U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations
|This report is an archived publication and may contain dated technical, contact, and link information|
Publication Number: FHWA-HRT-10-024
Date: April 2010
Development of a Speeding-Related Crash Typology
PDF Version (1.53 MB)
PDF files can be viewed with the Acrobat® Reader®
DEFINITION OF SR CRASHES
For all databases used in this study, the decision on whether or not a crash is SR (regardless of definition) is based on an examination of each vehicle in the crash. First, each vehicle is defined as SR or not based on the variables, which are described below. Then, if any vehicle in the crash is SR, the crash is considered SR. If all of the vehicles involved in a crash are coded as nonspeeding, then the crash is defined as not SR. In all other cases, it is unknown whether the crash is SR.
GES has three of these variables in its vehicle file: SPEEDREL (SR), VIOLATN (violations charged), and P_CRASH2 (critical event). The first variable is coded as "yes" when excessive speed by the driver is noted as a contributing factor on the police accident report or if a speeding violation has been issued. For the second variable, a speeding violation is charged when a driver travels over the speed limit or too fast for conditions. The third variable is coded "too fast for conditions" when drivers lose control of their vehicles due to driving too fast for conditions. Based on the definitions of these three variables, it was determined that SPEEDREL should be used to indicate SR crashes. Thus, if SPEEDREL is coded as "yes" for one of the vehicles involved in a crash, the crash is considered SR.
For FARS, it is theoretically possible to distinguish between the less restrictive and more restrictive definitions of SR. There are three variables which describe the violations that have been charged, and values of 21–25 and 29 all indicate some type of speeding violation. However, only 1.3 percent of drivers involved in a crash were charged a speeding violation. This is in contrast to data on SR in four related factor driver–level variables. In this instance, for 20 percent of the drivers, speeding was noted as a factor in the crash. Given that the sample of fatal crashes involving SR violations was small, it was decided that the driver–related variables combining "over speed limit" and "too fast for conditions" would be used to identify the SR crashes in FARS.
As indicated previously, FARS and GES only allow the use of the broader definition of SR, and HSIS data were used to allow for a more restrictive definition. It is noted that the issue here is more than just a technical one since the choice of definition can affect how the findings of the analyses are interpreted. Both definitions are based on the investigating officer's judgment made during the crash investigation. There is no exact measure of precrash speed available to the officer. It is perhaps true that fatal crashes would undergo a more detailed crash reconstruction, and thus, the precrash speed estimates would be more accurate; however, the level of accuracy is unknown. In addition, it is difficult to know whether an identified variable shows a true higher association with speed or whether the association shown is partially due to an officer bias in noting the "too fast for conditions" factor. Information from police officers indicates that the "too fast" designator may be systematically used in certain situations–if the condition is present, the officer is more likely to use the descriptor. For example, the officer may be more likely to use this designator in bad weather crashes or crashes on curves. Both are logical uses since drivers should reduce speeds during bad weather and should not drive too fast to maneuver around a sharp curve. However, if this systematic use is present, it may overstate the true effect of these variables, making it difficult to determine whether the overrepresentation of these factors is true or partly due to the effect of an inflated use of this descriptor in certain situations. If the latter is found, treatment programs oriented to these factors may not be as successful as if oriented to other characteristics where such a bias is not expected. The HSIS data were used to attempt to provide further clarity concerning whether such a possible bias matters and whether there are different findings when the two definitions are used for the same variable.
Data formats and data from all nine HSIS States were examined to determine the following: (1) whether available variables allowed the use of both the broader and the more restrictive definitions of SR and (2) whether or not the existing data provided ample sample sizes for use in both definitions. Data from North Carolina and Ohio met both these requirements but in different ways. Crash data in HSIS are divided into three linkable files: (1) crash variables, (2) vehicle variables, and (3) occupant variables. Just as with the GES and FARS data, a decision on whether or not a crash is SR is based on whether one or more vehicles indicated an SR factor.
The vehicle file for North Carolina contains three variables describing contributing factors, and the options for these factors include exceeded authorized speed limit and exceeded safe speed for conditions. The vehicle file for Ohio has only one contributing factor variable with unsafe speed and exceeded speed limit as two of the options. However, when the data were examined, the exceeded speed limit category was only checked for approximately 0.1 percent of the vehicles. Since the officer also recorded an estimate of the precrash speed for each vehicle and the posted speed limit at the crash location, these two variables could be compared to determine additional vehicles that are exceeding the posted limit, providing an adequate sample for analysis. It is noted, however, that because the definitional categories were defined differently in the two States, the resulting percentages for the two definitions differed (see table 1 and table 2).
Finally, note that the North Carolina and Ohio data contained in the HSIS files are for State system roads only and are under the control of the State transportation departments. In North Carolina, where there are no county roads systems, the State system roads include all roads except city streets which are not State–owned highways. In Ohio, the HSIS data do not include rural roadways owned by counties and non–State city streets. Thus, in general, the North Carolina and Ohio data are more rural in nature than the GES and FARS data since they do not include crashes on city streets. As will be noted in certain findings below, the differences in the databases affect some of the outcomes.
Table 1 shows the number of SR crashes in 2005 according to GES and FARS. Comparable data for Ohio and North Carolina are shown in table 2. Note that from this point forward, the more liberal definition combining both over speed limit and too fast for conditions will be referred to as the combined definition. In the tables, the combined definition is referred to as "Total SR."
Table 1. The number and percentage of SR and non–SR crashes in GES and FARS (2005).
Table 2. The number and percentage of SR and non–SR crashes in North Carolina (2002–2004) and Ohio (2003–2005).
As shown, the differences in the databases result in differences in the percentage of crashes that are considered SR even when the combined definition is used–20 percent for GES, 30 percent for FARS, 15 percent for North Carolina, and 11 percent for Ohio. As expected, the restricted over speed limit definition results in much lower percentages of SR for North Carolina and Ohio–3 percent for North Carolina and 7 percent for Ohio. The State percentages probably vary because of the differing definitional procedures used. The conclusion drawn is that all these databases are based on somewhat different crash populations. What will be of interest is whether findings based on categories within individual descriptor variables show any consistency across databases.
The findings also indicate that speeding appears to be a more significant factor in fatal crashes (FARS) than in total crashes (all others). This is a logical finding in that speeding is related to crash energy, which is related to injury. This is further examined in table 3 through table 5, which show the severity distributions of crashes from GES, North Carolina, and Ohio. Note that in all tables from this point forward, only crashes where SR can be defined as yes or no are included, and crashes where SR is unknown are omitted.
Table 3. Frequency and number/percentage of SR crashes regarding crash severity in GES.
Table 4. Frequency and number/percentage of SR crashes regarding crash severity in North Carolina.
Table 5. Frequency and number/percentage of SR crashes regarding crash severity in Ohio.
In all three databases, regardless of definition, the SR percentage appears to decrease with crash severity. The percentage for fatal crashes is 1.5 to 15 times higher than the percentage for crashes without injuries, depending on the database and definition. The total number of fatal crashes estimated by GES to be SR (8,689) is significantly smaller than the total number of SR crashes in the FARS database (11,553). As has been found by other studies, the total number of fatal crashes estimated by GES is smaller than the true total in FARS, providing further rationale for using both GES and FARS in this study.
Topics: research, safety, crash statistics, speed
Keywords: research, Safety, Speed, Speeding, Crash typology, Traffic safety, Highway Safety Information System, HSIS
TRT Terms: research, Safety