CHAPTER
5. CVISN SAFETY
BENEFITS
In 1998[1],
5,374 people were killed and approximately 127,000 were injured in crashes
involving approximately 412,000 large commercial motor vehicles (CMVs). The FMCSA has set as one of its primary
objectives the reduction of CMV-related fatalities and injuries by
50 percent by 2010. Although
new research, such as the Large Truck Crash Causation project (FMCSA 2001), will
help FMCSA better understand the causes of these crashes, vehicle safety defects
and driver violations of the Federal Motor Carrier Safety Regulations (FMCSRs)
are known to contribute to a portion of them (VNTSC
1999a).
The most important benefit expected from the deployment of CVISN
technologies, especially electronic screening and safety information exchange,
is a reduction in CMV‑related crashes through improved enforcement of the
FMCSRs. The principal hypothesis to
be tested is that CVISN technologies will help enforcement staff focus
inspection resources on high‑risk carriers. This will result in more out-of-service
(OOS) orders for the same number of inspections—thereby removing from service
additional trucks and drivers that would have caused crashes because of vehicle
defects and driver violations of safety regulations. A second hypothesis is that the
increased attention on high-risk carriers will encourage motor carriers to
improve their compliance with safety regulations. This indirect benefit is the number of
crashes that would have been caused by violations in safety regulations, but are
avoided due to improved compliance.
As outlined in the CVISN MDI Summary Evaluation Plan (Battelle 1998), the
safety benefits analysis addresses the following four
questions:
·
What is the impact of CVISN
on the numbers of crashes, injuries, and fatalities involving large
CMVs?
·
What is the impact of CVISN
on rates of driver and carrier compliance with the FMCSR?
·
To what extent does CVISN
help roadside safety enforcement officials identify high‑risk commercial
vehicles and motor carriers?
·
To what extent does CVISN
help roadside safety enforcement officials identify OOS
violators?
The CVISN safety benefits analysis was performed using a probability
model that predicts the number of crashes avoided under various scenarios. Each scenario is defined by specific
assumptions concerning the future deployment of CVISN. The probability model relates the number
of crashes avoided to several input parameters including the probability that a
CMV has an out‑of‑service (OOS) condition, the number of inspections performed,
historical rates at which OOS orders were issued, national crash/injury/fatality
rates involving large trucks, and probabilities that certain OOS conditions will
contribute to a crash. Estimates of
these inputs were obtained from the literature or from data collected in several
special studies conducted in states that had previously deployed—or were in the
process of deploying—CVISN safety information exchange and electronic screening
technologies. States that
participated in these studies were Connecticut, Kentucky, and
Oregon.
Section 5.1 contains an overview of our findings. Five scenarios are presented to
illustrate the safety benefits of CVISN under different deployment options and
assumptions concerning the potential outcomes. The technical approach discussion in
Section 5.2 describes the probability model and summarizes the design of the
special studies that were undertaken to obtain outcome measures used in the
model. Section 5.3 shows the
calculation of safety benefits for each scenario. Supporting analyses are presented in
Section 5.4.
Limitation of
Findings
The analysis contained in this chapter uses a probability model to
predict the number of truck-related crashes that would be avoided nationwide as
CVISN deployment expands. Although
the model can be justified by basic principles of probability, its application
relies on a variety of input parameters used to estimate impacts and benefits of
CVISN. Some of the parameters were
estimated using results from the open literature on crashes and highway
statistics, and others were estimated with data collected in special studies
involving participating CVISN states.
Both types of estimates are subject to errors of unknown
magnitude.
Some of the literature results were derived from a related FMCSA program
(VNTSC 1999a), which was reviewed by an expert panel (Nicholson 1998). The panel expressed concern that the
estimates of crash causation probabilities were based on limited data. However, no alternative approach was
recommended. Currently, FMCSA is in
the process of developing a new data collection program that has the potential
to fill this information gap.
Estimates of CVISN impacts and benefits obtained from the special CVISN
studies should also be used with caution.
There were few if any opportunities to replicate the studies in different
states in order to determine the statistical uncertainty of the estimates. In some cases the data limitations were
the result of CVISN deployment delays or reduced levels of deployment of
specific technologies.
Although additional data are needed to support these results, the safety
analysis presented in this chapter helps to illustrate how the deployment of
CVISN can affect highway safety.
The analysis can be easily modified as new data become
available.
5.1
Overview of
Results
This section contains an overview of the results of our safety benefits
analysis. Results were obtained by
combining analyses of selected roadside enforcement data with literature results
to project the benefits of these technologies under five CVISN roadside
enforcement (RE) deployment scenarios that incorporate the anticipated impacts,
and a baseline scenario representing current enforcement practices. The deployment scenarios are defined as
follows:
RE-0: Baseline—Pre-CVISN. Enforcement officers (inspectors) select
CMVs for inspection using personal experience and judgment, but without the aid
CVISN technologies.
RE-1: ISS with Manual
Pre-screening. Inspectors are equipped with laptop
computers containing Aspen and ISS.
CMVs are pre-screened based on weigh-in-motion (WIM) and/or visual
screening on a sorter ramp. Of the
CMVs directed to the fixed scale, officers use ISS to select vehicles for
inspection.
RE-2: ISS with Electronic
Screening. State deploys electronic screening with
safety snapshots at all major inspection sites. Motor carriers classified as “low-risk,”
based on SafeStat scores, enroll in the electronic screening program. Trucks from the low-risk carriers
(comprising approximately 52 percent of trucks on the road) are equipped with
transponders and allowed to bypass inspection sites. Inspectors use ISS in the manner
described in RE-1 to select vehicles for inspections from the remaining 48
percent of trucks in the high, medium, or unknown/ insufficient data risk
categories.
RE-3: ISS with Electronic
Screening and a Reduction in OOS Conditions Due to Improved FMCSR Compliance by
Motor Carriers. Motor carriers respond to targeted
enforcement by improving compliance with safety regulations. Specifically, we assume that the total
number of vehicle and driver OOS conditions will decrease by 25 percent due to
improved compliance. Enforcement is
conducted as in RE-2. (25 percent
is an assumed value to illustrate potential impacts of improved compliance on
crash reductions. The sensitivity
of this assumption is assessed by also performing the analysis with an assumed
reduction of 10 percent – referred to as scenario RE-3*. At this time, there is no statistical
evidence that targeted enforcement will have such effects on safety violation
rates.)
Deployment RE-1 represents a current application of ISS for vehicle
selection (e.g., Connecticut’s Greenwich and Union weigh stations). Scenarios RE-2 and RE‑3 (or RE-3*)
represent “feasible” situations that could occur as CVISN deployment
expands. Benefit/cost analyses for
these three scenarios are presented in Chapter 8. However, to illustrate the limits of the
direct and indirect benefits of CVISN roadside enforcement technologies, we
consider two additional hypothetical scenarios:
RE-4: 100 Percent Inspection Selection
Efficiency. State deploys electronic screening with
safety snapshots at all major inspection sites, and all trucks are equipped with
transponders. Safety analysis and
screening algorithms have progressed to the point that OOS violations can be
identified with near certainty.
Therefore, all inspections result in OOS orders. Although not expected to occur, this
scenario is used to illustrate the direct benefits of maximizing the efficiency
of roadside enforcement operations.
RE-5: 100 Percent Compliance with Safety
Regulations. In response to targeted enforcement,
violations of vehicle and driver safety regulations are eliminated. Although not expected to occur, this
scenario represents the maximum possible benefit (direct and indirect) of
improved enforcement.
Except as described in scenarios RE-3 and RE-5, the rates of compliance
with vehicle and driver safety regulations are based on estimates obtained in
the FMCSA’s National Fleet Safety Survey, or NFSS (Star 1997). The NFSS collected data from over 10,000
random Level I (driver and vehicle) inspections and estimated that 29
percent of all commercial vehicles and 5 percent of commercial vehicle
drivers were operating with OOS conditions. For all of these scenarios, it is
assumed that the numbers and types of inspections performed annually are
constant and equal to the numbers reported by FMCSA’s Motor Carrier Safety
Assistance Program (MCSAP) Quarterly Report Information System for Fiscal Year
1998 (FMCSA 1999)—the last year for which complete crash statistics are
available. Additional information
from the literature, such as annual crash/injury/fatality rates, numbers of
inspections performed and OOS orders issued, and crash causation statistics are
discussed in Sections 5.2 and 5.3, along with the methods and analyses used
to determine the crash reduction benefits of CVISN.
The impacts of CVISN technologies on roadside enforcement operations were
evaluated through special studies conducted in participating states. However, because CVISN is still in the
early stages of deployment, especially in the area of electronic screening with
safety snapshots, opportunities to evaluate these impacts directly were
limited. The following results,
obtained from CVISN pilot states, provide useful insight into these effects;
however, the degree to which these results are statistically representative of
future deployments could not be determined:
1. A
study of roadside inspection selection strategies at four Connecticut inspection
sites (two using ISS and two without ISS) demonstrated that using ISS, in
combination with manual prescreening, to select commercial vehicles for
inspection increases OOS orders by approximately 2 percent for the same number
of inspections—a 2 percent increase in inspection
efficiency.
2.
Analysis of this same inspection selection strategy under the added
assumption that “low-risk” carriers would be permitted to bypass the inspection
sites demonstrates that electronic screening, with full participation by all
low‑risk carriers, could increase inspection efficiency by more than 11
percent.
3. A
two-year study of the changes in safety compliance rates in Oregon, conducted
during the deployment of roadside screening and safety information exchange
technologies, failed to demonstrate that CVISN roadside deployment will increase
compliance with safety regulations.
However, in designing the study it was anticipated that advanced safety
screening technologies would be deployed during the second year; but deployment
of these systems was delayed, which made it difficult to observe the expected
impact on safety compliance. As
discussed in scenario RE-3, the 25 percent reduction (or 10 percent for scenario
RE-3*) in safety violations is assumed for illustration
purposes.
These estimated and assumed effects of CVISN deployment, along with
results from the literature, were applied to a crash avoidance model (described
in Section 5.2) to predict the numbers of truck-related crashes and associated
injuries and fatalities that would be avoided under each of the above roadside
enforcement scenarios. Table 5-1
summarizes the major results of this analysis. Supporting details are provided in
Section 5.3.
Table 5-1. Estimated
Safety Benefits of CVISN Under Selected Deployment Scenarios and
Assumptions
|
Scenario |
Description |
Numbers of Safety Events
Avoided1 |
Additional2 Safety
Events Avoided (CVISN Benefit) | ||||
|
Crashes |
Injuries |
Fatalities |
Crashes |
Injuries |
Fatalities | ||
|
|
Random
Selection |
3,765 |
1,160 |
49 |
|
|
|
|
RE-0 |
Baseline
(pre-CVISN) |
4,423 |
1,362 |
57 |
|
|
|
|
RE-1 |
ISS
with manual prescreening |
4,507 |
1,388 |
59 |
84
|
26 |
2 |
|
RE-2 |
RE-1 plus electronic bypass of
low-risk carriers |
5,012 |
1,544 |
65 |
589 |
181 |
8 |
|
RE-3 |
RE-2 plus 25% reduction in
safety violations |
14,368 |
4,425 |
187 |
9,945 |
3,063 |
130 |
|
RE-3* |
Same as RE-3 except with a 10%
reduction in safety violations |
8,755 |
2,697 |
114 |
4,332 |
1,335 |
57 |
|
RE-4 |
100% inspection selection
efficiency |
10,561 |
3,253 |
137 |
6,138 |
1,891 |
80 |
|
RE-5 |
100% compliance with safety
regulations |
42,436 |
13,070 |
552 |
38,013 |
11,708 |
494 |
1 In 1998, approximately
412,000 large trucks were involved in crashes resulting in 127,000 injuries and
5,374 fatalities. The estimated number of crashes avoided is based the
assumption that crashes are avoided when vehicles and drivers with safety
violations are placed out-of-service.
2 Compared to baseline scenario
(RE-0)
According to the model,
current roadside enforcement strategies (RE-0) are responsible for avoiding
4,423 truck-related crashes, which represents slightly more than 1 percent of
the 412,000 truck-related crashes that occur annually, based on 1998 crash
statistics (FMCSA 2000b).
Assuming that the numbers of injuries and fatalities are proportional to
the number of crashes, it is estimated that current roadside enforcement
activities are responsible for preventing 1,365 injuries and 57
deaths.
For reference, the numbers of crashes, injuries, and fatalities that
would be avoided if vehicles were randomly selected for inspection
(Random Selection) were also calculated and shown in Table 5-1. The differences between these numbers
and the baseline numbers can be used to estimate the benefits of current
inspection selection strategies, which include the training, knowledge, and
experience that the inspectors bring to the job.
The safety benefits of CVISN
are obtained by subtracting the numbers of crashes, injuries, and fatalities
avoided under the baseline scenario from the corresponding numbers under
scenarios RE-1 to RE-5. For
example, if ISS were used to select vehicles for inspection following manual
pre-screening on sorter lanes, as currently performed at two sites in
Connecticut, an additional 84 crashes, 26 injuries, and two fatalities could be
avoided. If electronic screening is
added, and all low-risk carriers enroll and are permitted to bypass inspections,
enforcement staff could focus inspections on more high-risk carriers. The increased numbers of OOS orders for
the same number of inspections would help avoid 589 additional truck‑related
crashes as well as 181 injuries and 8 fatalities. Although there is no direct evidence
concerning the degree to which safety compliance improves with enhanced
enforcement, scenarios RE-3 and RE-3* demonstrate the substantial safety
benefits that would occur if CVISN‑enhanced enforcement strategies helped to
encourage improved compliance with safety regulations.
Scenarios RE-4 and RE-5 are presented to illustrate the potential benefit
of CVISN under limiting conditions at full deployment. Clearly it is not realistic to expect
roadside enforcement to achieve 100 percent efficiency in selecting
vehicles with OOS conditions.
However, it is conceivable that advances in safety analysis, through
programs such as the Large Truck Crash Causation Study, combined with enhanced
data reporting capabilities offered by CVISN, would help to identify carriers
that might pose exceptionally high risks.
The additional 6,138 crashes avoided under scenario RE-4 represent the
maximum benefit of this enhanced enforcement capability. Of course, the maximum number of crashes
that could be avoided with any improvement related to safety compliance is
42,436, of which only 38,013 would be attributable to the indirect benefit of
CVISN or any other program that helps to increase compliance with safety
regulations. The resulting
reduction in injuries (11,708) and fatalities (494) represent 9 percent of the
numbers that occurred in 1998.
Thus, the analysis of this limiting condition demonstrates the maximum
degree to which CVISN can contribute to FMCSA’s goals of reducing the number of
injuries and fatalities from truck‑related crashes by 50 percent by the
year 2010.
The model used to calculate the number of crashes avoided is illustrated
in Figure 5‑1. The direct
effect of improved inspection efficiency (OOS orders
per 100 inspections) is represented by the contour lines, which
determine the number of crashes avoided for a given level of compliance. As inspection efficiency increases, the
number of crashes avoided increases up to a limit. Recall that scenario RE-4 produces the
maximum number of crashes avoided without increasing safety compliance. Note that the lower (realistic) limit on
inspection efficiency corresponds to random selection (percent OOS orders equals
the violation rate). The successive
lines, from bottom to top, represent the indirect impact of reducing the safety
violation rates by 25 percent, 50 percent, 75 percent, and 100 percent. The results (total crashes avoided) for
the scenarios described above are marked in Figure 5-1.
Figure 5-1: Number of Crashes Avoided versus Inspection
Selection Efficiency at Selected Levels of Reduction in Vehicle/Driver Safety
Regulation Violation Rates (Showing Approximate Locations of Estimates for
Selected Scenarios)
5.2
Technical
Approach
The CVISN safety benefits estimation methodology is based on a
probability model that relates the improvement in safety—as measured by the
numbers of crashes, injuries, and fatalities avoided—to the number of OOS orders
issued and other CMV safety parameters such as crash rates, violation rates, and
crash causation statistics. Data
were collected from published sources as well as new, special studies designed
to provide input to the model.
The four study questions presented on page 5-1 provide the basis for the
safety benefits estimation methodology.
While the first of these questions addresses the heart of the safety
benefits of CVISN, the particular CVISN technologies that are included in the
MDI achieve these safety benefits only through improvements in the enforcement
of vehicle and driver compliance with safety regulations. The remaining three questions, which
provide the main focus for the safety evaluation, examine the relationship
between CVISN deployment and its impact on enforcement practices. Results from the literature and new
studies were used to address the latter three questions. Once those questions were answered, a
statistical model was used to relate their answers to the question of the effect
of CVISN on crashes, injuries, and fatalities.
The second and third questions indicate that CVISN technologies might be
expected to help improve compliance with safety regulations in two ways, both
resulting from increased effectiveness of roadside inspection operations. The direct, but smaller, impact
is the removal of unsafe drivers and vehicles from the highways. The third question addresses the direct
effect by examining the impact that electronic screening and safety information
exchange technologies will have on the inspectors’ ability to select commercial
vehicles for inspection in the most efficient manner. The indirect effect, which is
expected to be much larger, occurs when drivers and carriers modify their
behavior in response to the improved, more targeted inspections. The second question addresses this
indirect effect. The hypothesis to
be tested is that carriers will expend resources to ensure that their vehicles
stay in compliance. Carriers with
good safety records (low-risk carriers) would expect to have a small probability
of being inspected. High‑risk
carriers will try to improve their safety rating to avoid increased
inspections.
The fourth question focuses on a different aspect of safety enforcement,
that of identifying violators of OOS orders. Safer Data Mailbox (SDM) is an
electronic database of inspection records that has been designed to provide
safety enforcement agencies with a national database of inspection
information. Queries to this
database provide information about any inspection of a vehicle within the past
45-day period. The extent to which
it is used by inspectors in the field will indicate the extent to which OOS
violators can be identified.
The following sections describe (1) the sources of data obtained from the
literature and from special studies that were conducted to quantify the impacts
of CVISN on roadside safety enforcement, and (2) the crash avoidance model used
to estimate CVISN safety benefits.
Data
Sources
Table 5-2 lists some key safety statistics obtained from the published
literature. Most of these data are
used in the crash avoidance analysis; others are provided for reference. According to FMCSA, 7.2 million large
trucks (>10,000 pounds gross vehicle weight) travel approximately
196 billion miles in the U.S. each year. In 1998, the last year for which
complete statistics are available, 412,000 trucks were involved in crashes,
resulting in approximately 127,000 injuries and 5,374 deaths. The corresponding rates per vehicle mile
traveled are derived from these values.
Other relevant statistics provided in Table 5-2 include the number
of commercial vehicle inspections performed in 1998 and the actual percentages
of OOS orders issued (25.5 percent of vehicles and 8.1 percent of
drivers). In 1996 FMCSA sponsored
the National Fleet Safety Survey (NFSS), in which 10,000 trucks were selected at
random for inspection in order to estimate the percentages of trucks and drivers
that operate with OOS conditions (i.e., violation rates). These estimates differ from the actual
OOS rates because inspectors choose vehicles for inspection based on vehicle
appearance and apply their knowledge and experience. The estimated violation rates reported
by the NFSS were 29 percent for vehicles and 5 percent for drivers
(Star 1997).
In order to determine the
impact of removing OOS violators from the roadway on the number of crashes, it
is necessary to estimate certain probabilities associated with crash
causation. Specifically, the
probabilities that a crash involving a large truck was caused by vehicle and
driver OOS conditions are needed.
FMCSA recognizes that information on the causes of large truck crashes is
lacking and, therefore, recently initiated the Large Truck Crash Causation Study
(2001), which will soon begin collecting the necessary data. In the meantime, this analysis uses the
best available information, which FMCSA used to evaluate the safety benefits of
roadside inspections (VNTSC 1999a).
These data issues are discussed more fully in the next section along with
the explanation of the crash avoidance model.
Table
5-2. Relevant Safety and
Safety Enforcement Statistics on Large Trucks
|
Statistic
Description |
Value |
Source
1 |
|
Number
of large trucks |
7.2
million |
Safety
Action Plan (FMCSA 2000a) |
|
Large
truck annual vehicle miles traveled (VMT) |
196
billion |
Safety
Action Plan (FMCSA 2000a) |
|
Number
of registered interstate motor carriers |
500,000 |
Safety
Action Plan (FMCSA 2000a) |
|
Number
of large trucks participating in electronic screening
(2000) |
150,000 |
What
Have We Learned (FHWA 2000) |
|
Large
trucks involved in crashes (1998) Injuries
from large truck crashes (1998) Fatalities
from large truck crashes (1998) |
412,000 127,000 5,374 |
1998
Crash Profile (FMCSA 2000b) |
|
Large
trucks involved in property damage-only crashes Large
trucks involved in injury-only crashes Large
trucks involved in fatal crashes |
318,000 89,000 4,935 |
1998
Crash Profile (FMCSA 2000b) |
|
Large
truck crash rate (truck crashes/100 million VMT) = 412,000 truck crashes/196 billion
VMT |
210.2 |
Derived |
|
Large
CMV injury crash rate (crashes/100 million VMT) =89,000 truck crashes/196 billion
VMT |
45.4 |
Derived |
|
Large
CMV fatal crash rate (crashes/100 million VMT) =4,935 truck crashes/196 billion
VMT |
2.5 |
Derived |
|
Commercial
vehicle (non-bus) inspections performed (1998) Commercial
vehicle (non-bus) driver inspections (1998) Total
CV (non-bus) inspections (driver and vehicle)
(1998) |
1,562,739 2,089,846 2,113,570 |
MCSAP
FY98 Data Report (FMCSA
1999) |
|
Percent
of vehicles placed OOS (1998) Percent
of drivers placed OOS (1998) Percent
of vehicles or drivers placed OOS (1998)
[estimated] |
25.5% 8.1% 30.4% |
MCSAP
FY98 Data Report (FMCSA
1999) |
|
Percent
of VMT with vehicle OOS conditions (1996) Percent
of VMT with brake-related OOS conditions (1996) Percent
of VMT with driver OOS conditions (1996) Percent
of VMT with vehicle or driver OOS conditions
(1996) |
29% 14% 5% 32% |
1996
National Survey (Star 1997) |
|
Percent
of large CMV crashes with vehicle OOS condition as contributing
cause |
4.6% |
Safety
Program Performance (VNTSC 1999a); OOS Criteria (Miller et al.
1996) |
|
Percent
of large CMV crashes with driver OOS condition as contributing
cause |
5.7% |
Safety
Program Performance (VNTSC
1999a) |
1 Full reference
citations are presented in Section 5.5.
While these
data provide much of the necessary information needed to estimate safety
benefits, additional data are needed to determine the current impact of CVISN
technologies on inspection efficiencies and compliance rates. Thus, three studies were conducted in
Oregon, Connecticut, and Kentucky to collect the necessary data for evaluating
the impact of CVISN deployment.
In Oregon, a two-part study was conducted. The first part of the study examined the impact of CVISN on carrier and driver compliance with the FMCSR. In the compliance study, conducted in conjunction with Oregon’s Green Light project, trucks were randomly selected for inspection using a statistical sampling plan. Approximately 1,200 vehicles were inspected at several locations in Northwest Oregon at four times spread over a period of 2 years. These data were used to estimate the change in violation rates over time as CVISN was deployed in the state and to provide information used in addressing the indirect effect of CVISN deployment currently.
The goal of the second part of the Oregon study was to quantify how
effectively roadside enforcement staff were able to target vehicles from
high-risk carriers with and without the CVISN technologies. The Oregon Department of Transportation
(ODOT) performed 508 inspections between June and September 1999 at
various locations in the Interstate 5 corridor in Oregon, some using ISS
and others without using ISS. The
proportions of high‑risk vehicles inspected were compared with and without ISS
to determine whether there were any differences. These results were used to examine the
direct effects of CVISN deployment currently. Appendix A.1 contains a detailed
discussion of the two Oregon studies.
The Connecticut Roadside Screening Study was conducted to estimate
the effectiveness of the CVISN safety information exchange deployment in
Connecticut, which consisted of ASPEN/ISS systems accessed from laptop
computers. The inspection operations of two
agencies, the Department of Motor Vehicles (DMV) and the Department of Public
Safety (DPS), were observed at four different weigh stations in the winter and
spring of 1999. Data were collected
from more than 10,000 vehicles entering these stations to characterize the
distribution of trucks at each location and to evaluate the inspection selection
process.
Following the roadside data collection, the motor carrier safety ratings
for every truck observed at the Connecticut sites were determined using the
SafeStat algorithm (VNTSC 1999b; VNTSC 1998). In addition, over 58,000 historical
records from inspections conducted between October 1995 and June 1999
were used to determine the distribution of inspected CMVs among risk
categories. As with the Oregon
inspection study, the proportion of high-risk CMVs inspected was estimated and
compared to the proportion of high-risk CMVs in the population to determine the
inspection efficiency conducted with laptops and ASPEN. These data were also used to estimate
the effects of using ISS in combination with manual pre-screening on the number
of OOS orders issued for a fixed number of inspections performed. Appendix A.2 provides a more detailed
discussion of the Connecticut Roadside Screening Study.
The Kentucky Screening Assessment Study was conducted to measure
changes in screening effectiveness at sites in Kentucky where CVISN's Electronic
Screening technology was deployed.
Of particular interest in the Kentucky study was to compare the
inspection efficiency between stations with and without electronic transponder
facilities that allow participating CMVs to bypass inspection stations. U.S. DOT identification numbers for over
150,000 CMVs that entered or bypassed one of five inspection stations
during a 5‑week period in September and October 1999 were
recorded. Risk categories were
determined from these vehicles to represent the distribution of CMVs among risk
categories for the population passing each station. Inspection records for 1998 at the same
stations were obtained and used to determine the distribution of CMVs among risk
categories for inspected vehicles.
The proportion of high-risk vehicles inspected at each inspection station
was compared to the proportion of high-risk vehicles in the population to
estimate the inspection efficiency.
Appendix A.3 provides more details about the Kentucky Screening
Assessment Study.
To address the fourth research question, regarding inspectors’ ability to
identify OOS order violators, a study was conducted of the SAFER Data Mailbox
(SDM) system. This study
examined the frequency and timeliness of inspection uploads and queries to SDM
to evaluate the potential for using SDM to catch OOS order violators. A separate DOT report provides a more
detailed description of the SAFER Data Mailbox study (Battelle
2000).
CVISN
Crash Avoidance Model
Ultimately, safety benefits will be realized only to the extent that
targeted inspections and improved compliance translate into reductions in
numbers of crashes. The premise of
targeted inspections is that, for the same number of inspections performed,
additional drivers and vehicles operating with OOS conditions will be removed
from the roadway. Furthermore, all
of the conditions leading to the OOS order will be fixed and “stay fixed” for a
period of time after the inspection.
Therefore, crashes that would have occurred during this period are
prevented because the OOS conditions that would have caused the crashes were
eliminated. The safety benefit of
CVISN is determined by comparing the number of crashes avoided under the
baseline scenario (i.e., with pre-CVISN roadside enforcement strategies and
technology) with the number of crashes avoided under each CVISN deployment
scenario. It is assumed under each
scenario that the corresponding numbers of injuries and fatalities avoided are
proportional to the number of crashes avoided.
The basic principle of the CVISN crash avoidance model, as well as
certain assumptions about how roadside enforcement affects crash rates, were
motivated by research on the Safe‑Miles model developed for FMCSA (formerly the
Office of Motor Carriers in the Federal Highway Administration) to estimate the
benefits of MCSAP, the Motor Carrier Safety Assistance Program (VNTSC
1999a). Although the model used in
the CVISN analysis is different from the one used in Safe-Miles, certain model
parameters, such as crash causation probabilities and the number of “safe miles”
a truck travels following an OOS order, are used in this analysis. It should be noted that the developers
of Safe-Miles (VNTSC 1999a), as well as an expert panel convened to review the
program (Nicholson 1998), identified certain limitations with the Safe-Miles
model. Some of their concerns are
relevant to the CVISN crash avoidance model, as discussed
below.
In its simplest terms, the number of crashes avoided can be written
as 
(1)
where
·
#OOSO is the number of OOS orders
issued, and
·
P(C,D|OOSC) is the probability of a
crash (C) with a contributing defect or driver safety violation (D), given that
a vehicle has the OOS condition (OOSC).
While the number of OOS orders issued is easily obtained, the probability
of a crash with a contributing defect that would have resulted in an OOS
condition is more complicated. We
start by representing the second term in (1) as a product of conditional
probabilities, so that the model for the number of crashes avoided can be
rewritten as
where
·
P(C|OOSC) is the probability of a
crash given that a vehicle has an OOS condition, and
·
P(D|C,OOSC) is the probability of a
contributing defect given that a vehicle is involved in a crash and has an OOS
condition.
Using Bayes Theorem, the middle term in Equation (2) can be rewritten
as
where
·
P(OOSC|C) is the probability that a
vehicle has an OOS condition given it is in a crash,
·
P(C) is the probability of a
crash, and
·
P(OOSC) is the probability that a
vehicle has an OOS condition.
Similarly, the last term can
be rewritten as
where
·
P(D|C) is the probability of a
contributing defect given that there was a crash, and
·
P(OOSC|D,C) is the probability that a
vehicle has an OOS condition given it has a crash with a contributing
defect.
The
last term, P(OOSC|D,C), is equal to 1 because we are assuming that the
vehicle defect or driver violation (D) is an OOS
condition.
In this analysis, we are only concerned with crashes that are avoided
because they would have been caused by a defect or driver violation that
resulted in an OOS order. Also, it
is generally assumed that the probability of a crash is proportional to the
number of vehicle miles traveled (VMT).
Therefore, the probability of a crash (among vehicles that would have
been operating with defects or driver violations) is estimated by the national
crash rate for large trucks (denoted by 8) multiplied by the number
of safe miles (SM) traveled as a result of “fixing” an OOS condition. This is the approach used in the
Safe-Miles program. The values of
SM used in the Safe-Miles program are 15,000 miles for vehicle OOS orders and
10,000 miles for driver OOS orders.
It should be noted that the expert panel reviewing the Safe-Miles program
was uncomfortable with these assumptions; but no alternative approach was
identified. The CVISN evaluation
team looked at an alternative approach to representing crash probabilities
following an inspection. It was
determined that the “safe miles” model was conceptually consistent with a more
rigorous approach that does not assume a fixed number of miles without OOS
conditions. However, both
approaches require data that currently do not exist. Therefore, it is recognized that this
portion of crash avoidance model should be updated as new information becomes
available from the Large Truck Crash Causation Project.
Combining Equations (2), (3), and (4) yields the following model for
crashes avoided:
Equation (5) is used in Section 5.3 to estimate the safety benefits
associated with various CVISN deployment scenarios. Under each scenario, 8, the national crash rate
for trucks, is 412,000 truck crashes divided by 196 billion vehicle miles
traveled (VMT), or 2.1 crashes per million miles traveled. Applying the same crash causation
probability estimates used in the Safe‑Miles program, we have P(D|C) is
equal to 0.046 for vehicle OOS conditions and 0.057 for driver OOS
conditions. The expert panel had
concerns about the accuracy of these estimates, so it is noted that these
estimates should also be updated as new information becomes available from the
Large Truck Crash Causation Project.
Additional data needed for this model include #OOSO, the number of OOS
orders issued nationally, and P(OOSC), the probability that a vehicle
will have an OOS condition. These
values depend on the particular roadside deployment scenario or enforcement
strategy under consideration.
5.3
Estimation
of CVISN Safety Benefits
In this section we present the calculations of the numbers of truck crashes, injuries, and fatalities avoided under each of the roadside enforcement scenarios described in Section 5.1. These calculations are based on Equation (5) and utilize specific assumptions defined by the scenarios. Results from special studies are presented as needed to justify some of the parameter estimates used in these models.
We begin by calculating the number of crashes that would be avoided were trucks to be selected for inspection randomly. This is not one of the roadside enforcement strategies being considered, nor is it a realistic strategy to employ. However, the calculation is useful for determining the contribution of the inspectors’ knowledge and experience during the vehicle selection process.
Under random inspections, the proportions of inspected vehicles and drivers that are given OOS orders are equal to corresponding FMCSR violation rates. Thus, by applying the results from the NFSS, 29 percent of the 1,562,739 vehicle inspections (453,194) would result in vehicle OOS orders (Star 1997). From Equation (5), the number of crashes that are avoided due to vehicle OOS orders when random inspections are performed is equal to
Similarly, 5 percent of the 2,089,846 driver inspections (104,492) would have resulted in driver OOS order leading to
crashes
avoided. Note that these two
numbers cannot be added to get the total number of crashes avoided because there
is some overlap in vehicle and driver OOS orders. To get an estimate of the total number
of crashes avoided, Table 5-2 shows that 29 percent of inpsections results in a
vehicle OOS order, 5 percent of inspections result in a driver OOS order, and 32
percent of all inspections results in an OOS order. Thus, 2 percent of inspections result in
both a driver and vehicle OOS order.
Equivalently, in 40 percent of the inspections where there is a driver
OOS order, there is also a vehicle OOS order. Because the impact of vehicle OOS orders
is greater than the impact of driver OOS orders, the number of crashes avoided
combined over vehicle and driver OOS orders can be determined by adding (a) the
number of crashes avoided due to vehicle OOS orders and (b) 60 percent of the
crashes avoided due to driver OOS orders.
Thus, the total number of crashes avoided with random inspections would
be 2,264 + (0.6*2,502) = 3,765.
Using the injury and fatality data in Table 5-2, there are on average 5,374/412,000 = 0.013 fatalities per crash and 127,000/412,000 = 0.308 injuries per crash. Therefore, if 3,765 crashes were avoided, it would be expected that 3,765*0.308 = 1,160 injuries would be avoided and 3,765*0.013 = 49 fatalities would be avoided. This relationship between the numbers of crashes, injuries, and fatalities is assumed to hold for all of the scenarios below.
Scenario
RE-0: Baseline –
Pre-CVISN
The calculation of crashes avoided in the baseline scenario is very similar to the calculation with random selection of vehicles, except instead of applying the results from the NFSS, we use the actual numbers of OOS orders for vehicles and drivers. In 1998, the reference year, 25.5 percent of the vehicles inspected were placed OOS, and 8.1 percent of the drivers received OOS orders.
Following the approach used with random selection, 25.5 percent of the 1,562,739 inspections (398,498) resulted in vehicle OOS orders. From Equation (5), the predicted number of crashes avoided due to vehicle OOS orders is equal to
Similarly, 8.1 percent of the 2,089,846 driver inspections (169,278) would have resulted in driver OOS order leading to
crashes
avoided.
Applying the 60 percent adjustment factor used under random selection,
the estimated number of crashes avoided is 1,991 + 0.6*4,053 = 4, 423. The corresponding numbers of injuries
and fatalities avoided are 1,362 and 57, respectively.
Note that the 1998 vehicle OOS rate of 25.5 percent is lower than the 29 percent violation rate estimated in the NFSS, and the 1998 driver OOS rate of 8.1 percent is higher than the 5 percent rate from the NFSS. This could be due to many factors, including individual or state‑specific inspection selection priorities or differences in truck traffic during scheduled versus randomly selected times. No specific explanation is available. Nevertheless, it is interesting to note that the estimated number of crashes avoided under normal (pre‑CVISN) inspection practices is 17 percent higher (4,423 versus 3,765) than the number that would be avoided under random selection of vehicles.
Scenario
RE-1: ISS with Manual
Pre-Screening
The primary direct impact of CVISN safety information exchange technologies is expected to be an increase in the efficiency of safety enforcement activities. In particular, it was expected that ISS would be used by safety enforcement staff to select vehicles and drivers for inspection based on a safety rating of the motor carrier and supplementary information on the carrier’s history involving inspections and safety incidents. However, because of the time and logistics involved in stopping a vehicle, entering identification numbers into the computer, and reviewing the data, ISS has not been used extensively as a tool for inspection selection. So, until ISS is integrated into electronic screening algorithms, or states develop other innovative ways to apply ISS as a selection tool (e.g., license plate readers or slow-down lanes with manual entering of identification numbers), the primary benefit of ISS will not be realized. Currently, most inspectors use Aspen/ISS after vehicle selection to help focus the inspection effort or adjust the level of inspection. Aspen is also used to record and transmit inspection results.
Fortunately, an opportunity to evaluate the use of ISS as a selection tool was made possible by the unique situation in one state. Connecticut, one of the first states to widely deploy laptop computers with Aspen and ISS, conducts a large number of inspections at four fixed weigh stations. Each station is equipped with a fixed scale, and all trucks are required to enter the station when it is open. Commercial vehicle inspectors are assigned at each station. However, at two of the stations, Danbury and Middletown, inspectors select vehicles for inspection using only judgment and experience. Inspections are then conducted with the aid of Aspen and ISS. At the other two sites, Union and Greenwich, all vehicles are pre-screened using weigh-in-motion results and quick visual inspections. Some trucks are allowed to bypass the fixed scale and return to the highway. The remaining trucks are sent to the fixed scale, and their identification numbers are entered into a roadside computer, which contains Aspen and ISS. The ISS information is then use to select vehicles for inspection.
During the spring and summer of 1999, a Screening Assessment Study was conducted at the four Connecticut weigh stations to evaluate the impact of ISS on the inspection selection process. Complete details on the study design, analysis plan, and findings are presented in Appendix 2. Also, a summary of the analysis supporting the major findings related to this crash avoidance analysis is provided in Section 5.4.
The primary finding relevant to scenario RE-1 is that when ISS is used in combination with manual pre-screening to select commercial vehicles for inspection (as currently performed at Union and Greenwich sites in Connecticut), the number of OOS orders issued for a fixed number of inspections will increase by 1.9 percent compared to sites that do not use ISS and manual pre-screening for inspection selection. Although this is a small increase in inspection selection efficiency, it is important to recognize that ISS is used to select vehicles for inspection after most of the vehicles have been eliminated during manual pre-screening. See Section 5.4 for additional discussion of these findings.
The calculation of the numbers of crashes, injuries, and fatalities avoided under this scenario is fairly straightforward. With a 1.9 percent increase in OOS orders, the number of crashes avoided under this roadside enforcement scenario is 1.019*4,423 = 4,507. This represents an increase of 84 crashes avoided compared to the baseline scenario. The corresponding number of injuries avoided is 1,388 (a difference of 93), and the number of deaths avoided 59 (a difference of 2). Although these benefits are fairly modest, they do not represent the full potential of ISS when it becomes integrated with electronic screening or other innovative roadside enforcement strategies. The following scenario helps to demonstrate some of this potential.
Scenario
RE-2: ISS with Electronic
Screening
As CVISN deployment expands and begins to integrate the use of ISS with electronic screening, roadside enforcement officials should be able to improve the efficiency with which they select high-risk CMVs for inspection. Currently, only a few states use ISS or similar tools in combination with electronic screening. However, even in these states, carrier enrollment in electronic screening is not sufficient to demonstrate any impacts on the inspection selection process. Therefore, to illustrate what could happen, the impact of using ISS with electronic screening was simulated using results from the Connecticut Screening Assessment Study. An analysis was performed under the scenario that (a) all states deploy electronic screening at all major inspection sites and (b) all of the motor carriers with SafeStat ratings in the low-risk category (representing approximately 52 percent of all trucks) choose to enroll in the electronic screening program.
Under this scenario, enforcement officials could choose to let the low-risk vehicles bypass the inspection site and focus all of their efforts on inspecting medium- and high‑risk carriers and carriers with insufficient safety data. It is assumed that ISS will be used with manual pre-screening, as in scenario RE-1, on the 48 percent of trucks that are not allowed to bypass the inspection site. Section 5.4 presents an analysis demonstrating that, under this scenario, the number of OOS orders will increase by 11.2 percent compared to the average number that would be achieved under scenario RE-1.
From here, the calculation of the numbers of crashes, injuries, and fatalities avoided under scenario RE-2 is straightforward. With an 11.2 percent increase in OOS orders (compared to RE-1), the number of crashes that can be avoided under RE-2 is 1.112*4,507=5,012. This represents an increase of 589 crashes avoided compared to the baseline scenario. The corresponding number of injuries avoided is 1,544 (a difference of 181), and the number of deaths avoided 85 (a difference of 9).
Scenario
RE-3 : ISS with Electronic
Screening and a 25 Percent Reduction in OOS Conditions
The preceding scenarios looked at the direct effects of CVISN deployment as it affects inspection selection efficiency. An additional, indirect effect of CVISN deployment will be to deter carriers from operating vehicles in unsafe conditions in violation of the FMCSRs. The increased compliance with the FMCSRs will result in fewer unsafe trucks on the road. This will also reduce the numbers of truck-related crashes, injuries, and fatalities.
Although the Oregon Compliance Rate Study (See Appendix A.1) was
conducted too early to determine if there will be a decline in FMCSR violation
rates as CVISN deployment expands, the potential impact of this effect was
investigated in scenarios RE-3 by assuming that targeted enforcement will
result in 25 percent fewer FMCSR violation rates.
The calculation of the number of crashes avoided under scenario RE-3 is divided into two parts. The first part involves determining the number of crashes avoided because there are 25 percent fewer trucks and drivers with safety violations on the road (the indirect effect). The second part involves determining the impact on inspection selection efficiency because there are fewer OOS violators to select for inspection.
It is assumed that a 25 percent reduction in FMCSR violation rates occurs
uniformly across all types of driver and vehicle violations, including those
that are likely to cause crashes.
Again, no crash causation data exist to support or refute this assumption
at this time. If the reduction
occurs in this manner, the number of crashes avoided would be equal to 25
percent of the number of crashes caused by vehicle defects and driver violations
before the improvement in safety compliance. Using the crash causation probabilities
employed in the Safe-Miles model, it is estimated that driver violations
contribute to 4.6 percent of truck-related crashes, and vehicle defects
contribute to 5.7 percent of these crashes. Therefore, assuming minimal cases in
which both driver and vehicle OOS conditions contributed to the same crash, the
number of crashes caused by OOS conditions is
412,000 *(0.046 +0.057) =
42,436.
From the discussion above, scenario RE-3 would result in a 25 percent
reduction in these crashes, or 10,609 crashes avoided due to the indirect
effect of enhanced roadside enforcement.
For the direct effect we consider how a 25 percent reduction in
FMCSR violation rates affects the number of crashes avoided due to roadside
enforcement. From Equation (5)
we see that a change in the violation rates affects the calculation in three
ways. First, the denominator,
P(OOSC), which represents the violation rate, will be reduced by 25
percent. Second, the number of OOS
orders that will be obtained with the same level of effort will decline, because
the proportion of CMVs with OOS orders will be smaller. It is assumed in this illustration that
compliance improves uniformly across all risk categories of CMVs and inspection
selection strategies at the roadside remain the same. Therefore, a 25 percent decline in
violation rate will result in a 25 percent decline in the number of OOS orders
issued. Third, if the violation
rate decreases by 25 percent, it is expected that the percent of crashes caused
by defects or driver violations, represented by P(D|C), will also
decrease because there will be fewer CMVs in violation, including those involved
in crashes. We assume this
probability will decrease by the same percentage; however, the data needed to
support this argument are not yet available. FMCSA’s Large Truck Crash Causation
Study may provide the necessary data.
The net effect of reducing FMCSR violation rates by 25 percent is that
for the same number of inspections performed there will be a 25 percent decrease
in the number of OOS orders issued.
Based on results for scenario RE-2, there will be (1-0.25)*5,012 = 3,759
crashes avoided through roadside enforcement with ISS and electronic
screening.
Combining direct and indirect effects yields 3,759 + 10,609 = 14,368
crashes avoided, which is an increase of 9,945 compared to the baseline
scenario. The corresponding number
of injuries avoided is 4,425 (a difference of 3,063), and the number of
fatalities avoided is 187 (a difference of 130).
Scenario
RE-3*: ISS with Electronic
Screening and a 10 Percent Reduction in OOS Conditions
Using the same approach as
for scenario RE-3, a 10 percent reduction in OOS conditions will result in 4,244
fewer crashes due to indirect effect and 4,511 fewer crashes from the direct
effect. Combining these yields
4,511 + 4,244 = 8,755 crashes avoided, which is an increase of 4,332 compared to
the baseline scenario. The
corresponding number of injuries avoided is 2,697 (a difference of 1,335), and
the number of fatalities avoided is 114 (a difference of
57).
Scenario
RE-4: 100 Percent Inspection
Selection Efficiency
The calculations of the number of crashes avoided under scenarios RE-1 and RE‑2 assumed that the increases in inspection selection efficiency would be proportional to baseline scenario for both the vehicle and driver inspections. If we assume that CVISN technology will advance to the point that every inspection will result in an OOS order, it would be necessary to make unfounded assumptions concerning the distribution of OOS orders attributable to vehicles versus drivers. Therefore, we make the simplifying assumption that all OOS orders will be for vehicle violations; but calculate the number of crashes avoided using the total number of inspections (Levels 1 through 5) performed in 1998. This yields a conservative estimate of
crashes avoided, which is 6,138 more than the baseline scenario. The corresponding numbers of injuries avoided is 3,253 (a difference of 1,891), and the number of deaths avoided 137 (a difference of 80).
The preceding calculations (as well as those in the preceding three scenarios) assume that the number of inspections performed annually is constant and equal to the number performed in 1998. The direct safety benefits of CVISN could be further improved by increasing the number of inspections that are performed annually.
Scenario
5: 100 Percent Compliance with
Safety Regulations
If CVISN deters all carriers from driving with OOS conditions, all crashes caused by OOS conditions would be eliminated. Thus, the number of crashes avoided is estimated by the number of trucks involved in crashes (412,000) times the probability that the crash is caused by a vehicle or driver OOS condition. As discussed under scenario RE-3, the estimated number of crashes avoided is
412,000*(0.46+0.57) = 42,436.
Because both vehicle and driver violations can be contributing causes to the same crash, this estimate represents an upper limit on the number of preventable crashes. This estimate represents an increase in 38,013 crashes avoided compared to the baseline scenario. The corresponding numbers of injuries avoided is 13,070 (a difference of 11,708), and the number of deaths avoided 552 (a difference of 494).
5.4
Key
Findings from the Connecticut Screening Assessment Study
The prediction of CVISN safety benefits under roadside enforcement scenarios RE‑1 to RE-3 relied on specific estimates of the improvement in inspection selection efficiency that are or could be directly attributable to CVISN deployment. The primary source of data for developing these estimates was the Connecticut Screening Assessment Study. This section presents the analyses that support these key findings.
As discussed in Section 5.2, the Connecticut Screening Assessment Study was conducted at four commercial vehicle weigh stations in Connecticut to evaluate the effectiveness of ISS for improving the inspection selection efficiency of roadside operations. Inspection selection efficiency is measured by the number of OOS orders issued per 100 vehicles inspected. Increased efficiency means that more unsafe vehicles or drivers will be removed from the highway for the same number of inspections performed. During 13 days of data collection, approximately 10,000 vehicle identification numbers were recorded for all trucks entering the four weigh stations. At two of the stations (Danbury and Middletown), vehicles are selected for inspection without the aid of ISS. At the other sites (Union and Greenwich), vehicles are pre-screened using weigh-in-motion (WIM) and visual inspection. Vehicles sent to the fixed scale for weighing are then screened for inspection using ISS ratings. Figure 5-2 shows the configuration of the Union facility.
Figure
5-2. Schematic of Connecticut’s Union
Facility with WIM Sorting
The vehicle identification numbers were used to characterize the distribution of trucks in terms of safety risk at each inspection site. This was achieved during the analysis phase by calculating the SafeStat score for each truck. SafeStat is an automated motor carrier safety status measurement system developed for FMCSA that combines current and historical safety data to measure the relative fitness of motor carriers (VNTSC 1999b; VNTSC 1998). In addition to the inspection results obtained during the data collection phase, results of over 58,000 inspections performed over a four-year period at these sites were analyzed.
The analyses performed with these data are summarized in Table 5-3. The SafeStat scores for the 10,000 trucks that entered the sites were used to estimate the distribution of trucks that would be inspected if vehicles were selected at random. This serves as a baseline which allows us to make valid comparisons of inspection selection strategies at each site. For example, at the Danbury site, which does not use ISS for vehicle selection, the distribution of trucks includes 8.6 percent high-risk vehicles (according to SafeStat scores) and 47.2 percent low‑risk vehicles. The actual inspection results show that inspectors are selecting more high‑risk (12.0 percent versus 8.6 percent) and fewer low-risk (36.1 percent versus 47.2 percent) vehicles for inspection then they would if vehicles were selected at random. Multiplying these percentages by the statewide OOS rate gives the expected number of OOS orders per 100 vehicles inspected within each risk category. The statewide OOS rate for low-risk carriers is 38 percent compared to rates of 42 percent to 63 percent for the other risk categories (Medium, Insufficient Data, and Unknown). The totals represent the expected number of OOS orders for a given inspection selection strategy. The inspectors at Danbury average 48.4 OOS orders per 100 inspections using their own judgment and experience to select vehicles for inspection. Random selection would produce only 46.76 OOS orders per 100 inspections. Combining the Danbury and Middletown results, we see that inspector judgment and experience produce 3.5 percent more OOS orders than random selection. Even though Connecticut’s OOS rates are much higher than the national average, the percent difference in these rates is consistent with similar findings from the National Fleet Safety Survey (1997).
The same calculations were performed with the data from the Greenwich and Union, which use ISS and manual pre-screening with WIM, in addition to judgment and experience, to make inspection selection decisions. This inspection selection process produces 5.4 percent more OOS orders than random selection. Using an odds ratio to adjust for differences in populations, we estimate that using ISS with manual pre-screening produces a net effect of 1.9 percent more OOS orders than would be achieved with inspector judgment and experience. This estimate was used in the model for crashes avoided under scenario RE-1.
To simulate the impact of electronic screening under full deployment, we assumed that all low-risk carriers would enroll and be permitted to bypass all inspection sites. Since no low‑risk carriers will be inspected, we assumed that inspectors would proportionally allocate the inspections among the other risk categories. The predicted number of OOS orders with electronic screening was then calculated in the same manner. The relevant finding is that by using electronic screening to eliminate the low-risk carriers (and thereby target high-risk carriers) can increase OOS orders by 11.2 percent. This estimate was used in the model for crashes avoided under scenario RE-2.