Traditional Approaches To Traffic Safety Evaluatio

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/303375853
Traditional Approaches to Trafﬁc Safety Evaluation (TSE): Application

Challenges and Future Directions
Conference Paper · May 2016

DOI: 10.1061/9780784479810.028
CITATIONS READS
0 203
3 authors:
S. M. SOHEL Mahmud Luis Ferreira

Bangladesh University of Engineering and Technology The University of Queensland
19 PUBLICATIONS 98 CITATIONS 255 PUBLICATIONS 2,730 CITATIONS
SEE PROFILE SEE PROFILE
Ahmad Tavassoli Hojati

The University of Queensland
35 PUBLICATIONS 289 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Freight modal research View project
Transport planning research View project
All content following this page was uploaded by S. M. SOHEL Mahmud on 25 January 2018.
The user has requested enhancement of the downloaded file.

Bridging the East and West 242
Traditional Approaches to Traffic Safety Evaluation (TSE):

Application Challenges and Future Directions
S. M. S. Mahmud1, L. Ferreira2, and A. Tavassoli3
1
PhD Researcher, School of Engineering, The University of Queensland (UQ),
Downloaded from ascelibrary.org by The University of Queensland Library on 06/14/16. Copyright ASCE. For personal use only; all rights reserved.
Brisbane, QLD 4072, Australia; Ph (+61) 469 911 070; email:

s.mahmud2@uq.edu.au / smsohelmahmud@gmail.com
2
Professor, School of Engineering, The University of Queensland (UQ), Brisbane
QLD 4072, Australia; Ph (+61) 431 880 208; email: l.ferreira@uq.edu.au
3
Post Doctoral Research Fellow, School of Engineering, The University of
Queensland (UQ), Brisbane, QLD 4072, Australia; Ph (+61) 411 868 530; email:
a.tavassoli@uq.edu.au
ABSTRACT
Traditional TSE approaches, based on chronological data including classical

before-after studies, mainstream/statistical modelling and/or person judgment based
approaches, has a long history. Despite the robustness of this established technique,
this has not always been appropriate, as crash statistics are frequently questioned.
Poor timeline, ethical issue, biasness and misjudgment also critical issues regarding
this. For a more qualified and comprehensive form of safety analysis, there is a need
for scientific methods that yield valid and reliable safety measures without the need
for (or in addition to) accident data. At the very outset of this process, there is a need
to pinpoint the confines of the current practices. This paper presents a critical review
of state-of-the-art traditional approaches of TSE. The strengths and application
challenges of those approaches which are also assessed based on methodological, as
well as empirical grounds. Finally, future directions for traffic safety evaluation
practices and researches are outlined.
Keywords: Traffic Safety Evaluation, Traditional approaches, Chronological data,

Proactive approaches, Surrogate measures
INTRODUCTION
Traffic Safety Evaluation (TSE) is an examination of the safety and performance of a

roadway facility and quantitative estimation of the effects of any action on crash
frequencies or severity at any stage of a project. The purpose for conducting a TSE is
to account for all roadway users when identifying potential safety issues and
opportunities for safety improvements (RTSSC, 2015). There are both quantitative
and qualitative benefits to conducting a TSE. Several qualitative benefits are defined
by various authors (e.g., Ogden, 1996; McGill, J. et al., 2005; Gibbs, M. et al. 2006;
LGED, 2015 etc.). Among those most prevalent are: 1) addressing potential safety
issue; 2) producing fewer and less severe crashes in future; 3) identifying low-
cost/high-value improvement options; 4) facilitating a culture of safety-conscious
© ASCE
Bridging the East and West

planning, design, implementation; 5) delivering a continuous progression of safety

knowledge and skill; 6) acting as a benchmark for safety issues on next projects; 7)
endorsing an efficient use of valuable time, money, and resources; 8) finally, provide
very important information for future safety decision making or for policy making
regarding safety. In addition, economic return due to the reduction of crash and
resulting adverse consequences are the major the quantifiable benefits of conducting
a TSE. However, several other quantifiable benefits have been documented, such as:
1) the elimination of re-construction costs; 2) the reduction in lifecycle costs; 3) the
reduction of societal costs due to collisions; and 4) the reduction of liability costs due
to safer roadway facilities.
There are different practical approaches for safety evaluation. One approach is based
on crash/accident data, statistical modelling and/or field observation. Another is
based traffic conflict observation coupled with micro and macroscopic traffic
simulation. The former is termed traditional and the second surrogate safety
evaluation. Both approaches have been described in detailed in the past. This paper
summarizes current practices of traditional approaches with its principle based on
methodological, as well as empirical grounds, requirements, advantages and
disadvantages. Issues related to factors affecting the results of evaluation are also
discussed with emphasis on problems related to historical data and database system.
Finally, through the critical review of the application challenges and considering the
technical advancement, such as automated image processing system, discrimination
of conflict (near miss crash) using microsimulation, future direction of traffic safety
evaluation practices and researches are pointed out.
The structure of the paper is a follows: The next section provides a brief outline on
the methodology of the study. This is followed by some important issues related to
traditional traffic safety evaluation which adversely affect the outcomes. Detailed
description of different traditional approaches with, advantages and disadvantages are
given in the third section with different sub-sections. Finally, a summary of the
application challenges of the traditional approaches and future directions for traffic
safety evaluation are provided.
ISSUES FOR TRAFFIC SAFETY EVALUATION (TSE)
Before coupled with after planning, designing and/or implementing of any facilities
or improvement measures, it is very important to evaluate its impact due to some
reasons. Firstly, it is important to assess the safety problems for different road users
to identify potential safety issues and opportunities for safety improvements.
Secondly, it is important to make sure whether or not this countermeasure is working
effectively for all users. Thirdly, sometimes one facility for specific safety reason
may invite another type of hazard. Last but not least, it is also important to observe
the community reaction, attitude and behavior towards the measure/initiative. This
follow-up is termed as traffic safety effectiveness evaluation as well as traffic safety
evaluation.
© ASCE

Despite huge benefit to traditional safety evaluation, there are several experimental
challenges related to casual factors. In addition to the problems related to the casual
factors, there are some other significant challenges and limitations in using the
historical crash data for traffic safety evaluation. A short description on both of those
factors is in below:
Casual factors
The performance of a treatment or group of treatment depends on two types of factor,

one is measurable and another one is non-measurable. The measurable factors can be
explained, understood and recognized by model while the non-measurable factors
cannot be easily explained or recognized. The factors that effect on evaluation are:
treatment effect, exposure effect, trend effect and random effect. A brief explanation
of those effects are given in the following sections. Further details can be seen in the
Council, 2009; Ogden, 1996; and LGED, 2015.
Treatment effect: The change of safety performance of a transportation facility due to

treatment is called treatment effect. To determine the net improvement/deterioration
in terms of safety performance, the treatment effect must be isolated from the other
causal factors.
Exposure effect: As the road environment is a dynamic thing, the exposure

characteristics and behavioral pattern could be changed due to new or reformed
facility. It is also established that the traffic volume and characteristics has a direct
relationship with the nature of accident. Therefore, exposure effect could be
significant if the remedial action changes the operations or capacity of the road.
Trend effect: It is a non-recognized factor and are not measured and understood. For
example, traffic composition (such as a higher/lower percentage of trucks or
pedestrians), driver composition (in terms of behavior, age, etc.), enforcement level,
weather conditions, etc. can be changed from the before period to the after period.
Random effect: It is known that crashes are rare and random events and with time,
however, in a particular site crash frequencies naturally fluctuate. The random effect
occurs because of a phenomenon called as regression-to-the-mean (RTM) bias in
statistics. The randomness of accident occurrence indicates that short term crash
frequencies alone are not a reliable estimator of long-term crash frequency.
Based on the definition of the above four effects, it can be concluded that even if no
safety treatment had been applied to the facility, it would have been likely to observe
a change in accident frequency from the before to the after periods. Consequently,
analysts must recognize the impact of each of these effects on their evaluation results
and must employ techniques that seek to minimize or account for these extraneous
effects.
© ASCE

Factors related to crash data

In the majority of safety evaluation studies, historical crash data are used as the
primary source of data and fundamental information to conduct safety evaluations.
However, despite the robustness of this established technique, this has not always
been appropriate, as crash statistics are frequently questioned in terms of their
availability, and quality. The most common limitations of using crash data as a single
measure to assess traffic safety are summarized as follows:
Reporting bias: Accuracy problems of accident data primarily stemming from

incomplete and inconsistent accident reports. Reporting criteria vary between
jurisdictions and thus accident experience is not comparable (Searles and Jamieson,
1983). While such information may be useful to individual agencies, it cannot be
used for identifying risks, selecting interventions, or measuring outcomes at an
aggregate level unless is it properly analyzed and disseminated (WHO, 2010).
Unavailability: It is a common that the crashes are not always properly reported and
this are hindering the availability and reliability of information used to conduct safety
studies. This problem is more serious in developing countries such as Bangladesh
(ADB, 1996; TRL, 2003; Hoque et al., 2007; Hoque & Mahmud, 2009). Some
countries consider reporting only those crashes that involve injuries or property
damage above a certain cost (Zegeer and Deen, 1978), while others may heavily
under-register some minor crashes (Berntman, 1994). Results from a number of
studies indicated that in developed countries under-reporting of fatalities was about 2
to 5 percent whereas 25 to 50 percent of those road crashes were underreported by the
police in developing countries (ADB, 1996, Hoque et al, 2007; James, 1991; Simpson,
1997; TRL, 2003; WHO, 2009).
Not fit for comprehensive assessment: There is growing recognition that effective
road safety management requires more comprehensive data than the existing crash
data (WHO, 2010). The success of the safety analysis is governed by the quality of
the accident information. Unfortunately, in many cases crash statistics do not provide
a complete picture of the road safety situation (Yang, 2012). Some information in the
accident record, particularly on accident severity and causes may not be very accurate,
as has been shown in comparisons of hospital and police data (Rosman & Knuiman,
1994). Crash data also do not capture information on risk factors such as helmet use
or speeding among the general population (ATC, 2011). Data such as corresponding
vehicle conditions and weather factors are also not easy to identify and obtain for
analysis (Pham et al., 2008).
Coding errors. These can occur throughout the process from the filling out of the
police accident report form to the data entry at the computer terminal. OECD has
pointed out the following three are the most common ones: Omission of entire
records, simple inaccuracies due to misclassification, misjudgement or misprinting
and delays in processing and recording information which may be different for
different data. (OECD, 1997). Typical problems range from errors in filling out the
police accident report form (e.g. getting the north point wrong) to errors in data
© ASCE

transcription or coding (O’Day, 1993). Furthermore, despite a uniform reporting

form, many reports are completed with varying levels of detail or may omit important
information.
Location errors. The location of an accident may easily be incorrect or imprecise in

the original police accident report form. In some cases, the location reference system
itself may be imprecise, which may mean that the exact location of the accident
cannot be accurately determined (O’Day, 1993). This has been acknowledged for a
long time as a fundamental deficiency in crash records. Such incomplete data can
thus jeopardize the success of safety analysis (Yang, 2012).
Discontinuities.) pointed out that the definitions or interpretations of field data may
be changed over time by those responsible for coding and recording, meaning that
data from one time period cannot be compared with that for another time period, or
worse, the user may be unaware of the presence of a discontinuity and as a result the
analyses may be incorrect (O’Day, 1993; Alam, 2003).
Long Collection Cycle: Due to the low frequency and random variations inherent in
traffic crashes, a relatively long observation duration is needed to gather sufficient
data to produce estimates or conduct evaluation study with acceptable statistical
properties (Chin and Quek, 1997). Zegeer and Deen (1978) suggested that up to 2
years of crash data are necessary to ensure reliability when diagnosing black spots. It
also takes several more years to prove the effectiveness of a specific improvement
based on crash records. Agencies responsible for data processing may not be
sufficiently resourced, and as a result it may be many months before information is
available for analysis. This means that countermeasure development is responding to
historical accident patterns which may be out of date (Alam, 2003).
Hidden problems. The implicit assumption in the process described here is that the
accident database is a good indicator of road safety problems. Generally this is
probably true (subject to the qualifications made in the previous paragraphs).
However, it may be masking other problems, for example if pedestrians avoid using
an area because of a perceived safety problem; in this case the safety issue has
resulted in lack of amenity rather than the occurrence of accidents (O’Day, 1993).
Therefore, development of unbiased, consistent, reliable, comprehensive quality and

easily accessible crash database together with the expansion of extensive coupled
with statistically sound models can lead to high quality safety evaluation (Yang,
2012). However, attempts to address the aforementioned problems, as well as some
other issues such as ethical concerns, regression-to-mean effect (Barker and Baguley,
2001) are still needed in the long run.
TRADITIONAL APPROACHES FOR TSE
Most of the current literature covering highway safety analysis, including ramp and
pavement marking safety studies, has been conducted using traditional methods (St-
© ASCE

Aubin, 2011). Those traditional approaches are carried out based on observed
historical accident data using various types of statistical methods and anticipatory
estimation studies based on safety audits (Yang, 2012). A list of those different
approaches are shown in Table 1.
Table 1. Different Traditional Approaches of Traffic Safety Evaluation (TSE)

Traditional Approaches Based on Historical Data

I. Before-after Observation
 Naïve before-and-after study
 Before-and-after study with yoked comparison
 Before-and-after study with comparison group
 Before-and-after study with the empirical bayes approach
II. Black Spot Identification Program
III. Mainstream/Statistical Modelling Approaches
 Descriptive model
 Predictive or analytical model
 Risk model
 Accident consequence model
Traditional Approaches Based on Person Judgment
I. Road safety audit (RSA)/inspection (RSI)
II. Safe system approach
Traditional Approaches Based on Historical Data
The traditional safety methods, based on the statistical analysis of historical accident
data, has a long history (St-Aubin, 2011). In these approaches, traffic safety is
commonly measured in terms of the number of traffic accidents and the consequences
of these accidents with regard to their outcome in terms of severity. These mainly
include the classical before-after observation study including black spot treatment
program and the cross-sectional study based on regression techniques.
Before-after Observation
One of the most common study designs employed in observational evaluation studies,
is a before-and-after study (Elvik, 2002). Inherent challenge in these studies is that
crashes are random and change from year to year, unlike laboratory experiments in
which the analyst can control many extraneous conditions. Other parameters that
affect the safety of a facility, such as traffic volume and weather conditions, change
over time. Consequently, specific evaluation techniques are required to account for
changes in order to estimate the true effects of safety improvements (Council, 2009).
There are four most commonly used approaches are found in literature to perform a
before-and-after study in order to evaluate the performance of a safety improvement
plan or an operational change on a transportation facility, namely: naïve study; study
with yoked comparison; with comparison group; and with Empirical Bayes approach.
© ASCE

A summary comprising major data requirements, advantages and disadvantages of

the before-after observation approaches are presented in Table 2 followed by a short
description.
Table 2. A summary of Before-after Observation Approaches

Approach Reference/s Data Advantages Disadvantages

name Requirements
Ogden, 1996;  Crash data during  Simple approach  Depends on reliable and consistent
Council, 2009. before and after  Few data required historical crash data
Naïve period (at least 3  Does not account Regression-to-Mean
before-and- years) (RTM) bias error
after study  Does not account time-trend error
 Could not able to separate the treatment
effect from the other effects
Ogden, 1996;  Crash data during  Relatively simple  Depends on reliable and consistent
Griffin and before and after approach historical crash data
Flowers, 1997; period(at least 3  Fewer data  It requires that the treatment site to be
Griffith,1999; years) required similar to the comparison site, which is
FHWA, 2002;  Similar almost impossible in real
Before-and-
comparison group  Unable to address the regression-to-the-
after study
that has one-to- mean bias issue
with yoked
comparison
one relationship  Treatment and comparison site has to be
one to one related
 Unable to deal with cases where the
comparison site has no history of accident
occurrence
Sayed, T., K.  Crash data during  The treatment  Depends on reliable and consistent
El-Basyouny before and after sites and historical crash data
and J. Pump, period for comparison group  Does not account RTM bias error
Before-and-
2006; Council, treatment and sites need to be  Cannot determine the treatment
2009. comparison sites similar, but a one- effectiveness if accident counts in either
after study
(at least 3 years). to one pairing is the before or the after period in the
with
 A comparison not required comparison group equal zero.
comparison
group
group that is in  Conformity check is required for
conformity with treatment and comparison group in before
the treatment period.
group in the
before period
Hauer , 1997;  Crash data during  Has strong  Highly sophisticated and complex method
Council, 2009; before and after theoretical compare to others
Persaud & period (at least 3 background and  High quality and diversified data needs
Before-and- Lyon, 2007; years). preferred by for the EB methodology
after study Elvik, 2008;  Suitable Safety researcher  Safety performance functions do not exist
with the Council, 2009; Performance  Account for all facilities being analysed
empirical Namjune & Function (SPF) regression-to-the-  Potential treatment sites could be
bayes Ihn, 2013. for the facility. mean effects overlooked due to the lack of proper
approach  It consider traffic reporting of crash data
volume changes
and time trends
impacts
Naïve before-and-after study: In this approach, accident counts in the before period
are used to predict the expected accident rate and, consequently, expected accident
counts if the safety treatment had not been implemented. The change in accident
counts between the before and the after conditions is considered the treatment effect
(Council, 2009). However, this technique is not recommended in practice.
Before-and-after study with yoked comparison: In this type of study, the treated
and untreated facilities are referred to as the treatment site and comparison site,
© ASCE

respectively (Griffin and Flowers, 1997). A group of similar facilities is selected so

that there is a one-to-one correspondence between each member of the comparison
group and the treatment group. This technique requires that the treatment site to be
similar to the comparison site, which is almost impossible in real world. Even, the
comparison site should not have undergone any geometric change or traffic control
improvement during the before and after periods (FHWA, 2002).
Before-and-after study with comparison group: The rationale behind this approach
is almost same as previous technique, except that there is no need for a one-to-one
matching between members of the comparison group and the treatment group in this
approach.
Before-and-after study with the empirical bayes approach: The Empirical Bayes
method is a statistical approach to determine the appropriate weighting to place on
each relevant factor to estimate accident outcomes for a treatment group. Safety
performance functions (SPFs) are used to estimate accident frequencies had the
treatment not been applied (Council, 2009; Hauer, 1997).
However, some researcher have argued that road safety improvements are
implemented for a variety of reasons and that a randomly high accident count is not
among the key selection criteria. These arguments arise question on the need for such
complex EB methodology. In addition, the Empirical Bayes Method has the
uppermost limit on research methodology (Namjune & Ihn, 2013). According to
Persaud & Lyon, 2007, there are a number of difficulties which, if not properly
resolved, will render this Empirical Bayes methodology just as invalid as the
conventional methods, resulting in a misuse of precious resources and a general lack
of faith in the method.
Black Spot Identification Program

Black spots are locations noted for a high incidence of crashes involving death and
injury (ANAO, 2007). In compare to the benefits that accrue over time, most of the
black spot treatment programs are relatively low cost and therefore provide
substantial economic returns (ATSB, 2004). Black Spot Programs have been used at
some stage by all leading road safety countries of the world including Canada, the
USA, New Zealand and much of Europe (Meuleners & Fraser, 2008). However, it is
a reactive approach fully based on the availability of historical data. Some of the
major drawbacks of related to Black Spot Identification Program are summarized
below:
Declining Black Spot Sites: Due to successive successful black spot treatment
programs, the opportunity to treat sites with high crash rates is of course diminishing
as the blackspots are steadily eliminated by remedial treatments (BTE, 2001; Roper
& Turner, 2008). Investigation of Victorian and New Zealand crash data (reported in
Turner, 2007) shows that only a third of fatal crashes occur in locations classed as
blackspots. New Zealand data also reveal that more than half of fatal crashes occurred
at locations where no other crashes had occurred in the previous five years. In
© ASCE

Bangladesh, of the total reported accidents, nearly 42 per cent fatal accidents
occurred on the national highway network and among those accidents, only 35
percent are occurring on defined hazardous road locations (Hoque et al., 2006). If
remedial attention is focused only on blackspots, the opportunity to prevent a large
proportion of crashes will be missed.
Lack of adequate crash databases: Blackspot programs rely on the use of crash data
to identify problem sites and this naturally means that crash data must be
comprehensive. As described earlier, due to deficient crash data, problematic sites
being overlooked when road authorities assemble lists of potential treatment sites
(Roper & Turner, 2008).
Causes of crashes are now better understood: Many years of research have resulted
in a comprehensive understanding of factors that contribute to the occurrence and
severity of crashes. This means that it is no longer necessary to identify high risk road
sections by waiting until crashes occur; instead, high risk sites can be detected and
treated before crashes occur ((Elvik et al., 2009; SWOV, 2007; Roper & Turner,
2008).
Mainstream/Statistical Modelling Approaches

Models are a restricted form of general theories, usually containing hypotheses and
assumptions used to examine these theories empirically (Hakim, et al., 1991). A
review on current practices on mainstream modelling approaches for safety
improvement was published by OECD (OECD, 1997) . The report proposes that the
process of formulating and testing models is critical for the improvement and
prioritisation of research related to traffic. In that report, mainstream modelling
approaches are classified by four types as follows:
 Descriptive Models
 Predictive Models for Aggregated Data
 Risk Models for Non-Aggregated Data
 Accident Consequence Models.
Data requirements, advantages, disadvantages with the important references related to

those types are summarized in Table 3. A brief description on principles with
assumptions of each types are also highlighted following the table.
Table 3. Advantages, Disadvantages of Modelling Approaches
Approach name Reference/s Data Advantages Disadvantages

Requirements
Archer, 2005;  Two principle  Most common  Quality data of both types is not
Rumar, 1985; sources of data: among always available.
OECD, 1997; traffic accident modelling  Uncertainties or inaccuracies in traffic
data and approaches data often complicate the
Descriptive model
exposure data. interpretation of the results.
 Do not cope with the inaccuracies
caused by systematic errors (bias) and
spread components.
© ASCE


Requirements
Conventional Hsiao, 1986;  Mainly crash,  Relatively easy  Cannot able to clearly explain
(multiple) McCullagh and traffic, and to implement in distributional characteristics derived
linear Nelder, 1983; geometric practice from random and discrete vehicle
regression Saccomanno & data/infrastructur crashes.
model Buyco, 1988; e etc.  Do not consider distributional
(cross- Miaou,1994; characteristics that are created by the
sectional Jovanis and randomness of collisions.
models, time Chang,1986.

series models,
combined
cross-
section/time-
series models)
Lord et al., 2005;  Different  Consider  The accuracy and reliability of these
Predictive or analytical model
Miaou, 1994; independent distributional models depends on the use of

Miaou & Lord., variables related characteristics explanatory variables. Unfortunately,
Statistical 2003; to crash created by the not enough good quality data were
modelling Oh et al., 2006; occurrence randomness of available for applying many
approach Hall, 2000; Austin including crash, collisions explanatory variables.
(Binomial, & Carson, 2002; traffic,  Productive way  Prevailing modelling approaches grasp
Poisson, Chin & Quddus, geometric/infrast of gaining deep from simply using crash rates which
Poisson- 2003; Kumara & ructure data understanding might give a false indication of the
gamma (or Chin2003; Lord et environment, into the process hazard.
negative al., 2007; and socio- of accident and  Many models are not flexible to use
binomial), OECD, 1997; economic etc. way it influence and also suffer for lack of a sound
zero-inflated Hauer,1986; Datta theoretical foundation.
Poisson and et al., 2001;  Many model cannot handle under-
negative Hankey et al., dispersion properly. Low sample size
binomial 1999; Stanton & mean and small sample size bias is
models (ZIP Salmon, 2009; also common. Most the cases,
and ZINB), Eenink et al., dispersion capacity cannot estimate
and 2008; Yang, 2012. properly.
multinomial  Generally fail to take into account
probability other significant explanatory factors
models) such as driver behaviour and errors.
 Difficult to judge the goodness of fit
of these models based on statistical
measures.
Brehmer, 1994;
Michon, 1989;
 Individual  Bottom-up
approach which
 Many risk models are often too
specific and context dependent.
variables such as
Williams, Paek
visual acuity,
identify risk  Huge lacking in consideration of
and Lund, 1995; factors and other important traffic system
reaction time,
Näätänen and determine their elements
personality etc.
Risk model Summala, 1976;
 road user task
mechanism at  Absence of connection between the
Molen and individual level aggregated theories and models on the
variables such
Bötticher, 1987; one hand and the individual theories
Shinar1978 disposition,
assimilation and and models on the other.
situation  Among many models only few are
considered as comprehensive models
University of In-depth data on  Describe  Lack of enough cases and a
Zurich, 1986; different type of accident representative database are the major
Hutchinson, 1986; accidents and consequences problems to draw general conclusions
Pham et al., 2008; injuries and their  Lack of a quantitative definition of
OECD, 1997. implications severity and the thresholds of the
 Provide detailed accident categories are more or less
information on arbitrary
Accident
consequence model
specific type of  Some significant factors such as
accident geometry have usually not been taken
 Biomechanical into account by these models
models help to  Difficult to use to some specific sites
detect, due to data extraction, processing, and
understand and system coverage limitation
eventually
prevent injury
© ASCE

Descriptive model: This is the most common approach. Model of Rumar (1985)
could be an appropriate example in this regard in which the traffic safety situation has
described in terms of exposure, risk and accident consequence. This methods are
normally designed to deal with that problems, for instance inaccuracies. But this can
only cope with type of inaccuracies which have a random character (OECD, 1997).
Therefore, have to be compensated for in other ways.
Predictive or analytical model: An Accident Prediction Model is a mathematical

formula which describes the relation between the level of safety of existing roads (i.e.
crashes, victims, injuries, fatalities etc.) and different related variables that explain
this level (road geometry including length, width etc. traffic characteristics, traffic
volume etc.) (Eenink et al., 2008). Two types of modelling approaches are found in
literatures for crash prediction:
 conventional (multiple) linear regression model
 statistical modelling approach
Conventional (multiple) linear regression model: Initial crash prediction models

used conventional (multiple) linear regression such as cross-sectional models, time
series models, combined cross-section/time-series models (panel data models) (Hsiao,
1986). Therefore, although comparatively these linear regression models are easy to
implement in practice, they do not usually consider distributional characteristics
created by the randomness of collisions.
Statistical modelling approach: To overcome these limitations, statistical modelling

methods have been introduced and considerable research has been conducted over the
last 20 years focused on predicting motor vehicle crashes on transportation facilities.
By adapting distributions into a way of measuring how often collisions might occur,
the probability of the number of crashes can be calculated. The range of statistical
models commonly in practice includes Binomial, Poisson, Poisson-gamma (or
negative binomial), zero-inflated Poisson and negative binomial models (ZIP and
ZINB), and multinomial probability models (Lord, Washington, & Ivan, 2005).
Initially, Poisson and negative binomial were proposed (Miaou, 1994; Oh et al.,
2006). Some researchers have developed and applied more advanced models, such as
zero inflated Poisson and zero altered probability models overcome problem due to
over-dispersion and excess-zeros which are commonly occurs in data processing
(Miaou, 1994; Hall, 2000; Austin and Carson, 2002; Lord et al., 2005; Oh et al.,
2006; Chin and Quddus, 2003; Kumara and Chin, 2003, Lord et al., 2007).
These types of modelling are most advantages when there are a large number of
experimental variables coupled with influences from various sources that are difficult
to control experimentally. But, the variables in the models are usually restricted to
some easily observable factors, such as AADT, speed, and geometric features. They
generally fail to take into account other significant explanatory factors such as driver
behaviour and errors which in fact are the critical causes of accidents (Hankey et al.,
1999; Stanton and Salmon, 2009). Therefore, significant effort still needed to address
© ASCE

these issues adequately for making real world decisions by using this models (Yang,
2012).
Risk models: These models identify and quantify risk factors that explain and predict
individual road-user behaviour and make safety assessments based of the risk-
reduction effect of various countermeasures. However, most of the model are not so
comprehensive in individual level. Often extreme theoretical concepts are found,

among them a few are suitable for meta level (on a global), others are on a micro (on
a very specific) level, this leads to a problematic output (see e.g. Näätänen and
Summala, 1976; Molen and Bötticher, 1987; Shinar1978; Michon, 1989; OECD,
1997).
Accident consequence model: Accident Consequence Models are used to describe

accident consequences and their implications and eventually aid in the development
of countermeasures to reduce injury consequences by identifying influential factors
such as those related to the roadway environment, vehicle safety and emergency
services; or alternatively, by promoting safety equipment or influencing driver
behaviour.
There are various ways to describe the injury level such as International
Classification of Diseases (ICD), Abbreviated Injury Scale (AIS), Injury Severity
Score (ISS). However, there is a lack of a quantitative definition of severity
(Hutchinson, 1986) and the thresholds of the accident categories are more or less
arbitrary.
Therefore, it is recognized that lack of congruency between models and theories at

different hierarchical levels of concept is one of the major problems of traffic safety
research, in particular those at the macroscopic and microscopic level. It is also
believed that the reason for these theoretical differences is attributable to the way in
which top-down and bottom-up research is carried out and the different aims and
goals of the two different approaches. Correspondingly, integration of model is likely
to require the amalgamation of many different scientific disciplines within an
integrated theoretical framework (OECD, 1997).
Traditional Approaches Based on Person Judgment
Identification of road engineering deficiencies to access the safety risk, nowadays is

one of the most powerful and widely accepted tools at the disposal of road safety
practitioners around the world. These are mainly known as proactive approaches by
which the safety of a road is evaluated on the basis of the features of the road and
roadside, without recourse to crash records (Job, 2012). Generally, these approaches
are based on engineering deficiencies identified by various means of audit and road
ratings. These approaches could be divided into two broad categories, Road Safety
Audit/Inspection and Safe System Approach. A summary with the advantages and
disadvantages of those approaches are given in Table 4.
© ASCE

Road Safety Audit (RSA)/Inspection (RSI)

Road Safety Audit (RSA) is a formal examination in which an independent, qualified
examiner’s report on the accident potential and safety performance is undertaken
(Austroads, 1994). It is not simply a 'check' but is more a vigorous and structured
process that requires a detailed examination of a road scheme, a written report from
the auditor, and a subsequent response by the project manager stating why
recommended actions have/have not been adopted (Hoque, 1997).
Training and accreditation of professionals in the auditing process are considered

vital if safety audit is to retain credibility as a powerful safety process (Appleton,
1996). It is also important that the safety auditors keep abreast of new developments
which will aid safety and should be aware of the principles of planning for safety of
the road network (Sabey, 1993).
Table 4. Summary of Traditional Approaches based on Person Judgement

Requirements
Job, 2012;  Accredited team  Proactive  Need skilled auditor but these skills are
Austroads, 1994; of professional approaches not easy to come by and there is need
Hoque, 1997; with sound  Detect design for continued safety audit training.
Ogden, 1996; knowledge on faults translated  Lack of accredited audit professionals,
Appleton, 1996; transportation into reality. accreditation body and training facility
Sabey,1993; engineering and  Do not depends in many countries.
YANG et al., safety on accident  Level of success relies heavily upon the
2012; iRAP, 2012. database auditors’ experience and individual
Road Safety Audit
preferences.
(RSA)/Inspection
(RSI)  This approach only deal with
engineering features and do not take
into account other important
explanatory factors such as user’s
behaviour and error
 iRAP only consider the road and
roadside features. Behaviour,
enforcement, vehicle fitness are out of
focus of this methodology
Mooren et al.,  Team of skilled  Proactive  Fully person dependent, which is not
2009; Tingvall, professionals approaches do free from biasness and misjudgement
1998; ATC, 2011;  Stakeholders not wait for  Mainly assesses the risk of road on the
Safe system MoT, 2010; collaboration accident basis of road and roadside features that
approach WHO, 2010a; and support occurrence. are present, other issues are out of focus
Roper & Turner,  Holistic  Gaining the support of different
2008; Chen & approach stakeholders is important but difficult to
Meuleners, 2011. gain.
With the principle of RSA, another tools have been developed for evaluating roads
and acquired enormous international recognition which is termed as the Road
Assessment Program in its various guises (iRap, euroRAP, etc.). iRAP is in use in
over 70 counties with over 500,000km of road assessed (iRAP, 2012). Though, iRAP
is an important (and increasingly powerful) tool to commence the risk identification
and treatment assessment process, it only consider the road and roadside features.
Other issues like user behaviour, enforcement, vehicle fitness are out of focus of this
methodology, whereas these issues are the major contributing factors of accident.
© ASCE

Safe System Approach

This is relatively a new approach based on the risk inherent in road and roadside
features. At present, it is working well for road safety (Mooren et al., 2009, Tingvall,
1998), and in consequence is increasingly being adopted internationally. The
principles are endorsed by Commission for Global Road Safety (2008), and form the
explicit basis of the Australian National Road Safety Strategy 2011-2020 (ATC,
2011), the New Zealand Road Safety Strategy: Safe Journeys (MoT, 2010), and the
Global Plan for the Decade of Action for Road Safety 2011-2020 (WHO, 2010a).
This approach has grown from the safety audit process (particularly audits of existing
roads) and the movement has occurred for a number of reasons. ‘Firstly, Australia has
adopted the ‘Safe System’ approach which includes the need to address all locations
where fatal or serious crashes may occur and not just those where crashes have
previously occurred. Secondly, there are legal liability issues (stemming from the
removal of non-feasance) with a need for road authorities to know where risks lie on
their roads. Thirdly, there is a decreasing number of treatable crash blackspots,
especially in rural areas. By just treating black spots, a large proportion of crashes are
neglected. Finally, based on research over the last few decades, far more is now
known about the road elements that influence the level of risk, and this information is
able to be used to estimate which sites are likely to be of higher risk’ (Roper &
Turner, 2008). However, in this approach an observer assesses the risk of road on the
basis of road and roadside features that are present. Therefore, this method is also
fully person dependent, which is not free from biasness and misjudgment.
CONCLUSIONS
Summary
This paper has reviewed and summarized the prevailing practices of traditional
approaches of road safety evaluation to understand different approaches, their
principles and requirements for effective, efficient and reliable outcomes. Advantages
and disadvantages of those approaches are highlighted.. From this comprehensive
examination, it is clear that traditional road safety evaluation approaches, whether
based on historical data or field observation, have significant limitations and
application challenges. First of all, the traditional approaches to safety evaluation has
very often been to “wait and see”, i.e. safety countermeasures are not considered until
the accident situation becomes unacceptable.
Most of the traditional approaches are heavily reliant on historical crash data,
whereas many of the crashes are not reported consistently in crash databases. In
addition to the problems related to the underreporting and inconsistent recording,
there are some other issues of crash data which are significant challenges. Firstly,
accumulation of sufficiently large number of sample data is necessary to obtain a
reliable result from analysis which are statistically sound. As a crash is a random
event, a long time period (around 3 to 5 years) is needed to get a statistically
significant sample of data. This problem become worse when this sample is
segregated by different features, such as by location, types, time of occurrence etc. in
© ASCE

order to get more understanding of crash causation factors. Secondly, in many

countries, the reported crash data are not readily available according to the
requirements. Many important missing parameters may have significant influence in
crash causation, as well as modelling approaches for evaluation. Thirdly,
underreporting of crashes and lack of consistent detailed information in the reported
records, which is a common problem in many jurisdictions, often restrict the ability to
understand the chain of events of crash occurrence process. Finally, historical crash
record based these approaches of safety evaluation might be unethical in nature as it
demands sufficiently large number of crashes before any preventive or corrective
measures.
These problems have led to the development of some proactive approaches based on
personal judgment. However, these approaches are fully person dependant, which is
not free from biasness and misjudgement. They mainly rely on engineering features
and do not take into account other important explanatory factors, such as driver error
which may be one of the major causes of crashes.
Future directions
In order to overcome the shortcomings and challenges of historical data based safety
evaluation approaches, many ways of employing non-accident data have been
suggested. These include travel speeds, speed variances, erratic manoeuvres, and
traffic conflicts (Debnath, A. K., et al., 2014). Among these surrogate measures, one
of the most popular and widely used forms of non-accident information is traffic
conflict, which are defined as critical incident not necessarily involving collisions.
Since the 1960’s, significant numbers of studies have been carried out to develop,
validate and apply different conflict measurement indicators for the evaluation of
traffic safety in different road geometric and operation condition as a substitute of
historical database approaches in traffic safety research (e.g., Hayward, 1971; Hyden,
1987; Allen, B.L. et al., 1978; von Buseck et al., 1980; Chin, H.C, et al. 1991;
Minderhoud and Bovy, 2001; Archer, 2005; FHWA, 2008; Ozbay et al., 2008; Yang,
2012; Cao et al., 2014). Among those Time-to-Collision (TTC), Time-to-Accident
(TA), Post-Encroachment Time (PET), Potential Index for Collision with Urgent
Deceleration (PICUD), Critical Gap (CG), Crash Potential Index (CPI), are most
prevalent.
The most appealing aspect of this approach is that conflict data can be gathered
within a shorter time period thereby eliminating the need for waiting longer time to
accumulate sufficient number of crash data. This approach analyses the safety
problem proactively before the collision. Thus, the analysis will be not only less
affected by time-dependent factors, but also the ethical problem associated with the
need of significant crash record is not present. The effectiveness of any safety
program can also be assessed in a shorter period of time (Chin & Quek, 1997). In
spite of the significant benefits, there are some issues related to surrogate safety
measures based on traffic conflict technique (TCT) which have hindered the wider
use of this approach. For example, discrimination of conflicts from observation,
validation of process and optimum threshold values to be used. However, the
© ASCE

development of advanced automated video analysis systems, coupled with

improvement in different micro-simulation modelling, have offered a wider
perspectives in understanding and application of TCT for safety evaluation (for
further details, refer to Archer, 2005; Yang, 2012; Laureshyn et al., 2010).
Therefore, in order to perform a more comprehensive form of safety evaluation and to

assess and predict levels of traffic safety at specific types of traffic facilities, TCT
based safety evaluation with the application of automatic discrimination of traffic
conflict from traffic steam video footage coupled with advanced simulation model
would be a better alternative in future. This alternative approach could be more
informative, more resource effective, more reliable and might be a foundation for
predictive modelling without the need for (or in addition to) accident data in order to
estimate safety impact with an acceptable level of statistical accuracy.
REFERENCES
ADB (1996). Road Safety Guidelines for the Asian and Pacific Region. Asian
Development Bank (ADB), Manila.
Alam, M.J.B. (2003). “Use of Accident Data and Accident Database.” Accident
Research Centre (ARC), BUET, Dhaka, Bangladesh
Allen, B.L., Shin, B.T., and Cooper, P.J. (1978). “Analysis of traffic conflicts and
collisions.” Report No. TRR 667, Transportation Research Board, Washington, D.C.
Appleton, I. (1996). “Progress with introduction of road safety audit in Australia
and New Zealand.” Joint 18th ARRB Transport Research conference and Transit
New Zealand Land Transport symposium ‘Roads96’.
Archer, J. (2005). “Indicators for traffic safety assessment and prediction and
their application in micro-simulation modelling: A study of urban and suburban
intersections.” Doctoral thesis, Sewden.
ATC (2011). National Road Safety Strategy 2011 – 2020. Australian Transport
Council,(https://infrastructure.gov.au/roads/safety/national_road_safety_strategy/files
/NRSS_2011_2020.pdf, accessed on 10 November 2015).
Austin, R. D. & L Carson, J. (2002). “An alternative accident prediction model
for highway-rail interfaces.” Accident Analysis & Prevention, 34, 31-42.
ANAO (2007). “The National Black Spot Programme: The Auditor-General
Audit.” Report No. 45 2007-07 Performance Audit, Australian National Audit Office
(ANAO), Commonwealth of Australia, Canberra.
ATSB (2004). “Road Safety in Australia: A Publication Commemorating World
Health Day 2004.” Australian Transport Safety Bureau (ATSB), Canberra
Austroads, (1994). “Road safety audits. Publication AP-30/94.” Austroads,
Sydney.
Barker, J., Baguley, C., (2001). A road safety good practice guide. Transport
Research Laboratory, Department for Transport, Local Government and the Regions,
Great Britain.
Berntman, M., (1994). “Metoder för insamling av uppgifter om svårt
trafikskadade – några källoroch tekniker.” Bulletin 6. Department of Traffic Planning
and Engineering. Road Construction, Lund University, Lund, Sweden.
© ASCE

Brehmer, B. (1994). “Psychological Aspects of Traffic Safety.” European

Journal of Operational Research, 75, pp. 540-552.
BTE, (2001). “The Blackspot Program 1996 – 2002. An evaluation of the first
three years.” Bureau of Transport Economics, Report no. 104: Canberra, Australia.
Cao, X., et al. (2014). “Automatic discrimination of traffic conflicts during lane
changing manoeuvres.” ARRB Conference, 26th, 2014, Sydney, New South Wales,
Australia.
Chen, H. and L. Meuleners (2011). "A literature review of road safety strategies
and the safe system approach." Curtin-Monash Accident Research Centre (C-MARC),
Curtin University, Australia, July, 2011.
Chin, H. & Quddus, M. (2003). “Applying the random effect negative binomial
model to examine traffic accident occurrence at signalized intersections.” Accident
Analysis & Prevention, 35, 253-259.
Chin, H.C., Quek, S.T. and Cheu, R.L. (1991). “Traffic conflicts in expressway
merging.” Journal of Transportation Engineering, Vol.117, No.6, pp.633-643.
Chin, H.C., Quek, S.T., 1997. “Measurement of traffic conflicts.” Safety Science
26, 169-185.
Council, T. S. (2009). “Before-and-After Study Technical Brief.” Institute of
Transportation Engineers.
Datta, T.K., Schattler, K.L., Kar, K., (2001). “Using GIS to analyze statewide
traffic crash data in Michigan.” AET European Transport Conference, Homerton
College, Cambridge, UK.
Debnath, A. K., et al. (2014). “Proactive safety assessment in roadwork zones: a
synthesis of surrogate measures of safety.” 2014 Occupational Safety in Transport
Conference, Gold Coast, Australia.
Eenink, R., Reurings, M., Elvik, R., Cardoso, J., Wichert, S., & Stefan, C. (2008).
“Accident prediction models and road safety impact assessment: recommendations
for using these tools.” Institute for Road Safety Research, Leidschendam.
Elvik, R. (2002). “The importance of confounding in observational before-and-
after studies of road safety measures.” Accident Analysis & Prevention, 34(5), 631-
635.
Elvik, R. (2008). “The predictive validity of empirical Bayes estimates of road
safety.” Accident Analysis & Prevention, 40(6), 1964-1969.
Elvik, R., Vaa, T., Erke, A., & Sorensen, M. (Eds.). (2009). “The handbook of
road safety measures.” Emerald Group Publishing
FHWA (2008). “Surrogate Safety Assessment Model and Validation: Final
Report.” Publication No. FHWA-HRT-08-051, Federal Highway Administration
(FHWA): Washington, D.C. JUNE 2008.
FHWA (2002). “Safety Effectiveness of Intersection Left- and Right-Turn
Lanes.” Report No. FHWA-RD-02-089. Washington, DC: U.S. DOT, FHWA, 2002.
Gibbs, Margaret et al. (2006). “Road Safety Audits: Case Studies.” Opus
Hamilton Consultants Ltd, Vanasse Hangen Brustlin Inc, Federal Highway
Administration. December, 2006.
Griffin, L.I. and R.J. Flowers (1997). “A Discussion of Six Procedures for
Evaluating Highway Safety Projects.” Washington, DC, USA: U.S. Department of
Transportation (U.S. DOT), Federal Highway Administration (FHWA), 1997.
© ASCE

Griffith, M. (1999). “Safety Evaluation of Continuous Rolled-In Rumble Strips

Installed on Freeways.” Transportation Research Record, No. 1665 (1999): 28–34.
Hakim, S., Shefer, D., Hakkert, A., & Hocherman, I. (1991). “A critical review of
macro models for road accidents.” Accident Analysis & Prevention, 23(5), 379-400.
Hall, D. B. (2000). “Zero-Inflated Poisson and Binomial Regression with Random
Effects: A Case Study.” Biometrics, 56, 1030-1039.
Hankey, J.M., Wierwille, W.W., Cannell, W.J., Kieliszewski, C.A., Medina, A.,
Dingus, T.A., Cooper, L.M., (1999). “Identification and evaluation of driver errors:
Task c report, driver error taxonomy development.” Project no. Dtfh-61-97-c-00051.
Virginia Tech Transportation Institute, Blacksburg, VA.
Hauer, E., (1986). “On the estimation of the expected number of accidents.”
Accident Analysis and Prevention, 18 (1), pp. 1-12.
Hauer, E. (1997). “Observational Before/After Studies in Road Safety. Estimating
the Effect of Highway and Traffic Engineering Measures on Road Safety.” Pergamon
Press/Elsevier Science Ltd., Oxford/UK.
Hayward, J.C. (1971). Near misses as a measure of safety at urban intersections.
Dissertation, The Pennsylvania State University.
Hoque M.M., (1997). “Road Safety Audit in Developing Countries.”
Transportation Research Group, Department of Civil and Environment Engineering,
University of Southampton, UK.
Hoque, M. M., Rahman, M. F., Ashrafuzzaman, M., & Sarkar, S. (2006).
“Observational studies of hazardous road locations on national highways in
Bangladesh.” 22nd ARRB Conference, 2006, Canberra, Australia
Hoque M.M., Mahmud S. M. S., Siddique C.K.A. (2007). “Road Safety in
Bangladesh and Some Recent Advances.” 14th International Conference on Road
Safety on Four Continents, 14-16 November 2007, Bangkok, Thailand.
Hoque M. M., Mahmud S. M. S. (2009). “Road Safety Engineering Challenges in
Bangladesh.” 13th Conference of the Road Engineering Association of Asia and
Australasia (REAAA), 23-26 September 2009, Songdo Convensia, Incheon, Korea,
Paper Topic-Road Safety, code RS 6-30.
Hsiao C. (1986). “Analysis of panel data.” Econometric Society Monographs no
11, Cambridge University Press, Cambridge.
Hutchinson T.P. (1986). “Statistical modelling of injury severity with special
reference to driver and front seat passengers in single-vehicle crashes.” Accident
analysis and prevention. New York. Vol. 18, No 2, p.157-167.
Hyden, C. (1987). The development of a method for traffic safety evaluation: the
Swedish traffic conflicts technique. Lund Institute of Technology, Lund.
iRAP (International Road Assessment Program) (2012). “Vaccines for Roads
(2nd ed.).” iRAP: Hampshire, UK.
James, H., (1991). “Under-reporting of road traffic accidents.” Traffic
Engineering and Control 32 (12), pp. 574-583.
Job, R. S. (2012). “Advantages and disadvantages of reactive (black spot) and
proactive (road rating) approaches to road safety engineering treatments: When
should each be used?” Australasian Road Safety Research Policing Education
Conference, 2012, Wellington, New Zealand.
© ASCE

Jovanis, P. & Chang, H. (1986). “Modeling the relationship of accidents to miles

traveled.” Transportation Research Record, 1068, 42-51.
Kumara, S. & Chin, H. (2003). “Modeling accident occurrence at signalized tee
intersections with special emphasis on excess zeros.” Traffic Injury Prevention, 4, 53-
57.
Laureshyn, A., Svensson, A., and Hyden, C. (2010). “Evaluation of traffic safety
based on micro-level behavioural data: Theoretical framework and first

implementation.” Accident Analysis and Prevention, 42: 1637–1646. doi:10.1016/
j.aap.2010.03.021. PMID:20728612.
LGED (2015). Rural Road Safety Manual for LGED. Local Government
Engineering Department (LGED), Bangladesh, June 2015
Lord, D., Washington, S. P., & Ivan, J. N. (2005). “Poisson, Poisson-gamma and
zero-inflated regression models of motor vehicle crashes: balancing statistical fit and
theory.” Accident Analysis & Prevention, 37(1), 35-46.
Lord, D., Washington, S. & Ivan, J. (2007). “Further notes on the application of
zero-inflated models in highway safety.” Accident Analysis & Prevention, 39, 53-57.
Meuleners, L., & Fraser, M. (2008). “Review of the Wa state black spot program:
a literature review of Australian and international black spot programs.” Centre for
Population Health Research, School of Public Health, Curtin University of
Technology, Perth.
McCullagh, P. and Nelder, J. (1983). “Generalized linear models.” Chapman and
Hall, New York.
McGill, J. et al. (2005). Road Safety Audit Guidelines. Science Applications
International Corporation, Synectics Transportation Consultants, Kittelson and
Associates, CTRE, Penn State, Federal Highway Administration. Washington, D.C.
February, 2005.
Miaou, S.-P. (1994). “The relationship between truck accidents and geometric
design of road sections: Poisson versus negative binomial regressions.” Accident
Analysis & Prevention, 26, 471-482.
Miaou, S.-P., Lord, D., 2003. “Modeling traffic crash-flow relationships for
intersections: dispersion parameter, functional form, and Bayes versus Empirical
Bayes.” Transportation Research Record 1840, 31–40.
Michon, J.A. (1989). “Explanatory Pitfalls and Rule-Based Driver Models.”
Accident Analysis and Prevention, 21, 4, pp. 341-353.
Minderhoud, M.M., and Bovy, P.H. (2001). “Extended time-to-collision measures
for road traffic safety assessment.” Accident Analysis and Prevention, 33: 89–97.
doi:10.1016/S0001-4575(00)00019-1. PMID:11189125.
Molen V.D.H.H. and A.M.T. Bötticher (1987). “Risk models for traffic
participants - A concerted effort for theoretical operationalizations”. Rothengatter,
J.A. and de Bruin, R.A. (eds). Road users and traffic safety. Assen, Van Gorcum, NL.
Mooren, L, Grzebieta, R., Job, R.F.S. Williamson, A. (2011). “Safe System-
International Comparisons of this Approach.” Australasian College of road Safety
Conference, Melbourne, September, 2011
MoT (New Zealand) (2010). Safe Journeys: New Zealand’s Road Safety Strategy
2010-2020. Wellington: Minsitry of Transport (MoT), (New Zealand)
© ASCE

Namjune, P., & Ihn, L. Y. (2013). “Comparison of Empirical Bayes Method and
Before-After Study Method.” Eastern Asia Society for Transportation Studies (Vol.
9).
Näätänen R. and H. Summala (1976). “Road User Behavior and Traffic
Accidents.” North- Holland Publishing Company, Amsterdam.
O’Day J (1993) “Accident data quality.” National Cooperative Highway Research
Program. Synthesis of Highway Practice 192. 48 p. (Transportation Research Board,

Washington, DC)
OECD. (1997). Road safety principles and models: review of descriptive,
predictive, risk and accident consequence models. Retrieved from OECD Road
Transport Research, Paris
Ogden K. W. (1996). Safer Roads: A Guide to Road Safety Engineering. Institute
of Transport Studies, Department of Civil Engineering, Monash University,
Melbourne, Australia.
Oh, J., Washington, S. P. & Nam, D. (2006). “Accident prediction model for
railway-highway interfaces.” Accident Analysis & Prevention, 38, 346-356.
Ozbay, K., et al. (2008). “Derivation and validation of new simulation-based
surrogate safety measure.” Transportation Research Record: Journal of the
Transportation Research Board(2083): 105-113.
Persaud, B., & Lyon, C. (2007). “Empirical Bayes before–after safety studies:
lessons learned from two decades of experience and future directions.” Accident
Analysis & Prevention, 39(3), 546-555.
Pham, M.H., Mouzon, O., Chung, E., El Faouzi, N.E., (2008). “Sensitivity of road
safety indicators in normal and crash cases.” 10th International Conference on
Application of Advanced Technologies in Transportation, Athens, Greece.
Roper, P., & Turner, B. (2008). “Why do we need to take a risk assessment based
approach in road safety?” 23rd, ARRB Conference, 2008, Adelaide, South Australia,
Australia.
Rosman, D.L. and Knuiman, M.W. (1994). “A comparison of hospital and police
road injury data.” Accident analysis and prevention 26(2) 215-222.
RTSSC (2015). “Traffic Safety Evaluations”.
http://www.ugpti.org/rtssc/resources/tse.php, Accessed on 30 October 2015
Rumar, K. (1985). “The Role of Perceptual and Cognitive Filters in Observed
Behaviour.” In: L. Evans sand R.C. Schwing, (Eds.), Human Behaviour and Traffic
Safety, New York: Plenum Press.
Sabey B.E. (1993). Safety audit procedures and practice. Traffex’93, PTRC,
London.
Saccomanno, F. & Buyco, C. (1988). “Generalized loglinear models of truck
accident rates.” Traffic accident analysis and roadway visibility.
Sayed, T., El-Basyouny, K., & Pump, J. (2006). “Safety evaluation of stop sign
in-fill program.” Transportation Research Record: Journal of the Transportation
Research Board, (1953), 201-210
Searles, B. and Jamieson, J. (1983), “Accident Data: Other Sources.” Proceedings
of the Workshop on Traffic Accident Evaluation, Monash University, Australia.
Shinar D. (1978). “Psychology on the road: The human factor in safety.” Wiley,
London/New York.
© ASCE

Simpson, H., (1997). “National hospital study of road accidents. Road Accidents
Great Britain:1996 -The Casualty Report.” Department of the Environment, Transport
and the Regions, Government Statistical Service.
St-Aubin P. G. (2011). “Traffic Safety Analysis for Urban Highway Ramps and
Lane-Change Bans Using Accident Data and Video-Based Surrogate Safety
Measures.” Master of Engineering Thesis, Department of Civil Engineering and
Applied Mechanics McGill University, Montréal. December 2011.

Stanton, N.A., Salmon, P.M., (2009). “Human error taxonomies applied to
driving: A generic driver error taxonomy and its implications for intelligent transport
systems.” Safety Science 47 (2), pp. 227-237.
SWOV, (2007). “The blackspot approach.” SWOV Factsheet. SWOV,
Leidschendam, the Netherlands. September 2007.
http://www.swov.nl/rapport/Factsheets/FS_Blackspots.pdf.
Tingvall, C. (1998). “The Swedish Vision Zero and how Parliamentary approval
was obtained.” Australasian Road Safety Research, Policing and Education
Conference, Wellington, New Zealand, 1998
TRL, (2003). Bangladesh Road Crash Costing Discussion Document, TRL,
Crowthome, UK, April, 2003.
Turner, B (2007). “Automatic collection of safety related road and roadside data.”
Australasian Road Safety Research, Policing and Education Conference, Melbourne,
Australia.
University of Zurich And Swiss Federal Institute of Technology (1986). “The car-
pedestrian collision, injury reduction, accident reconstruction, mathematical and
experimental simulation.” Interdisciplinary Working Group for Accident Mechanics.
Zurich. 151 p.
von Buseck, C., Evans, L., Schmidt, D., and Wasielewski, P., (1980). "Seat Belt
Usage and Risk Taking in Driving Behavior," SAE Technical Paper 800388, 1980,
doi:10.4271/800388
WHO (2009). Global status report on road safety: time for action. World Health
Organization, 2009.
WHO (2010). Data systems: A Road Safety Manual for Decision-Makers and
Practitioners. Geneva World Health Organization, 2010
WHO (2010a). Global Plan for the Decade of Action for Road Safety 2011-2020,
World Health Organization (WHO), , Geneva
Williams, A.F., Paek, N.N., and Lund, A.K., (1995). “Factors that Drivers say
Motivate Safe Driving Practices.” Journal of Safety Research, 26 (2), pp.119-124
Yang, H. (2012). “Simulation-based evaluation of traffic safety performance
using surrogate safety measures”. Doctoral dissertation, Rutgers University-Graduate
School-New Brunswick.
Yang, H., Ozbay, K. & Bartin, B. (2012.) “Application of Simulation-Based
Traffic Conflict Analysis for Highway Safety Evaluation.” 12th WCTR, July 11-15,
2010 – Lisbon, Portugal
Zegeer, C.V., Deen, R.C., (1978). “Traffic conflict as a diagnostic tool in
highway safety.” Transportation Research Record. Journal of the Transportation
Research Board No. 667, pp. 48-55.
© ASCE
View publication stats Bridging the East and West

Traditional Approaches To Traffic Safety Evaluatio

Uploaded by

Copyright:

Available Formats

Traditional Approaches To Traffic Safety Evaluatio

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Traditional Approaches To Traffic Safety Evaluatio

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

Traditional Approaches to Trafﬁc Safety Evaluation (TSE): Application

Conference Paper · May 2016

S. M. SOHEL Mahmud Luis Ferreira

SEE PROFILE SEE PROFILE

Ahmad Tavassoli Hojati

Freight modal research View project

Transport planning research View project

The user has requested enhancement of the downloaded file.

Traditional Approaches to Traffic Safety Evaluation (TSE):

Brisbane, QLD 4072, Australia; Ph (+61) 469 911 070; email:

Traditional TSE approaches, based on chronological data including classical

Keywords: Traffic Safety Evaluation, Traditional approaches, Chronological data,

Traffic Safety Evaluation (TSE) is an examination of the safety and performance of a

Bridging the East and West

planning, design, implementation; 5) delivering a continuous progression of safety

ISSUES FOR TRAFFIC SAFETY EVALUATION (TSE)

Bridging the East and West

The performance of a treatment or group of treatment depends on two types of factor,

Treatment effect: The change of safety performance of a transportation facility due to

Exposure effect: As the road environment is a dynamic thing, the exposure

Bridging the East and West

Factors related to crash data

Reporting bias: Accuracy problems of accident data primarily stemming from

Bridging the East and West

transcription or coding (O’Day, 1993). Furthermore, despite a uniform reporting

Location errors. The location of an accident may easily be incorrect or imprecise in

Therefore, development of unbiased, consistent, reliable, comprehensive quality and

TRADITIONAL APPROACHES FOR TSE

Bridging the East and West

Table 1. Different Traditional Approaches of Traffic Safety Evaluation (TSE)

Traditional Approaches Based on Historical Data

Traditional Approaches Based on Historical Data

Bridging the East and West

A summary comprising major data requirements, advantages and disadvantages of

Table 2. A summary of Before-after Observation Approaches

Approach Reference/s Data Advantages Disadvantages

Bridging the East and West

respectively (Griffin and Flowers, 1997). A group of similar facilities is selected so

Black Spot Identification Program

Bridging the East and West

Mainstream/Statistical Modelling Approaches

Data requirements, advantages, disadvantages with the important references related to

Table 3. Advantages, Disadvantages of Modelling Approaches

Approach name Reference/s Data Advantages Disadvantages

Bridging the East and West

Approach name Reference/s Data Advantages Disadvantages

models, time Chang,1986.

Miaou, 1994; independent distributional models depends on the use of

Bridging the East and West

Predictive or analytical model: An Accident Prediction Model is a mathematical

Conventional (multiple) linear regression model: Initial crash prediction models

Statistical modelling approach: To overcome these limitations, statistical modelling

Bridging the East and West

comprehensive in individual level. Often extreme theoretical concepts are found,

Accident consequence model: Accident Consequence Models are used to describe

Therefore, it is recognized that lack of congruency between models and theories at

Traditional Approaches Based on Person Judgment

Identification of road engineering deficiencies to access the safety risk, nowadays is