHE taking place on the gore area and separators, i.e., areas between frontage roads and main trafficways, were not included. National Highway Traffic Safety Administration’s (NHTSA) definition of “passenger vehicles” was adopted, which include various sizes of passenger cars, utility vehicles, vans, and pickup trucks (24, 25). The reason for adopting this definition was that NHTSA has a precise mapping of vehicles in terms of body types as shown in Appendix 1. In addition, it provides tables linking specific vehicle make, model, and model-year to each body type (26). Furthermore, vehicle classification schemes and type definitions are standardized among the three national crash databases administered by NHTSA, which make the analysis results from these databases more comparable. It is well recognized that roadside geometry data are largely unavailable in the existing crash databases. This crash data analysis study is, however, not intended to establish the relationship between rollover probability and steepness of slopes. This relationship is to be established in a separate study using the vehicle dynamics simulations. Through simulations, vehicle roadside encroachment characteristics, such as encroachment speed and angle, and driver control input, such as braking and steering, can be controlled and made uniform across all vehicle types considered. Such controls allow the comparison to be made, and relationships to be established under the same encroachment conditions and driver maneuvering behaviors for all vehicle types. In contrast, such uniform encroachment conditions and driver control inputs across vehicle types cannot be completely assured with the crash data analysis (even if the roadside geometry data are available for the analysis). It is also important to note that the lack of roadside geometry data does not prohibit or bias the crash data analysis to identify vehicle types more likely to rollover for the following reason. Although individual vehicles may experience different sideslope ratios when they

24 encroach on the roadside, the distribution of sideslope ratios experienced collectively by a particular type of passenger vehicles, say, 4-door sedans, can be reasonably assumed to be the same as the distribution experienced by other types of passenger vehicles, say, compact utility vehicles. Putting it differently, sideslope ratios do not discriminate against the types of vehicles and, thus, when treating encroached vehicles of a particular vehicle type as a group, the probability distribution of sideslope ratios experienced by one group, though unknown, can be reasonably assumed to be the same as other vehicle groups. Under this mild assumption, the comparison of rollover risks of encroached vehicles between vehicle types is unbiased with respect to their exposure to sideslope ratios. The rest of the chapter is organized as follows:  3.2 Crash Databases  3.3 Classification of Vehicles  3.4 Risk Measures and Statistical Methods  3.5 Analysis Procedures and Results  3.6 Conclusions Two appendices have been included in the report for this chapter. Appendix 1 provides a list of vehicle body types used by NHTSA. Appendix 2, which is structured into three sections, is intended to provide a more detailed technical description of the risk measures and statistical methods used in the crash data analysis than what has been provided in Section 3.4. The two appendices are organized as follows:  Appendix 1. Classification of Vehicle Body Types by NHTSA  Appendix 2. Risk Measures and Statistical Methods o A2.1 Risk Measures and Statistical Inferences o A2.2 Dealing with Crash Data Limitations o A2.3 Dealing with Sampling Errors in GES Data 3.2 CRASH DATABASES Following national crash databases administered by NHTSA were reviewed with a goal of understanding their capability to provide the necessary data to achieve the analysis objective of identifying vehicle types more likely to rollover on slopes:  Fatality Analysis Reporting System (FARS) (24, 25)  National Automotive Sampling System (NASS) General Estimates System (GES) (26, 28)  NASS Crashworthiness Data System (CDS) (28, 29) FARS database contains a census of all crashes of motor vehicles traveling on public roadways in which a person died within 30 days of the crash. Data in GES and CDS come from a probability sample of police-reported motor vehicle crashes of all severity levels, including property damage only (PDO) crashes, injury crashes, and fatal crashes. GES covers all vehicle types, while CDS focuses on light-duty vehicles. All three databases are collected and organized annually.

25 For over 30 years, FARS has been the most referenced source for U.S. fatal crash data. Since 1988, GES has been an essential source for non-fatal injury crash data. Both systems have used fairly similar sets of data elements (variables) to describe crashes, but the differences in the systems’ coding details have required separate software, documentation, coding, and analysis. In 2009, NHTSA began a three-phase process to standardize and unify FARS and GES data definitions and coding, simplify crash data entry and analysis, and reduce costs and errors. The crashes investigated in CDS have been collected and published annually since 1979. Unlike FARS and GES data, which are described as police-reported level data, CDS data are considered “in-depth” crash data in that crashes are acquired and studied by trained investigators who examine vehicles, injuries, and the crash site in some detail to reconstruct the event and in some cases to estimate impact speed and the resulting velocity change. Relative to the other two databases, the main strengths and limitations of each of the three databases are described next. 3.2.1 FARS Database The main strengths of FARS data are:  Census -- no statistical sampling errors involved as in GES and CDS data  Good data quality (in terms of its timeliness, accuracy, and completeness)  All vehicle types (including passenger cars, light trucks, utility vehicles, vans, medium and large trucks, buses, and motorcycles)  Coding ditch and culvert as two separate fixed objects struck by encroached vehicles. Note that, to be discussed shortly, with the exception of 2010 GES data, ditches and culverts have been coded as a single type of roadside objects in GES and CDS and are therefore not distinguishable in analysis. Main limitations of FARS data are:  Fatal crashes only (representing more severe crashes and thus, on average, is representative of more severe encroachment conditions)  Providing rather limited data about the roadway and roadside where crashes occurred (as in other national crash databases). Note that, relatively, it contains more detailed information about the vehicles and occupants involved in the crash, and it includes a good number of variables describing the movement and collision of each involved vehicle prior to and during the crash. 3.2.2 GES Database NASS GES data are obtained from a probability sample of police-reported traffic crashes, covering all vehicle types and injury severity levels. About 55,000 crash cases per year have been sampled in recent years (from an estimated crash population of about 5.8 million per year). GES survey is designed to sample from a predetermined set of geographical areas and crash strata. The survey divides the U.S. into 1,195 geographical areas, called primary sampling units (PSUs). It is a multi-stage survey with a total sample of 60 PSUs across the U.S. in the first stage, from which approximately 400 police agencies are surveyed in the second stage. The third and final stage is a sampling of crashes from several groups of crashes, called crash strata, which are determined by severity of the injured, type of vehicles involved, and tow status of the

26 vehicles involved. Crashes have been grouped into six strata since 2002. Within each of these crash strata, a systematic sample of crashes is selected. More specifically, every k-th crash listed in the stratum is selected, where the integer k is dependent on a predetermined sampling ratio which varies by stratum. The survey design is a composite design typically described as a stratified multi-stage cluster-sampling design. Under the design, some crash sub-populations are sampled more intensively than others on purpose to reduce data collection cost and to achieve certain levels of statistical precision when data are used to estimate some population parameters of policy and legislative interests. The resulting sample can be very different from the overall population from which it is drawn and about which we wish to make inferences. Statistically, it is a biased sample by design, in which crashes are selected with unequal probabilities. In addition, sample cases are not statistically independent because of the sampling from geographical clusters. Conventional statistical methods, which are based on simple random sampling (SRS) or equal selection probability sampling assumption, are not appropriate for analyzing complex survey data, such as GES and CDS data. Specialized survey-based statistical methods are needed. This has been discussed extensively in the statistical literature since the 1950s. Many textbooks on the statistical aspects of the sampling design and related data analysis are available. The book by Lohr (2010) is an excellent example (30). More specifically, statistical analysis results of sampled data from a complex survey are meaningless if the bias in the sampling design is not properly accounted for in the analysis. In addition, the statistical uncertainty of an estimate cannot be properly determined if the clustering and stratified nature of the data are not taken into account in the inferential procedures. For GES, as well as CDS, NASS statisticians have developed sampling (or case) weights to correct for selection bias, as well as biases incurred during implementation. Each sampled crash case is assigned a sampling weight indicating the number of crashes in the target population the crash represents. These sampling weights make it possible for users to compute unbiased estimates of descriptive population statistics for variables of interest. In addition, these sampling weights and other survey design variables provided in the database, such as PSU and crash stratum, allow the uncertainty of an estimate due to sampling design to be statistically assessed. In summary, the general strengths of GES data are:  All vehicle types (note that CDS focuses on light duty vehicles only)  A probability sample of about 55,000 crash cases per year, which is about 10 times that of the CDS data (and thus estimates will have smaller sampling errors in general) Main limitations include:  Subject to statistical sampling errors (when compared to FARS data)  A biased sample by design (data need to be analyzed with specialized, survey-based statistical method)  Police-reported level crash data (not as detail as CDS data)  Limited roadway and roadside data (as is the case with FARS and CDS data)  With the exception of 2010 data, coding of ditch and culvert as a single type of roadside objects in the database makes it impossible to distinguish them.

27 3.2.3 CDS and Associated Databases Crashes investigated in CDS are a probability sample of police reported crashes in the U.S. A CDS crash must fulfill the following requirements: must be police reported, must involve a harmful event (property damage and/or personal injury) resulting from a crash, and must involve at least one towed passenger car or light truck or van in transport on a trafficway. Every crash, which meets these conditions, has a chance of being selected. About 5,000 crash cases per year have been sampled in recent years. As in GES, CDS survey is designed to sample from a predetermined set of geographical areas and crash strata. It uses the same 1,195 PSUs as GES. But it sampled only 24 of the PSUs in the first stage, from which a number of police jurisdictions are surveyed in the second stage. Note that 27 PSUs were selected from 2002 to 2007. The final stage is a sampling of crashes from 10 crash strata, which are determined by severity of the injured, type and model-year of the vehicle involved, disposition and hospitalization of the injured, and tow status of the vehicles involved. As in GES, the survey design is typically referred to as a stratified multi-stage cluster-sampling design. The discussions above regarding the limitations of sampling and the need to use specialized survey- based statistical methods to analyze the GES data apply to the CDS data as well. CDS contains sample crash cases with a wide variation of sampling weights and some cases have extremely large sampling weights. Let’s take CDS 2008 data as an example. There are 5,160 crash cases in total with an average weight of 408.4. Sampling weights, however, range from 1.2 to 73,136.2. That is, some sample cases represent close to 1 case in the real- world, while some represent over 73,000 cases. In contrast, GES is relatively more efficient in sampling design. For the 55,946 sampled cases in 2008 GES, their sampling weights vary from 1.7 to 1,935.8, with an average weight of 88.6. The consequence of having such a large disparity in sampling weights among sample cases in CDS is that it usually results in large variances in estimated sample statistics. Using sample statistics calculated from such data to infer population parameters can be highly unreliable and should be used with extra caution. When deemed appropriate from engineering considerations, using more years of GES and CDS data in the analysis to reduce the sampling errors of estimates is desirable. As indicated earlier, unlike GES data, which are police-reported level data, CDS data are considered “in-depth” crash data in that crashes were acquired and studied by trained investigators who examined vehicles, injuries, and the crash site in some detail to reconstruct the event and in some cases to estimate impact speed and the resulting velocity change (delta-V). As in FARS and GES, CDS contains limited roadway and roadside data on the crash site. However, the information contained in the scene diagrams, narratives, and scene and vehicle photographs prepared by the investigators can sometimes provide additional data. As in GES, when coding fixed objects struck by the involved vehicles, CDS has used the same code for ditch and culvert. It is therefore not possible to tell from the data whether a vehicle involved in a collision with an object coded as “ditch/culvert” actually struck a ditch or a culvert. As indicated earlier, FARS allows this distinction to be made, while GES starts to make this distinction in 2010. For CDS data, narratives, scene diagrams, and photos provided by the investigators can be manually reviewed to distinguish the two types of objects. The review process is quite time-consuming, however.

28 NCHRP 16-05 project manually reviewed 12 years of SVROR crashes involved ditch/culvert as FHE from 1997 to 2008 (21). The number of sampled cases involved ditches as the FHE was 321 cases in total, in which 263 cases involved roadside ditches and 58 cases involved median ditches. Of the 263 cases struck roadside ditches as the FHE, 109 cases (41% of the cases) rolled over. When these ditch cases are grouped by vehicle type, the available sample cases for each vehicle type are simply too small to be useful for the type of analyses of interest to this study. Note that, due to the inherent randomness of the occurrence of crashes, irrespective of how vehicles are classified, at least 400 slope-related rollover crashes need to be observed for the rollover risk of any vehicle type to be assessed with an acceptable statistical precision. This will be shown and discussed in a later section. In addition, a larger number of rollover crashes are required to achieve the same level of precision if more sources of statistical uncertainties, other than the inherent randomness of crashes, are introduced into the collection of crash data. One such source of uncertainties is the clustered sampling design used by GES and CDS. There are two databases developed or to be developed from the CDS data: NCHRP Project 17-22 and 17-43 databases (31, 32). Each database contains a small subset of the CDS sample cases. However, more detailed roadway and roadside geometric data are available for sampled cases than originally contained in the CDS database. In 17-22 database, these additional roadway and roadside geometric data are collected retrospectively from supplemental field data collection efforts. These databases also provide estimates of vehicle roadside encroachment and impact characteristics, e.g., encroachment speed and angle distributions at the point of departure, using crash reconstruction techniques. As a subset of CDS crash cases, the available number of relevant crashes in these two databases is small. In summary, an obvious strength of CDS data is that it is an in-depth crash data. Its limitations include:  A probability sample of about 5,000 crash cases per year (which is about 1/10 of GES sample cases)  Subject to statistical sampling errors (as in GES)  A biased sample by design (which needs to be analyzed with specialized survey-based statistical methods)  Having a wide variation of sampling weights in sampled cases (which generally results in unreliable estimates of population parameters)  Light duty vehicles only -- passenger cars, utility vehicles, vans, and pickup trucks  Limited roadway and roadside data (as in FARS and GES)  Coding of ditch and culvert as a single type of roadside objects makes it impossible to distinguish them. (However, unlike GES data, narratives, scene diagrams, and photos provided by the investigators are available for each crash case and can be manually reviewed to distinguish the two types of objects.)  Too small a sample to conduct the type of analyses of interest to this study Based on the review presented in this section, the crash data analysis efforts had focused on FARS and GES data. To be presented later, the analysis results from FARS and GES were found to be consistent in general. However, the results from FARS data had a considerably

29 better statistical quality than those from GES data and, therefore, the final analysis had relied more heavily on the results from FARS data. 3.3 CLASSIFICATION OF VEHICLES The premise of this data analysis is that vehicle type is a risk factor affecting a vehicle’s rollover probability when the vehicle runs off the road and encroaches on a sideslope during the excursion. The first step of the analysis was therefore to decide on an appropriate vehicle classification scheme to use in the analysis based on available crash data, statistical consideration, and project resources. As discussed earlier, the focus of the analysis, and thus the vehicle classification scheme, was on passenger vehicles. As shown in Appendix 2, irrespective of how vehicles are classified, at least 400 slope- related rollover crashes need to be observed for the rollover risk of any vehicle type to be assessed with an acceptable statistical precision (because of the inherent randomness of the occurrence of crashes). A larger number of rollover crashes are required to achieve the same level of precision if more sources of statistical uncertainties, other than the inherent randomness of crashes, are introduced into the collection of crash data. One such source of uncertainties, as discussed earlier, is the clustered sampling design used in GES to collect the data. When comparing rollover risks using crash data, most previous studies have classified vehicles by body type. For example, Viner grouped vehicles into five body types: passenger cars, utility vehicles, pickups, vans, and medium/heavy trucks (2). Classifying vehicles by body type was attractive for several reasons. First of all, body type is easy to understand. Secondly, NHTSA has a clear definition for each body type listed in Appendix 1 and provides detailed tables linking specific vehicle make, model, and model-year to each body type. Thirdly, body type is a standardized variable used in the national crash databases administered by NHTSA, including FARS, GES, and CDS, which makes the analysis results from these databases more comparable. Table 3.1 shows slope-related rollover crash frequencies and distribution by NHTSA’s vehicle body type based on 7 years of FARS data from 2004 to 2010. Note that a more detailed discussion of how the slope-related rollover crashes were identified from the database will be given in a later section (Section 3.5). As the table shows, even with a total of 11,703 slope- related rollover crashes, the crash frequencies for most vehicle body types are still quite low and their relative rollover risks on slopes could not be assessed reliably. Vehicle body types with more than 400 slope-related rollover crashes are as follows: 02 – 2-Door Sedan/Hardtop/Coupe 04 – 4-Door Sedan/Hardtop 14 – Compact Utility (ANSI D-16 Utility Vehicle Categories “Small” and “Midsize”) 15 – Large Utility (ANSI D-16 Utility Vehicle Categories “Full Size” and “Large”) 20 – Minivan 30 – Compact Pickup (Gross Vehicle Weight, GVWR, < 4,500 lbs) 31 – Standard Pickup (4,500 lbs ≤ GVWR < 10,000 lbs) Two other vehicle classification schemes were also considered: (1) by vehicle model and (2) based on physical characteristics of vehicles such as dimension and mass. The number of rollover crashes needed to maintain a good statistical quality in assessing the rollover risk made

30 the analysis at the vehicle model level infeasible. Grouping vehicles based on vehicle characteristics, such as wheelbase, track width, center of gravity (C.G.) height, and curb weight, required manual extractions of vehicle data from other data sources, such as Expert AutoStats (33). Considering the available project resources, extracting such vehicle characteristics data manually for a large number of crash cases needed in this study was not deemed affordable. It was determined that NHTSA’s vehicle body type classification was a good choice for this analysis. In addition, NHTSA’s definition of passenger vehicles, based on vehicle’s body type and its trailing units (if any), was adopted for the analysis (see, e.g., page 534 in Appendix C of NHTSA, 2011a (24)). The choice to focus on passenger vehicles, as discussed earlier, excluded crashes that involved single-unit straight trucks, medium and heavy trucks, large buses, and motorcycles. Furthermore, the seven vehicle types with more than 400 slope-related rollover crashes mentioned above, together with the large van, i.e., 21 – Large Van – Including van-based buses, were chosen as the focus of the analysis. Note that, despite its underlying statistical limitation, large vans were included for comparison purposes. Table 3.2 shows slope-related rollover crash frequencies and distribution by NHTSA’s vehicle body type based on GES data from 2004 to 2010. Again, a more detailed discussion of how the slope-related rollover crashes were identified from the database will be given in a later section (Section 3.5). Vehicle body types that have the greatest number of sampled cases are:  4-Door sedan: 1,104 cases  Compact utility: 1,092 cases  Standard pickup: 593 cases  Compact pickup: 405 cases  2-Door sedan: 293 cases On average, each sampled vehicle represents about 76 crashes in the target crash population of interest. As discussed in Appendix 2, with the exception of 2010 data, ditch and culvert are reported as a single type of roadside objects struck by a vehicle and are therefore not distinguishable. This type of rollover crashes is therefore unidentifiable from GES data (except for the 2010 data). Thus, when using GES data, the number of slope-related rollover crashes is only estimable up to a constant at best. See Appendix 2 for more technical discussion. From the 2010 GES data, for which ditch and culvert are reported as two separate objects, it was estimated that among the SVROR crashes of interest to this study, which involved ditch or culvert as FHE, about 10% of the crashes struck culverts and 90% struck ditches. A check on FARS data indicated that the percentage breakdown was about 30% culverts and 70% ditches.

Table 3.1. Slope-related rollover crash frequencies and distribution by NHTSA’s vehicle body type: FARS data (Data Source: FARS, 2004-2010; Posted Speed Limit: 45-75 mph) Body Type Index Rollover Frequency Distribution in Percent Vehicle Body Type Description 1 78 0.67 Convertible (excludes sunroof, T-bar) 2 975 8.33 2-Door Sedan/Hardtop/Coupe 3 237 2.03 3-Door/2-Door Hatchback 4 2,574 21.99 4-Door Sedan/Hardtop 5 51 0.44 5-Door/4-Door Hatchback 6 150 1.28 Station Wagon (excluding van and truck-based) 7 0 0.00 Hatchback, number of doors unknown 8 8 0.07 Sedan/Hardtop, number of doors unknown (since 1994) 9 23 0.20 Other or Unknown automobile type (since 1994) 10 4 0.03 Auto-Based Pickup 14 2,528 21.60 Compact Utility (ANSI D-16 Utility Vehicle Categories: “Small” and “Midsize”) 15 608 5.20 Large Utility (ANSI D-16 Utility Vehicle Categories: “Full Size” and “Large”) 16 185 1.58 Utility Station Wagon 19 2 0.02 Utility Unknown Body 20 514 4.39 Minivan 21 173 1.48 Large Van−Including van-based buses 22 2 0.02 Step Van or Walk-In Van 28 1 0.01 Other Van Type (Hi-Cube Van) 29 0 0.00 Unknown Van Type 30 1,005 8.59 Compact Pickup (Gross Vehicle Weight, GVWR, < 4,500 lbs) 31 2,546 21.76 Standard Pickup (4,500 lbs ≤GVWR < 10,000 lbs) 32 18 0.15 Pickup with Slide-In Camper 39 9 0.08 Unknown (pickup style) Light Conventional Truck Type 40 12 0.10 Cab Chassis-Based (includes light stake, light dump, light tow, rescue vehicles) All 11,703 100.00 All Passenger Vehicles

Table 3.2. Slope-related rollover crash frequencies and distribution by NHTSA’s vehicle body type: GES data (Data Source: GES, 2004-2010; Posted Speed Limit: 45-75 mph) Body Type Index No. of Sampled Cases Estimated No. of Crashes Estimated Distribution (%) Vehicle Body Type Description 1 23 1,243 0.37 Convertible (excludes sunroof, T-bar) 2 293 21,308 6.37 2-Door Sedan/Hardtop/Coupe 3 112 8,154 2.44 3-Door/2-Door Hatchback 4 1,104 82,312 24.61 4-Door Sedan/Hardtop 5 20 1,667 0.50 5-Door/4-Door Hatchback 6 69 4,919 1.47 Station Wagon (excluding van and truck-based) 7 0 0 0.00 Hatchback, number of doors unknown 8 0 0 0.00 Sedan/Hardtop, number of doors unknown (since 1994) 9 164 12,041 3.60 Other or Unknown automobile type (since 1994) 10 1 35 0.01 Auto-Based Pickup 14 1,092 79,802 23.86 Compact Utility (ANSI D-16 Utility Vehicle Categories: "Small" and "Midsize") 15 194 14,629 4.37 Large Utility (ANSI D-16 Utility Vehicle Categories: "Full Size" and "Large") 16 45 3,172 0.95 Utility Station Wagon 17 11 523 0.16 3-Door Coupe 19 5 300 0.09 Utility Unknown Body 20 162 11,134 3.33 Minivan 21 71 6,600 1.97 Large Van–Including van-based buses 22 0 0 0.00 Step Van or Walk-In Van 24 0 0 0.00 Van-based school bus 25 0 0 0.00 Van-based transit bus 28 10 457 0.14 Other Van Type (Hi-Cube Van) 29 2 112 0.03 Unknown Van Type 30 405 34,090 10.19 Compact Pickup (Gross Vehicle Weight, GVWR, < 4,500 lbs) 31 593 50,114 14.98 Standard Pickup (4,500 lbs ≤ GVWR < 10,000 lbs) 32 2 112 0.03 Pickup with Slide-In Camper 39 10 1,049 0.31 Unknown (pickup style) Light Conventional Truck Type 40 1 76 0.02 Cab Chassis-Based (includes light stake, light dump, light tow, rescue vehicles) 45 0 0 0.00 Other Light Conventional Truck Type (includes stretched suburban limousine) 48 9 619 0.19 Unknown Light Truck Type (not a pickup) 49 0 0 0.00 Unknown Light Vehicle Type (automobile, utility vehicle, van or light truck) All 4,398 334,468 100.00 All Passenger Vehicles

33 3.4 RISK MEASURES AND STATISTICAL METHODS Three commonly used risk measures in scientific applications are composition ratio, relative risk, and odds ratio. The first two measures were selected by this study to assess the effect of vehicle types on slope-related rollover risks. As discussed in Appendix 2, the odds ratio measure could not be used by this study because of the limitation of crash data. Composition ratio (CR) is also referred to as representation ratio or disproportionality ratio in some applications. CR is used in this analysis to identify vehicle types that are overrepresented in slope-related rollover crashes. Putting the statistical uncertainty aside, a vehicle belonging to a vehicle type of a higher CR value is more likely to rollover on slopes than a vehicle belonging to a vehicle type of a lower CR value. This particular measure was used in the slope rollover study by Viner (2). Relative risk (RR) is synonymous with risk ratio in many publications. RR can be used to compare the slope-related rollover risk (or probability) of one vehicle type to that of another (e.g., compact utility to 4-door sedans) or between groups of vehicle types (e.g., passenger cars to pickup trucks). To be presented later, one of the usages of RR in this analysis is to compare the rollover risk of a particular type of vehicles – a study group, say large utility vehicles, to that of all the rest of the passenger vehicle types – a comparison group. That is, all other non-large utility passenger vehicle types are combined into a single comparison group. A serious limitation of using the CR measure is that its statistical uncertainty is hard to estimate and not reported in practice. This limitation prevents the analysts from asking questions such as “What is the probability that the difference in CR values between two vehicle types could occur simply by chance (when in fact there is no difference between the two vehicle types)?” The statistical uncertainty of the RR measure, on the other hand, can be estimated with a reasonable degree of accuracy and is widely reported in applications. A non-technical introduction of how these two risk measures are applied in the analysis is given next. More detailed technical discussions are provided in Appendix 2. 3.4.1 Composition Ratio (CR) Vehicle CR is derived by dividing the proportion of a specific type of vehicles in passenger vehicles involved in slope-related rollover crashes by the proportion of the same type of vehicles in passenger vehicles involved in all slope-related crashes, which include both rollover and non-rollover crashes. It can be expressed in an equation as Composition Ratio (for Type v Vehicles) = CR (v) = (Proportion of Type v Vehicles in Passenger Vehicles Involved in Slope-Related Rollovers Crashes) ÷ (Proportion of Type v Vehicles in Passenger Vehicles Involved in All Slope-Related Crashes) A CR value of 1.0 indicates equal representation, values between 0 and 1 indicate underrepresentation, and values greater than 1 indicate overrepresentation. For example, if 20 percent of passenger vehicles involved in slope-related rollover crashes is Type v vehicles and only 10 percent of passenger vehicles involved in all slope-related crashes is Type v vehicles, the CR for the Type v vehicles is 20/10 or 2.0, indicating that Type v vehicles are overrepresented in slope-related rollover crashes among passenger vehicles.

34 3.4.2 Relative Risk (RR) RR is a measure of association commonly used in epidemiological and biomedical studies to compare the chance of a disease developing among individuals of a population exposed to a certain factor compared with a similar population not exposed to the factor. It is usually used in the statistical analysis of binary outcomes (which are rollover and non-rollover in this study) where the outcome of interest usually has a relatively low probability of occurrence. RR for a certain outcome (rollover on slopes) to occur is the ratio of the incidence rate (rollover probability) among individuals (vehicles encroached on slopes) with a given risk factor (a particular vehicle type) to the incidence rate among those without the risk factor (other vehicle types). As indicated, RR can be used to compare the slope-related rollover risk of different study group-comparison group settings, including (1) one vehicle type to that of another, (2) between groups of vehicle types, and (3) one particular vehicle type to that of all the rest of the passenger vehicle types. As an example, the following equation compares the rollover risk of two vehicle types with Type u vehicles as the study group and Type v vehicles as the comparison group. Crashes Related-Slope inver That Rollo Vehicles Passenger vType of Proportion Crashes Related-Slope inver That Rollo Vehicles Passengeru Type of Proportion ),(RRVehicles) vType vs.Vehicles u (Type Risk Relative vu The interpretation of RR values between a study group and a comparison group is as follows:  An RR = 1 means there is no difference in risk between the two groups.  An RR < 1 means the event (rollover on slopes) is less likely to occur in the study group than in the comparison group.  An RR > 1 means the event (rollover on slopes) is more likely to occur in the study group than in the comparison group. Let’s take RR (Compact Utility, Non-Compact Utility Vehicles) = 2.4 as an example. This RR value indicates that a compact utility vehicle is 2.4 times as likely to rollover on slopes as a vehicle from the rest of the passenger vehicle types. More detailed technical descriptions of these two risk measures are provided in Appendix 2, including the calculation of 95% confidence interval and sample size requirements to achieve an acceptable statistical precision.

35 3.5 ANALYSIS PROCEDURES AND RESULTS The number of all slope-related crashes and the number of slope-related rollover crashes are not readily available from the existing crash databases for several reasons, including the lack of roadside geometric data, the complexity of run-off-road crash events, and the limitation of the sequence-of- events data coded for each crash. For example, none of the databases provides variables that allow slope-related rollover crashes to be directly identified and, as a consequence, they need to be inferred from existing variables in the database. Thus, as in previous studies, an important task of the data analysis was to identify variables in the existing databases that can be used to determine whether encroaching on slopes was a likely pre-event that contributed to the rollover. Moreover, these variables should allow the crash frequencies to be estimated in some way so that the risk measures presented in the last section, and associated statistical inferences discussed in Appendix 2, can be derived. Furthermore, the comparison of rollover probability between vehicle types from the derived risk measures should be pertinent to the study and statistically valid. Considering the roadside design issues of interest to this study, an initial data screening using the following criteria was conducted to select a subset of crashes for further analysis:  Crash Type: SVROR crashes  Vehicle Type: passenger vehicles (using NHTSA’s definition)  Roadway Type: posted speed limit between 45 mph and 75 mph (72 km/h and 121 km/h) inclusive  Crash Location: FHE took place outside the shoulder and on either roadside or median areas After the initial screening, the number of slope-related rollover crashes was estimated as the sum of the following two types of rollover crashes, for which encroaching on slopes was likely to be a major pre-event that contributed to the rollover:  Crashes with rollovers as the FHE  Rollover crashes that involved ditches as the FHE and either ditch or rollover was cited as the most harmful event (MHE) This estimation procedure for slope-related rollover crashes is similar to that adopted by Viner (2). For GES data, with the exception of 2010 data, ditch and culvert are reported as a single type of roadside objects and are therefore not distinguishable. These types of rollover crashes are therefore unidentifiable from GES data (except for the 2010 data). Thus, when using GES data, the number of slope-related rollover crashes is only estimable up to a constant (assuming that the second type of rollover crashes is a fixed percentage of the first type of rollover crashes). To estimate the number of slope-related crashes, the number of fixed-object crashes was used as a surrogate measure to estimate relatively how often slope-related crashes occurred for each type of vehicles. Specifically, the expected number of slope-related crashes was assumed to be proportional to the expected number of fixed-object crashes for each vehicle type, and the proportionality constants were assumed to be the same for all vehicle types. In other words, a vehicle type with a greater number of fixed object crashes was expected to have proportionally more roadside encroachments and thus more crashes involving roadside slopes. This is a reasonable assumption in that roadside conditions, such as slopes and fixed objects, do not discriminate against any vehicle type. The use of fixed-object

36 crashes by vehicle type as a surrogate to infer the distribution of slope-related crashes by vehicle type was previously used in Viner’s study as well. From the crashes that met the initial screening criteria, fixed-object crashes were identified as those crashes that collided with any of the following fixed objects as FHE and MHE:  guardrail face and end  concrete, cable, and other traffic barriers  bridge rail and parapet  luminary support and utility pole  tree  wall FHE and MHE were allowed to involve different types of fixed objects above. Note that by restricting crashes to those that collided with fixed objects as MHE allowed crashes that involved non- collision types of harmful events, such as fire and explosion, as MHE to be eliminated. There are good reasons to suggest that the probability for these non-collision types of harmful events to occur may vary by vehicle type. Thus, eliminating these non-collision types of harmful events should create a cleaner data set which allowed vehicle type to be better evaluated as a slope-rollover risk factor. A small sensitivity study was conducted to understand the changes of risk measures when additional fixed objects, such as boulder, building, culvert, and other fixed objects, were added to the fixed-object list above. It was found that the addition of these fixed objects had insignificant effects on the risk measures of interest and did not lead to any changes in inferential conclusions. The crash data analysis efforts had focused on FARS and GES data from 2004 to 2010. Slope- related rollover crash frequencies and distribution by body type have been presented earlier in Tables 3.1 and 3.2. The data analysis focused on the eight major body types discussed earlier. Analysis results based on composition ratios and relative risks are presented in Section 3.5.1 and Section 3.5.2, respectively. Within each section, results from FARS data are presented first, which are followed by the results from GES data. 3.5.1 Composition Ratio Distributions of slope-related rollover crashes and all slope-related crashes, as well as the CR measures, estimated from FARS data are presented in Table 3.3. The distributions are presented for each of the eight major body types and a catch-all type including all other passenger vehicle types. The distributions and CR ratios are also graphically presented in Figures 3.1 and 3.2, respectively. As indicated earlier, since the analysis was based on fatal crashes only, the results are expected to represent more severe roadside encroachment and driver control input conditions.

37 Table 3.3. Distributions of slope-related rollover crashes and slope-related crashes by vehicle body type and associated vehicle composition ratios: FARS data (Data Source: FARS, 2004-2010; Posted Speed Limit: 45-75 mph) Body Type Slope-Related Rollover (%) Slope-Related Crashes (%) Composition Ratio 2-Door Sedan 8.33 13.97 0.60 4-Door Sedan 21.99 39.26 0.56 Compact Utility 21.60 10.17 2.13 Large Utility 5.20 2.17 2.39 Minivan 4.39 3.60 1.22 Large Van 1.48 1.13 1.31 Compact Pickup 8.59 7.53 1.14 Standard Pickup 21.76 14.64 1.49 Other Passenger Vehicles 6.67 7.53 Total 100.01 100.00 Figure 3.1. Distributions of slope-related rollover and crashes by vehicle body types - FARS data. 2-D r S ed an 4-D r S ed an Co mp ac t U tili ty La rg e U tili ty Mi niv an La rg e V an Co mp ac t P ick up St an da rd Pi ck up Slope-Related Rollover Slope-Related Crashes 0 5 10 15 20 25 30 35 40 D is tr ib ut io n in Pe rc en t Vehicle Body Type Crash Type Distributions of Slope-Related Rollover and Slope-Related Crashes by Vehicle Body Type (FARS: 2004-2010)

38 Figure 3.2. Vehicle composition ratios by vehicle body types - FARS data. Table 3.4. Distributions of slope-related rollover crashes and slope-related crashes by vehicle body type and associated vehicle composition ratios: GES data (Data Source: GES, 2004-2010; Posted Speed Limit: 45-75 mph) Body Type Slope-Related Rollover (%) Slope-Related Crashes (%) Composition Ratio 2-Door Sedan 6.37 9.31 0.68 4-Door Sedan 24.61 32.75 0.75 Compact Utility 23.86 12.96 1.84 Large Utility 4.37 2.71 1.61 Minivan 3.33 3.44 0.97 Large Van 1.97 0.68 2.91 Compact Pickup 10.19 6.59 1.55 Standard Pickup 14.98 11.20 1.34 Other Passenger Vehicles 10.31 20.36 Total 100.00 100.00 Putting the statistical uncertainty aside, the CR ratios suggest that compact and large utility vehicles are greatly overrepresented and standard pickups and large vans are somewhat overrepresented in slope-related rollover crashes under more severe vehicle roadside encroachment conditions. Passenger cars, both 2-door and 4-door sedans, are greatly underrepresented. 0.60 0.56 2.13 2.39 1.22 1.31 1.14 1.49 0.0 0.5 1.0 1.5 2.0 2.5 C om po si tio n R at io 2-D r S ed an 4-D r S ed an Co mp ac t U tili ty La rg e U tili ty Mi niv an La rg e V an Co mp ac t P ick up St an da rd Pi ck up Vehicle Body Type Composition Ratio by Vehicle Body Type (Data Source: FARS, 2004-2010; Posted Speed Limit: 45-75 mph)

39 Table 3.4 shows the same results from GES data. Recall that the statistical uncertainty associated with the estimates from GES is much greater than that from FARS data. The estimates are therefore less reliable. Despite the statistical uncertainty, the CR ratios suggest that compact and large utility vehicles are again overrepresented, and passenger cars, both 2-door and 4-door sedans, are underrepresented, which are generally consistent with the results from FARS data. Recall that large vans were included for comparison purposes only because of its sample size and thus statistical limitation. 3.5.2 Relative Risk Relative risks of slope-related rollovers for one particular vehicle type to that of all the rest of the vehicle types and the associated confidence interval from FARS data are presented in Table 3.5 and Figure 3.3. The top three vehicle body types that are more likely to rollover on slopes are compact utility vehicles, large utility vehicles, and standard pickup trucks. This is consistent with the results based on the CR values, as presented above. The RR values for utility vehicles indicate that a utility vehicle is about 2.4 times as likely as a vehicle from the rest of the passenger vehicle types to rollover on slopes. In addition, a standard pickup truck is about 1.6 times as likely as a vehicle from the rest of the passenger vehicle types to rollover on slopes. Table 3.5. Relative risks of slope-related rollovers for one particular vehicle type to that of all the rest of the vehicle types and associated confidence interval: FARS data (Data Source: FARS, 2004-2010) Body Type Number of Slope- Related Rollovers Surrogate Measure for Number of Slope- Related Crashes Relative Risk 95% Confidence Interval − Lower Limit 95% Confidence Interval − Upper Limit 2-Door Sedan 975 1365 0.56 0.52 0.60 4-Door Sedan 2,574 3,835 0.44 0.42 0.46 Compact Utility 2,528 993 2.44 2.33 2.54 Large Utility 608 212 2.47 2.28 2.68 Minivan 514 352 1.23 1.13 1.34 Large Van 173 110 1.32 1.13 1.53 Compact Pickup 1,005 735 1.16 1.08 1.23 Standard Pickup 2,546 1,430 1.62 1.55 1.69 Other Passenger Vehicles 780 736 Total 11,703 9,768 Relative risks of slope-related rollovers for one particular vehicle type to that of 4-door sedans and the associated confidence interval from FARS data are presented in Table 3.6 and Figure 3.4. Relative to the 4-door sedans, the top three vehicle body types that are more likely to rollover on slopes are again compact utility vehicles, large utility vehicles, and standard pickup trucks. Also, a

40 utility vehicle is about 4 times as likely as a 4-door sedan to rollover on slopes, while a standard pickup truck is about 2.6 times as likely. Table 3.6. Relative risks of slope-related rollovers for one particular vehicle type to that of 4- door sedans and associated confidence interval: FARS data (Data Source: FARS, 2004-2010) Body Type Number of Slope-Related Rollovers Surrogate Measure for Number of Slope-Related Crashes Relative Risk 95% Confidence Interval − Lower Limit 95% Confidence Interval − Upper Limit 2-Door Sedan 975 1365 1.06 0.99 1.15 4-Door Sedan 2,574 3,835 1.00 ------ ------ Compact Utility 2,528 993 3.79 3.59 4.00 Large Utility 608 212 4.27 3.91 4.67 Minivan 514 352 2.18 1.98 2.39 Large Van 173 110 2.34 2.01 2.73 Compact Pickup 1,005 735 2.04 1.89 2.19 Standard Pickup 2,546 1,430 2.65 2.51 2.80 Other Passenger Vehicles 780 736 Total 11,703 9,768

Figure 3.3. Relative risks of slope-related rollovers for one particular vehicle body type to that of all the rest of the body types and associated confidence intervals: FARS data. Relative Risks of Slope-Related Rollovers for One Particular Body Type to That of the Rest of the Vehicle Types and Associated Confidence Limits (Data Source: FARS, 2004-2010; Posted Speed Limit: 45-75 mph) 0.0 1.0 2.0 3.0 4.0 5.0 2-Dr Sedan 4-Dr Sedan Compact Utility Large Utility Minivan Large Van Compact Pickup Standard Pickup Vehicle Body Type R el at iv e R is k (v s. A ll O th er T yp es C om bi ne d)

Figure 3.4. Relative risks of slope-related rollovers for one particular vehicle body type to that of 4-door sedans and associated confidence intervals - FARS data. Relative Risks of Slope-Related Rollovers for One Particular Body Type to That of 4-Door Sedans and Associated Confidence Limits (Data Source: FARS, 2004-2010; Posted Speed Limit: 45-75 mph) 0.0 1.0 2.0 3.0 4.0 5.0 2-Dr Sedan 4-Dr Sedan Compact Utility Large Utility Minivan Large Van Compact Pickup Standard Pickup Vehicle Body Type R el at iv e R is k (v s. 4 -D oo r S ed an s)

43 Table 3.7 and Figure 3.5 show the relative risks of slope-related rollovers for one particular vehicle type to that of all the rest of the vehicle types and the associated confidence interval from GES data. As can be seen from the figure, the statistical uncertainty associated with the estimate is much greater than that obtained from FARS data for a particular vehicle type. As discussed in Appendix 2, this is largely due to the nature of the sampling design of GES data. Aside from the RR for large vans, which is highly statistically uncertain, the top three vehicle body types that are more likely to rollover on slopes are compact utility vehicles, large utility vehicles, and compact pickup trucks. This ranking is highly uncertain however. Considering the high statistical uncertainty involved, this result is generally considered consistent with that from FARS data. Table 3.7. Relative risks of slope-related rollovers for one particular vehicle type to that of the rest of the vehicle types and associated confidence interval: GES data (Data Source: GES, 2004-2010) Body Type Estimated Number of Slope- Related Rollovers Surrogate Measure for Number of Slope- Related Crashes Estimated Relative Risk 95% Confidence Interval − Lower Limit 95% Confidence Interval − Upper Limit 2-Door Sedan 21,308 131,751 0.66 0.52 0.83 4-Door Sedan 82,312 463,250 0.67 0.56 0.81 Compact Utility 79,802 183,350 2.10 1.87 2.37 Large Utility 14,629 38,364 1.64 1.27 2.12 Minivan 11,134 48,675 0.97 0.71 1.33 Large Van 6,600 9,571 2.95 2.08 4.26 Compact Pickup 34,090 93,240 1.61 1.37 1.87 Standard Pickup 50,114 158,439 1.40 1.19 1.64 Other Passenger Vehicles 34,595 288,040 Total 334,584 1,414,680

44 Figure 3.5. Relative risks of slope-related rollovers for one particular vehicle body type to that of the rest of the body types and associated confidence intervals - GES data. Relative Risks of Slope-Related Rollovers for One Particular Body Type to That of the Rest of the Vehicle Types and Associated Confidence Limits (Data Source: GES, 2004-2010; Posted Speed Limit: 45-75 mph) 0.0 1.0 2.0 3.0 4.0 5.0 2-Dr Sedan 4-Dr Sedan Compact Utility Large Utility Minivan Large Van Compact Pickup Standard Pickup Vehicle Body Type R el at iv e R is k (v s. A ll O th er T yp es C om bi ne d)

45 3.6 CONCLUSION Based on the observed frequency of slope-related rollover crashes from 2004 to 2010 FARS data, five top ranked vehicle models for each body type are presented in Table 3.8. For each of the five ranked model, the model-year that has the greatest number of slope-related rollover crash frequency and is 2001 or newer model is listed in the table. Vehicle models older than 2001 were not considered since one of the objectives of this research is to update the traversability guidelines based on the current vehicle fleet. Since 2001, NHTSA has provided estimates of rollover risk annually for some popular passenger vehicle models. Estimates are available to the public on NHTSA’s website as part of its 5-star safety ratings consumer information program. Table 3.1 shows the estimated rollover risk for a number of vehicle models by drive system (i.e., 2-wheel and 4-wheel drives). As shown, it includes most of the vehicle models in each of the 8 vehicle body types recommended in the crash data analysis for further comparisons with the MASH vehicles. To be explained next, only 2004 and newer models are covered in the table. NHTSA’s estimates are based on rollover risk models developed in its New Car Assessment Program (NCAP) (34). Estimates for 2001 to 2003 vehicle models were based solely on the static stability factor (SSF) measurement of the vehicle. Note that SSF of a vehicle is the ratio of one half its track width to its center of gravity (C.G.) height. Beginning with 2004 vehicle models, the rollover risk model was developed using a logistic regression technique with two predictors: SSF and an indicator variable representing tip-up performance under a dynamic Fishhook maneuver test with heavy load (the so-called FH test). The regression parameters of the rollover risk model were estimated using 96,000 single-vehicle crashes of 25 vehicle make- models, which included 20,000 rollovers. These crashes occurred between 1994 and 1998 in 6 states. It is important to note that the rollover risk estimated by NHTSA is a measure of all tripped and untripped rollovers, which can occur both on and off the traveled way. Rollovers on slopes, which are of interest to this study, are only a part of the estimated rollover risk. Several general observations about the rollover risks listed in Table 3.9 can be made:  Except for a very small number of vehicle models, the difference in rollover risk is rather small between a 2-wheel drive model and a 4-wheel drive model.  Large vans appear to have the highest rollover risk, followed by large utility and compact utility vehicles.  Relative to passenger cars (including 2-door and 4-door sedans), large vans are about 2.6 times more likely to rollover, while utility vehicles are about 2 times and pickup trucks and minivans are about 1.7 times more likely to rollover. In general, these estimates are consistent with the findings from the crash data analysis.

Table 3.8. Top ranked vehicle models and model-years based on slope-related rollover crash frequencies: FARS data (Data Source: FARS, 2004-2010; Posted Speed Limit: 45-75 mph) Vehicle Body Type Frequency Rank Vehicle Make Vehicle Model Model-Year (2001 on) 2-door sedan,hardtop,coupe 1 Chevrolet Cavalier 2003 2-door sedan,hardtop,coupe 2 Ford Mustang/Mustang II 2003 2-door sedan,hardtop,coupe 3 Chevrolet Monte Carlo (1995 on) 2001 2-door sedan,hardtop,coupe 4 Pontiac Grand AM 2001 2-door sedan,hardtop,coupe 5 Honda Civic/CRX, del Sol 2004 4-door sedan, hardtop 1 Honda Accord 2004 4-door sedan, hardtop 2 Ford Taurus/Taurus X 2001 4-door sedan, hardtop 3 Toyota Camry 2005 4-door sedan, hardtop 4 Pontiac Grand AM 2002 4-door sedan, hardtop 5 Chevrolet Impala/Caprice 2004 Compact utility 1 Ford Bronco (thru 1977)/Bronco II/Explorer/Explorer Sport 2002 Compact utility 2 Chevrolet S-10 Blazer/TrailBlazer (2002 only) 2002 Compact utility 3 Jeep (Includes Willys/Kaiser-Jeep) Cherokee (1984-on) 2001 Compact utility 4 Toyota 4-Runner 2001 Compact utility 5 GMC Jimmy/Typhoon/Envoy 2004 Large utility 1 Chevrolet Fullsize Blazer/Tahoe 2004 Large utility 2 Ford Expedition 2003 Large utility 3 GMC Fullsize Jimmy/Yukon 2001 Large utility 4 Ford Excursion 2002 Large utility 5 Lincoln Navigator 2001 Minivan 1 Dodge Caravan 2002 Minivan 2 Chevrolet Astro Van 2001 Minivan 3 Ford Windstar 2003 Minivan 4 Chevrolet Venture 2004 Minivan 5 Pontiac Trans Sport/ Montana/SV6 2001 Large van 1 Ford E-Series van/Econoline (e.g., E-350 Van) 2004 Large van 2 Chevrolet G-series van (e.g., Express 1500 Van) 2004 Large van 3 Dodge B-Series van/Ram Van/Ram Wagon 2001

Vehicle Body Type Frequency Rank Vehicle Make Vehicle Model Model-Year (2001 on) Large van 4 GMC G-series van/Savana Van 2005 Large van 5 Dodge Sprinter 2004 Compact pickup 1 Ford Ranger 2003 Compact pickup 2 Chevrolet S-10/T-10 Pickup 2002 Compact pickup 3 Toyota Tacoma 2003 Compact pickup 4 Dodge Dakota 2003 Compact pickup 5 GMC S15/T15/Sonoma 2003 Standard pickup 1 Ford F-Series pickup 2001 Standard pickup 2 Chevrolet C, K, R, V-series pickup/Silverado 2003 Standard pickup 3 Dodge Ram Pickup 2001 Standard pickup 4 GMC C, K, R, V-series pickup/Sierra 2003 Standard pickup 5 Toyota Tundra 2002

Table 3.9. Risk of rollover for a number of vehicle models by drive system (Data Source: New Car Assessment Program, NHTSA) Vehicle Body Type Vehicle Make Vehicle Model Model Years (2004+) Risk of Rollover 2-Wheel Drive 4-Wheel Drive 2-door sedan,hardtop,coupe Chevrolet Cavalier 2004-05 11.30% ------ 2-door sedan,hardtop,coupe Ford Mustang 2005-10 8.70% ------ 2-door sedan,hardtop,coupe Chevrolet Monte Carlo 2006-07 10.30% ------ 2-door sedan,hardtop,coupe Pontiac Grand AM 2004-05 10.90% ------ 2-door sedan,hardtop,coupe Honda (Non-Acura) Civic 2006-10 10.30% ------ 4-door sedan, hardtop Honda (Non-Acura) Accord 2008-10 9.50% ------ 4-door sedan, hardtop Ford Taurus 2008-09 10.50% 11.30% 4-door sedan, hardtop Toyota Camry 2007-10 10.70% ------ 4-door sedan, hardtop Pontiac Grand AM 2004-05 10.90% ------ 4-door sedan, hardtop Chevrolet Impala 2006-10 11.30% ------ Compact utility Ford Explorer 2006-10 23.70% 22.80% Compact utility Chevrolet TrailBlazer 2004-09 20.40% 19.10% Compact utility Jeep Cherokee ------ ------ ------ Compact utility Toyota 4-Runner 2004-09 21.90% 20.40% Compact utility GMC Envoy 2004-10 20.40% 19.10% Large utility Chevrolet Tahoe 2006-10 24.60% 22.80% Large utility Ford Expedition 2006-10 21.20% 19.80% Large utility GMC Yukon 2006-10 24.60% 22.80% Large utility Ford Excursion ------ ------ ------ Large utility Lincoln Navigator 2008-10 21.20% 19.10% Minivan Dodge Caravan 2006-07 16.90% ------ Minivan Chevrolet Astro Van ------ ------ ------ Minivan Ford Windstar ------ ------ ------ Minivan Chevrolet Venture ------ ------ ------ Minivan Pontiac Montana SV6 2006 20.40% ------ Large van Ford E-350 12 Passenger Wagon 2008-10 27.90% ------ Large van Chevrolet Express 1500 Passenger Van 2008-10 27.90% 22.80%

Vehicle Body Type Vehicle Make Vehicle Model Model Years (2004+) Risk of Rollover 2-Wheel Drive 4-Wheel Drive Large van Dodge Ram Van/Ram Wagon ------ ------ ------ Large van GMC Savana Van 2006-07 28.30% ------ Large van Dodge Sprinter ------ ------ ------ Compact pickup Ford Ranger 2008-09 21.90% 26.70% Compact pickup Chevrolet S-10 ------ ------ ------ Compact pickup Toyota Tacoma 2006-09 14.70% 19.10% Compact pickup Dodge Dakota 4-Dr 2005-09 19.10% 17.90% Compact pickup GMC Sonoma ------ ------ ------ Standard pickup Ford F150 Regular Cab 2004-08 17.40% 19.80% Standard pickup Chevrolet Silverado 1500 2007-08 19.10% 18.50% Standard pickup Dodge Ram 1500 2005-08 17.90% 19.80% Standard pickup GMC Sierra 2005-06 15.90% 18.50% Standard pickup Toyota Tundra 2007-09 20.40% 19.80% Note: “------“ indicates “not available” or “not applicable.”

Next: Chapter 4. Comparison of Vehicle Characteristics »
Guidelines for Slope Traversability Get This Book
×
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

TRB’s National Cooperative Highway Research Program (NCHRP) has released a pre-publication version of Research Report 911: Guidelines for Slope Traversability, which includes guidelines for determining the traversability of roadside slopes considering the characteristics of the current passenger vehicle fleet.

As part of development of this report, researchers performed full-scale traversability tests and compared the performance of the vehicles with the simulations performed for the same test conditions.

Rollovers are the leading cause of fatalities in single vehicle ran-off-road (SVROR) crashes. Analysis of six years of data from the National Automotive Sampling System Crashworthiness Data System indicates that 31% of SVROR crashes result in a rollover. Approximately 75% of these rollover crashes are initiated by vehicles digging into the ground on embankments or in ditches after encroaching onto the roadside.

Development of NCHRP Research Report 911 was prompted by concern that some roadside slope conditions that have for many years been considered traversable for passenger cars may not be traversable for light trucks. With the steadily increasing percentage of light trucks in the vehicle fleet, further research was needed to determine what should be considered as safe sideslope conditions for today’s vehicle fleet. Proper assessment of slope traversability may help reduce the number of rollover crashes and associated fatalities.

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  6. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!