Highway Safety: Research Continues on a Variety of Factors That  
Contribute to Motor Vehicle Crashes (31-MAR-03, GAO-03-436).	 
                                                                 
Nearly 6.3 million motor vehicle crashes occurred in the United  
States in 2001, or one crash every 5 seconds. On average, a	 
person was injured in these crashes every 10 seconds, and someone
was killed every 12 minutes. Since the 1970s, progress has been  
made in reducing the number of fatalities and injuries on our	 
nation's roads. From 1975 through 2001, fatalities decreased from
44,525 to 42,116, while the rate of fatalities per 100 million	 
vehicle miles traveled decreased from 3.35 to 1.51. However, the 
decline in fatalities has leveled off in recent years. In the	 
1970s, Indiana University conducted one of the most significant  
studies to date on the factors that contribute to motor vehicle  
crashes. This study examined human, environmental, and vehicle	 
factors that contribute to crashes. As requested, this report	 
provides more recent information from data, experts, and studies 
about the factors that contribute to motor vehicle crashes and	 
information about major ongoing and planned Department of	 
Transportation research into factors that contribute to crashes. 
-------------------------Indexing Terms------------------------- 
REPORTNUM:   GAO-03-436 					        
    ACCNO:   A06520						        
  TITLE:     Highway Safety: Research Continues on a Variety of       
Factors That Contribute to Motor Vehicle Crashes		 
     DATE:   03/31/2003 
  SUBJECT:   Data collection					 
	     Highway safety					 
	     Motor vehicle safety				 
	     Motor vehicles					 
	     Traffic accidents					 
	     Traffic violations 				 
	     Transportation research				 
	     Transportation safety				 
	     NHTSA Crashworthiness Data System			 
	     NHTSA Fatality Analysis Reporting System		 
	     NHTSA General Estimates System			 

******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO Product.                                                 **
**                                                              **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced.  Tables are included, but    **
** may not resemble those in the printed version.               **
**                                                              **
** Please see the PDF (Portable Document Format) file, when     **
** available, for a complete electronic file of the printed     **
** document's contents.                                         **
**                                                              **
******************************************************************
GAO-03-436

Report to Congressional Requesters

United States General Accounting Office

GAO

March 2003 HIGHWAY SAFETY Research Continues on a Variety of Factors That
Contribute to Motor Vehicle Crashes

GAO- 03- 436

Many factors combine to produce circumstances that may lead to a motor
vehicle crash* there is rarely a single cause of such an event. Three
categories of factors contribute to crashes: human factors, roadway
environment factors, and vehicle factors. Human factors involve the
actions taken by or the condition of the driver of the automobile,
including speeding and violating traffic laws, as well as being affected
by alcohol or drugs, inattention, decision errors, and age. Roadway
environment factors include the design of the roadway,

roadside hazards, and roadway conditions. Vehicle factors include any
failures that may exist in the automobile or design of the vehicle. Human
factors are seen as the most prevalent, according to data, experts, and
studies, in contributing to crashes, followed by roadway environment and
vehicle factors.

Agencies within the Department of Transportation have research projects
underway or planned that address the factors that contribute to crashes.
For example, the Federal Motor Carrier Safety Administration and the
National Highway Traffic Safety Administration are conducting a

study on the causes and contributing factors to large truck crashes. In
addition, the National Highway Traffic Safety Administration is conducting
a 100- Car Naturalistic Driving Study and the Drive Atlanta Study. The
100- Car Naturalistic Driving Study involves collecting data from vehicles
equipped with sensors and cameras to obtain better information on crashes
and near misses. The Drive Atlanta Study involves collecting data from
1,100 vehicles equipped with data recorders to develop information about
how excessive speed contributes to crashes. In addition, the
Transportation Research Board has proposed a broad, 6- year, $180 million
research program focused on making significant improvements in highway
safety. This study, among other things, would involve installing sensors
and other data collection devices on over 5,000 vehicles. Fatality
Statistics, 1975* 2001

HIGHWAY SAFETY

Research Continues on a Variety of Factors That Contribute to Motor
Vehicle Crashes

www. gao. gov/ cgi- bin/ getrpt? GAO- 03- 436 To view the full report,
including the scope and methodology, click on the link above. For more
information, contact Peter Guerrero, (202) 512- 2834, guerrerog@ gao. gov.
Highlights of GAO- 03- 436, a report to

congressional requesters

March 2003

Nearly 6. 3 million motor vehicle crashes occurred in the United States in
2001, or one crash every 5 seconds. On average, a person was injured in
these crashes every 10 seconds, and someone was killed

every 12 minutes. Since the 1970s, progress has been made in reducing the
number of fatalities and injuries on our nation*s roads. From 1975 through
2001, fatalities decreased from 44, 525 to 42, 116, while the rate of
fatalities per 100

million vehicle miles traveled decreased from 3. 35 to 1.51. However, the
decline in fatalities

has leveled off in recent years. In the 1970s, Indiana University
conducted one of the most significant studies to date on the

factors that contribute to motor vehicle crashes. This study examined
human, environmental, and vehicle factors that contribute to crashes. As
requested, this report provides more recent

information from data, experts, and studies about the factors that
contribute to motor vehicle crashes

and information about major ongoing and planned Department of
Transportation research into factors that contribute to crashes.

Page i GAO- 03- 436 Traffic Crash Causation Letter 1 Results in Brief 2
Background 3 Human, Roadway Environment, and Vehicle Factors Contribute to

Motor Vehicle Crashes 6 Federal Research Directed at Better Understanding
of Factors That Contribute to Crashes 32 Agency Comments and Our
Evaluation 36 Appendix I Objectives, Scope, and Methodology 38

Analyzing NHTSA Data 38 Identifying Studies 40 Interviewing Federal
Officials and Experts 41 Ongoing and Planned Transportation Research 41
Appendix II Tri- Level Study of the Causes of Traffic Accidents 42
Objectives, Scope, and Methodology for the Tri- Level Study 42 Results of
the Tri- Level Study 43 Appendix III Roadway Design Features 46

Figures

Figure 1: Fatality Statistics, 1975* 2001 4 Figure 2: Crash Causes Found
by the Tri- Level Study 5 Figure 3: Speeding Drivers in Fatal Crashes, by
Age and Gender, 1997* 2001 8 Figure 4: Drivers in Alcohol- Related Fatal
Crashes, by Age and Gender, 1997* 2001 12 Figure 5: Inattentive Drivers
Involved in Crashes by Age, 1997* 2001 15 Figure 6: Number and Rate of
Driver Involvement in Fatal Crashes

by Age, 1997* 2001 18 Figure 7: Fatality Rates by Type of Road System,
2001 22 Figure 8: Vehicle Crash Rates, 2001 29 Figure 9: Passenger Vehicle
Rollovers, 2001 30 Figure 10: Factors Contributing to Crashes Identified
by the TriLevel Study 44 Figure 11: Impact of Access Points on Traffic
Crashes 50 Contents

Page ii GAO- 03- 436 Traffic Crash Causation Abbreviations

AAA former American Automobile Association AASHTO American Association of
State Highway and Transportation

Officials BAC blood alcohol content CDS Crashworthiness Data System DOT
Department of Transportation FARS Fatality Analysis Reporting System FHWA
Federal Highway Administration FMCSA Federal Motor Carrier Safety
Administration F- SHRP Future Strategic Highway Research Program GES
General Estimates System

NHTSA National Highway Traffic Safety Administration SUV sport utility
vehicle VMT vehicle miles traveled

This is a work of the U. S. Government and is not subject to copyright
protection in the United States. It may be reproduced and distributed in
its entirety without further permission from GAO. It may contain
copyrighted graphics, images or other materials. Permission from the
copyright holder may be necessary should you wish to reproduce copyrighted
materials separately from GAO*s product.

Page 1 GAO- 03- 436 Traffic Crash Causation March 31, 2003 The Honorable
Carl Levin

United States Senate The Honorable George V. Voinovich United States
Senate

Nearly 6.3 million motor vehicle crashes occurred in the United States in
2001, or one crash every 5 seconds. On average, a person was injured in
these crashes every 10 seconds, and someone was killed every 12 minutes.
While there have been significant improvements in motor vehicle safety
over the past several decades, decreases in injuries and fatalities have

leveled off since the early 1990s. In the 1970s, Indiana University
conducted a major study that examined the human, environmental, and
vehicle factors that contribute to traffic crashes. 1 You asked us to (1)
provide more recent information on the factors that contribute to motor
vehicle crashes, and (2) identify major ongoing and planned Department of
Transportation research into factors that contribute to crashes.

To provide information on factors that contribute to motor vehicle
crashes, we obtained and analyzed crash data from three Department of
Transportation databases. In addition, we interviewed experts from
academia, insurance organizations, and advocacy groups. To identify recent
studies on factors that contribute to motor vehicle crashes, we conducted
a literature search, explored the Transportation Research Information
System, and reviewed periodicals. This effort resulted in numerous studies
being identified on various aspects of motor vehicle crashes. We then,
with input from a number of experts and agency

officials, judgmentally selected studies that would provide additional
information on the particular factors being discussed. For each of the
selected studies that are used in this report, we determined whether the
studies* findings were generally reliable. We evaluated the methodological
soundness of the studies using common social science and statistical
practices. To identify the major ongoing and planned research into factors

1 J. R. Treat et al., Tri- Level Study of the Causes of Traffic Accidents
(Washington, D. C.: Institute for Research in Public Safety, May 30,
1979), for the U. S. Department of Transportation. United States General
Accounting Office Washington, DC 20548

Page 2 GAO- 03- 436 Traffic Crash Causation that contribute to crashes, we
interviewed officials from the National Highway Traffic Safety
Administration, the Federal Highway

Administration, and the Transportation Research Board. Appendix I provides
more details on our scope and methodology.

Many factors can combine to produce circumstances that lead to a motor
vehicle crash* there is rarely a single cause of such an event. Three
categories of factors contribute to crashes: human factors, roadway
environment factors, and vehicle factors. Human factors involve the
actions taken by or the condition of the driver of the motor vehicle,
including speeding and violating traffic laws, as well as being affected
by alcohol or drugs, inattention, decision errors, and age. Roadway
environment factors that contribute to, or are associated with, crashes
include the roadway design (for example, medians, narrow lanes, the lack
of shoulders, curves, access points, or intersections); roadside hazards
(for example, poles, trees, or embankments adjacent to the road); and
roadway conditions (for example, rain, ice, snow, or fog). Vehicle factors
include any vehicle- related failures that may exist in the automobile or
design of the vehicle. In general, human factors are considered to be the
most prevalent factors contributing to crashes, followed by roadway
environment and vehicle factors.

Various agencies within the Department of Transportation have research
projects underway or planned that address the factors that contribute to
crashes. For example, the Federal Motor Carrier Safety Administration and
the National Highway Traffic Safety Administration are studying the causes
of, and factors contributing to, large truck crashes. In addition, the
National Highway Traffic Safety Administration*s 100- Car Naturalistic
Driving Study involves collecting data from vehicles equipped with sensors
and cameras to obtain better information on crashes and near misses.
Another project, the Drive Atlanta Study, involves collecting data from
1,100 vehicles equipped with data recorders to develop information about
how speeding contributes to crashes. A number of follow- on studies to
these efforts are also being considered. In addition, the Transportation
Research Board has proposed a broad, 6- year, $180 million research
program focused on making significant improvements in highway safety. This
program, among other things, could involve installing sensors and

other data collection devices on over 5,000 vehicles. The final phase of
the research program would use the results of the instrumented vehicle
study to identify countermeasure improvements. Results in Brief

Page 3 GAO- 03- 436 Traffic Crash Causation We provided copies of a draft
of this report to the Department of Transportation for its review and
comment. In discussing this report,

agency officials provided technical clarification and information, which
we incorporated in the report as appropriate. In addition, National
Highway Traffic Safety Administration officials provided information
comparing light truck and passenger car crash rates, which we also
incorporated in the report.

Since the 1970s, progress has been made in reducing the number of
fatalities and injuries on our nation*s roads, but the numbers are still
significant. From 1975 through 2001, annual fatalities decreased from
44,525 to 42,116, or about 5 percent. During the same period, the fatality
rate per 100 million vehicle miles traveled, a common method of
measurement, dropped from 3.35 to 1.51, or about 55 percent. This
reduction in fatalities was considerable, given the growth in the number
of drivers and vehicles on the road. For example, from 1975 through 2001,
licensed drivers increased from about 130 million to about 191 million,
and the number of registered vehicles increased from about 126 million to
about 221 million. Figure 1 shows the yearly number of fatalities and the
rate of fatalities per 100 million vehicle miles traveled. Injury and
propertydamage- only crashes also fell, going from about 6.8 million in
1988, the earliest year of available data, to about 6.3 million in 2001.
Background

Page 4 GAO- 03- 436 Traffic Crash Causation Figure 1: Fatality Statistics,
1975* 2001

The fatal, injury, and property- damage- only crashes have significant
economic cost. The National Highway Traffic Safety Administration (NHTSA)
recently calculated the economic costs for motor vehicle crashes in 2000
at more than $230 billion, or the equivalent of over $800 for every person
living in the United States. NHTSA*s estimate of economic costs includes
productivity losses, property damage, medical costs, rehabilitation costs,
travel delay, legal and court costs, emergency services, insurance
administration costs, and costs to employers.

One of the most significant studies to date on the factors that contribute
to motor vehicle crashes was the Tri- Level Study of the Causes of Traffic
Accidents, conducted in the 1970s by the Indiana University at Bloomington
Institute for Research in Public Safety. Referred to as the TriLevel

study, it investigated how frequently various factors contributed to
traffic crashes. According to NHTSA officials, the Tri- Level study has
been the only study in the past 30 years to collect large amounts of on-
scene

crash causation data. To provide researchers with insight into the factors
that contribute to traffic crashes, collision data were collected on three
levels, each providing an increasing level of detail, including 13,568
policereported crashes; 2,258 crashes investigated by on- scene
technicians; and 420 crashes investigated in depth by a multidisciplinary
team. The study assessed causal factors as either definite, probable, or
possible. The study found that crashes were caused by human (or driver-
based) factors,

Page 5 GAO- 03- 436 Traffic Crash Causation environmental (roadway or
weather- related) factors, or vehicle- related factors. As shown in figure
2, the study concluded that the human factors

were definite or probable causes in about 93 percent of crashes, while
environmental and vehicle factors contributed to about 33 and 13 percent,
respectively. See appendix II for a more detailed discussion of the
TriLevel

study. Figure 2: Crash Causes Found by the Tri- Level Study

NHTSA*s mission is to reduce deaths, injuries, and economic losses
resulting from motor vehicle crashes. As part of this responsibility,
NHTSA conducts or sponsors research into the causes of motor vehicle
crashes. NHTSA also conducts research on driver behavior and traffic
safety to develop more efficient and effective means to improve safety.
Three principal databases provide information about traffic crashes: the
Fatality Analysis Reporting System (FARS), the Crashworthiness Data System
(CDS), and the General Estimates System (GES). The FARS database

contains information provided by the states on all vehicle crashes that
result in the death of an occupant or nonmotorist within 30 days of the

Page 6 GAO- 03- 436 Traffic Crash Causation incident. The CDS database
contains information from a detailed sample of about 4,000 minor, serious,
or fatal tow- away crashes, annually. To

obtain this information, teams of trained crash investigators visit the
crash site and collect data on such elements as the damage to the vehicle
and interior locations struck by the occupants. The GES database contains

information from a nationally representative sample of police accident
reports. This is NHTSA*s largest crash database, with information
collected on over 50, 000 crashes each year.

The Federal Highway Administration*s (FHWA) safety mission is to reduce
highway fatalities and injuries through development and implementation of
a program of nationally coordinated research and technology

innovations. Research is conducted in areas that address FHWA*s highway
safety goals related to roadway departure, intersections, and pedestrians.
FHWA is also conducting research in a number of areas that will partially
focus on crash causation, including rollovers, speed management,
intersection safety, and pedestrian and bicyclist safety. FHWA annually
produces a highway statistics report, which consists of data on motor
fuel, motor vehicles, driver licensing, highway- user taxing, state and
local government highway finance, highway mileage, and federal aid for
highways. FHWA also maintains a database, called the Highway Safety
Information System. The system uses data on crash, roadway, and traffic
variables collected by eight states to analyze a number of highway safety
problems. These analyses range from identifying basic problems, to
identifying the size and extent of a safety issue, to modeling efforts
that attempt to predict future crashes from roadway characteristics and
traffic factors.

Motor vehicle crashes are complex events that rarely have a single cause.
For example, it would be challenging to identify a single cause of a crash
that occurred on a narrow, curvy, icy road when an inexperienced driver,
who had been drinking, adjusted the radio or talked on a cell phone. It
would likely be the combined effect of a number of these factors that
contributed to the crash. In examining the causes of motor vehicle
crashes, a number of experts and

studies identified three categories of factors that contribute to crashes*
human, roadway environment, and vehicle factors. Human factors involve the
actions taken by or the condition of the driver of the automobile,
including speeding and other traffic violations, as well as the effects of
alcohol or drugs, inattention, decision errors, and age. Roadway
environment factors that contribute to or are associated with crashes
Human, Roadway

Environment, and Vehicle Factors Contribute to Motor Vehicle Crashes

Page 7 GAO- 03- 436 Traffic Crash Causation include the design of the
roadway (for example, medians, lane width, shoulders, curves, access
points, or intersections); roadside hazards (for

example, poles, trees, or embankments adjacent to the road); and the
roadway conditions (for example, rain, ice, snow, or fog). Vehicle factors
include vehicle- related failures and vehicle design issues that
contribute to a crash. In general, human factors are considered to be the
most prevalent

factor contributing to crashes, followed by roadway environment and
vehicle factors. Although this report discusses these categories
separately, they should be viewed in terms of how they can concurrently
contribute to an unstable situation that results in a crash.

Human factors involve actions taken by or the condition of the driver of
the vehicle. They are considered the most prevalent factors by data,
experts, and studies in traffic crashes. Human factors that can contribute

to crashes include speeding and other traffic violations, as well as the
effects of alcohol or other drugs, inattention, driver decision errors,
and age.

Driving either faster than the posted speed limit or faster than
conditions would safely dictate can contribute to traffic crashes.
Speeding reduces a driver*s ability to steer safely around curves or
objects in the roadway, extends the distance necessary to stop a vehicle,
and increases the distance a vehicle travels while the driver reacts to a
dangerous situation.

According to our analysis of NHTSA*s databases, from 1997 through 2001,
speeding was identified as a contributing factor in about 15 percent of
all crashes and about 30 percent of all fatal crashes. In addition, almost
64,000 lives were lost in speeding- related crashes. 2 As shown in figure
3, we found that for every age category of drivers involved in fatal
crashes, males were more likely than females to be involved in a fatal
speed- related crash. In addition, younger drivers, regardless of sex, are
the most likely to be involved in a speed- related fatality. From 1997
through 2001, 36 percent of male drivers and 24 percent of female drivers
16 to 20 years old who were involved in fatal crashes were speeding at the
time of the crash. The

2 NHTSA defines a crash as speed- related if the driver was charged with a
speed- related offense or if an officer indicated that the driver was
racing, driving too fast for conditions, or exceeding the posted speed
limit. Human Factors Contribute

to Motor Vehicle Crashes Speeding

Page 8 GAO- 03- 436 Traffic Crash Causation percentage of speeding-
related fatal crashes decreases with increasing driver age. 3 Figure 3:
Speeding Drivers in Fatal Crashes, by Age and Gender, 1997* 2001

A 1998 study by NHTSA and FHWA indicates that fatal crashes increased in
states that raised speed limits. 4 When Congress enacted the National
Highway System Designation Act of 1995 (P. L. 104- 59), which repealed the
national maximum speed limit, the Secretary of Transportation was required
to study the impact of states* actions to raise speed limits above

3 Some analyses in this report discuss fatality data associated with
specific factors. It should be noted that other elements, in addition to
the factor discussed, might have also contributed to the fatalities. These
would include circumstances such as the use of safety belts or other
occupant protection measures. 4 NHTSA/ FHWA, Report to Congress: The
Effect of Increased Speed Limits in the PostNMSL

Era (Washington, D. C.: February 1998).

Page 9 GAO- 03- 436 Traffic Crash Causation 55 and 65 miles per hour. The
study found that states with increased speed limits in 1996 experienced
approximately 350 more Interstate fatalities

than would have been expected based on historical trends* about 9 percent
above expectations. Concurrently, the Interstate fatalities experienced in
states that did not increase speed limits in 1996 were consistent with
pre- 1996 trends. The Insurance Institute for Highway Safety also assessed
the effects of speed limit increases. 5 Its researchers found an increase
in fatalities for a 9- month period in 1996 on Interstate highways and
freeways, as compared with the previous 6 years* about 16 percent in 12 of
the states that had raised maximum speed limits to at least 70 miles per
hour by March 1996. In contrast, occupant fatalities increased only 4
percent on Interstate highways and freeways in the comparison group of
states that did not raise speed limits. However, both of these studies are
limited because they cover short time periods.

According to a Transportation Research Board official, studies have
confirmed a direct relationship between speed and crash severity. 6 Once a
crash has occurred* that is, a vehicle has hit another vehicle or a

stationary object* the vehicle undergoes a rapid change in speed. While
the vehicle decelerates rapidly, its occupants continue to move at the
vehicle*s speed prior to impact until they are stopped by striking the
interior of the vehicle, by impact with objects external to the vehicle if
ejected, or by being restrained by a safety belt or an airbag that
deploys.

According to the FHWA Director of the Office of Safety Programs, while
absolute speed clearly relates to injury and fatality outcomes, speeding
is the real issue. The Director pointed out that despite their lower
volumes, almost half of all speeding- related fatalities occur on local or
collector roads* low- speed roads found in residential and business areas.
In addition, the Director said that speed variance is also a factor. When
vehicles driving down a particular roadway are traveling at very different
speeds, the probability of a crash increases. The relative
crashinvolvement rate increases for vehicles that are traveling above or
below the average speed of traffic.

5 Charles M. Farmer, Richard A. Retting, and Adrian K. Lund, Effect of
1996 Speed Limit Changes on Motor Vehicle Occupant Fatalities (Washington,
D. C.: Insurance Institute for Highway Safety, October 1997). This study
focused on 12 states that raised maximum speed limits to at least 70 miles
per hour between December 8, 1995, and April 1, 1996.

6 The Transportation Research Board is a unit of the National Research
Council, a private, nonprofit institution that is the principal operating
agency of the National Academy of Sciences and the National Academy of
Engineering.

Page 10 GAO- 03- 436 Traffic Crash Causation Drivers who fail to follow
prescribed traffic control laws also contribute to crashes. This includes
running red lights or failing to stop at stop signs.

Our analysis of NHTSA*s data found that from 1997 through 2001, about 36
percent of motor vehicle crashes occurred at traffic control devices. Of
those crashes, 59 percent occurred at traffic lights while an additional
28 percent occurred at stop signs.

A study performed by the Insurance Institute for Highway Safety and the
Preusser Research Group identified characteristics of red light* running
crashes and the drivers involved. 7 It found that drivers* noncompliance
with traffic control devices, such as traffic signals and stop signs, is a
major cause of motor vehicle crashes. The study examined the prevalence of
red light* running crashes on a national basis to identify the

characteristics of such crashes and the drivers involved. 8 The study
estimated that almost 260,000 red light- running crashes occurred in 1996,
of which 809 resulted in fatalities. It also found that, as a group, red
light

runners involved in crashes were more likely than other drivers to be
younger than age 30, to be male, to have prior moving violations and
convictions for driving while intoxicated, to have invalid driver*s
licenses, and to be reported by police as having consumed alcohol prior to
the crash.

According to an official from Northwestern University, red light* running
might also partly reflect driver frustration with poor traffic operations.
For example, a driver might feel the need to speed through a red light
because of previous experience of being held at that light too long, or of
being subjected to a series of unsynchronized stop lights. A 1999 study
funded by DaimlerChrysler Corporation surveyed over 5,000 people regarding
their

behavior at red lights. 9 The study found that those respondents who
reported speeding up to beat a red light would most often do so because
they were in a rush and wanted to save time.

7 Richard A. Retting and Allan F. Williams for the Insurance Institute for
Highway Safety, and Robert G. Ulmer for the Preusser Research Group,
*Prevalence and Characteristics of Red Light Running Crashes in the United
States,* Accident Analysis and Prevention, vol. 31 (1999), pp. 687* 94.

8 The study reviewed intersection crashes in both the Fatality Analysis
Reporting System and the General Estimates System during the 5- year
period from 1992 through 1996. 9 Bryan E. Porter, Thomas D. Berry, Jeff
Harlow, and Tancy Vandecar, A Nationwide Survey of Red Light Running:
Measuring Driver Behaviors for the *Stop Red Light Running* Program, June*
August 1999. Traffic Control Violations

Page 11 GAO- 03- 436 Traffic Crash Causation Alcohol and other drugs are
contributing factors in many motor vehicle crashes. It is illegal in every
state and the District of Columbia to drive a

motor vehicle while under the influence of, impaired by, or with a
specific blood content of alcohol or drugs. In addition, all states but
Massachusetts have blood alcohol laws that make it illegal to drive with a
specified level of alcohol in their blood. 10 As of January 2003, 17
states had set the standard at .10 percent blood alcohol content (BAC)*
the level at which a person*s blood contains 1/ 10 of 1 percent alcohol.
11 The remaining states have more stringent laws, setting the limit at .08
percent BAC. According to NHTSA, on average, a 170- pound man reaches .08
percent BAC after consuming five 12- ounce beers (4.5 percent alcohol by
volume) over a 2- hour period. A 120- pound woman reaches the same level
after consuming three beers over the same period.

In analyzing NHTSA*s databases, we found that from 1997 through 2001,
there were about 76,000 alcohol- related fatal crashes (41 percent of all
fatal crashes), 980,000 alcohol- related injury crashes (10 percent of all
injury crashes), and 2.3 million alcohol- related crashes (7 percent of
all crashes). 12 During this 5- year period, nearly 85,000 people died in
alcoholrelated crashes. Eighty- six percent of these fatalities occurred
in crashes where the highest recorded BAC was .08 percent or above, while
14 percent occurred in crashes where the highest recorded BAC was between
.01 percent and .07 percent. In addition, we found that male drivers were
more likely to be involved in alcohol- related fatal crashes than female
drivers. Figure 4 shows that, for each age category, there were a greater
number of male than female drivers in fatal crashes that involved alcohol.

10 Blood alcohol content of .08 percent in Massachusetts is evidence of
alcohol impairment, but it is not illegal per se. 11 Of these states,
Louisiana, New York, and Tennessee have .08 percent BAC laws that will be
effective during the latter half of 2003. 12 NHTSA indicates that a
fatality is alcohol- related if it occurred in a crash where any one of
the actively involved persons in the crash had a BAC of .01 percent or
greater. Alcohol and Other Drugs

Page 12 GAO- 03- 436 Traffic Crash Causation Figure 4: Drivers in Alcohol-
Related Fatal Crashes, by Age and Gender, 1997* 2001

While research has shown that everyone*s driving is impaired at blood
alcohol levels of .10 percent and higher, recent research has shown that
lower levels of alcohol also affect performance. In a study by the
Southern California Research Institute, 168 test subjects were tested at
zero BAC; then at the highest BAC for their drinking classification of
either light, moderate, or heavy drinker; and then at .02 percent BAC
intervals, as their alcohol levels decreased. 13 For this study, the
researchers defined impairment by comparing the subjects* performance on a
given test while under the influence of alcohol versus their performance
on the same test after being given a placebo. According to the resulting
report, alcohol impaired the driving- related skills for these volunteers
at .02 percent BAC,

13 H. Moskowitz, M. Burns, D. Fiorentino, A. Smiley, and P. Zador, Driver
Characteristics and Impairment at Various BACs (Los Angeles, CA: Southern
California Research Institute, August 2000).

Page 13 GAO- 03- 436 Traffic Crash Causation the lowest tested alcohol
level. The magnitude of impairment increased consistently at BACs through
.10 percent, the highest level tested. 14 According to a Southern
California Research Institute official, this study is

significant because it provided important, previously unknown findings
that certain driving- related skills are impaired at any departure from
zero BAC.

A recent study by Westat examined the relative risk of fatal crash
involvement as a function of the BAC of fatally injured or surviving
drivers. 15 By combining crash data from FARS with exposure data from the
1996 National Roadside Survey, the researchers determined that, in
general, the relative risk of involvement in a fatal passenger vehicle
crash increased steadily with increased driver*s BAC. 16 For example, the
study found that a .02 percent BAC increase among 16- through- 20- year-
old male drivers was estimated to more than double the relative risk of a
fatal single- vehicle crash injury. The study also found that among
drivers aged 21 through 34, those with a BAC of .03 percent have twice the
risk of fatalities as compared with drivers with zero BAC. Furthermore,
among drivers aged 21 through 34, those with a BAC of .10 percent have
over 10 times the risk of a fatality compared with drivers with zero BAC.

All states restrict driving while under the influence of, being impaired
by, or being incapable of safely driving because of illegal drugs or
prohibited substances in the driver*s body. As of January 2003, eight
states have statutes that make it unlawful for a driver to have any amount
of an illegal drug or prohibited substance in his or her body while
operating a motor vehicle, regardless of how the drug affects the driver*s
driving ability. 17 14 The study subjects were examined only as their BAC
was declining and, according to the

study, the results would underestimate the magnitude of impairment
expected during alcohol consumption when BAC was rising. 15 P. L. Zador,
S. A. Krawchuk, and R. B. Voas, Relative Risk of Fatal Crash Involvement
by BAC, Age and Gender (Rockville, MD: Westat, April 2000). 16 The 1996
National Roadside Survey was a national survey of weekend, nighttime
drivers in the 48 contiguous states. The survey consisted of interviewing
and breath- testing over 6,000 noncommercial four- wheel vehicle operators
between September and November 1996.

17 These eight states include Arizona, Georgia, Indiana, Illinois, Iowa,
Minnesota, Rhode Island, and Utah. The Georgia Supreme Court has
determined that the Georgia statute, Ga. Code Ann. 40- 6- 391( a)( 6)
(2002), is an unconstitutional denial of equal protection. See Love v.
State, 271 Ga. 398 (1999). Accordingly, the enforceability of the Georgia
statute is questionable.

Page 14 GAO- 03- 436 Traffic Crash Causation Additional states have
varying legislation that also allows zero tolerance to driving under the
influence of drugs. 18 Studies have shown that drugs can affect driving-
related skills. For

example, a study by Maastricht University, the Netherlands, indicated that
the combined use of marijuana and alcohol impairs driving performance. 19
For a small number of subjects who were somewhat frequent users of

marijuana, the study found that either marijuana doses alone or alcohol
alone impaired the subjects* test- driving performances. However, subjects
who used marijuana in combination with alcohol demonstrated

impairment in several aspects of driving performance. Another study by
Maastricht University also found the combined use of marijuana and alcohol
to produce similar effects on a small, limited group of subjects.

The study showed that under the influence of low doses of either marijuana
or alcohol, the drivers were less able to detect peripheral traffic and
instead focused on the central driving task. 20 Driver inattention occurs
when there is a delay in recognition of

information needed to safely accomplish the driving task. Two categories
of driver inattention are distraction and drowsiness. Drivers may become
distracted when they direct their attention elsewhere because of some
occurrence inside or outside of the vehicle. NHTSA defines four categories

of distraction: visual distraction (for example, looking away from the
roadway), auditory distraction (for example, responding to a noise, such
as a ringing cell phone), biomechanical distraction (for example, manually
adjusting the radio volume), and cognitive distraction (for example, being
lost in thought). Many distracting activities that drivers engage in can
involve more than one of these components. Driver drowsiness is also a
type of driver inattention, in that a tired or fatigued driver may exhibit
behaviors typically associated with inattentive drivers.

18 Examples of zero tolerance to driving under the influence of drugs
include laws that prohibit drug addicts or habitual users of drugs from
driving vehicles (found in California, Colorado, Idaho, Kansas, and West
Virginia) or statutes that make it illegal for minors to drive with any
amount of a prohibited drug in their bodies (found in North Carolina and
South Dakota). 19 Hindrik W. J. Robbe and James F. O*Hanlon, Marijuana,
Alcohol, and Driving

Performance (The Netherlands: Institute for Human Psychopharmacology, July
1999). 20 C. T. J. Lamers and J. G. Ramaekers, Visual Search and Urban
City Driving under the Influence of Marijuana and Alcohol (The
Netherlands: Maastricht University, June 2001).

Specifically, both studies examined the effects of delta- 9-
tetrahydrocannabinol (THC), the primary active ingredient of cannabis
(marijuana). Driver Inattention

Page 15 GAO- 03- 436 Traffic Crash Causation Our analysis of 1997 through
2001 data from NHTSA found that, overall, about 2.5 million drivers of
passenger vehicles that were towed away from crashes were identified as
inattentive. Of these, about 1.3 million were

distracted, about 871,000 *looked but did not see* (an aspect of being
inattentive), and about 348,000 were sleepy or asleep. In addition, about
7.6 million drivers were identified as *attentive* at the time of the
crash. 21 We also conducted a more detailed analysis of inattentive
drivers. As

figure 5 shows, overall, more drivers between ages 16 and 44 were involved
in inattentive- type crashes than drivers aged 45 and above. More drivers
aged 16 to 20 were inattentive than any other age group.

Figure 5: Inattentive Drivers Involved in Crashes by Age, 1997* 2001

Note: This includes only those drivers involved in crashes where at least
one passenger vehicle had to be towed away.

21 About 6 million were identified as *unknown* or *no driver present.*

Page 16 GAO- 03- 436 Traffic Crash Causation We also analyzed NHTSA*s
databases to determine specific sources of distraction. We found that some
outside person, object, or event was

identified as contributing to 27 percent of the distractions. Other common
sources of distractions included another occupant in the vehicle, followed
by adjusting a radio, cassette, or CD.

A recent study by the AAA Foundation for Traffic Safety analyzed 1995
through 1999 NHTSA crash data on driver attention status and sources of
distraction and found that 8 percent of drivers were identified as
distracted, 5 percent as *looked but did not see,* and 2 percent as sleepy
or asleep, while 49 percent of the drivers were identified as attentive at
the time of the crash. 22 The remaining 36 percent were either unknown or
had no driver present. Without the unknowns, the percentage of drivers
identified as distracted increases to 13 percent. The study also
identified specific sources of distraction. Some external person, object,
or event

caused almost 30 percent of such distractions. Drowsiness and fatigue are
also aspects of inattention and can contribute to crashes. Drowsiness is a
basic physiological state, brought about by the restriction or
interruption of sleep. It also results from natural changes in the body*s
level of alertness during each 24- hour sleep- wake cycle. According to
the National Sleep Foundation, our internal body clocks program us to be
sleepy twice a day: first during the early morning hours between midnight
and dawn, and again between 1: 00 p. m. and 4: 00 p. m. For the driver,
the main effect of drowsiness or fatigue is a progressive withdrawal of
attention from the road and traffic demands, leading to impaired
performance behind the wheel. Drivers can become so fatigued that they are
slow to perceive risky situations and are unable to respond quickly enough
to avoid a crash. Fatigue can also arise because of medication or illness.

According to an official from the National Sleep Foundation, studies have
shown that sleep- deprived individuals are less likely to be able to
concentrate on the task at hand. In addition, as people get tired they
engage in behaviors that lead to other distractions, such as smoking,
drinking or eating, turning up the radio, or employing other *tricks* to
try to stay awake. The official also told us that the foundation*s
national polls 22 Jane C. Stutts, Donald W. Reinfurt, Loren Staplin, and
Eric A. Rodgman, The Role of Driver Distraction in Traffic Crashes
(Washington, D. C.: University of North Carolina for the AAA Foundation
for Traffic Safety, May 2001). This study examined drivers involved in

crashes where at least one vehicle had to be towed away.

Page 17 GAO- 03- 436 Traffic Crash Causation and international studies
support the perspective that driver fatigue is a much larger problem than
what the federal statistics show. Recently, the

National Sleep Foundation conducted a telephone survey and found that
about 51 percent of the respondents reported that they had driven a car or
another vehicle while feeling drowsy, and about 17 percent had dozed off

while driving within the past year. 23 The study found that male
respondents were more likely than female respondents to say they had
driven while feeling drowsy. In addition, respondents with children and
respondents aged 18 to 29 were at the highest risk for driving while
feeling drowsy. The study also found that older respondents, 65 and over,
are less likely to drive drowsy or to fall asleep at the wheel.

Driver decision errors involve misjudgments made while driving. These
include improperly judging stopping distances, improperly judging
distances of cars traveling behind the vehicle, and other misjudgments of
distance between cars that result in a crash. Decision errors also include
crashes that result from traveling the wrong way on a one- way street. A
driver decision error differs from an error that may have resulted from
inattention or distraction because the driver sees a hazard, such as an

oncoming car making a left- hand turn, but makes the wrong decision
concerning the proper action to take. A recent study by Veridian
Engineering examined unsafe driving acts in

severe crashes in four sites across the United States to determine the
specific driver behaviors and unsafe driving acts that lead to crashes,
along with the situational, driver, and vehicle characteristics associated
with these behaviors. 24 According to the study, in 717 of the 723 crashes
analyzed (99 percent), a driver behavioral error caused or contributed to
the crash. Of the 1,284 drivers involved in these crashes, 732 drivers (57
percent) contributed in some way to the cause of their crashes.

There is a strong relationship between a driver*s age and the likelihood
of being involved in a crash. While age, in itself, would not be the cause
of 23 The National Sleep Foundation commissioned WB& A Market Research to
conduct the 2002 *Sleep in America* telephone poll of 1, 010 adults at
least 18 years old between October 1 and December 10, 2001. The margin of
error is plus or minus 3.1 percent. 24 D. L. Hendricks, M. Freedman, P. L.
Zador, and J. C. Fell, The Relative Frequency of Unsafe Driving Acts in
Serious Traffic Crashes (Washington, D. C.: Veridian Engineering, Westat,

Inc., and Star Mountain, Inc., January 2001). A sample of 723 crashes
involving 1,284 drivers was investigated at four different sites in the
country between April 1, 1996, and April 30, 1997. Driver Decision Errors

Age

Page 18 GAO- 03- 436 Traffic Crash Causation the crash, some of the
characteristics displayed at various ages can lead to a higher probability
of being involved in traffic crashes. Our analysis of NHTSA*s databases
found that younger and older drivers

become involved in a greater number of crashes, especially fatal crashes,
than do other age groups. Figure 6 shows that drivers aged 16 through 20
and those aged 75 or more have a greater chance of being involved in fatal
crashes per vehicle mile traveled (VMT) than do other age groups. 25
Figure 6: Number and Rate of Driver Involvement in Fatal Crashes by Age,
1997*

2001

25 For this VMT analysis, we used data from 2001 National Household Travel
Survey. The National Household Travel Survey consists of household- based
travel surveys conducted every 5 years by DOT. Survey data are collected
from a sample of U. S. households and

expanded to provide national estimates of trips and miles by travel mode,
purpose, and a host of other characteristics. The survey collects
information on daily, local trips and on long- distance travel in the
United States.

Page 19 GAO- 03- 436 Traffic Crash Causation According to the Insurance
Institute for Highway Safety, teenagers* crash rates are
disproportionately high mainly because of the drivers* youth

combined with driving inexperience. A recent study by the Insurance
Institute for Highway Safety showed that the age factor plays out in a
more risky driving style among adolescents. 26 The study, which reviewed
and summarized other research on the risks associated with younger
drivers, found that increased crash risk comes immediately on licensure
and drops very rapidly in the first few months. Compared with older
drivers, this study concluded that young people are more likely to drive
at excessive speeds, follow too closely, violate traffic signs and
signals, overtake other vehicles in a risky manner, allow too little time
to merge, and fail to yield

to pedestrians. Risky driving leads young people into hazardous
situations, and inexperience makes it more difficult to cope with such
situations. The researchers also found that driving at night is associated
with an increased risk of serious crashes for young drivers. The driving
task is more difficult for young drivers when it*s dark, and the risky
driving that involves younger drivers, generally associated with
recreational activities, is more likely to occur at night. Fatigue and
alcohol are also more likely to contribute to younger drivers* crashes
during nighttime hours. The study also found that there is a heightened
crash risk when teenage drivers have passengers in their vehicles. The
study found that this increased risk is present only for teenage drivers,
and it increases incrementally with each additional passenger.

Older drivers are also at increased risk, because the elderly have higher
rates of fatal crashes per vehicle mile traveled than all but the youngest
drivers. According to a recent study by the Insurance Institute for
Highway Safety, this is largely attributable to their increased fragility.
27 In the study, fragility started at age 60 to 64 and continued to rise
with advancing age. In addition, a recent study by Dr. Leonard Evans found
that given similar crash severity, older drivers are more likely to
sustain fatal injuries than younger drivers. 28 The author suggested that
if populations of 70- year- old

26 A. F. Williams and S. A. Ferguson for the Insurance Institute for
Highway Safety, *Rationale for Graduated Licensing and the Risks It Should
Address,* Injury Prevention, vol. 8 (2002), pp. 9* 16.

27 Guohua Li, Elisa R. Braver, and Li- Hui Chen, Exploring the High Driver
Death Rates per Vehicle- Mile of Travel in Older Drivers: Fragility versus
Excessive Crash Involvement (Arlington, VA: Insurance Institute for
Highway Safety, August 2001).

28 Leonard Evans, *Age and Fatality Risk from Similar Severity Impacts,*
Journal of Traffic Medicine, vol. 29, 2001, pp. 10* 19.

Page 20 GAO- 03- 436 Traffic Crash Causation males and 20- year- old males
were subjected to the identical mixes of blunt trauma, the population of
older males would sustain over two times more

fatalities. A similar comparison of female populations would yield almost
two times more fatalities for older females. In addition, a literature
review conducted by the University of Michigan Transportation Research
Institute found that older drivers are more likely to suffer from medical
disabilities that could impair their driving, and they may use medications
that could affect their driving performance. 29 The study also found that
with increasing age, most drivers experience some loss of visual
perception and decreased cognitive and psychomotor functions. For example,
a 1988 AAA Traffic Safety Foundation study tested a small group of
volunteers and found that older adults with less joint flexibility
exhibited poorer driving ability than those with wider ranges of motion.
30 In addition, according to the FHWA Director of the Office of Safety
Programs, frailty is not the sole factor in older driver fatality rates,
noting that drivers 85 and older have more than twice the overall crash
rate of middle- aged drivers aged 40 through 44.

The roadway environment is generally cited as the second most prevalent
factor contributing to crashes by data, experts, and studies. It can be
defined as those factors external to the driver and the vehicle that
increase the risk of a crash. Roadway environment factors that contribute
to, or are associated with, crashes include the design of the roadway (for
example, medians, narrow lanes, the lack of shoulders, curves, access
points, or intersections); roadside hazards (for example, poles, trees, or
embankments adjacent to the road); and the roadway conditions (for
example, rain, ice, snow, or fog).

The principal guidance on roadway design is the American Association of
State Highway and Transportation Officials* (AASHTO) Policy on Geometric
Design of Highways and Streets. This guidance provides recommendations on
constructing the nation*s roadways, including such features as the
sharpness of curves, the slope of roadways, the width of

29 David W. Eby, Deborah A. Trombley, Lisa J. Molnar, and Jean T. Shope,
The Assessment of Older Drivers* Capabilities: A Review of the Literature
(Ann Arbor, MI: University of Michigan Transportation Research Institute,
August 1998). 30 Kenard McPherson, Jeffrey Michael, Andrew Ostrow, and
Peter Shaffron, Physical Fitness and the Aging Driver, Phase I
(Washington, D. C.: AAA Foundation for Traffic Safety, 1988). Roadway
Environment

Contributes to Motor Vehicle Crashes

Roadway Design

Page 21 GAO- 03- 436 Traffic Crash Causation lanes, and the design of
medians and barriers. In general, different functional road systems are
constructed for specific purposes. For

example, Interstate highways are intended for high mobility and therefore
have limited access points, while local roads are designed for increased
access, which can limit mobility. Design principles generally suggest that
as average daily traffic increases, additional design elements should be
adopted that increase safety, including wider lanes, paved shoulders, and
clear zones (areas free of roadside hazards next to the roadway).

Based on FHWA*s data, we found that fatal crashes were more frequent on
rural roads than on urban roads. 31 In 2001, rural roads handled only
about 40 percent of all vehicle miles traveled, yet more than 60 percent
of all fatalities occurred on these roads. Figure 7 shows that fatality
rates are higher on rural roads in comparison with urban roads, regardless
of the road type.

31 The term *urban* is used to denote the federal- aid legislation
definition of an area. Such areas include, at a minimum, a census place
with an urban population of 5, 000 to 49, 999, or a designated urbanized
area with a population of 50,000 or more. Rural areas are those areas
outside urban areas.

Page 22 GAO- 03- 436 Traffic Crash Causation Figure 7: Fatality Rates by
Type of Road System, 2001

Note: The urban Interstate fatality rate also includes fatalities from
other urban freeways and expressways.

A recent FHWA study developed relationships between roadway features and
crash rates on two- lane rural highways. 32 For this study, FHWA developed
predictive models to estimate the safety impacts of roadway design
features. Studies have found that the following roadway design features
can affect crash rates. Appendix III contains additional information on
roadway design features.

 Medians * Medians (that is, the physical separations between opposite
lanes of traffic) provide a recovery area for out- of- control vehicles
and reduce head- on crashes by separating traffic driving in opposite
directions.

32 D. W. Harwood, F. M. Council, E. Hauer, W. E. Hughes, and A. Vogt,
Prediction of the Expected Safety Performance of Rural Two- Lane Highways,
FHWA- RD- 99- 207 (December 2000).

Page 23 GAO- 03- 436 Traffic Crash Causation  Lane width * Wider lanes
may reduce crashes by allowing for greater separation between vehicles
traveling in adjacent lanes as well as

providing additional space to recover from near- crash situations. 
Shoulders * Wide roadway shoulders that are paved provide an

opportunity for drivers to recover from errors that cause a vehicle to
stray out of a lane.

 Curves * Curves have been shown to contribute to crashes, whether
horizontal curves (left or right) or vertical curves (up and down). Crash
rates on curves are associated with their design features (including
degree, length, and angle) and cross- sectional curve elements (lane
width, shoulder size and type, and median characteristics).

 Access points * As the number of access points, or locations where
vehicles can gain entry to the roadway, increases, the more likely it is
that a traffic crash will occur.

 Intersections * Intersections, or at- grade locations where vehicles may
transfer between roads, are among the most complex roadway designs a
driver encounters. This is the result of increased points of conflict
between vehicles, and between vehicles and pedestrians.

FHWA*s Chief Highway Safety Engineer told us that it is important that a
roadway be designed to allow a driver the time and space to make and
recover from various errors without crashing. For example, two- lane rural
roads are often characterized by sharp horizontal and vertical curves,
narrow lanes, no shoulders or narrow ones, and roadside hazards such as
utility poles or trees adjacent to the road. These design elements can be
associated with higher fatality rates. This contrasts with the multilane
highways, which generally have gradual horizontal and vertical curves,
wider lanes and shoulders, and wide, clear zones adjacent to the road.

FHWA*s Director of Office of Safety Research and Development pointed out
that there are some data limitations associated with crashes and roadway
design. For example, the Director noted that NHTSA*s crash databases
contain very limited data on roadway design features at the crash location
or immediately preceding the crash location. Accordingly, detailed
analysis comparable to what is possible for the driver is not

possible for the roadway. The Director also stated that efforts are
underway to provide the means to more precisely locate the point of a
crash and to relate that location to detailed roadway and roadside
information databases.

Page 24 GAO- 03- 436 Traffic Crash Causation Roadside hazards are physical
features that a vehicle can crash into if it leaves the roadway. Each
year, about 14,000 persons are killed and almost

1 million persons are injured when vehicles run off the road and crash.
Many of these deaths and injuries result from crashes into poles and
trees, which are often located close to the edge of the roadway.

Our analysis of NHTSA*s data found that 16 percent of all crashes from
1997 through 2001 involved striking a roadway object as the first
property- damaging or injury- producing event in the crash. In addition,
we found that that in these crashes, posts or poles were the most common
fixed objects for a vehicle to hit after leaving the roadway (about 20
percent), followed by ditches (14 percent), trees (14 percent), and
guardrails (11 percent).

The Washington State Transportation Center conducted a study examining
roadside crashes on a single section of roadway. 33 Models were created to
predict the frequency and severity of run- off- the- road crashes related
to a variety of roadway environmental factors. The study found, for
example, that both a decreased distance from the outside shoulder edge to
roadside objects and an increased number of trees near the roadway
increased the

likelihood of a crash. Overall, the study supported the enlargement of
roadside recovery space to decrease the occurrence and severity of runoff-
the- road crashes.

The chairperson of AASHTO*s Task Force on Roadside Safety addressed the
importance of roadside hazards. The AASHTO official said that, in order of
preference, the four methods for addressing roadside hazards are to (1)
remove it, (2) relocate it, (3) redesign it, and (4) shield the roadside
hazards (for example, a guardrail or impact barrier).

Roadway conditions can contribute to crashes through both road surface
conditions and reduced visibility. Surface conditions that can impair a
driver*s ability to control the vehicle include standing water, snow, ice,
and oil, in addition to such road surface features as holes, ruts, paved
edge drop- offs, and worn surfaces. Crashes can also result when
visibility is

somehow reduced, preventing a driver from receiving the proper visual 33
Jinsun Lee and Fred Mannering, Analysis of Roadside Crash Frequency and
Severity and Roadside Safety Management, Washington State Transportation
Center (December 1999). Roadside Hazards

Roadway Conditions

Page 25 GAO- 03- 436 Traffic Crash Causation driving cues. Reduced
visibility can occur because of weather- related events or the presence or
absence of natural or artificial lighting.

Surface conditions. Common road surface conditions that can create
slippery roads are rain, snow, and ice. Slippery road conditions lead to a
loss of friction between a vehicle*s tires and the roadway. This loss of
friction may lead to the reduced controllability of the vehicle,
ultimately resulting in a crash.

Our analysis of NHTSA*s data from 1997 through 2001 found that about 23
percent of crashes occurred when road surface conditions were either wet,
snowy, slushy, or icy. In addition, a recent study by Iowa State*s Center
for Transportation Research and Education examined the weather*s impacts
on safety. 34 The researchers examined the impact of more severe winter
storms on volume, safety, and speed characteristics on seven segments of
Interstate highways in Iowa. Their analysis of 54 storm events concluded
that crash rates increased by over 1,000 percent during

winter storm events with high snowfall rates.

Reduced visibility. Reduced visibility can occur during nighttime hours
(including dawn and dusk) and during weather- related events, such as fog,
rain, or snow. Reduced visibility can decrease a driver*s ability to
receive

the proper visual cues to successfully navigate the road. Our analysis of
NHTSA*s data found that overall, while 15 percent of all crashes took
place under limited light conditions, about 34 percent of all traffic
fatalities occurred at that time. Although other factors are involved
during nighttime crashes, such as alcohol or fatigue, the reduction of
visual cues for the driver also appears to play a role. The primary
purpose of roadway lighting is to provide increased visibility of the
roadway and its immediate environment, to allow a person to drive more
efficiently and safely. An FHWA study examined the impact of lighting
options on urban freeways in Minnesota. 35 The study used data on crashes,
roadways, and traffic volume to compare the safety of

34 Keith K. Knapp, Dennis Kroeger, and Karen Giese, The Mobility and
Safety Impacts of Winter Storm Events in a Freeway Environment Final
Report, Iowa State Center for Transportation Research and Education
(February 2000).

35 Federal Highway Administration, Comparison of the Safety of Lighting
Options on Urban Freeways, FHWA Public Roads On- Line (Autumn 1994).

Page 26 GAO- 03- 436 Traffic Crash Causation continuously lighted urban
freeways with that of urban freeways having interchange lighting only.
Using data from between 1985 and 1990, the study determined that 12
percent more crashes occur on sections with

interchange- only lighting than on road sections with continuous lighting,
assuming all other factors remain the same. The study concluded that there
was a positive relationship between urban freeway lighting and highway
safety.

Weather- related phenomena can also inhibit driver visibility. While fog
crashes are proportionally small compared with all other crashes, they can
involve numerous vehicles in a chain- reaction pileup. A recent example of
this occurred in Wisconsin in October 2002, where a fog- related crash

involving 51 vehicles resulted in 10 deaths. The National Transportation
Safety Board has concluded that major fog- related incidents generally
occur because drivers have not maintained uniform reduced speeds during
times of limited visibility.

According to NHTSA*s Director of the Office of Human- Centered Research,
the significance of adverse weather, including both slippery roads and
reductions in driver visibility, is not fully understood because there are
no measurements (for example, vehicle miles traveled under adverse weather
conditions) available to make comparisons between crash rates under
various conditions. A researcher at the University of Michigan*s
Transportation Research Institute said that pedestrian- related crashes
are particularly sensitive to light conditions. The researcher pointed out
that, unlike vehicles and roads that may have lighting or reflective
markings, pedestrians are generally not highly visible and are more likely
to be involved in crashes during nighttime hours.

Vehicle factors can contribute to crashes through vehicle- related
failures and vehicle design characteristics (attributes that may increase
the likelihood of being involved in certain types of crashes). While such
recent events as the number of crashes involving tire separations have
highlighted the importance of vehicle factors, it is generally shown by
data and studies and believed by experts that vehicle factors contribute
less often to crashes than do human or roadway environment factors.

Two types of vehicle- related failures can contribute to traffic crashes:
equipment- related and maintenance- related. Equipment- related failures
include both original manufacturer and aftermarket- installed vehicle
equipment that function improperly. If not corrected, some
equipmentrelated

failures might lead to the loss of a vehicle*s handling capabilities,
Vehicle Factors Contribute

to Motor Vehicle Crashes Vehicle- Related Failures

Page 27 GAO- 03- 436 Traffic Crash Causation resulting in traffic crashes.
The widely publicized tire separations are a recent example of an
equipment- related failure. Equipment- related failures

can be identified by the manufacturer or by NHTSA, and may result in a
recall. In 2002, NHTSA reported 413 recalls involving over 18 million
vehicles, over 1 million pieces of equipment, about 675,000 tires, and
over 1 million child safety seats. NHTSA*s Director of the Office of
Defects Investigations told us that its investigations have identified 143
fatalities associated with recalls from 1990 through 2000. Maintenance-
related failures result from an operator*s improper maintenance of vehicle
components, which may impair the function of the vehicle*s equipment.
Examples of maintenance- related failures include inadequate tire tread
depth, worn brakes, unchecked or unchanged vehicle fluids, and
underinflated tires.

Our analysis of NHTSA*s data found that from 1997 through 2001, there were
about 778,000 crashes in which police identified that a specific vehicle-
related failure might have contributed to the crash. Where these failures
were identified, brake systems and tires were identified most frequently,
at 29 percent and 27 percent, respectively. Data is not collected by NHTSA
in a manner that provides information on whether these

crashes were caused by equipment or maintenance- related failures. One
vehicle factor that NHTSA believes may contribute to crashes is
underinflated tires. In 2001, NHTSA conducted a study that found that 27
percent of passenger cars and 33 percent of light trucks were being driven
with one or more underinflated tires. To reduce this problem, Congress

passed the Transportation Recall Enhancement, Accountability, and
Documentation Act of 2000 (P. L. 106- 414), which will require motor
vehicles to be equipped with a tire- pressure monitoring system to warn
the driver if a tire is significantly underinflated. In May 2002, NHTSA
issued part one of a two- part final rule requiring this system. It
requires that between November 1, 2003, and October 31, 2006, auto
manufacturers phase in one of two different tire- monitoring systems. The
second rule, which has yet to be finalized, is scheduled to be issued
March 1, 2005.

Several officials told us that vehicle- related failures and their effect
on crashes are difficult to quantify. For example, NHTSA*s Chief of
Information Services stated that the central problem with identifying
vehicle factors is that police officers are not necessarily qualified to
identify vehicle defects.

The design of a vehicle has been shown to affect handling in particular
types of maneuvers. For example, high- performance sports cars have very
Vehicle Design

Page 28 GAO- 03- 436 Traffic Crash Causation different handling
characteristics from those of sport utility vehicles (SUV). Recent changes
in the composition of the nation*s vehicle fleet, in

part attributable to the purchase of many SUVs, have resulted in an
overall shift toward vehicles with a higher center of gravity (more top-
heavy), which can roll over more easily than some other vehicles. Rollover
crashes are particularly serious because they are more likely to result in
fatalities. NHTSA has developed rollover ratings for vehicles by
calculating their static stability. This factor is a static metric that is
determined by dividing a vehicle*s track width, or distance between wheels
from side to side, by twice the height of its center of gravity.

As shown in figure 8, our analysis of NHTSA*s 2001 data showed that vans
were the least likely to be involved in a crash, with about 432 crashes
per 100 million vehicle miles traveled (VMT). 36 Passenger cars were the
most likely to be involved in a crash, with a rate of 655 crashes per 100
million VMT. The figure also shows that both vans and SUVs had the lowest
fatal crash rate, at 1.9 and 2.3 fatal crashes per 100 million VMT,
respectively.

36 For this VMT analysis we used data from the 2001 National Household
Travel Survey.

Page 29 GAO- 03- 436 Traffic Crash Causation Figure 8: Vehicle Crash
Rates, 2001

In 2001, rollover crashes killed 10,118 occupants in passenger cars,
pickup trucks, SUVs, and vans. This represents almost one- third of the
year*s 31,875 occupant deaths in these types of vehicles. Figure 9 shows
the percentage of rollover occurrence by vehicle type in 2001. Passenger
cars were the vehicle type least likely to roll over in a crash; passenger
cars rolled over in about 2 percent of all crashes, and rolled over nearly
16 percent of the time in fatal crashes. In comparison, our analysis shows
that SUVs were over three times more likely to roll over in a crash than
were passenger cars; that is, occurring in almost 6 percent of all
crashes. The proportion of SUVs that rolled over in fatal crashes was over
twice as high when compared with passenger cars. In 2001, SUVs rolled over
in fatal

crashes over 35 percent of the time.

Page 30 GAO- 03- 436 Traffic Crash Causation Figure 9: Passenger Vehicle
Rollovers, 2001

The National Transportation Safety Board (NTSB) recently examined
rollovers in 15- passenger vans from 1991 through 2000. 37 The NTSB found
that 15- passenger vans with 10 to 15 passengers had a rollover rate about
three times greater than that of vans seating 5 or fewer passengers. In
addition, NTSB found that the 15- passenger vans carrying 10 to 15
passengers rolled over in 96 of the 113 single- vehicle crashes (85
percent). However, they also found that the vans rolled over only 28
percent of the time, or 69 times out of the 244 single- vehicle crashes,
when there were

fewer than 5 occupants in the van. Additional analysis showed that higher
speeds were also strongly correlated with a greater chance of rollovers.
NTSB recommended that 15- passenger vans be rated by NHTSA for rollover
propensity. Although NHTSA has established a rollover resistance

37 National Transportation Safety Board, Evaluation of the Rollover
Propensity of 15- Passenger Vans, NTSB SR- 02/ 03 (October 2002).

Page 31 GAO- 03- 436 Traffic Crash Causation rating system and is
currently developing dynamic rollover tests, 15- passenger vans will not
be evaluated for rollover propensity because they

exceed the weight criteria for the testing program. A study by the
Insurance Institute for Highway Safety examined singlevehicle rollover
crashes. 38 The study concluded that the combined rollover crash rate for
pickup trucks and SUVs was more than twice the rate for passenger cars.
The higher rollover rate for pickup trucks and SUVs was present even when
taking into consideration a variety of crash circumstances, including
location, roadway alignment, and the driver*s age. The study concluded
that both pickup trucks and SUVs are more prone to rollover crashes than
are passenger cars.

A recent NHTSA study addressed rollovers from 1991 through 2000. 39 One of
its findings was that of all vehicle types considered in the study, SUVs
are the only type in which the number of occupant fatalities in rollover
crashes exceeds the number of occupant fatalities in nonrollover crashes;
in 2000, nearly two- thirds of SUVs* occupant fatalities occurred in
rollover crashes. One of the report*s conclusions was that, despite
declines in

passenger car occupant fatalities, the increasing influence of light truck
fatal crashes in general, and rollover crashes in particular, is
instrumental in maintaining the level of traffic crash fatalities. NHTSA*s
Division Chief of Math Analysis stated that reducing rollovers is one of
the NHTSA Administrator*s top five priorities.

In commenting on a draft of this report, NHTSA provided an analysis
comparing the crash rates for both passenger cars and light trucks using
VMT based on FHWA*s Highway Statistics Series. 40 The analysis indicates
that passenger cars had a lower fatal crash rate at 1.73 per 100 million
VMT, as compared with a rate of 2.13 for light trucks. With regard to the

38 Charles Farmer and Adrian Lund, *Rollover Risk of Cars and Light Trucks
after Accounting for Driver and Environmental Factors,* Accident Analysis
and Prevention,

vol. 34 (2002), pp. 163* 73. The study examined all single- vehicle fatal
crashes for 4 years, along with single- vehicle injury crashes involving
rollovers from three states by vehicle type for 4 years. The study used
vehicle registration as a means to make comparisons among vehicle types.

39 William Deutermann, Characteristics of Fatal Rollover Crashes, DOT HS
809 438 (April 2002). 40 This VMT analysis used the FHWA*s Highway
Statistics Series. FHWA obtains its data on vehicle miles traveled by
counting the number and types of vehicles passing particular

points around the country.

Page 32 GAO- 03- 436 Traffic Crash Causation vehicle involvement rate in
all crashes, NHTSA*s analysis shows that passenger cars had a rate of 423
crashes per 100 million VMT, which is

slightly higher than that for the light trucks (401 crashes per 100
million VMT).

In February 2003, the Alliance of Automobile Manufacturers, a trade group
that represents the three major U. S. automobile manufacturers and a
number of foreign manufacturers, published analyses examining occupant
fatality rates by vehicle type. One analysis used registered vehicles as a

method to compare fatality rates between vehicle types. Its results
indicate that in 2001, SUVs had a slightly higher occupant fatality rate
than had passenger cars* 16.25 and 15.70 per 100, 000 registered vehicles,

respectively. The alliance points out, however, that 72 percent of people
killed in SUV rollover crashes were not wearing safety belts, which can
reduce a driver*s risk of fatal injury in a rollover by 80 percent. They
further stated that in 2000, 35 percent of SUV single- vehicle rollover
fatalities were alcohol- related.

Various modal agencies within the Department of Transportation have
research projects underway and planned that address aspects of crash
causes. For example, the Federal Motor Carrier Safety Administration and
NHTSA are conducting a study on the causes and contributing factors to
large truck crashes. NHTSA is also funding the 100- Car Naturalistic

Driving Study, which involves collecting data about crashes and near
misses from 100 vehicles equipped with sensors. Further, NHTSA is funding
a project called the Drive Atlanta Study that involves collecting data
from 1,100 vehicles equipped with data recorders. In addition to possible
follow- on research on the above projects, planned research includes a
Transportation Research Board proposal for a 6- year program that would,
among other things, involve installing sensors and other data collection
devices on over 5,000 vehicles.

In 1999, Congress established the Federal Motor Carrier Safety
Administration (FMCSA) within DOT and mandated that it study the causes of
and contributing factors in large truck crashes. In 2001, large truck
crashes resulted in about 5,000 fatalities and 131,000 injuries. FMCSA
partnered with NHTSA to implement the 4- year, $18 million Large Truck
Crash Causation Study. The study*s goal is to develop a greater
understanding of the factors leading to large truck crashes, so that
costeffective countermeasures can be developed to decrease the number and
severity of these crashes. Federal Research Directed at Better

Understanding of Factors That Contribute to Crashes

Large Truck Crash Causation Study

Page 33 GAO- 03- 436 Traffic Crash Causation To conduct this study, FMCSA
and NHTSA built on the existing crash investigation system that NHTSA had
established to collect data for the

Crashworthiness Data System database. For this effort, researchers at 24
locations collect information on a sample of large truck crashes by
visiting the crash sites shortly after they occur and completing a
response protocol that was developed for this project. NHTSA*s Director of
the National Center on Statistics and Analysis told us that the most
informative crash causation data is often collected at the site of the
crash while the vehicles and participants are still present.

Cooperative agreements were established between the police, FMCSA, and
NHTSA to use an established, on- scene investigative approach. These
cooperative agreements were based on previous agreements set up between
NHTSA and police for data collection for the CDS database, but they were
modified to accommodate the other parties involved and a faster time frame
for the crash investigations. A NHTSA official stated that this
multiagency partnering is important for the success of the study, and that
establishing rapid notification procedures requires the cooperation of

state and local police along with their police dispatch personnel. The
researchers expect to investigate at least 1,000 crashes by the end of
2003. FMCSA and NHTSA officials said that the results will yield findings
about critical pre- crash events, the reasons for these events, and
relative

risks in truck crashes. They also said that this information should
significantly help to create proven countermeasures to decrease the number
and limit the severity of truck crashes.

As a follow- on to this study, NHTSA requested $10 million in its fiscal
year 2004 budget to begin a National Motor Vehicle Crash Causation Survey.
This study would develop and conduct a nationally representative effort to
collect on- scene crash causation data. The Large Truck Crash Causation
Study would be used as the model for the proposed study. The on- scene
methodologies and procedures developed for the Large Truck Crash Causation
Study would also be applicable to this proposed effort. NHTSA officials
said that start- up costs and implementation timing would be reduced by
making use of the infrastructure in place for the truck study, which is
scheduled to complete data collection by the end of 2003.

NHTSA is currently conducting the 100- Car Naturalistic Driving Study,
whose purpose is to help develop better crash- avoidance warning systems.
This 1- year, $3 million driving research study involves collecting data
from 100 vehicles equipped with various sensors and cameras. NHTSA has One
Hundred- Car

Naturalistic Driving Study

Page 34 GAO- 03- 436 Traffic Crash Causation partnered with FHWA,
Virginia, and the Virginia Polytechnic Institute and State University
(Virginia Tech) to fund the study. Virginia Tech is

responsible for conducting the study. NHTSA has equipped 100 cars (80
individually owned and 20 leased) with five video cameras and a variety of
sensors to track proximity and relationships to other vehicles and
objects. In addition, the vehicles have sensors that detect glare and
whether the driver is using a cell phone in the car. Volunteers will use
the vehicles for their everyday driving in the metropolitan Washington, D.
C., area for the duration of the study, which began in early 2003. The
cameras and sensors are to provide data for studying crashes as well as
near misses. In the event of a crash, NHTSA will send a team of
researchers to the site to investigate.

NHTSA officials told us that they are considering a follow- on to this
study, if it is successful. An expanded version of the study could include
a representative sample of up to 10, 000 equipped cars around the country.
The official said that after completion of the initial study, researchers
should have greater knowledge about which sensors and equipment provided
the most relevant information on contributing factors to motor vehicle
crashes, and would install only that equipment in the larger fleet of
vehicles. NHTSA told us that they might seek funding from auto
manufacturers and other entities to supplement their funding. Later this
year, NHTSA will begin a 2- year, $3.1 million Drive Atlanta

Study, which involves installing data recorders in 1,100 vehicles to
develop information on situations and circumstances where excessive speed
contributes to crashes. Drive Atlanta is primarily funded by a $1.9
million contract with NHTSA and $1.2 million from Safety Intelligence
Systems, Inc. FHWA is also contributing money to the study. The private
company is providing the development costs and is prototyping and testing
the

MACBOX, the data recorder that will be used by the Georgia Institute of
Technology to conduct the study.

In this study, the data recorder information will be combined with three
other types of data. Data will be contributed by the Atlanta Traffic
Management Center on prevailing traffic conditions, the National Oceanic
and Atmospheric Administration on weather, and the Georgia Department

of Transportation on roadway characteristics. According to a program
official, this combination of data will enable the researchers to know
when and where the driving occurred, what were the posted speed limits
along the drivers* routes, what were the roads* characteristics, and
numerous Drive Atlanta Study

Page 35 GAO- 03- 436 Traffic Crash Causation other data. The researchers
plan to create speed profiles for all of the study*s participants at the
conclusion of the study, to examine exactly how

speed is involved in crashes. NHTSA estimates that at least 100 crashes
will occur over the next 2 years involving these vehicles.

In the Transportation Equity Act for the 21st Century (P. L. 105- 178),
Congress requested that the Transportation Research Board conduct a study
to determine the goals, purposes, research agenda and projects,
administrative structure, and fiscal needs for a new strategic highway
research program. In response to this request, a committee of highway
industry leaders was formed to develop recommendations. The committee
engaged in an outreach process to gather input from the highway community
regarding strategic priorities and promising research approaches. The
committee*s report was published in October 2001 and recommended a 6-
year, $450 million to $500 million Future Strategic Highway Research
Program (F- SHRP) focused on the following areas: (1) accelerating the
renewal of America*s highways; (2) making a significant

improvement in highway safety; (3) providing a highway system with
reliable travel times; and (4) providing highway capacity in support of
the nation*s economic, environmental, and social goals. FHWA contributed
$1.5 million over fiscal years 2002 and 2003 toward the F- SHRP planning
activities.

The F- SHRP objective of *making a significant improvement in highway
safety* includes three major areas: (1) methodology development using
existing data, (2) large- scale research studies of multiple factors
related to the risk of collisions and casualties for high priority roadway
safety issues, and (3) analysis of the field data for countermeasure
implications. A key aspect of this project is the use of in- vehicle and
roadside technologies to gather data to examine crash rates and pre- crash
conditions on a large

scale to perform risk analyses. The study recommended that $180 million to
$200 million be committed to this safety objective.

The F- SHRP safety plan includes using the data from NHTSA*s national
crash databases as well as other studies that have used instrumented
vehicles and roadside technologies. The F- SHRP study would involve
collecting data from 5,000 to 6, 000 instrumented vehicles and roadside
technologies for over 2 to 3 years. According to the contractor who
developed the implementing plan for the F- SHRP safety goal, analysis of
the previous study data will enable the F- SHRP researchers to first test
risk measures and analysis methods before implementation of the F- SHRP
field study. The contractor said that NHTSA*s 100- Car Naturalistic
Driving Future Strategic Highway

Research Program

Page 36 GAO- 03- 436 Traffic Crash Causation Study and Drive Atlanta Study
and FHWA*s Road Departure Study would be good sources of the type of
instrumented vehicle data the early

methodology projects need. 41 The proposed F- SHRP instrumented vehicle
study would involve data collection for two high- priority highway safety
problems: run- off- the- road and intersection crashes. The fleet of
instrumented vehicles would be split between at least two geographic
areas, with volunteer drivers using the vehicles for their everyday
driving. The final phase of the research would be to use the results of
the largescale

instrumented vehicle study to identify appropriate countermeasure
improvements.

We provided copies of a draft of this report to the Department of
Transportation for its review and comment. In discussing this report,
NHTSA and FHWA officials provided technical clarification and information,
which we incorporated in the report as appropriate. In addition, NHTSA
provided information comparing light truck and passenger car crash rates,
which we also incorporated in the report.

As arranged with your offices, unless you publicly announce its contents
earlier, we plan no further distribution of this report until 30 days
after the date of this letter. At that time, we will send copies of this
report to cognizant congressional committees and to the Honorable Norman
Y. Mineta, Secretary of Transportation; the Honorable Dr. Jeffrey W.
Runge, Administrator of the National Highway Traffic Safety
Administration; and the Honorable Mary E. Peters, Administrator of the
Federal Highway Administration. We will also make copies available to
others upon request. In addition, the report will be available at no
charge on the GAO Web site at http:// www. gao. gov.

41 In the Road Departure Study, the University of Michigan Transportation
Research Institute will develop and test a new crash avoidance warning
system in 11 passenger cars. The system, designed to prevent road
departure and run- off- the- road crashes, will alert the driver when the
vehicle begins to wander off the road or when the vehicle is traveling too
fast for an upcoming curve. Agency Comments

and Our Evaluation

Page 37 GAO- 03- 436 Traffic Crash Causation If you have questions about
the report, please contact me at (202) 512- 2834. Key contributors to this
report were Michele Fejfar, Glenn C.

Fischer, Bonnie Pignatiello Leer, Sara Ann Moessbauer, Elsie Picyk,
Beverly Ross, and Glen Trochelman.

Peter Guerrero Director, Physical Infrastructure Issues

Appendix I: Objectives, Scope, and Methodology

Page 38 GAO- 03- 436 Traffic Crash Causation To provide information on the
factors that contribute to motor vehicle crashes, we obtained and analyzed
crash data from National Highway

Traffic Safety Administration (NHTSA) databases, obtained and reviewed
research studies on the topic, and interviewed a variety of experts and
federal officials. To identify major ongoing and planned Department of
Transportation (DOT) research into factors that cause crashes, we obtained
documents from and interviewed officials of NHTSA and the Federal Highway
Administration (FHWA).

For each factor contributing to traffic crashes, we obtained and analyzed
data for calendar years 1997 through 2001 from NHTSA*s crash reporting
systems* the most recent 5- year period for which these data are
available. Our analysis involved the use of three of NHTSA*s databases:
the Fatality Analysis Reporting System, the General Estimates System, and
the Crashworthiness Data System. Each database contained different levels
of crash data.

 Fatality Analysis Reporting System (FARS) * This database provides
information on all traffic- related fatalities. A crash must result in the
death of an occupant or nonmotorist within 30 days of the incident to be
included in this database. Each of the states provides the data to NHTSA
in a standardized format. The states generally obtain this information
through data from reports that police officials prepare at the scene of
the crash as well as state vehicle registration files, state driver
licensing files,

state highway department data, death certificates, coroner or medical
examiner reports, hospital medical records, and emergency medical service
reports. NHTSA created the database to identify traffic safety problems,
develop and implement countermeasures, and evaluate vehicle safety
standards and highway safety programs. We used this database to

present overall information on traffic deaths as well as to provide an
understanding of crashes involving speed, alcohol, age, and vehicle
design. It should be noted that while fatality data is useful in
understanding crashes, other factors, in addition to those involved in
causing the crash, might have contributed to the fatality. This would
include such factors as whether safety belts or other occupant protection
measures were used and operated properly.

 General Estimates System (GES) * This database is created from a
nationally representative sample of police accident reports completed for
crashes. Other criteria necessary for inclusion in the database are that
the crash must involve at least one motor vehicle traveling on a traffic
way, and that the crash must result in property damage, injury, or death.
This database was created to identify traffic safety problem areas,
provide a Appendix I: Objectives, Scope, and

Methodology Analyzing NHTSA Data

Appendix I: Objectives, Scope, and Methodology

Page 39 GAO- 03- 436 Traffic Crash Causation basis for regulatory and
consumer initiatives, and form the basis for cost and benefit analyses of
traffic safety initiatives. This is NHTSA*s largest

crash database, with information collected on over 50,000 randomly sampled
police accident reports each year. We analyzed the GES data to provide
information on speed, traffic control violations, roadside hazards,
roadway conditions, and vehicle defects. In addition, for the analysis in
which we compared fatalities with other types of crashes (reduced
visibility, vehicle design, age), we combined FARS and GES data.

 Crashworthiness Data System (CDS) * This database contains information
from a detailed sample of about 4,000 minor, serious, and fatal crashes
annually. The criterion necessary for inclusion in the database is that at
least one passenger vehicle must be damaged severely enough to require
towing from the crash site. Teams of trained crash investigators visit the
crash site and collect data elements such as vehicle crash damage and
interior vehicle locations that the occupants struck. The investigators
also generally locate and interview crash victims and review medical
records to determine the types of crash- related injuries. A goal of this
database includes having the ability to examine the crashworthiness of
vehicles; that is, how vehicles perform in crashes with respect to
protecting their occupants. We used the CDS database to provide
information on crashes involving driver inattention.

In commenting on a draft of our report, NHTSA officials said that the
FARS, GES, and CDS databases, although providing useful information, rely
on data from police accident reports or on data collected days or weeks
after the crash, making it difficult to obtain causation data. Therefore,
NHTSA relies on the Indiana Tri- Level study data, which is almost a
quarter of a century old. They noted that since the Tri- Level study was
completed, cars, drivers, highways, technology, and lifestyles in the
United States have changed dramatically. As previously discussed, NHTSA
has proposed to develop and conduct a nationally representative survey to
collect on- scene crash causation data* the National Motor Vehicle Crash
Causation Survey. NHTSA officials indicated that these on- scene, real-
time data are needed to best understand crash causation.

In addition to using these three databases, for some analyses we also
calculated frequency rates using vehicle miles traveled. We used vehicle
miles traveled data from two different sources* the 2001 FHWA Highway

Statistics Series data and the 2001 National Household Travel Survey. For
example, we used FHWA*s vehicle miles traveled data in examining crash
rates by road type. FHWA obtains its data on vehicle miles traveled by
counting the number and types of vehicles passing particular points

Appendix I: Objectives, Scope, and Methodology

Page 40 GAO- 03- 436 Traffic Crash Causation around the country. Because
FHWA*s statistics do not include data on age or gender, we used vehicle
miles traveled from the 2001 National

Household Travel Survey for some analyses. 42 This survey of about 26,000
households in the United States was conducted from March 2001 through May
2002. It provides data on personal travel behavior at the national level

to use as a benchmark for a variety of applications. Although the overall
response rate for sampled households was low (41 percent), there are few
other sources for information on U. S. travel patterns. We used the most
recent available data, preliminary release (version 1), to get estimates
of annual vehicle miles traveled for our tables on rates of driver
involvement and types of vehicles involved in crashes.

We assessed the reliability of FARS, CDS, and GES by reviewing existing
information about the data and performing electronic tests of the data.
There are certain limitations associated with using these databases for
our

analysis. For example, the source of the GES data is police accident
reports that are prepared at the scene of the crash. Although the GES has
procedures to ensure that data reflect information in the accident
reports, we did not verify the accuracy of the accident reports
themselves. In addition, since GES, CDS, and the National Household Travel
Survey are based on samples, any estimates derived from these databases
are subject to sampling errors. A sampling error indicates how closely the
results of a particular sample would be reproduced if a complete count of
the

population were taken with the same measurement methods. The estimated
sampling errors (at the 95 percent confidence level) do not exceed plus or
minus 11 percentage points.

To identify recent studies on factors that contribute to motor vehicle
crashes, we conducted a literature search, explored the Transportation
Research Information System, and reviewed periodicals. This effort
resulted in numerous studies being identified on various aspects of the
motor vehicle crashes. We then, with input from a number of experts and
officials from NHTSA and FHWA, judgmentally selected studies that would
provide additional information on the particular factors being discussed.
For each of the selected studies that are used in this report, we
determined whether the study*s findings were generally reliable. To do so,
we

evaluated the methodological soundness of the studies using common 42 2001
National Household Travel Survey User*s Guide, Version 1( Preliminary
Release) January 2003. Identifying Studies

Appendix I: Objectives, Scope, and Methodology

Page 41 GAO- 03- 436 Traffic Crash Causation social science and
statistical practices. For example, we examined each study*s methodology,
including its limitations, data sources, analyses, and conclusions.

In conducting this review we interviewed a wide variety of federal
officials and other experts. Within DOT, we interviewed officials from the
Volpe Center, the National Highway Traffic Safety Administration, and the
Federal Highway Administration. We also spoke with individuals affiliated
with academic institutions, including the University of North Carolina*s
Highway Safety Research Center, the University of Michigan*s
Transportation Research Institute, Northwestern University*s Center for

Public Safety, Texas A& M University*s Texas Transportation Institute, and
the Johns Hopkins School of Public Health. In addition, we interviewed
officials from the Insurance Institute for Highway Safety. We also spoke
with automobile industry representatives at the Alliance of Automobile
Manufacturers, a trade group that represents the three major U. S.
automobile manufacturers and a number of foreign manufacturers. We spoke
with officials from the Transportation Research Board, Advocates for
Highway and Auto Safety, AAA (formerly the American Automobile
Association), American Association of State Highway and Transportation
Officials (AASHTO), the National Sleep Foundation, the Midwest Research
Institute, and the Southern California Research Institute. In general, the
officials and experts provided information about major factors that
contribute to motor vehicle crashes and research on these factors.

To identify major ongoing and planned DOT research into factors that
contribute to motor vehicle crashes, we interviewed officials from NHTSA,
FHWA, and the Transportation Research Board. These agencies have a great
deal of ongoing and planned research on a wide variety of motor vehicle
safety issues, such as research to mitigate accident severity and safety
system issues. However, to respond to this objective, we selected

ongoing and planned studies that (1) represented major research studies,
(2) examined multiple factors contributing to crashes, (3) examined causal
factors rather than countermeasures, and (4) collected original data,
rather than analyzed existing data. We also obtained documents describing
the research projects and reviewed federal budgetary documents on the
projects.

We performed our review from July 2002 through February 2003 in accordance
with generally accepted government auditing standards. Interviewing
Federal

Officials and Experts Ongoing and Planned Transportation Research

Appendix II: Tri- Level Study of the Causes of Traffic Accidents

Page 42 GAO- 03- 436 Traffic Crash Causation Researchers at the Indiana
University Bloomington*s Institute for Research in Public Safety conducted
the Tri- Level Study of the Causes of Traffic

Accidents from 1972 through 1977. The study investigated how frequently
various human, environmental, and vehicle factors were involved in traffic
crashes. According to NHTSA officials, the Tri- Level study has been the
only study in the past 30 years to collect on- scene crash causation data.
The study, conducted for NHTSA, incorporated 13,568 police- reported
crashes, including on- scene investigation of 2,258 crashes, and an in-
depth investigation of 420 crashes. The investigation teams assessed
causal

factors as definite, probable, or possible. The in- depth team identified
human errors as definite or probable causes in 93 percent of the crashes,
environmental factors in 34 percent, and vehicle factors in 13 percent. In
20 percent of the crashes studied in depth, no definite cause could be
identified.

Indiana University conducted the study to satisfy a broad range of NHTSA*s
needs for data on traffic crash causation. Two of the main objectives for
the research were to:

 Identify those factors that are present and serve to initiate or
influence the sequence of events resulting in a motor vehicle crash.

 Determine the relative frequency of these factors and their causal
contribution within a defined crash and within the driving population.

Researchers collected collision data on three levels (A, B, and C), each
providing an increasing amount of detail. Data collection for level A
involved examining police reports for 13,568 crashes and collecting other
baseline data, such as vehicle registration files, driver license files,
roadway inventories, and local surveys. For level B, teams of technicians
conducted on- site investigations of 2,258 crashes immediately following
their occurrence. For level C, a multidisciplinary team conducted
independent, in- depth investigations of 420 of the crashes. The crashes
investigated on- scene and in- depth were generally representative of all
police- reported crashes occurring in Monroe County, Indiana, during the

study period. In the clinical assessments of crash causation in Monroe
County, a traffic crash was viewed as the last event in a chain of events
and conditions that preceded it. A crash cause was defined as an event or
condition but for which the crash would not have occurred. Emphasis was
placed on events Appendix II: Tri- Level Study of the Causes of Traffic
Accidents Objectives, Scope,

and Methodology for the Tri- Level Study

Appendix II: Tri- Level Study of the Causes of Traffic Accidents

Page 43 GAO- 03- 436 Traffic Crash Causation and conditions that
immediately preceded the crash because they may be viewed as the final
links of a casual chain that culminates in the crash.

According to the study, during in- depth investigations, the researchers
attempted to acquire as much relevant information as possible, and then
made clinical case- by- case determinations of the causal factors
involved, based on all of the information obtained. An assessment system
permitted each identified factor to be evaluated as definitely, probably,
or possibly

involved as either a causal or severity- increasing factor. A causal
factor was defined as a factor necessary or sufficient for the occurrence
of the crash; had the factor not been present in the crash sequence, the
crash would not have occurred. A severity- increasing factor was defined
as a factor that was neither necessary nor sufficient for the occurrence
of the crash, but its removal from the crash sequence would have lessened
the speed of the initial impact. The causal assessment process for each
crash involved two major steps: first, identifying relevant deficiencies
of drivers, vehicles, and the driving environment that were present in the
crash sequence; and second, assessing the investigation team*s certainty
that the crash would not have occurred had each deficiency been corrected
to its minimally acceptable state.

In addition, data on Monroe County drivers, vehicles, roads, and crashes
were compared with available national data. It was found that for Monroe
County the severity distribution of reported crashes; the proportion of
crashes occurring on dry, wet, or snow- or ice- covered roads; the
proportion occurring in urban or rural areas; and the age distribution of
the vehicles were nearly the same as for the United States as a whole. The
most notable difference was that young drivers were overrepresented.
However, the effects of this overrepresentation on the overall causal
results were found to be minimal. Thus, it was found that while the
results from Monroe County, Indiana, do not represent the United States as
a whole in a statistical sense, they indicate factors that are likely to
be important on a national level and their relative involvement.

Human factors were the most frequently implicated of the three categories,
and vehicle factors the least frequently implicated. As figure 10 shows,
the in- depth team concluded that human factors were definite causal
factors in 71 percent of the crashes; environmental factors in 13 percent;
and vehicle factors in 4 percent. Similarly, the in- depth team concluded
that these same three categories were definite or probable

causal or severity- increasing factors in 93 percent, 34 percent, and 13
percent, respectively. Results of the TriLevel

Study

Appendix II: Tri- Level Study of the Causes of Traffic Accidents

Page 44 GAO- 03- 436 Traffic Crash Causation Figure 10: Factors
Contributing to Crashes Identified by the Tri- Level Study

The on- site team concluded that human factors were definite causal
factors in 64 percent of the crashes; environmental factors in 19 percent;
and vehicle factors in 4 percent. The on- site team concluded that these
same three categories were definite or probable causal or
severityincreasing factors in 90 percent, 35 percent, and 9 percent,
respectively. The in- depth team could not establish a definite cause for
20 percent of the crashes they investigated, while the on- site
technicians could not establish a cause for 26 percent of the crashes.
However, the on- site team identified one or more probable causes in
nearly all the crashes. Also, more than one factor was implicated as a
cause in many of the crashes.

The study categorizes human direct causes based on an
informationprocessing model of the driver as vehicle controller. This
model assumes that drivers are continuously engaged in perceiving and
comprehending information, making decisions, and taking actions to achieve
necessary control responses. The *perception* and *comprehension*
categories were combined as *recognition errors* because of the difficulty
in distinguishing Human Factors

Appendix II: Tri- Level Study of the Causes of Traffic Accidents

Page 45 GAO- 03- 436 Traffic Crash Causation errors in these functions
through crash investigation. A *critical nonperformance* category was
added to reflect instances where a driver

ceases to perform as an information processor. A *noncrash* category was
included to accommodate any intentional crash involvements. Recognition
errors were cited as the most prevalent human causal factor, followed by
decision errors, performance errors, and critical nonperformance errors.

More specific human- direct- cause categories were grouped in the causal
hierarchy, under these major headings. With regard to specific human
errors, improper lookout was cited as the most prevalent error. Other
specific human errors cited included excessive speed, inattention,
improper evasive action, and internal distraction. The researchers
separately recorded human conditions and states that

impeded the ability of the driver to function as an information processor.
These factors, which included fatigue, driver experience, and alcohol
impairment, were viewed as potential *reasons behind the reasons.* Alcohol
impairment was cited as the most prevalent human condition, followed by
other drug impairment and fatigue.

The study categorized environmental factors as involving highway- related
factors, slick roads, or other ambience- related factors. Among these,
highway- related factors predominated; the in- depth team identified them
as definite causes in 7 percent of crashes. Slick roads were definite
causes

in 4 percent of crashes, and other ambience- related factors in 2 percent.
More specific environmental causes were defined under these three broad
headings. The most commonly cited specific environmental factors were view
obstructions and slick roads.

The study categorized vehicle factors according to major vehicle systems,
and then according to more specific categories. The most commonly cited
deficiency in these systems was with the brake system, followed by the
tires and wheels. The most commonly cited vehicle deficiency causal factor
was gross brake failure, followed by inadequate tread depth. Environmental
Factors

Vehicle Factors

Appendix III: Roadway Design Features Page 46 GAO- 03- 436 Traffic Crash
Causation A number of roadway design studies and experts we spoke with
addressed how various aspects of roadway design might contribute to
traffic crashes.

These included medians, lane widths, shoulders, curves, access points, and
intersections.

Medians are physical separations between opposite lanes of traffic that
provide a recovery area for out- of- control vehicles. They also serve to
separate traffic driving in opposite directions, thereby minimizing their
interactions and likelihood of being involved in catastrophic head- on
crashes. Some considerations regarding medians include their presence or
absence (that is, divided vs. undivided roads), the width of the median,
and whether a barrier is placed in the median.

An analysis of NHTSA*s databases showed that from 1997 through 2001, 44
percent of all traffic fatalities occurred on undivided, rural, two- lane
roads. This represents 73 percent of all traffic fatalities in rural
areas. In urban areas, 35 percent of traffic fatalities occurred on two-
lane undivided roadways. In addition, a study conducted by the Kentucky
Transportation Center examined the impact of converting two- lane
undivided rural roads to four- lane divided roads at 25 locations. 43 They
found that, on average, there was a reduction in crash rate after the
road*s conversion from a two- lane undivided rural road to a four- lane
divided road.

According to experts with whom we spoke, medians provide safety benefits
by allowing vehicles enough room to recover from various vehicle or human
factors that could contribute to a crash. A study published in the

Transportation Research Record used Highway Safety Information System data
from Illinois and Utah to assess the relationship between median width and
crash rates. 44 The study was based on a total of 3,055.1 miles of
roadway, with speed limits of at least 35 miles per hour. The study
attempted to isolate only the median*s width as the predictive factor for
crash rates, but it acknowledged that there could be other elements
influencing the crash rates as well. Overall, the study concluded that
crash rates decrease with increasing median widths greater than 25 to 30
feet,

43 Kenneth Agent and Jerry Pigman, Safety Impacts of Rural Road
Construction, Kentucky Transportation Center KTC- 01- 01( February 2001).
44 M. W. Knuiman, F. M. Council, and D. W. Reinfurt, The Effect of Median
Width on Highway Accident Rates, Transportation Research Record 1401,
1993. Appendix III: Roadway Design Features

Medians Presence or Absence of Medians

Width of Medians

Appendix III: Roadway Design Features Page 47 GAO- 03- 436 Traffic Crash
Causation and increasing widths continue to provide additional benefits up
to widths of approximately 65 to 80 feet.

Some experts told us that although the installation of median barriers can
reduce head- on crashes, their presence may increase the number of total
crashes. This might occur because the median barrier reduces the amount of
space a vehicle has to recover within the median. The Washington State
Department of Transportation recently conducted a study of cross- median
crashes on multilane and divided state highways with full- access control.
45 One goal of the study was to revise the guidelines for the installation
of

median barriers. The study examined cross- median crashes from 1996
through 2000 from a sample of 677 miles of road. Using a benefit- cost
analysis, the study recommended installing median barriers on all full-
access control, multilane highways with posted speed limits of 45 miles
per hour or greater where the median width was 50 feet or less.

Wider lanes increase the separation between vehicles traveling in adjacent
lanes as well as provide additional space to recover from near- crash
situations. In a recent study, FHWA addressed the relationship between
lane width and crashes on two- lane rural highways based on expert
assessments and previous studies. 46 The study included an analysis of the
combined effects of lane width and average daily traffic on crash rates,
and it predicted that lane width has only a slight impact on crash rates
at low volumes of traffic. 47 However, the study also predicted that at
high- average daily traffic volumes, the two- lane rural roads with 9-
foot lanes have a 50 percent greater chance of having crashes than have
similar roads with 12- foot lanes. In discussing lane width with experts,
we were told by one academic researcher that while wider lanes provide
additional space between vehicles, wider lanes may give drivers an
increased

perception of safety resulting in higher rates of speed, possibly leading
to other safety problems. 45 Washington State Department of
Transportation, Median Treatment Study on Washington State Highways (March
2002). 46 D. W. Harwood, F. M. Council, E. Hauer, W. E. Hughes, and A.
Vogt, Prediction of the Expected Safety Performance of Rural Two- Lane
Highways, FHWA- RD- 99- 207 (December 2000).

47 This factor applies to single- vehicle run- off- the- road, multiple-
vehicle same- direction sideswipe crashes, and multiple- vehicle opposite-
direction crashes. Existence of Barriers Lane Widths

Appendix III: Roadway Design Features Page 48 GAO- 03- 436 Traffic Crash
Causation Roadway shoulders provide a clear space for drivers to recover
from errors. A recent FHWA study examined the relationship between
shoulder

width, average daily traffic, and crash rates for two- lane rural highways
and predicted that, in general, at low- average daily traffic rates,
shoulder width only slightly affects the crash rate but as the average
daily traffic rate increases, so does the influence of shoulder width on
crash rates. 48 For example, at high- average daily traffic volumes, the
study predicted

that a 50 percent greater number of crashes occur on two- lane rural
highways with no shoulders than on similar roads with 6- foot shoulders.
Experts told us that while wider shoulders are generally better than
narrow ones, the benefits that shoulders provide are also influenced by
the material from which they are constructed. A researcher at the
University of North Carolina*s Highway Research Safety Center told us that
paved shoulders are associated with fewer crashes at lower rates than
those with gravel or grass shoulders. Another expert pointed out that soft
shoulders can lead to a loss of vehicle control both through the uneven
edges between the driving lane and the shoulder or through a differential
of friction between the driving lane and the shoulder.

Curves have been shown to contribute to crashes, whether horizontal curves
(left or right) or vertical curves (up and down). Various elements of
curves may affect the likelihood of a crash, including features of the
curve (for example, degree, length, and angle of the curve) and cross-
sectional curve elements (for example, lane width, shoulder size and
type.) A 1991 FHWA study identified factors more strongly associated with
curves than adjacent straightaways in Washington State. 49 These factors
included a higher percentage of fatal crashes, head- on and opposite
sideswipe crashes, fixed- object and rollover crashes, crashes at night,
and crashes involving drinking drivers. Vertical curves have also been
associated with higher crash rates, though, according to an American
Association of State

Highway and Transportation Officials chairperson, not as much as compared
with horizontal curves. An important design element regarding 48 D. W.
Harwood, F. M. Council, E. Hauer, W. E. Hughes, and A. Vogt, Prediction of
the

Expected Safety Performance of Rural Two- Lane Highways, FHWA- RD- 99- 207
(December 2000).

49 C. Zegeer, J. Stewart, F. Council, and D. Reinfurt, Cost Effective
Geometric Improvements for Safety Upgrading of Horizontal Curves, Federal
Highway Administration Report FHWA- RD- 90- 021 (October 1991). Roadway
Shoulders

Curves

Appendix III: Roadway Design Features Page 49 GAO- 03- 436 Traffic Crash
Causation vertical curve safety is the need to provide drivers with
adequate stopping sight distance.

Access points are locations where vehicles enter a roadway, such as
residential and business driveways and exit and entrance ramps on
highways. A 1998 study completed for the Minnesota Department of
Transportation found that as access points to roads increase, so do the
number of crashes. 50 For example, on four- lane urban conventional
roadways, with no left turns, the researchers found that in Minnesota
there were an average of 2.22 crashes per million vehicle miles traveled
when there were from zero to ten access points per mile. (See fig. 11.)
However, the rate of crashes increased to 7.38 when the number of access
points was greater than 50 per mile. Additionally, traffic safety experts
supported the conclusion that more access points generally lead to higher
crash rates.

50 Minnesota Department of Transportation, Statistical Relationship
Between Vehicle Crashes and Highway Access (August 1998). Access Points

Appendix III: Roadway Design Features Page 50 GAO- 03- 436 Traffic Crash
Causation Figure 11: Impact of Access Points on Traffic Crashes According
to FHWA, intersections are among the most complex roadway designs a driver
encounters. A recent report for NHTSA found that in 2001, intersection and
intersection- related crashes represented 22.5 percent of

total fatal crashes and 43 percent of overall crashes. 51 There are four
major crash types at intersections: crossing, rear- end, improper lane
changing, and pedestrian and bike. Multiple factors contribute to
intersection crashes, including: poor physical design, inadequate traffic
engineering, failure of driver licensing and education to train drivers in
negotiating

intersections, and driver disregard for traffic control devices. For
example, a poorly designed intersection might provide inadequate sight
distance, which could limit a driver*s response time to react to vehicles
or pedestrians at that intersection. Additionally, incorrectly timed or
inconspicuous traffic control devices can also contribute to a crash.

51 National Highway Traffic Safety Administration, Traffic Safety Facts,
2001. Intersections

(545006)

The General Accounting Office, the audit, evaluation and investigative arm
of Congress, exists to support Congress in meeting its constitutional
responsibilities and to help improve the performance and accountability of
the federal government for the American people. GAO examines the use of
public funds; evaluates federal programs and policies; and provides
analyses, recommendations, and other assistance to help Congress make
informed oversight, policy, and funding decisions. GAO*s commitment to
good government is reflected in its core values of accountability,
integrity, and reliability.

The fastest and easiest way to obtain copies of GAO documents at no cost
is through the Internet. GAO*s Web site (www. gao. gov) contains abstracts
and fulltext files of current reports and testimony and an expanding
archive of older products. The Web site features a search engine to help
you locate documents using key words and phrases. You can print these
documents in their entirety, including charts and other graphics.

Each day, GAO issues a list of newly released reports, testimony, and
correspondence. GAO posts this list, known as *Today*s Reports,* on its
Web site daily. The list contains links to the full- text document files.
To have GAO e- mail

this list to you every afternoon, go to www. gao. gov and select
*Subscribe to daily E- mail alert for newly released products* under the
GAO Reports heading.

The first copy of each printed report is free. Additional copies are $2
each. A check or money order should be made out to the Superintendent of
Documents. GAO also accepts VISA and Mastercard. Orders for 100 or more
copies mailed to a single address are discounted 25 percent. Orders should
be sent to: U. S. General Accounting Office 441 G Street NW, Room LM
Washington, D. C. 20548 To order by Phone: Voice: (202) 512- 6000

TDD: (202) 512- 2537 Fax: (202) 512- 6061

Contact: Web site: www. gao. gov/ fraudnet/ fraudnet. htm E- mail:
fraudnet@ gao. gov Automated answering system: (800) 424- 5454 or (202)
512- 7470 Jeff Nelligan, managing director, NelliganJ@ gao. gov (202) 512-
4800

U. S. General Accounting Office, 441 G Street NW, Room 7149 Washington, D.
C. 20548 GAO*s Mission Obtaining Copies of

GAO Reports and Testimony

Order by Mail or Phone To Report Fraud, Waste, and Abuse in Federal
Programs Public Affairs
*** End of document. ***