REAL WORLD CRASH EVALUATION OF VEHICLE

STABILITY CONTROL (VSC) TECHNOLOGY

G. Bahouth
Pacific Institute for Research and Evaluation
Calverton, MD

ABSTRACT
This study quantifies the effect of Vehicle Stability Control
(VSC) in reducing crash involvement rates for a subset of vehicles in
the US fleet. Crash rates for a variety of impact types before and after
VSC technology was implemented are compared. Police-reported
crashes from six available US state files from 1998-2002 were
analyzed including 13,987 crash-involved study vehicles not equipped
with the technology and 5,671 crashes of vehicles equipped with VSC
as a standard feature. Overall, an 11.2% (95% CI: 2.4%, 21.1%)
reduction in multi-vehicle frontal crash involvement was identified
for VSC-equipped vehicles. A 52.6% (95% CI: 42.5%, 62.7%)
reduction in single-vehicle crash rates was found.

In recent years, numerous technological advances have improved the

active safety of vehicles. These technologies—anti-lock brakes,
traction control, active yaw-control, and active-roll control to name a
few—are being developed to assist drivers with crash avoidance.
Such active safety technologies may reduce the severity of a crash or
even eliminate the crash from occurring.

The primary function of VSC is to assist the driver in

maintaining control of the vehicle during sudden maneuvers or
adverse weather conditions. VSC can be classified as an active yaw-
control technology, which also uses the functions of anti-lock

49th ANNUAL PROCEEDINGS

ASSOCIATION FOR THE ADVANCEMENT OF AUTOMOTIVE MEDICINE
September 12-14, 2005
brakes and traction control. With VSC, a deceleration force and an
appropriate inward moment are produced to help prevent front-wheel-
skid and improve course-tracking performance. Outward moments
also may be applied to avoid rear-wheel-skid and maintain vehicle
posture.

Figure 1. Effect of Vehicle Stability Control (VSC) During

Oversteer and Understeer Scenarios

VSC technology evaluates an occupant’s steering input

compared to the true attitude of the vehicle. If differences are
detected, the stability control system will utilize the vehicle braking
system and engine power to compensate for these differences. In an
oversteer situation, control of the vehicle’s rear end is lost. In this
case, the VSC system would mainly apply the outside, front brake to
redirect the vehicle to its intended course. In an understeer scenario,
the front end of the car tends to slide out. To correct for this, the VSC
system will mainly apply the inside, rear brake to redirect the vehicle
to its intended course.

Similar systems have been introduced by other vehicle

manufacturers under names such as Electronic Stability Control
(ESC), Vehicle Dynamic Control (VDC), or AdvanceTrac.

Starting with the 1998 model year, certain vehicles were

equipped with VSC as standard equipment. Table 1 lists these vehicle
models and the year that the VSC was first implemented in each.
Other safety technologies and core vehicle structural attributes were
generally similar before and after the date of implementation of the
VSC systems. As such, the performance changes in these vehicles
under real-world driving conditions should be measurable when
comparing crash rates before and after the implementation of VSC
technology.

20
 Table 1. Toyota and Lexus Vehicle Models and Model Years
with and without VSC as Standard Equipment
Final Year VSC
Make and Model w/o VSC Standard
Passenger Cars
Lexus LS430/400 1997 1998
Lexus GS430/400/300 1997 1998
SUVs
Toyota Land Cruiser 1999 2000
Lexus LX470 1999 2000
Toyota 4Runner 2000 2001
Lexus RX300 2000 2001

Several previous studies directly relate to this work. In 1995,

Evans (1998) conducted a statistical review of state accident data to
assess the effectiveness of anti-lock braking systems. He proposed
methods for exposure control based on circumstances where the
technology should be most effective versus crashes where anti-lock
brakes (ABS) is not. For these analyses, dry weather crashes were
used to control for exposure whereas counts of wet weather frontal
crashes were assessed for vehicles with and without the technology.
Like the analysis of VSC technology presented here, the
Evans study of anti-lock braking included vehicles that were not
equipped with the technology in one year, followed by an introduction
as standard equipment in the next model year. Methods to correct for
the influence of vehicle age on crash occurrence and other
confounding factors were applied during the Evans study.
Regarding the need for active control systems, several
researchers have reported on the pre-impact characteristics of crashes.
Based on German insurance data, Langwieder (1999) reported that
25-30% of crashes involved some type of pre-impact skidding. In
67% of those cases, he noted that this skidding occurred over a 40-70
meter range (131-230 ft.). Sferco et al. (2001), using the European
Accident Causation Survey, determined that yaw-stability control
systems would have influence in 67% of fatal crashes and 42% of
injury accidents where “loss of vehicle control” was classified as the
cause of the crash. For all types of crashes, this corresponded to 34%
of fatal crashes and 18% of injury crashes.
Tingvall et al. (2003) studied the performance of ESP-
equipped vehicles in Sweden from 2000-2002, using analysis
techniques similar to Evans’ study on anti-lock braking systems.
Rear-end impacts in dry weather conditions were selected as the crash
scenario least sensitive to the ESP technology. This provided a
normalization criterion for crash exposure and calculation of ESP
effectiveness. The study did not account for the increased exposure of

21
older-model-year vehicles. The overall effectiveness of ESP was
calculated to be 22.1 ± 21%. On wet roads, the effectiveness was 31.5
± 23.4%, and in ice and snow, it was 38.2 ± 26.1%.

In 2004, Dang (2004) and the National Highway Traffic

Safety Administration (NHTSA) released an evaluation note
indicating that a 35% (95% CI: 29%, 41%) reduction in single-vehicle
crashes for passenger cars was found for ESC-equipped vehicles. A
67% (95% CI: 60%, 74%) reduction for SUVs was shown. These
reductions were established using US State Crash Data files. For fatal
crashes using FARS data, NHTSA reported a 30% reduction (95% CI:
10%, 50%) in passenger car single-vehicle crashes and a 63% (95%
CI: 44%, 81%) fatality reduction for SUVs.

A 2004 study by Farmer (2004) and the Insurance Institute for

Highway Safety (IIHS) identified significant reductions in single
vehicle and fatal crashes for vehicles equipped with stability control
technology. Using US state crash data, a 7% (95% CI: 3%, 10%)
reduction in overall crash involvement for stability-control-equipped
vehicles was reported. A 41% (95% CI: 33%, 48%) reduction in
single-vehicle crash risk was observed, and a 34% (95% CI: 21%,
45%) overall fatal crash risk reduction was found. The method used
vehicle registration counts as a control and accounted for vehicle age
to evaluate the effectiveness of stability control systems for a series of
vehicle makes and models.

Masami Aga and Akio Okada (2003) used Japanese field data
from the Institute for Traffic Accident Research and Data Analysis
(ITARDA) to investigate the performance of VSC for three Toyota
vehicles. Based on the data resources used, the study concluded that
VSC was most effective in reducing single-vehicle crashes (35%
reduction) and vehicle-to-vehicle frontal impacts (30% reduction). It
also indicated that more severe crashes, based on vehicle damage
extent, would experience an even greater reduction. This study was
based on 3 common passenger vehicle platforms in Japan including
980,000 vehicle years without VSC and 390,000 vehicle years with
VSC.

The analysis presented here expands on the work of Aga and

Okada to evaluate Toyota vehicles in a larger sample of crashes
occurring in the United States. The analysis described below also
benefits from a larger number of vehicle platforms and model years in
service for inclusion in the study. The goal of this study is to identify
reductions in certain crash types for VSC-equipped vehicles using
high-volume crash data.

22
METHODS
This study identifies the rate of crash involvement for the
subset of vehicles shown in Table 1. US state crash data were used to
evaluate crash involvement of the study vehicles by crash type.
Adjustments were made to account for influential factors including
vehicle exposure to crash involvement as well as the influence of
vehicle age on the likelihood of involvement for certain crash types.
Final crash odds are compared for platforms before and after the
addition of VSC technology as a standard feature.

DATA SOURCES – NHTSA’s State Crash Data files and the

Federal Highway Administration’s (FHWA’s) State Highway Safety
Information System (HSIS) were used for this study. The state data
files are a census of crashes whose severity exceeds that required to
file a police report. State files are compiled by each state DOT and
adapted for use by the NHTSA and the FHWA.

State crash files offer the largest census of crashes occurring

within a given region of the United States; however, data elements
and their definitions vary considerably between state files. Table 2
identifies states, an average annual vehicle involvement count, and
available data years that were used for this analysis. These states were
selected due to the availability of vehicle make, model, and model
year information. For each state listed in Table 2, with the exception
of Texas, vehicle identification numbers (VIN) were decoded to
identify appropriate vehicles for the study. Texas provides unique
vehicle codes and model year information within its crash files. These
codes were used to identify the crash-involved study vehicles of
interest for the Texas data.
Table 2. US State Crash File Availability and
Annual Vehicle Crash Records Collected
Available Vehicle Crash Records
State Data Years (annual average)
Florida 1998-2001 496,944
Illinois 2000-2003 746,995
Maryland 1998-2002 186,141
Missouri 1998-2001 356,868
Texas 1998-2001 587,400
Utah 1998-2002 99,339

During the analysis of state-collected crash data files, several

features prevent simple aggregation of data files across states. First,
the variable sets collected are not common across all states. As an
example, the definition of crash direction is classified according to

23
clock direction for Kansas, whereas Florida, Illinois, Maryland,
Missouri, and Utah categorize 18 potential damage regions for each
vehicle. New Mexico does not code an indicator of crash direction.
Texas codes a 3-digit vehicle damage type that contains indicators of
location of damage and degree of damage for each vehicle according
to the Vehicle Damage Scale for Crash Investigators by the National
Safety Council.

Although each state’s presentation of direction can be

classified into general vehicle body regions (i.e., front, left side, right
side, and rear), slight variations in the definitions of each could
influence crash proportions when compared across states.

A second influential characteristic of the state-collected data is

the varied reporting criteria for Police Accident Reports (PARs). For
example, Florida PARs include the following: a crash causing at least
one fatality, a crash causing a personal injury, a crash involving
damage to an attended vehicle, a crash causing damage to property
where a vehicle has left the scene, a crash where at least one driver is
under the influence of alcoholic beverages or chemical substances, or
a crash deemed severe enough to report by the investigating officer.
Police-reported crashes in Illinois must have property damage of $500
or more, a fatality, or a personal injury associated with the crash.
Maryland crashes reported by police must have at least one vehicle
involved in the crash towed away from the scene, a personal injury, or
a fatality associated with the crash. In Texas, a police report is
required for a crash resulting in at least $250 of property damage, a
personal injury, or a fatality associated with the crash.

Because rear-impact crashes do not often result in occupant

injury or very high levels of property damage, significant variations in
crash population could result due to only slight variations in state
crash reporting criteria.

For this analysis, crash rates have been calculated separately

by state to avoid possible inconsistencies due to varied sampling
criteria. Further, clustering of crashes within geographic regions with
common weather and roadway types may influence crash-
involvement rates as well. For these reasons, the state samples have
been analyzed independently; however, data files have been
aggregated across available years within each state for this analysis.
Care has been taken to ensure that variable definitions and case
collection criteria remain constant from year to year for each state
over the period evaluated.

24
 Table 3 shows registration counts for study vehicles by study
state. Registration counts were derived from the RL Polk National
Vehicle Population Profile (NVPP) dataset by state and partitioned
based on the presence of VSC as shown.
Table 3. Vehicle Registration and Crash Involvement
Counts for Study Vehicles

Study Vehicle Study Vehicle

Registration Count Crash Count*
(Calender Year 2001) (1998-2002)
State Pre VSC Post VSC Pre VSC Post VSC
Florida 52,302 55,752 3,356 1,181
Illinois 22,491 23,991 3,405 2,115
Maryland 12,864 12,048 1,320 479
Missouri 5,720 5,018 660 199
Texas 51,607 46,206 4,081 1,288
Utah 3,496 2,618 658 183
Total 155,383 150,828 13,987 5,671
* crash counts may include different data years per state
(see Table 2)

Study vehicle crash counts listed in Table 3 were compiled for

multiple years for each state used. Some variation exists in available
data years for each state so that high-volume crash states, including
Florida and Texas, may be somewhat underrepresented due to missing
datasets at the time of this analysis.

CRASH TYPES AFFECTED BY VEHICLE STABILITY

CONTROL – VSC systems act to correct conditions involving a loss
of driver control. These conditions may be brought on by excessive
steering inputs or adverse weather conditions. To monitor historical
crash data to recognize the influence of VSC systems, crash types that
are associated with loss of control have been isolated.

As reported by the NHTSA Fatality Analysis Reporting

System (FARS), loss-of-control fatal crashes are most often
associated with single-vehicle crashes, multivehicle frontal crashes.
Side-impact and rear-impact crashes involve a loss of control for the
struck vehicle infrequently, but loss of control by the striking vehicle
occurs often. This study will consider the influence of VSC on
multivehicle frontal crashes and single-vehicle crashes in all
directions.
CALCULATION OF CRASH ODDS – Crash odds were
calculated using the “induced exposure” method as applied by

25
Tingvall et al. (2003). Using this method, crash counts where VSC is
assumed to be influential are divided by a population of crashes where
the technology is assumed to have no effect. The unaffected group
acts as a population control.

As described above, VSC technology helps drivers to maintain

their intended course during difficult driving situations. For this
reason, VSC is assumed to be influential for single-vehicle crashes
andmultivehicle frontal crashes where loss of control is often a key
factor leading to a crash. Frontal crashes are defined as any crash where
frontal damage occurred in the study vehicle or the initial impact point was in
the vehicle front regardless of damage region in the struck vehicle (i.e.
frontal crashes were not limited to head-on collisions). Counts of these
crash types that may be positively affected by VSC technology are the
numerator of the crash odds calculation shown in Equation 1.

In the denominator of Equation 1, crashes that should not be

affected by VSC technology are used as a control group. This crash
count will indicate variations in exposure of a particular make, model,
and model year to potential crash events. Rear-impact crash counts
will serve as this proxy for the number of vehicles in service and the
frequency that these vehicles are driven (i.e., exposed to potential
crashes).

NF(VSC )
CrashOddsVSC = (1)
NR(VSC )

Where:

NF(vsc ) = Count of Multivehicle Frontal or Single-

Vehicle Crashes for post VSC vehicles
NR(vsc) = Count of Rear Crashes for post VSC-equipped
vehicles as struck vehicle

For rear-impact crashes (where the VSC-equipped vehicle is

the struck vehicle), the stability of the struck vehicle is assumed to
have little to no influence on the likelihood of being impacted. The
frequency of rear-impact involvement gives an indication of the
relative population of vehicles and conditions when the vehicle is
driven without being influenced by the technology in question.

For comparison with Equation 1, the odds ratio for crash

involvement may be calculated for vehicles without the technology in
the same way (see Equation 2).

26
 NF( preVSC )
CrashOdds preVSC = (2)
NR( preVSC )

Where:

NF(preVSC) = Count of Multivehicle Frontal or Single-

Vehicle Crashes for study vehicles pre VSC
NR(preVSC) = Count of Rear Crashes for study vehicles pre
VSC as struck vehicle

If the ratios calculated using Equations 1 and 2 are

significantly different than one another, where other conditions
besides the implemented technology remain the same, then
technology can be considered influential. If VSC is effective in aiding
drivers to keep in control of their vehicles and avoiding a particular
crash scenario, the overall magnitude of CrashOddsVSC should be less
than CrashOddspreVSC. The odds ratio that results when CrashOddsVSC
is divided by CrashOddspreVSC indicates the overall effect of the
technology in reducing a given crash type. If the technology is helpful
in reducing crash occurrence, the odds ratio calculated using Equation
3 will be less than 1.

CrashOdds(VSC )
OddsRatioVSC: preVSC = (3)
CrashOdds( preVSC )

INFLUENTIAL FACTORS – As noted by Evans during an

analysis of ABS systems, vehicle age (or model year effect) is known
to significantly affect crash involvement. In the 1998 study of anti-
lock brake effectiveness, Evans reports that vehicles only one year
older showed nearly 9% higher crash involvement than their younger
counterparts. A more recent study by Poindexter (2003) indicates that
these effects are much less pronounced based on NASS/GES data;
however, an increase in crash risk does exist.

For this study, it is important to account for the influence of

vehicle age on the ratio of the study conditions relative to rear-impact
crashes as described by Equations 1 and 2. If the age effect of
vehicles influences rates of crash involvement in the numerators or
denominators of these equations, then the effect must be accounted
for to accurately assess the contribution of VSC.

Figure 2 shows the relative percentage of frontal versus side

versus rear crashes within the six state data files in use for this study.
These trends indicate an increase in the proportion of frontal crashes

27
for the population of reported crashes versus vehicle age. Conversely,
the proportion of rear impacts reported appears to decrease with
vehicle age. Caution must be used in interpreting the trends shown in
Figure 2. It is possible that vehicles are involved in frontal crashes
more frequently as they age; however, this trend also may result due
to decreasing numbers of reported rear impacts as vehicles age. This
may be caused by vehicle attrition rate or a reduction in rear-impact
crash cases reported to police where a PAR was filed. The effect is
accounted for in the analysis presented here.

Vehicle Crash Case Percentages by Vehicle Age

(US State Data for 6 Study States)
60%

50%

Frontal
Percent of Vehicles

40% Side
Rear

30%

20%

10%

0%
1 2 3 4 5 6 7 8 9 10
Vehicle Age (years)

Figure 2. Vehicle Crash Ratios by Crash Mode Based on Vehicle Age.

To adjust for this effect of vehicle age on the ratio of frontal

versus rear-crash case counts, a second odds ratio may be calculated.
The ratio shown in Equation 4 identifies the odds of crash
involvement relative to rear impacts for a vehicle that is N years old.

NF( Nyrs )
CrashOdds Nyrs = (4)
NR( Nyrs )

Where:

NF(Nyrs) = Count of Multivehicle Frontal or Single-Vehicle

Crashes for vehicles N years old
NR(Nyrs) = Count of Rear Crashes for vehicles N years old

By dividing the odds of crash involvement for a vehicle that is

one year old by the ratio shown in Equation 4, the odds ratio of crash

28
occurrence for a vehicle N years old compared with a vehicle only
one year old is found (see Equation 5).

CrashOdds(1 yr )
OddsRatio1 yr:Nyrs = (5)
CrashOdds( Nyrs )

Finally, to account for the influence of vehicle age on the odds

of crash occurrence with and without VSC technology implemented,
crash odds are adjusted for each age group beyond one-year-old
vehicles. Each odds ratio calculated per Equation 3 is adjusted as
shown in Equation 6. This calculation was performed for study
vehicles by vehicle age per state reviewed. This age adjusted odds
ratio is the measure presented in the results section below.

CrashOdds(VSC )
CrashOdds( preVSC )
OddsRatio( Age)VSC: preVSC = (6)
CrashOdds(1 yr )
CrashOdds( Nyrs )

STANDARD ERROR ESTIMATES – The standard error of

the odds ratios are calculated as shown in Equation 7. This is the
standard error of the log odds ratio.

8
1
σ ln ( R ) = ∑n
i =1
(7)
i

Where ni = age adjusted crash count per VSC and crash

involvement category from Equations 1 and 2. Based on equation 7, it
can be seen that small counts of either the case populations
(multivehicle frontal or single vehicle) or control crash populations
(rear impacts) will significantly increase the standard error of the odds
ratios presented. All confidence limits presented below are calculated
for the 95% range.

RESULTS

In calculating odds for each crash type discussed above, data

were aggregated across multiple years for each state during
calculations. Each state’s crash data was partitioned by crash direction
for VSC-equipped and non-VSC-equipped vehicles. Also, single-
vehicle crash counts were tabulated for calculations below.

29
 CRASH ODDS – Table 4 shows the odds of crash
involvement for multivehicle frontal crash impacts. Each value
reported represents the crash odds where the percent reduction in
crash occurrence with VSC technology can be determined using the
following equation.

%reduction = 100 * (1 − OddsRatiovsc:preVSC ) (8)

As an example, the odds ratio for multivehicle frontal crash

involvement relative to rear impacts with VSC is .819 for Illinois as
shown in Table 4. This indicates an overall 18.1% reduction in crash
rate per Equation 8 for this state. The 95% confidence limits indicate
that, based on available information, one can be 95% certain that the
true rate reduction falls between the lower bound of 72.9% and the
upper bound of 92.1%. If this confidence interval spans 1.0, no
significant effect can be observed in the data regardless of the
reported odds ratio. Findings from Florida, Missouri, and Utah were
non-significant based on available data for multivehicle frontal
crashes relative to rear impacts.

As indicated previously, odds ratios based on data aggregated

across multiple states are not possible due to differences in inclusion
criteria and variable definitions. To summarize the odds ratios
reported across the multiple state files queried, a total odds ratio is
presented based on a weighted average of each state’s results. The
contribution of each state to this weighted average is proportional to
the number of study vehicles registered in each state. Confidence
intervals were calculated by considering each estimate as a simple
random sample. In doing so, the standard deviation of the individual
state estimates can be computed. From these standard deviations, the
95% confidence limit estimates were derived. This method, as
proposed by Kahane (1989), takes into account sampling error within
states as well as state to state differences in crash definitions. It
results in a very conservative estimate of the confidence limits. The
total reduction in crash rates for multivehicle frontal crashes is 11.8%
(95% CI: 2.4%, 21.1%) as shown in Table 4.

30
 Table 4. Multivehicle Frontal Crash Odds Relative to Rear Impact Crashes
Odds Ratio SE Lower Upper
State Front:Rear LN-OR 95% CI 95% CI
Florida 0.975 0.078 0.837 1.135
Illinois 0.819 0.059 0.729 0.921
Maryland 0.729 0.128 0.567 0.936
Missouri 1.048 0.183 0.731 1.501
Texas 0.885 0.079 0.758 0.973
Utah 0.902 0.191 0.621 1.311
Total 0.882 0.037 0.820 0.947

Table 5 indicates an average single-vehicle crash rate

reduction of 52.6% (95% CI: 42.5%, 62.7). This category includes
crashes where no other vehicles were involved during the impact.
This category of crashes was found to include the largest population
of injured occupants across all state data files. All odds ratios reported
by state show statistically significant reductions in single-vehicle
crash occurrence.
Table 5. Single-Vehicle Crash Odds Relative to Rear Impact Crashes
Odds Ratio SE Lower Upper
State Single:Rear LN-OR 95% CI 95% CI
Florida 0.554 0.145 0.417 0.736
Illinois 0.571 0.119 0.452 0.720
Maryland 0.251 0.217 0.164 0.384
Missouri 0.473 0.313 0.256 0.873
Texas 0.391 0.126 0.306 0.501
Utah 0.521 0.275 0.304 0.892
Total 0.474 0.065 0.417 0.538

DISCUSSION

The method used during this study to evaluate crash-reduction

rates due to VSC considers vehicle exposure and the influence of
vehicle age on the likelihood of crash occurrence. The method
considers all crash occurrences relative to rear-impact crashes. This
approach requires consistent reporting of rear impacts for each group
of crashes analyzed. For this reason, it is not possible to report
aggregate crash reductions across multiple state files. This point is
important during the analysis of state data files for VSC as well as
other topics of research.

A second considerable factor that was accounted for in this

analysis was the effect of vehicle age on reported frontal versus rear
impacts. The age adjustment presented here should account for this
effect. It should be noted that 84% of the vehicles analyzed during
this study were between 1 and 3 years old. For this reason, the effect
of age on reported crashes was not extremely large in this population;
therefore, an adjustment was made for this effect. Some Lexus sedans,

31
as early as model year 1996 (i.e., up to 5 years old), were present in
the pre-VSC-equipped crash populations.

The positive effects of VSC technology in reducing crash

occurrences for frontal multivehicle crashes and single-vehicle
crashes are significant and encouraging. However, the true benefit of
this effect must be considered in the context of property damage
reduction, injury reduction, and fatality reduction.

As previously presented by Aga and Okada (2003), VSC

technology will increase the limits where a driver will remain in
control of his vehicle. However, the technology will not prevent all
loss-of-control crashes across all severities from occurring.
Accordingly, a 52% reduction in single-vehicle crashes may not
directly translate to the same reduction in single-vehicle fatal crashes
or single-vehicle injury crashes. Although a significant number of
injury and fatal crashes will be avoided, a portion of these high-
severity loss-of-control events could be beyond the limits of dynamics
correctable by VSC systems. Further analysis of the nature of crashes
that are reduced by VSC technology is necessary and will be a topic
of future analyses.

The US State Crash Data provides a resource to evaluate high

volume police reported crash information. These datasets rely on
information observed by police officers on-scene and on verbal
accounts given by crash involved occupants. In some cases, crash
reports are completed hours after a crash scene has been cleared. The
completeness and accuracy of the data elements coded could be
questionable for these reasons. Further, police are not required to
perform a detailed crash investigation or reconstruction which may
also lead to missing information or incorrect assumptions made by
officers. It is understood and reasonable that the primary objective of
police is to ensure the safety of the public and to restore roadways to
normal operation as quickly as possible. Collection of data and
reporting may become secondary in some situations.

The results presented here are similar to findings by other

researchers who used different analytical techniques. Farmer (2004)
indicated a 41% reduction in single-vehicle crashes controlling for
exposure using vehicle registration data. NHTSA reported a 35%
reduction for passenger cars and a 67% reduction for SUVs. The
above analysis indicates a 52% reduction in single-vehicle crash rates
for VSC-equipped vehicles using rear-impact crash counts as a
control. Differences in the findings may be due, in part, to differences
in study vehicle populations. Also, the above studies included
different study states where varied weather and driving situations

32
could influence the relative number of times where stability was a
factor in the occurrence of crashes. Results from each analysis
indicate a quantifiable benefit in reducing crashes with these systems
in place.

CONCLUSIONS

This study identified the reduction in crash involvement for

vehicles equipped with VSC compared with crash involvement for the
same vehicle platforms before the technology was offered as standard
equipment. Rear-impact crashes were used as a control group to
account for vehicle exposure. This study also accounted for the
influence of vehicle age on the likelihood of involvement in certain
crash types.

Overall, it was shown that there is an 11.8% (95% CI: 2.4%,

21.1%) reduction in multivehicle frontal crashes for the VSC-
equipped vehicles included in the study population. A 52.6% (95%
CI: 42.5%, 62.7%) reduction in single-vehicle crash rates was shown
for these vehicles.

Single-vehicle or run-off-road crashes account for more than

37% of the traffic fatalities on US roadways. A subset of these 14,000
fatal crashes occurs due to a loss of vehicle control. Findings of this
study indicate that a significant portion of fatal and nonfatal single-
vehicle crashes may be prevented with widespread implementation of
stability control technology. Further analysis is required to fully
quantify the ratio of property damage only, injury crashes, and fatal
crashes that are reduced with the implementation of VSC.

ACKNOWLEGEMENTS

The authors wish to thank Toyota Motor Corporation for its support
for this project. Additionally, special thanks are due to the FHWA, the
NHTSA, and each state DOT that provided data and technical support
for the analyses used in this paper.

REFERENCES

Aga, M, Okada, A, Analysis of Vehicle Stability Control (VSC)’s

Effectiveness from Accident Data, Enhanced Safety of Vehicles
Conference, Paper #541, Nagoya, Japan; 2003.

Dang, JN. Preliminary Results Analyzing the Effectiveness of

Electronic Stability Control (ESC) Systems NHTSA Evaluation Note,
DOT HS 809-790; September 2004.

33
Evans, L. Antilock Brake Systems and Risk of Different Types of
Crashes in Traffic, Enhanced Safety of Vehicles Conference, Paper
#98-S2-O-12, Windsor, Canada; 1998.

Farmer, CM. Effect of electronic stability control on automobile crash

risk, Traffic Injury Prevention, 5:317–325, 2004.

Kahane, CJ. The Effectiveness of Center High Mounted Stop Lamps:

A Preliminary Evaluation. NHTSA Technical Report, DOT HS 807
076, 1987.

Langwieder, K., Characteristics of Car Accidents in the Pre-Crash

Phase, JSAE Spring Convention Proceedings, Paper #9932539; 1999.

Sferco, R, Page, Y, LeCoz, J, Fay, P. Potential Effectiveness of

Electronic Stability Programs (ESP) – What European Field Studies
Tell Us, ESV, Paper #2001-S2-O-327, Netherlands; 2001.

Tingvall, C, Krafft, M, Kullgren, A, Lie, A, The Effectiveness of ESP

(Electronic Stability Programme) in Reducing Real Life Accidents,
Enhanced Safety of Vehicles Conference, Paper #261, Nagoya, Japan;
2003

34