Linkping Studies in Science and Technology

Dissertation No. 1100

Road Safety Development Index (RSDI)

Theory, Philosophy and Practice

Ghazwan Al-Haji

Department of Science and Technology

Linkping University, SE-601 74, Norrkping, Sweden

Norrkping, Sweden 2007

 Ghazwan Al-Haji, 2007

Dissertation Number: 1100

ISBN: 978-91-85715-04-6
ISSN: 0345-7524
Printed by:
LiUTryck, Linkping, Sweden, 2007
Distributed by:
Linkping University
Department of Science and Technology (ITN)
Campus Norrkping
SE-601 74, Norrkping, Sweden
Tel: +46 11 36 30 00
Fax: +46 11 36 32 70
http://www.itn.liu.se

ii

ABSTRACT
This dissertation develops, presents and tests a new international tool, the so-called
Road Safety Development Index (RSDI), which indicates in a comprehensive and easy
way the severity of the road safety situation in a specific country and/or in comparison
with other countries. There are three pillars of outcomes involved in the framework of
RSDI. One pillar is the People focus (road user behaviour). The second is the System
focus (safer vehicles, safer roads, enforcement, management, etc). The third is the
Product focus in terms of accident death rates. This thesis analyses each of these pillars.
In addition, RSDI links the key national practices of road safety to each other and to the
end-results (accident death rates). The study suggests a master-list of performance
indicators to be implemented for assessing road safety level in a country and for RSDI
building. Based on the master-list, a short key list of performance indicators is chosen
and classified into two primary categories that correspond to two groups of countries:
LMCs Less Motorised Countries and HMCs Highly Motorised Countries. RSDI
aggregates the key performance indicators into one single quantitative value (composite
index). Four main objective and subjective approaches are used to calculate RSDI and
determine which one is the best. One approach uses equal weights for all indicators and
countries, whereas the other approaches give different weights depending on the
importance of indicators. Two empirical studies were carried out, in different parts of
the world, to determine the applicability of this tool in real world applications. The first
empirical study comes from eight European countries (HMCs). The second empirical
study comes from five Southeast Asian countries (LMCs). The RSDI results from this
study indicate a remarkable difference between the selected countries even at the same
level of motorisation and/or with close accident death rates. The unavailability of
comparable and useful data are problems for deeper analysis of RSDI, especially the
index should be as relevant as possible for different parts of the world. The empirical
and theoretical assessments prove that RSDI can give a broader picture of the whole
road safety situation in a country compared to the traditional models and can offer a
simple and easily understandable tool to national policy makers and public.
Key Words: Road safety, RSDI, international benchmarking, national development,
policy makers, ranking, composite indices, key performance indicators, macro- models.

iii

SAMMANFATTNING
Denna avhandling utvecklar, presenterar och testar ett nytt internationellt verktyg, det s
kallade Road Safety Development Index (RSDI), vilket p ett begripligt och
lttillgngligt stt beskriver trafikskerhetslget i ett visst land jmfrt med andra
lnder. Resultatet av RSDI utgrs av tre grundpelare. Den frsta pelaren r Fokus p
mnniskor (vgtrafikbeteende). Den andra r Fokus p systemet (skrare fordon, skrare
vgar, beivrande, management, osv). Den tredje pelaren r Fokus p produkten med
avseende p antal dda per fordon och per invnare. Arbetet analyserar var och en av
dessa tre pelare. RSDI kopplar dessutom samman de viktigaste nationella praxisarna
och erfarenheterna med varandra och till slutresultaten (antal ddsfall). Studien freslr
en lista med de viktigaste indikatorerna p hur olika lnder vidtar tgrder fr
trafikskerheten. Grundat p denna master-lista kan en kort lista med de viktigaste
indikatorerna skapas och klassificeras i tv huvudkategorier fr tv typer av lnder:
LMC lnder med lg andel fordon och HMC lnder med hg andel fordon. RSDI
aggregerar de viktigaste performance-indikatorerna till ett enda kvantitativt mtt (ett
sammansatt index). Fyra olika objektiva och subjektiva huvudangreppsstt anvnds fr
att berkna RSDI och bestmma vilket av dem som r det bsta. En metod anvnder sig
av lika stora vikter fr alla indikatorer och lnder, medan en annan metod ger olika
vikter beroende p indikatorernas betydelse. Tv empiriska studier genomfrdes i olika
delar av vrlden fr att bestmma tillmpligheten av detta verktyg i verkliga situationer.
Den frsta empiriska studien kommer frn tta lnder i Europa (HMC-lnder). Den
andra empiriska studien har gjorts i fem lnder i Sydostasien (LMC-lnder). Resultaten
frn detta RSDI tyder p en anmrkningsvrd skillnad mellan de valda lnderna, ocks
om andelen bilgare och/eller andra variabler fr trafikskerhet hlls konstanta. Bristen
p jmfrbara och anvndbara data medfr problem vid en djupare analys av RSDI fr
olika delar av vrlden. De empiriska och teoretiska skattningarna visar att RSDI kan ge
en bredare bild av hela trafikskerhetssituationen i ett land jmfrt med traditionella
modeller och kan erbjuda ett enkelt och lttfrsteligt verktyg fr de nationella
beslutsfattarna liksom fr allmnheten.

Acknowledgments
The dissertation has now reached the end of a long and enjoyable journey. On this
occasion I would like to acknowledge many people for their help along the way.
First and foremost, my deepest gratitude is to my thesis advisor, Prof. Kenneth Asp, for
his constant support, encouragement and advice during my doctoral studies. It was a
privilege to work with you and benefit from your broad knowledge, management and
international experience. Next, I would like to express my special thanks to Prof. Jan
Lundgren who has reviewed my thesis and enriched it with his insightful comments and
valuable advices. Thank you also for facilitating everything to finish my thesis.
My sincere appreciation goes to my projects members and colleagues. To Per Lindskog,
Malin Eriksson, Ing-Marie Eriksson, Johanna Emilsson and Lars Ohlsson. We shared
offices, discussions, projects, papers, conferences and success. Thank you for a great
time and mutual experience! I am grateful to Prof. Kre Rumar, for his inspiring ideas
and fruitful discussions, especially during the RetsNet project work. To Arne Karyd, my
office mate, and Somharutai Bootjan for our interesting discussions. To Di Yuan for
your review and comments on my dissertation during slutseminarium Final seminar.
To ke Sivertun and Imad Ali from IDA department for our joint works within the
projects Spider and Globesafe. My acknowledgement goes also to Prof Christer Hydn
and Dr. se Svensson at Lund University for their feedback and constructive
suggestions on my thesis work. I would also like to thank the administrators and
directors of the ITN department for your great kind help in different matters.
This thesis has also profited from the collaboration with international institutions such
for instance ADB (Asian Development Bank) and GRSP (Global Road Safety
Partnership) during ASNet project. I would specifically like to mention the following
consultants: Charles M. Melhuish (ADB), Alan Ross (ADB), Michael Goodge (ADB),
David Silcock (GRSP) and Andrew Downing (GRSP). My thanks also go to several
people from overseas whom I met through the projects: RetsNet, ASNet, TechTrans and
Spider. Special thanks to Prof. Valentine Silyanov and Sr. Scientist Anatolyi Utkin at the
State Technical University (MADI) in Moscow, Russia; to Prof. Heru Sutomo and Dr.
Arif Wismadi at Gadja Mada University, Yogyakarta, Indonesia; and to Prof. Nguyen
Xuan Dao and Ms. Trinh Thuy Anh at the University of Communication and
Transportation, Hanoi, Viet Nam. Thank you for making our joint work and
applications interesting and valuable.
My deepest gratitude and love go to my parents in Syria, my brothers and sisters and
their families, for supporting me at all times. I express my sincere thanks to my older
sister Nadia for her frequent contact. Last, but certainly not least, I would like to thank
my beloved wife Mai for her encouragement and support, and our son Aghyad who has
enriched my Swedish vocabulary with his new words he learned from Kindergarten and
who often took my place in front of the computer.
Norrkping, Sweden, March 2007
Ghazwan Al-Haji

vii

List of Abbreviations
ADB
ASEAN
ASNet
EU
GDP
Globesafe
GRSP
HDI
HMCs
KPIs
LMCs
NGO
OECD
RSDI
SIDA
UN
UNDP
VRU
WB
WDI
WHO

Asian Development Bank

Association of South East Asian Nations
Regional Traffic Safety Network to Ten South East Asian Countries
European Union
Gross Domestic Product
Global Road Safety Database
Global Road Safety Partnership
Human Development Index
Highly Motorised Countries
Key Performance Indicators
Less Motorised Countries
Non-Governmental Organisation
Organisation For Economic Cooperation and Development
Road Safety Development Index
The Swedish International Development Agency
United Nations
United Nations Development Programme
Vulnerable Road Users
World Bank
World Development Indicators, World Bank
World Health Organisation

ix

Table of Contents
ABSTRACT................................................................................................................................................... III
SAMMANFATTNING................................................................................................................................... V
ACKNOWLEDGMENTS ........................................................................................................................... VII
LIST OF ABBREVIATIONS ...................................................................................................................... IX
LIST OF FIGURES ................................................................................................................................... XIII
LIST OF TABLES ......................................................................................................................................XIV

CHAPTER 1: INTRODUCTION ................................................................................................................1

1.1
1.1.1
1.1.2
1.2
1.3
1.4
1.5
1.6
1.7

BACKGROUND THE ROAD SAFETY PROBLEM ..............................................................................1

Road safety is a global problem..............................................................................................2
International benchmarking of road safety ............................................................................3
PURPOSES AND RESEARCH QUESTIONS .........................................................................................6
RESEARCH METHODS ....................................................................................................................7
SCOPE AND LIMITATIONS ..............................................................................................................9
THESIS STRUCTURE .......................................................................................................................9
CONTRIBUTIONS, PUBLICATIONS AND SIGNIFICANCE OF THIS STUDY ........................................12
DEFINITIONS OF TERMS ...............................................................................................................15

CHAPTER 2: LITERATURE REVIEW..................................................................................................19

2.1
PART ONE: PREVIOUS RESEARCH ON INTERNATIONAL BENCHMARKING OF ROAD SAFETY .........20
2.1.1
The first generation: Linking motorisation, traffic risk and personal risk ..........................22
2.1.2
The second generation: Linking traffic risk, motorisation and personal risk with time .....25
2.1.3
The third generation: The need for increased integration with many variables involved ..28
2.1.4
The fourth generation: Linking product, practices and strategic benchmarking................29
2.1.5
Summary from literature review (part 1)..............................................................................30
2.2
PART 2: PREVIOUS RESEARCH REGARDING ON MULTIDIMENSIONAL AGGREGATION .................30
2.2.1
Based on composite indices...................................................................................................31
2.2.2
Business excellence models...................................................................................................34
2.3
CONCLUSIONS .............................................................................................................................36
CHAPTER 3: THE THEORETICAL FRAMEWORK OF MACRO- PERFORMANCE
INDICATORS IN ROAD SAFETY IN ROAD SAFETY.......................................................................37
3.1

STAGE ONE: IDENTIFYING THE MACRO-INDICATORS IN RELATION TO RISK, EXPOSURE AND

CONSEQUENCES .........................................................................................................................................39

3.1.1
What are the Exposure, Risk and Consequences?................................................................40
3.1.2
Correlation between the quantified macro factors and road accidents...............................43
3.2
STAGE TWO: FINDING A LIST OF MACRO-DIMENSIONS IN ROAD SAFETY .....................................49
3.3
STAGE THREE: CRITERIA FOR SELECTING MACRO-PERFORMANCE INDICATORS .........................52
3.3.1
Sample of survey and the multidimensional index................................................................54
3.3.2
Quantitative versus qualitative indicators............................................................................54
3.3.3
IT supports the macro-performance indicators....................................................................54
3.4
BUILDING A MASTER-LIST OF MACRO-PERFORMANCE INDICATORS IN ROAD SAFETY ................54
CHAPTER 4: THE CONCEPTUAL FRAMEWORK OF RSDI ..........................................................59
4.1
4.2
4.3
4.4
4.5

INTRODUCTION............................................................................................................................59
WHAT IS RSDI? ..........................................................................................................................59
THE EXPECTED BENEFITS OF USING RSDI ..................................................................................61
RSDI QUALITY CRITERIA ............................................................................................................62
THE PRINCIPLES AND PHILOSOPHY BEHIND RSDI ......................................................................62

xi

4.6
4.7
4.8
4.9
4.10

THE RSDI MODEL FROM AN ECONOMIC PERSPECTIVE ...............................................................68

THE PROCESS OF RSDI DEVELOPMENT .......................................................................................70
SELECTING THE RIGHT INDICATORS TO BE ADDED INTO RSDI...................................................71
IDENTIFYING SHORT-TERM KEY LIST OF INDICATORS FOR ROAD SAFETY PERFORMANCE ..........74
CONCLUSION ...............................................................................................................................78

CHAPTER 5: METHODOLOGICAL APPROACHES.........................................................................79

5.1
5.2
5.3
5.4
5.5
5.5.1
5.5.2
5.5.3
5.5.4
5.6
5.7

INTRODUCTION............................................................................................................................79
CONSTRUCTION OF RSDI............................................................................................................81
NORMALISING THE INDICATORS .................................................................................................82
WEIGHTING THE VARIABLES .......................................................................................................83
COMBINING THE CHOSEN INDICATORS INTO RSDI BY USING DIFFERENT APPROACHES .............84
Approach 1: Using Simple Average......................................................................................84
Approach 2: Expert Judgements ...........................................................................................86
Approach 3: Subjective weights based on previous experience...........................................87
Approach 4: Principal Components Analysis.......................................................................89
POSSIBLE APPLICATIONS AND ILLUSTRATION OF RSDI..............................................................90
SUMMARY ...................................................................................................................................92

CHAPTER 6: EMPIRICAL APPLICATION 1: APPLYING RSDI TO HMCS IN EU....................93

6.1
6.2
6.2.1
6.3
6.4

BACKGROUND- ROAD SAFETY IN THE EU...................................................................................93

DATA AND INDICATORS INCLUDED IN RSDI...............................................................................95
Limitation and quality of data...............................................................................................95
CALCULATION OF RSDI..............................................................................................................98
SUMMARY OF RESULTS .............................................................................................................103

CHAPTER 7: EMPIRICAL APPLICATION 2: APPLYING RSDI TO LMCS IN ASIA ............. 105

7.1
7.2
7.3
7.4

BACKGROUND- THE ROAD SAFETY SITUATION IN SOUTHEAST ASIA ........................................105

DATA AND LIMITATIONS ...........................................................................................................106
CALCULATION OF RSDI............................................................................................................108
SUMMARY OF RESULTS .............................................................................................................112

CHAPTER 8: ANALYSIS AND DISCUSSION OF RESULTS ......................................................... 113

8.1
AN EMPIRICAL ASSESSMENT: RESULTS AND DISCUSSION ...........................................................114
8.1.1
Why does RSDI differ between countries?......................................................................... 116
8.1.2
Can RSDI indicators and results be generalised to all countries? ................................... 119
8.1.3
Can we combine the results from the two empirical studies together?............................. 120
8.1.4
Which approach (method) of RSDI is the best?................................................................. 121
8.2
THEORETICAL ASSESSMENT: THE S.W.O.T ANALYSIS ..................................................................123
8.3
SUMMARY .................................................................................................................................125
CHAPTER 9: CONCLUSIONS AND FUTURE WORK .................................................................... 127
9.1
9.2
9.3

SUMMARY OF RESEARCH APPROACH ........................................................................................129

CONCLUSIONS ABOUT EACH RESEARCH QUESTION...................................................................129
OVERALL CONCLUSIONS AND FUTURE WORK ...........................................................................135

REFERENCES.......................................................................................................................................... 137

xii

List of Figures
FIGURE 1.1: MOTORISATION, PERSONAL RISK AND TRAFFIC RISK IN DIFFERENT REGIONS ..........................3
FIGURE 1.2: RESEARCH APPROACH FOR PROMOTING BOTH THEORY AND PRACTICE ...................................8
FIGURE 1.3: STRUCTURE OF THE THESIS.....................................................................................................11
FIGURE 2.1: EVOLUTION OF ROAD SAFETY BENCHMARKING TOWARDS INTEGRATED BENCHMARKING ...21
FIGURE 2.2: THE INFLUENCING FACTORS ON THE DEVELOPMENT CURVE OF ROAD SAFETY......................24
FIGURE 2.3: ILLUSTRATION OF ROAD SAFETY PROFILES ...........................................................................29
FIGURE 3.1: THE PYRAMID OF ROAD SAFETY INDICATORS AND LEVELS OF AGGREGATION ......................38
FIGURE 3.2: ROAD SAFETY PROBLEM DESCRIBED BY THREE-DIMENSIONAL CUBE ....................................42
FIGURE 3.3: THE SELECTED DIMENSIONS IN ROAD SAFETY ........................................................................51
FIGURE 3.4: CRITERIA TO IMPROVE INDICATOR QUALITY AND IMPORTANCE. ...........................................57
FIGURE 4.1: RSDI CONCEPTUAL FRAMEWORK (OVERALL ROAD SAFETY PERFORMANCE)........................64
FIGURE 4.2: THE INPUT-OUTPUT-OUTCOME-RSDI FRAMEWORK...............................................................67
FIGURE 4.3: SHARING THE LONG-TERM VISION OF RSDI...........................................................................74
FIGURE 5.1: AN ILLUSTRATION OF THE COMBINED AGGREGATED RESULTS WITH WEIGHTS INTO RSDI...89
FIGURE 5.2: AN EXAMPLE OF ILLUSTRATION OF THE THREE RSDI LEVELS AND DIMESNIONS ..................91
FIGURE 5.3: AN EXAMPLE OF ILLUSTRATION OF THE THREE RSDI LEVELS AND DIMESNIONS ..................91
FIGURE 6.1: SCORE PLOT OF THE FIRST TWO PRINCIPAL COMPONENTS ..................................................103
FIGURE 7.1: SCORE PLOT OF THE NEW SCORES OF PC1 AND PC2 IN THE SAMPLE OF COUNTRIES ...........111
FIGURE 8.1: COMPARING THE RSDI SCORES BETWEEN THE SELECTED SAMPLES OF COUNTRIES ...........116
FIGURE 8.2: AN ILLUSTRATION OF THE DIFFERENCES BETWEEN VARIOUS PILLARS OF RSDI .................118
FIGURE 8.3: AN ILLUSTRATION OF THE DIFFERENCES BETWEEN VARIOUS DIMENSIONS OF RSDI ..........119
FIGURE 8.4: RSDI PERFORMANCE DEVELOPMENT ...................................................................................120

xiii

List of Tables
TABLE 1.1: THE RELATION BETWEEN THE RESEARCH QUESTIONS, STRATEGIES AND ANSWERS..................9
TABLE 3.1: MACRO-FACTORS INFLUENCING EXPOSURE, RISK AND CONSEQUENCES.................................48
TABLE 3.2: THE MASTER LIST OF MACRO ROAD SAFETY INDICATORS AND DIMENSIONS...........................56
TABLE 4.1: APPLYING ECONOMIC PERFORMANCE PRINCIPLES TO RSDI ...................................................69
TABLE 4.2: THE SUGGESTED LIST OF KEY PERFORMANCE INDICATORS FOR LMCS AND HMCS ...............77
TABLE 5.1: THE SUBJECTIVE WEIGHTING RESULTS IN A NUMERICAL SCALE OF RSDI ..............................87
TABLE 5.2: ALTERNATIVE CHOICES OF WEIGHTING THE PILLARS OF RSDI...............................................88
TABLE 6.1: ROAD SAFETY DATA AND KEY INDICATORS FOR THE SELECTED HMCS IN EU.......................97
TABLE 6.2: SELECTING THE WEIGHTS OF THE INDICATORS OF RSDI.........................................................99
TABLE 6.3: RSDI SCORES USING SIMPLE AVERAGE TECHNIQUE AND SUBJECTIVE THEORIES..................100
TABLE 6.4: THE EIGENVALUE ANALYSIS OF THE NORMALISED INDICATORS...........................................101
TABLE 6.5: THE ADJUSTMENT PROCEDURES OF PCA WEIGHTS TO MATCH THE SCALE OF THE RSDI .....102
TABLE 6.6: THE RSDI SCORES AND RANKS FROM THE EMPIRICAL ANALYSIS (1) ...................................104
TABLE 7.1: THE KEY PERFORMANCE INDICATORS AND DIMENSIONS IN THE SELECTED COUNTRIES .......107
TABLE 7.2: SELECTING THE WEIGHTS OF THE INDICATORS OF RSDI.......................................................109
TABLE 7.3: RSDI SCORES USING SIMPLE AVERAGE TECHNIQUE AND SUBJECTIVE THEORIES..................110
TABLE 7.4: EIGENVALUE ANALYSIS OF THE COVARIANCE MATRIX OF NORMALISED VALUES ................111
TABLE 7.5: THE ADJUSTMENT PROCEDURES OF PCA WEIGHTS TO MATCH THE SCALE OF THE RSDI .....112
TABLE 7.6: THE RSDI SCORES AND RANKS FROM THE EMPIRICAL ANALYSIS AND APPROACHES ...........112
TABLE 8.1: THE RSDI SCORES AND RANKS FROM THE TWO EMPIRICAL APPLICATIONS ..........................114
TABLE 8.2: SUMMARY OF THE STRENGTHS AND WEAKNESSES IN THE FOUR APPROACHES .....................122

xiv

Chapter 1: Introduction

Chapter 1

Introduction
The purpose of this introductory chapter is to provide the reader with a
brief background of this dissertation research and to define the purposes
as well as the research questions. Furthermore, the outline of the
dissertation is presented.

1.1

Background the road safety problem

It is a common perception among the public that road accidents are a problem, resulting
in death, injury or property damage. Unfortunately, many people do not fully realise the
size of the problem. It is clear that while many people, especially in the developing
countries, have the general idea that driver error is the main cause of the problem, they
have no idea that a several causes and factors contribute to the problem as well. Their
understanding, in most cases, is limited because they have no clear measurement(s) that
can show them the size of the problem in a simple and adequate way.
When a policymaker decides which actions those need to be taken nationally, it has to
be based on some sort of statistical measurements. When road users want to know their
accident risks, this should also be based on understandable statistical measurements.
Unfortunately, most present measurements that are used to address the scale of road
safety problem in a country or city are mainly based on death rates (deaths per vehicle
or per person). These rates are often too complex to be understood by ordinary people
and, in some cases, by policy makers since the scale of these measurements is not
uniform, and vary from one study to another and the results are mostly in a decimal
number. Furthermore, the death rates say little about achievement of a country or its
progress towards a certain goal. Therefore, it has always been of interest to researchers
to develop a measurement that provides public and policy makers with a clear
understanding of the causes and magnitude of traffic accidents in their countries.

1.1.1

Road safety is a global problem

As the number of motor vehicles is continuously increasing globally, and road traffic
accidents are causing more and more deaths and injuries. The World Health
Organisation (WHO) statistics (Peden M et al., 2004) have shown that almost 1.26
million people are killed in road accidents each year worldwide and an additional 50
million people are estimated injured. Nearly half of them are seriously injured or
disabled. Due to the unreliability and under-reporting of data in most countries, these
figures are still under-estimated. Road accidents are the eighth leading cause of death in
the world today, and the WHO (WHO, 2004) estimates they will become the worlds
third leading cause of death by the year 2020 if no effective actions and efficient
measures are taken.
Annually, the national cost of road accidents is estimated between one and three percent
of a countrys Gross National Product (GNP). This cost is a considerable waste of
resources and it also has negative effects on the development of every country, and
especially in the low-income countries.
All counties suffer from the road accident problem. Yet the size of the problem is
different from one country to another, because countries vary widely in their
development levels, road safety systems and experiences. According to Jacobs et al.
(2000), the majority of road deaths and injuries occur mostly in developing and
transitional countries. Highly Motorised Countries (HMCs) have sixty percent of the
total motor vehicle fleet but their contribution of the total global road accident deaths is
only fourteen percent. Several studies (i.e. OECD, 2002a) have shown that the total
number of road deaths in HMCs has been declining or stabilising during recent decades,
whereas the situation in in Less Motorised Countries (LMCs) remains severe and the
total number of deaths continues to increase.
Al-Haji (2001) has performed an international comparative study across different
regions in terms of motorisation (vehicles per person), personal risk (deaths per person),
and traffic risk (deaths per vehicle). There it was stated (Figure 1.1) that highly
developed countries have the lowest risk records and high motorisation, while Africa
has the lowest motorisation and a high traffic risk. Southeast Asia, Africa and the
Middle East run the highest risk of being killed in terms of personal safety. However,
the study recommended that such comparison should not be taken too seriously, since
there are differences within the same region concerning for instance: motorisation,
population, education, health, investment level on road safety measures, etc. In order to
achieve adequate results in comparisons, international comparisons have to be carried

out between similar countries or regions at the same level of development, motorisation
and with similar type of transport system as much as possible.
70
60
50
40
30
20
10
0
Motorisation

Personal Risk

Traffic Risk

Developed Countries (5)

Eastern Europe (2)

South East Asia (5)

Middle East (5)

Latin America (3)

Africa (6)

Notes: Motorisation is taken in this figure as the number of vehicles per 100 persons.
Personal Risk is deaths per 100,000 inhabitants.
Traffic Risk is deaths per 10,000 vehicles.
( ) = Number of selected countries.

Figure 1.1: Motorisation, personal risk and traffic risk in different regions in 1995 (Al-Haji, 2001)

Additionally, the characteristics and nature of the road safety problem differs between
countries. For instance, the majority of road accident injuries and deaths in LMCs are
the vulnerable road users (pedestrians, cyclists, motorcyclists), whereas, due to the high
number of car owners, car occupants account for most of the victims in HMCs (Asp et
al., 1998). Therefore, road safety priorities are different from country to country and
from region to region in accordance to their accident characteristics, nature, causes and
challenges.
1.1.2

International benchmarking of road safety

International benchmarking as a term has widely been used in many fields of research.
There are many examples of applications of international benchmarking and ranking
between countries, ranging from health care (i.e. NHS, 1999), higher education (i.e.
CHEMS, 1998), innovation (i.e. IRE, 2005), business management (i.e. Spendolini,
1992), management in government (Pollitt et al, 1994), public sector (i.e. Dorch &
Yasin, 1998), and many more. The main idea of benchmarking is to compare
achievements between countries (or organisations) and to learn from each other.
There are several definitions of benchmarking, depending on the type of activity and the
target group. In general, these definitions deal with some common topics such as:

comparison, sharing information and best practices. Two examples of definitions are
highlighted below:
Benchmarking is a powerful technique that provides practical learning through
comparing measurements, policies or outcomes, across industries, sectors,
policies, products or services. The essence of benchmarking is the process of
identifying the highest standards of excellence for products, services or
processes and then making the improvements necessary to reach those
standards. (IRE, 2005).
A continuous systematic process for evaluating the products, services and work
of organisations that are recognised as representing best practices for the
purpose of organisational improvement (Spendolini, 1992).
Similarly, in terms of road safety, many countries recognise the importance of
international benchmarking to measure their own achievements with similar countries or
countries that have already passed through similar stages of challenges and
development. This comparison allows countries to identify their problems and improve
their performance in road safety based on existing practices and lessons in other
countries. In general, these benchmarking models intended to answer:

Which country performs better than another?

Why is a specific country more successful than others?
How and what measures a successful country has used to improve its road safety
work.
What actions have to be taken to improve road safety performance in a country
in future?

A number of benchmarking models in road safety has already being developed and they
range from relatively simple to highly complex models depending on the number of
indicators involved, details of data and complexity of methods used in calculations and
analysis. These benchmarking models in road safety can be classified into four broad
categories as follows (Al-Haji & Asp, 2006b):
1. Product Benchmarking is used to compare death rates.
2. Practices Benchmarking is used to compare activities related to human-vehicleroad performance (e.g. seat belts use, crash helmets use, motorways level, etc.)
3. Strategic Benchmarking is used to compare National Road Safety Programme
(NRSP), management and organisational framework.

4. Integrated Benchmarking is used to compare countries in terms of the three

previous types of benchmarking altogether. Road safety performance in a country
is seen and perceived within a holistic context.
Although numerous research and applications have been carried out in the first three
types in macro benchmarking of road safety, little research has investigated the fourth
type (Integrated Benchmarking). The major possible obstacles in constructing an
integrated benchmarking model are:

Misunderstanding what integrated benchmarking means.

The lack of data from different countries, especially in LMCs.
May be very expensive in terms of money and the time taken to collect and
analyse data from many countries.

However, today data is more accessible. Faster computers are developing rapidly as
well, which simplifies the work and analysis of a large amount of road safety data that
was not available before. This development makes the work in the integrated
benchmarking easier, bringing it closer to reality.
Additionally, the term Sustainable Development has become more popular and
applied in different sectors of research. This term simply means integrating several
efforts at the same time for maximising the development of a specific sector. Examples
are Sustainable Transport, Sustainable Environment and Sustainable Health Care.
To date, little research has reported on Sustainable Road Safety Development, which
needs to be emphasised in research. One of the major challenges to sustain road safety
in a country is that the traditional measurements used regarding the first three types of
benchmarking are not powerful enough to model the complexity of road safety situation
in a country. Therefore, an integrated benchmarking tool can contribute to a sustainable
improvement in road safety in the country and bring all relevant concepts together. Van
Vliet & Schermers (2000) is one of few studies that examined the issue of sustainable
road safety by developing an integrated national strategy in the Netherlands.
The benchmarking models in general are mostly based on two types of measurements,
from a statistical point of view, for the overall performance of a country. The first type
is to develop a set of national Key Performance Indicators (KPIs) that measure the
country performance from different aspects; and the second type is the composite
indices that combine many key indicators KPIs into a single value.
There is a considerable number of studies (beyond road safety sector) that has
highlighted the importance and usefulness of having composite indices as a tool for

making integrated benchmarking, (see the review of Saisana & Tarantola, 2002 and
Andrew, 2004). Many composite indices have been developed internationally and used
in different aspects of life to indicate progress or achievements between countries. Such
examples are: Human Development Index (HDI), which was developed by the United
Nations and the Overall Health System Index by the World Health Organisation.
Business, in general, has also recognised the importance of having a multi-dimensional
index for managing, assessing, controlling and sustaining the performance of the
company/organisation (Ahmed & Rafiq, 1998). This has led to the development of
quality benchmarking and excellence models, which are today quite popular and are
widely accepted and used in modern economic and business benchmarking research at
national and international levels. Such examples are Total Quality Management and the
European Foundation for Quality Management (EFQM, 2006).
To date, no similar multi-dimensional (composite) index has been developed and used
for benchmarking road safety issues. Most attempts in past research have focused on
improving and implementing KPIs in road safety (i.e. ETSC, 2003). Simple safety
indicators have been used and developed by a number of international institutions and
databases such International Road Federation (IRF) and International Road Traffic and
Accidents Database (IRTAD). However, unfortunately, many databases overlap to a
large extent and they are not so detailed enough since many important indicators and
data are not available for a large number of countries in the world. One reason for the
lack of data, among many others, is that there is no single list of performance indicators
in road safety, universally accepted, collected and used.
1.2

Purposes and research questions

This dissertation develops an integrated benchmarking model called Road Safety

Development Index (RSDI). As the background has pointed out that road safety
represents complex phenomena where a high number of accident factors of human,
vehicle, road, environment, regulations are involved. There is a need for developing an
easy understandable tool for policy makers and public that quantitatively measures road
safety. Therefore, the main overall goal of this dissertation is summarised as follows:
The overall goal:
To create an international benchmarking model that indicates and communicates
in a comprehensive and easy way the severity of the road safety situation in a
specific country and/or in comparison to other countries in time.

As one can see, the overall purpose is quite broad where it is not possible to collect data
from all countries and for each key indicator. This needs enormous work and research.
To narrow the research, this dissertation seeks to contribute to the initial research
purpose by answering seven general research questions:
The first research question is:
What are the most commonly used benchmarking models for road safety?
The second research question is:
What are the most commonly used performance indicators for benchmarking
road safety internationally?
The third research question is:
What are the key performance indicators in road safety that can be applied
uniformly for most countries?
The fourth research question is:
What are the knowledge, criteria and methodologies that must be addressed
when aggregating the indicators into one single index?
The fifth research question is:
Do death rates have to be replaced with the new index, or should it be a
supplementary part, or be part of the new index?
The sixth research question is:
How can the new index be applied internationally for a sample or more of
countries?
The seventh and final research question is:
How can the applicability (usefulness) of this index be checked and evaluated?
1.3

Research Methods

There are many ways to design research. The best research approach for this study
would mix quantitative and qualitative methods by combining a qualitative philosophy
together with quantitative data and statistical procedures, in order to find whether the
quantitative results hold true with qualitative philosophy or vice versa.
Due to the absence of earlier models and methods in road safety in the area of
international aggregated tools, this study relies on knowledge and theories from other

sectors of science. This type of research method can be seen as deriving concept
hierarchies or as a part of Hermeneutics Cycle art of interpretation (Routio, 2006).
However, we have to be aware that applying any outside philosophy for the purpose of
research, may lead to better analysis, but not necessarily to deeper understanding if
there is no clear interpretation and analysis. Thus, the imported concepts must be
translated to road safety language and actual data of this study.
Furthermore, the dissertation attempts to develop a new theory and apply it to real world
applications on the basis of previous models and actual data. Therefore it is necessary to
choose a suitable research strategy that can promote both theory and practice. One
example of this type of work is the approach shown in Figure 1.2.

Figure 1.2: Research approach for promoting both theory and practice (Routio, 2006)

Following the above discussion, the process of the research strategies can be organised
into four practical stages: theory development, model building, empirical studies and
model assessment. These four stages correspond to the Deming PDCA cycle (e.g.
Watson, 1993): plan, do, check, and act; which are widely applied as a research model
of benchmarking process (e.g. between organisations). As a result, the research
questions can be now addressed in relation to these four research strategies as Table 1.1
shows.

Table 1.1. The relation between the research questions, strategies and answers:
Research strategies

Research questions

Research answers are addressed in

Theory development

Question 1, 2 and 3

Chapter 2 and 3

Model building

Question 3, 4 and 5

Chapter 4 and 5

Empirical studies

Question 6

Chapter 6 and 7

Model assessment

Question 7

Chapter 8

1.4

Scope and limitations

Answering the research questions, especially in building a new international concept

with a vision of real world applications are restrained by several limitations which are
summarised as follows:

1.5

The scope of thesis is carried out on a macroscopic level (national level) and is
not applied on a microscopic level (local level). Therefore, the distinction
between macro and micro performance indicators has to be described clearly.
This thesis will clearly describe the linkage between the performance indicators
(e.g. between practices and safety product in terms of death rates or between
process indicators and end-results indicators) in order to eliminate
misunderstandings.
The area of research is wide and it needs to focus on a narrow scope (e.g. small
number of countries), which may skew the results and will not show the full
usefulness of methodologies and the obtained index. Therefore, we must be
cautious in interpreting the final results.
Selection of interesting performance indicators is the most important step for
any benchmarking. However data is not always available or reliable for the
chosen indicators in a large number of countries.
Thesis structure

The dissertation consists of nine chapters, as illustrated in Figure 1.3. The structure
follows from the purposes of this study and from the stated research questions in
Section 1.2 and Table 1.1. First part of the thesis is more theoretical, particularly
Chapter 2, 3, 4 and 5, describes the conceptual framework and design of RSDI with
support from literature review. The second part is more practical, particularly Chapter 6,
7 and 8, which aims to put RSDI into practice and to find some answers to the
usefulness of RSDI in real applications of international benchmarking and ranking
between countries.

The purpose of Chapter two, Literature Review, is to explore what research has been
done in international benchmarking of road safety and what research needs to be done in
the same area. This chapter will also review the literature, which has already undertaken
the challenge of aggregating performance indicators for international benchmarking
(e.g. international composite indices, business management models). This discussion
provides answers to the first research question and partly to the second and fourth
question as well.
Chapter three aims to identify a master-list of performance indicators, which can be
useful for international benchmarking purposes in road safety as well as for RSDI
building. This work will be the starting point for finding a key list in the next chapter.
This chapter includes elaboration of the answers to the second research question and
partly to the third question.
The aim of Chapter four, The Conceptual Framework and Design of RSDI, is to
introduce, develop and describe the conceptual framework of RSDI including the key
concepts and philosophy behind RSDI that describes the overall performance of road
safety in a country. For instance, any successful international benchmarking in road
safety has to link practices to the end results. This chapter attempts also to identify a
short-term list of key performance indicators for road safety that can be applied
uniformly for most countries in each of LMCs and HMCs. This chapter does not only
provide a strong evidence of the necessity to have such a tool, but it also demonstrates
the mechanism of how to design such tool and use it. This chapter deals mainly with the
research questions 3, 4 and 5.
Chapter five discusses different methods that can be used for normalising the indicators,
weighting them and combining the weighted indicators into the index (RSDI). This
chapter provides answers to the fourth research question.
The purpose of Chapters six and seven, Empirical Studies, is to present two empirical
studies from LMCs and HMCs (answer to the sixth research question). The first
empirical study comes from eight European countries. The second empirical study
comes from five Southeast Asian countries. The collected data and selected indicators
are presented; also, some background about road safety situation from the selected
countries is presented.
Chapter eight discusses the results, outcomes and experiences, with focus on the results
from the two applications that have been conducted in Chapters 6 and 7 (empirical
assessment). In addition, there is a summary of the Strengths, Weaknesses,

10

Opportunities, and Threats of RSDI by using the SWOT analysis (theoretical

assessment). This chapter gives answers to the seventh and final research question.
In Chapter nine, a short summary of the study is presented. The conclusions are drawn
by summing up the answers in relation to the defined research questions at the
beginning of the study. At the end, the limitations of the present research will be
discussed and finally, an indication of the need for further work is given.

Chapter one:
Introduction

Chapter two:
Literature Review

The RSDI Theoretical foundation, Design and Methodologies

Chapter three:
Macro-performance
Indicators

Chapter four:
The Conceptual
Framework of RSDI

Chapter five:
RSDI Methodological
Approaches

Empirical Studies
Chapter six:
Applying RSDI to
HMCs

Chapter seven:
Applying RSDI to
LMCs

Chapter eight:
Discussion

Chapter nine:
Conclusions

Figure 1.3: Structure of the thesis.

Further, for any successful benchmarking research, the repetition of the whole process
of research strategies (four stages: plan, do, check, and act) is quite necessary to keep a
sustainable development. Thus, it is necessary to link the conclusions (Chapter 9) from
this research to the conceptual framework (Chapter 4) for possible continuous
improvement in the future.

11

1.6

Contributions, publications and significance of this study

This dissertation work provides four major contributions. The first contribution is the
design and development of the RSDI, which assesses and compares the road safety
achievements in different countries in a comprehensive and easily accessible form. The
second contribution is the development of a master list of macro performance indicators
that can be used for different purposes of international benchmarking in road safety. The
third contribution of the thesis is the development of a key list of applicable
performance indicators, categorised into three pillars and nine dimensions. This key list
can be used for assessing road safety in most countries worldwide based on data
availability, quality and importance. The fourth main contribution is an in-depth
analysis of the knowledge and criteria that are required in the selection and aggregation
of performance indicators in road safety, supported by a literature review.
These contributions are useful to national policy-makers and the general public in
helping them to make the magnitude of the problem easy to understand and to increase
their awareness of this phenomenon. In addition, these contributions can be useful for
researchers and traffic engineers who have an interest in collecting and analysing traffic
accident data.
The dissertation is designed as a monograph; however some parts of this thesis have
been published in other publications. The papers listed below are given in chronological
order, the earliest first. These papers are also presented in the reference list.
Paper 1: Traffic Accidents Reduction Strategy, Best Practices from European States, in
Proceeding of the International Conference in Traffic & Its Contemporary Issues,
Kuwait, May 12-14, 2007.
(with K. Asp)
Paper 2: New Tools for Assessing and Monitoring National Road Safety Development,
in Proceeding of the 2nd International Road Safety Conference, pp.31-34, Dubai United Arab Emirates, November 6 -7, 2006.
(with K. Asp)
Paper 3: The Evolution of International Road Safety Benchmarking Models: Towards a
Road Safety Development Index (RSDI), The International Journal Science &
Technology for Highways, 2006, Vol.3, pp.74-83.
(with K. Asp)

12

Paper 4: Road Safety in Southeast Asia- Factors Affecting Motorcycle Safety, in

Proceeding of the ICTCT international workshop on measures to assess risk in traffic,
Campo Grande, Brazil, March 21- 23, 2005.
(with P. Lindskog)
Paper 5: Applying Road Safety Development Index (RSDI) for Big Cities, in
Proceeding of the 6th International Conference: Traffic Safety Management for Big
Cities, pp. 218-222, St. Petersburg, Russia, September 14 - 15, 2004.
(with K. Asp)
Paper 6: Developing Road Safety Development Index, in Proceeding of the ICTCT
international workshop on improving safety by linking research with safety policies and
management, Soesterberg, the Netherlands, October 30.-31, 2003.
This paper was the point of departure of the development of RSDI, where the first
version of the framework of RSDI was first presented and published.
Paper 7: Road Safety Perspective in Arab Countries- Comparative Study and Analysis
of Progress, in Proceeding of the SORIC 02 Conference (Safety on Roads), pp. 116121, Bahrain, October 21-23, 2002.
(with K. Asp)
In addition, there are two papers under submission for possible publication in an
academic journal.
Paper I: A Composite Index for International Road Safety Benchmarking (RSDI): A
New Tool for the 21st Century.
(with K. Asp)
Paper II: Road Safety International Benchmarking: Results of Applying the Road
Safety Development Index (RSDI) to Less-Motorised Countries in Southeast Asia.
(with K. Asp)
Further, parts of this dissertation were published in the authors Licentiate thesis:
Towards a Road Safety Development Index (RSDI) - Development of an International
Index to Measure Road Safety Performance. Licentiate Thesis No. 1174, ISBN 9185299-70-7, Linkping University, Sweden, 2005.

13

Since the RSDI was first published in year 2003 (see paper 5), there has been an
increasing international interest in reviewing and testing RSDI through different studies,
projects and international bodies, for instance:
Study I: Planzer, R., (2005). Traffic Safety in Latin America and the Caribbean, Actual
Situation and Challenges, UNECLAC The United Nations Economic Commission for
Latin America and the Caribbean, ISBN: 92-1-322783-3, pp 27 & pp 28.
This study is a comparative research in road safety in Latin America and the Caribbean,
the full RSDI conceptual framework and its methodology were reviewed, presented and
translated to Spanish.
Study II: SafetyNet, (2005b). State of the art Report on Road Safety Performance
Indicators, European Commission, SWOV, Netherlands, pp16 and pp80
The SafetyNet is a major European project. It started year 2004 and is sponsoring by the
European Commission. This project aimed to collect, harmonise and analyse traffic
safety data in EU including the 10 new European member countries. The SafetyNet
study referred to RSDI within the context of developing road safety performance
indicators.
Study III: Capitulo, A. (2005), Cooperation Agreement in the framework of the project
Observatory for road safety, OROS, R8-B2-04, pp 9.
This European project acknowledged RSDI within the context of improving road safety
in towns.
Study IV: Fang, S (2006), Research on the safety evaluation index system of road
networks, Journal of Safety Science and Technology, Vol.2 No.2, P.34-38
Moreover, some international organisations have shown considerable interest in this
RSDI tool. For instance, the concept of RSDI was first introduced in 2002 in a seminar
hosted by the Swedish International Development Agency (SIDA) and attended by
representatives of the World Health Organisation (WHO) and Karolinska Institute (one
of the leading medical universities in Sweden and Europe). Likewise, the EMBARQ
organisation of the WRI Centre for sustainable Transport at the World Bank has
acknowledged RSDI as a benchmarking tool of interest for the development of Bus
Rapid Transit (BRT) systems in different cities, mainly in developing countries. More
recently, the RSDI has inspired the development of similar tools, for instance the ETSC

14

(European Transport Safety Council) has launched the EU Road Safety Performance
Index (PIN) at the EU summit of transport ministers in Brussels on June 2006.
Nevertheless, the author has been involved in a variety of international research projects
that contributed directly or indirectly to the development of this dissertation (e.g. data
selection, data collection, country road safety profile, etc.). These projects are:
Project 1: RetsNet Regional Traffic Safety Network, started year 2000. It aimed to
strengthen the cooperation and technology transfer in road safety between five southern
African countries (Botswana, Malawi, Namibia, South Africa and Zimbabwe) and
Sweden.
Project 2: ASNet project The ASEAN Road Safety Network, started in October 2003.
It is designed as an Internet networking system to strengthen the regional cooperation
related to traffic safety in the ten ASEAN countries: Brunei, Cambodia, Indonesia,
Laos, Malaysia, Myanmar, the Philippines, Singapore, Thailand and Vietnam.
Project 3: TechTrans project started in December 2003. It aims to establish a
sustainable virtual resource centre at the State Technical University (MADI) in
Moscow, Russia. This project developed e-learning courses and applications in the field
of road safety to Russian universities.
Project 4: SPIDER project, The Swedish Program for ICT in Developing Countries
started in 2004. This project aims to create programmes and applications for higher
education adapted to the needs of developing countries. The project focuses on three
developing countries for cooperation with: Burkina Faso, Sudan and Vietnam.
Project 5: Globesafe Global Road Safety Database. It is an Internet-based tool that
collects, harmonises and analyses the road safety data for the purpose of global
comparisons.
1.7

Definitions of terms

There are several terms concerning road safety issues and their applications. It is not
possible to give a precise definition of all the terms used in the following chapters,
because it would run to many pages and discussions. It may be useful however, to
briefly discuss the key terms and their meanings in this study. Readers should bear in
mind that the meaning of each term depends on the context and the subject of
discussion.

15

In this thesis, the term "Road Safety" is often used instead of "Traffic Safety" because
this thesis focuses on road traffic safety only including road user safety and vehicle
safety. The term "Traffic Safety" is a general term and could refer to the safety of all
traffic modes: air traffic, rail and road.
Some studies are not comfortable with the term "accidents" in the sense of describing
road safety problems as accidents that happen by chance. They prefer to use the term
"crash" instead of accidents. However, this thesis keeps using the term accidents
because it is still widely used in most current literature. Accidents can be simply
classified as fatal, serious, slight and damage only. The term "casualties" means both
deaths and injuries.
The term "macroscopic data" is used to describe the data from a national level e.g.
number of road accident deaths in the country, whereas "microscopic data" refers to
detailed information at local level e.g. number of deaths by location of accident, number
of deaths by type of vehicle, time, etc.
The distinction between developed and developing countries is a difficult matter. The
term "developing countries" is rather misleading because all countries are developing
today. Developing countries are officially classified in a human development index
(according to the United Nations) or on an economic basis. Whether we call this group
developing countries, low-income countries, less developed countries, underdeveloped
countries, third world, south, or other names; there is no precise definition of the term
"developing countries". Many international studies in road safety, e.g. Jacobs et al.
(2000), consider "vehicle ownership" as the most appropriate criteria to classify
countries. However, one has to note that the number of vehicles cannot be considered as
a sign of road safety development in a country as this depends on the whole transport
system and not only on the number of vehicles. To give an example, many European
countries are currently discussing whether to stop or reduce the increasing level of
motorisation. Similarly, for the purpose of this study and because of the lack of
something better, the study primarily deals with this issue by using vehicle ownership
(motorisation) as the most appropriate criterion to classify countries internationally. The
term Less Motorised Countries (LMCs) refers to countries with low and medium rates
of vehicle ownership (e.g. less than 500 vehicles per 1,000 population), while Highly
Motorised Countries (HMCs) refers to countries with high rates of vehicle ownership
(e.g. more than 500 vehicles per 1,000 population). The HMCs are mainly located to
countries in North America, Western Europe and Japan; while LMCs are the remaining
countries.

16

The term "composite index" is used throughout this thesis to mean the combination of
several indicators. The "indicator" itself is not any measurement and it is normally used
to show the level of performance in the country.
Terms such as "performance", "continuous improvement" and "success" are crucial
words and should be better defined in order to be easily understood by the readers. The
term performance in this dissertation refers to the end results that can be benchmarked
against similar end-results from other countries. The term "continuous improvement"
focuses on improving activities and practices (i.e. road user behaviour). The word
"success" refers to countries that fulfilled their performance targets. For further
reference to these quality terms, see for instance (Tangen, 2005).
The term "Master-list of performance indicators" refers to a long list of desired
indicators which can be useful for international benchmarking purposes in road safety,
whereas the term "Key list of performance indicators" is used to refer to a short list of
realistic indicators that have acceptable level of availability and quality worldwide.

17

18

Chapter 2: Literature Review

Chapter 2

Literature Review
The purpose of this chapter is to explore what research has been done in
international benchmarking of road safety and what research needs to be
done. This review will also move beyond the road safety sector through
literature that has already undertaken the challenge of aggregating
performance indicators for international benchmarking (e.g.
international composite indices, business management models).

The literature review will be discussed in two parts as follows:

The first part (Section 2.1), aims to provide an overview of the major work related
to international benchmarking of road safety, which has been done in the past and very
recently, and outlines the evolution of road safety benchmarking models towards
todays understating of integrated benchmarking approach. This review also attempts to
address the limitations of the traditional models of measuring the overall performance of
a country. Also, it will assess the first research question (stated in previous chapter):
What are the most commonly used benchmarking models for road safety?
The second part (Section 2.2), deals with the question of how the study has chosen
certain types of methodologies and framework for the purpose of building a composite
index. There are many models, in different aspects of science, which have addressed
and designed interesting conceptual frameworks, concerning aggregated indices, in
which some of them can be translated for the purpose of this study and RSDI building.
This part also will review literature related to the fourth research question (as stated in
previous chapter):
What are the knowledge, criteria and methodologies that must be addressed when
aggregating the performance indicators of a country into one single model for the
purpose of international benchmarking?

19

Each part concludes with a brief summary, and the end of the chapter gives a general
summary of the entire review.
2.1

International benchmarking of road safety

Road safety models can be microscopic (disaggregated level), macroscopic (aggregated

level) or mesoscopic (include some disaggregated variables on aggregated level). At all
levels, road safety models can be either explanatory (explain the development) or
forecasting (for short or long-term predictions). A review of these studies is reported,
among others, by (Hakim, 1991; Hakkert & McGann, 1996; OECD, 1997; Van den &
Wets, 2003, Turner et al., 2004). The focus of this literature review is on the major
macroscopic models that used for benchmarking road safety development
internationally.
The road safety situation is a complex issue and there is high number of accident factors
involved. A large amount of research has investigated the characteristics of these factors
and road safety outcomes such as Haddon's Matrix (1972), which aimed to identify the
basic road safety elements: driver, vehicle, road design, environment and their
interrelation between each other. Rumar (1999) has described the road safety problem
and factors as a function of three basic dimensions: exposure, accident risk and
consequences. These research activities have resulted in a large number of applications
and analysis at both micro and macro-level.
Work on international benchmarking has found significant interest and inspiration
among both researchers and practitioners. The major objective of such work is to
encourage countries to assess their development with other countries and learn practices
from each other. A number of benchmarking models are already being developed and
they range from relatively simple models to highly complex ones depending on the
number of indicators involved, details of data and complexity of methods used in
calculations and analysis. In road safety benchmarking between countries, four types of
models, are generally used (Al-Haji & Asp, 2006b):
1. Product Benchmarking is used to compare road accident death rates.
2. Practices Benchmarking is used to compare activities related to human-vehicleroad performance (e.g. seat belts use, crash helmets use, motorways level, etc.)
3. Strategic Benchmarking is used to compare National Road Safety Programme
(NRSP), management, enforcement and organisational framework.
4. Integrated Benchmarking is used to compare countries in terms of the three
previous types of benchmarking altogether.

20

Based on the above classifications, the evolution of road safety benchmarking models
can be divided into four stages or generations, which are simplified in the following
description and illustration of generations in figure 2.1:

The first generation is characterised by models that compare countries road safety
performance in terms of risk and exposure indicators such as accident death rates
and motorisation (Product Benchmarking). These models are cross-sectional
models, where international data were observed in the same year.
The second generation takes time into account. Theses models benchmark the road
safety product over time series. These models are useful in monitoring the trends in
road safety in countries and indicate the direction of progress ahead.
The third generation has realised the need for increased integration between
product (accident death rates) and other indicators in the same model (e.g. key
practices and strategies).
The fourth generation focuses on the three types of benchmarking: Product,
Practices and Strategic Benchmarking together.

Most of the early benchmarking models are still in use and being applied in different
studies. However, today computers are developing rapidly, which simplifies the analysis
of a large amount of road safety data that was not available before. This development
has made the work in the third and fourth generations easier and bringing it closer to
reality. Furthermore, picking up ideas (i.e. performance indicators) from the first three
generations was useful in reaching the fourth generation of integrated benchmarking.

Strategic
Benchmarking
Product
Benchmarking

Product
Benchmarking

Integrated
benchmarkin
Practices
Strategic
Benchmarking
Benchmarking
Practices
Benchmarking

The first, second and third generation

The fourth generation

Time

Figure 2.1: The road safety benchmarking evolution towards integrated benchmarking

21

2.1.1

The first generation: Linking motorisation, traffic risk and personal risk

An early study in 1949, R. J. Smeed compared twenty countries, mostly European for
the year 1938, where the author developed a regression model (log-linear model) and
found an inverse (or negative) relationship between the traffic risk (fatality per motor
vehicle) and the level of motorisation (number of vehicles per inhabitant). This
regression represented the best estimates of the mean values of traffic risk for each
given value of motorisation (what is called least square). This shows that with annually
increasing traffic volume, fatalities per vehicle decrease. Smeed concluded that fatalities
(F) in any country in a given year are related to the number of registered vehicles (V)
and population (P) of that country by the following equation:
F/V = (V/P)-
where

(2.1)

F is the number of fatalities in road accidents in the country

V is the number of vehicles in the country
P is population
is 0.003 and is 2/3

This formula became popular and has been used in many studies. It is often called as
Smeeds formula or equation despite some authors preferring to call it a law.
This nonlinear relationship can be translated to a linear one by taking the logarithms of
the two sides: log Y = log + log X , where Y is F/V and X is V/P.
The number of fatalities can be derived from Smeeds formula as: F = c.V.P, where c,
, are parameters and they are estimated from data by using the least square method.
For the Smeed data (year 1938) the formula was: F = 0.0003 P2/3 V1/3
Personal Risk (fatalities per population) is obtained by multiplying both sides of
Smeeds equation (2.1) by V/P as follows: F/P = a (V/P)1-b or F/P = 0.0003(V/P)1/3
Since 1949, many studies have been discussed on the basis of Smeeds equation (2.1) or
made reference to this formula. Some authors followed the equation of estimating the
regression parameters (, ) of the data by calculating the country road safety
performance in comparison to other countries; see Jacobs and Hutchinson (1973),
Jacobs (1982), Haight (1983), Mekky (1985). These studies found that Smeeds formula
can give a close estimation of the actual data and it can be applied to different sample
sizes of countries and years with the use of different values of and . Jacob and
Fouracre (1977) applied this formula to the same sample of countries used by Smeed for
the years 1968-1971 and it was found that the formula remains stable. Jacobs and

22

Hutchinson (1973) examined the data for 32 less and high motorised countries from the
year 1968. Al-Haji (2001) compared 26 countries around the world with different levels
of development and motorisation. The results from this study support Smeeds view of
the relation between motorisation and fatality rates. The correlation was high, 96% of
the variations are explained for the low motorised countries and 93% for the highly
motorised countries. Mekky (1985) found that the equation significantly captures the
relationship between motorisation and traffic risk for the Rich Developing Countries
(RDCs). Similarly, this conclusion was also reached by the study Al-Haji & Asp (2002),
which used cross sectional data of road safety in Arab countries.
The number of registered vehicles has been replaced by the total vehicle kilometre
driven in many late studies (e.g. Silvak, 1983; Fred, 2001). This measure (vehicle
kilometre driven) was not available at the time of Smeeds study. According to
Koornstra & Oppe (1992), the development of motorisation (referring to the number of
vehicle kilometres per year) follows an increasing S-shaped curve. Furthermore, Timo
(1998) has carried out a cross-sectional comparative study in many Eastern and Western
European countries to examine the development in the number of fatalities in relation to
the development levels of mobility. The study has shown that when the mobility reaches
the saturation level, as happened in many Western Europe countries, the decrease in the
number of fatalities has stopped or fluctuated only slightly.
Additionally, some other studies have tried to explain why the curve of development
(fatality rates) declines as has been noted in many countries and shown in Smeeds
formula. The studies have analysed the factors and measures that influence the
development of the curve of road safety. A review of these studies is reported by Elvik
& Vaa (2004) and Hakim (1991). Besides, Minter (1987) and Oppe (1991b) showed
that Smeeds law is a result of a national learning process over time. The development
in society at the national level is the result of the developments at the local level. In
other words, the individuals (road users) can learn by experience in traffic where they
improve their driving skills and knowledge, while society as a whole can learn by better
national policy and action plans. Figure 2.2 illustrates these factors on the development
curve of road safety where the long-term trends are based mainly on repeated crosssection surveys from different countries for different years. An early level of
motorisation, first leads to an increase in traffic risk, but not necessarily with the same
high growth in personal risk. However, later at a medium level of motorisation, traffic
and personal risks increase and both values are high. At the third stage of higher
motorisation, when a country is completely motorised, traffic and personal risks
decrease. The change between the three stages, as mentioned above, is due to better
engineering of vehicles and roads and greater understanding of the system by the road
users.

23

Fatalaties
Rate [per
vehicle (1)
and per
person (2)]

Learning and
society force

(1)

Engineering force
Economy force

(2)
Low

Medium
LMCs

High

Motorisation

HMCs

Figure 2.2: The influencing factors on the development curve of road safety
(Adapted from several works)

Conversely, at the same time, many studies have criticised Smeeds model because it
only concentrates on the motorisation level of country and ignores the impact of other
variables, see (Broughton, 1988), (Andreassen, 1985), (Adam, 1987), where according
to Smeeds model, population and vehicles are the only country values, that influence
the number of fatalities. This means that road safety measures have no meaning because
road fatalities can simply be predicted from population and vehicle numbers in any
country and any year. Andreassen (1985) criticised the models accuracy because there
would always be a decline in traffic risk for any increase in the number of vehicles, but
generally in non-linear way. Andreassen proposed relating fatalities to (V)B4 where B4 is
a parameter highly related to each particular country, even to countries with a similar
degree of motorisation. Furthermore, Smeeds study analysed data for one year. It was a
cross-sectional analysis with no time series analysis (Adam, 1987). Smeeds formula
expected the downtrend in fatalities rate but not the number of absolute fatalities, which
occurred in the HMCs in the seventies (Broughton, 1988). In other words, the trend
failed to fit with the real figures in HMCs. Broughton has concluded that: Smeeds
formula has no generally validity
In later years Smeed (in Oppe, 1991a) has commented on some of these remarks that:
We must be guided by the data and not by our preconceived ideas...The
number of fatalities in any country is the number that the country is prepared to
tolerate

24

Also, Haight (in Andreassen, 1985) has referred to Smeeds equation that:
When the formula disagrees with the observations we tend to assume that the
particular area under investigation is safer or less safe than it ought to be
Regardless of whether one agrees or disagrees with Smeeds model, the fact remains
that Smeeds model gave a simplified and fairly good representation of traffic risk and
motorisation in different parts of the world during the earlier stages of road safety
development.
To summarise, the models previously mentioned (in general) are in some way based on
regression models or multiple regression models or quadratic regression models. They
employ a small number of indicators (motorisation, personal risk and traffic risk) to
check the goodness of fit to data from different countries and to find the appropriate
related equation(s) for making comparisons between the chosen countries.
2.1.2

The second generation: Linking traffic risk, motorisation and personal risk with
time

In this generation, many benchmarking models have been developed to describe and
predict safety development between countries on the basis of time series models and
theories. They relate the variables to a function of time to determine the long run change
in safety development over time either in a monthly form or annually. These models
attempt to find the smoothed curves to the time series data between countries.
Koornstra (1992) has shown that motorisation is considered to be dependent on time
and the relationship between deaths and population should include time. The author
found the following formula for approximating the number of fatalities from country to
another in a particular year:

Ft = zVt xVt wk ( max ) c 1

V
t k

where

(2.2)

Ft is the number of fatalities for a country in a year t,

Vt is the number of vehicle kilometres travelled in the year t,
Vmax is the maximum number of vehicle kilometres,
k is the time lag in years, and
x, w, z, y, and c are constants

Oppe (1989) assumes that fatality rates follow a negative exponential learning function
in relation to the number of vehicle kilometres and time. This method has been found to
be most effective when the components describing the time series behave slowly over
time as follows:

25

ln (Ft/Vt) = ln (Rt) = t +
Or, equivalently:

Rt = et +
where

(2.3)

the ln function is the natural logarithm,

Ft is the number of fatalities for some country in a year t,
Vt is the number of vehicle kilometres travelled in that year,
Rt is Ft/Vt and
, are constants

This means that the logarithm of the fatality rate decreases (sign of improvement) if is
negative proportional with time. This model is called the negative exponential learning
model, where is supposed to be less than zero. Both and are the parameters to fit.
Additionally, Oppe (1991a) assumes that the amount of vehicle kilometres per year is
related to time and it is assumed that traffic volume will develop over time by a logistic
function of a saturation model. This assumption indicates that the growth rate of traffic
volume is a percentage of the ratio between the traffic already existing and the
remaining percentage of Vm as follows:

ln(

Vt
) = t +
Vm Vt

Or, equivalently:
Vt =
where

Vm
1 + e ( t + )

(2.4)

Vt is the number of vehicle kilometres travelled in that year, and

Vm is the maximum number of vehicle kilometres

This formula shows that countries with a large should have a fast growth in traffic.
The traffic volume will increase quickly first and at the end it will reach its saturation
level, which differs from country to country.
Oppe has applied the two formulas (2.3 and 2.4) to data from six highly motorised
countries over the time period 1950-1985. It was found that both formulas describe the
data fairly well. It concluded that the development in road safety is a result of the
development (learning) of the traffic system in the country, which is more or less
similar to Smeeds conclusions. However, Oppes theory in estimating the remaining
growth of traffic is questionable, particularly when we know that many European
countries are currently discussing the possibility of stopping or reducing the increase
rate of motorisation. It is uncertain whether the number of fatalities can be predicted
simply from the fitted curves or from the number of vehicle-kilometres. The question is

26

therefore whether this decreasing equation (2.3) assumes that the fatality rate is reduced
to zero in the end or not, and in this case what is the predicted year for one particular
country according to its current level of mobility? Besides, what will happen to the
expected number of fatalities if the countrys trend towards full motorisation is realised?
Adams (1987) has stated a similar relation between fatalities (F) and vehicle kilometres
(V), which was presented: Log (F/V) = a + by where y = year(1985). Broughton
(1988) has tested this logarithmic model on data from Britain between 1950 and 1985
and the model fitted well with actual data. In the same study, Broughton applied the
same model to data from four western countries: U.S.A (1943-85), West Germany
(1965-85), Norway (1947-85) and New Zealand (1948-83). It was found again that this
model reflects the data pretty well.
Additionally, Navin (1994) has extended Smeeds equation into a three-dimensional
function including the traffic risk, personal risk and motorisation as shown below. The
study has applied the function to time series data from Canada (1910-1990), United
States (1906-1991) and United Kingdom (1905-1990). The results have shown that the
proposed relationship fit with the actual data.

T = Tf e
where

M
M0

(2.5)

M0 is the value of motorisation at maximum personal risk,

Tf is the point where the exponential curve meets the T-axis,
T is the traffic risk, fatalities per number of vehicles, and
M is the motorisation, vehicles per population

Another technique, the singular value decomposition method, is a useful tool in

comparing road safety trends between different countries (Broughton, 1991; Oppe,
2001a). This technique investigates the similarities and dissimilarities between different
groups of countries regarding fatality trend. They compared various time series of data
of countries jointly to investigate the correlation between these series. This technique is
useful in classifying the countries that are similar in accident patterns to each other.
The more detailed time series data have led to advanced and sophisticated ways of
fitting a curve to data, especially with the current use of computer packages. For
example, auto-regressive integrated moving average (ARIMA) techniques are used to fit
and forecast the time series that are changing fairly quickly; see Frits et al. (2001).
Scott (1986) has applied this method to model the accidents in England (seasonal and
annual data). Similarly, Oppe (2001b) has applied this method to a model that predicts
the accident data development from Poland (1980-2010) with some sort of comparison
to similar development in other European countries.

27

2.1.3

The third generation: The need for increased integration with many variables
involved

Many researchers have realised the importance of having more indicators involved in
the same model rather than relying on few variables in terms of people, vehicle, deaths
and time. The road safety level in a country is a result of the whole development in
society (e.g. health, education, enforcement, engineering, etc.) and therefore there
should be more variables included in benchmarking models.
In the study by Fieldwick (1987), speed limits were included in the same model by
Smeed for developing its accuracy further. Kopits & Cropper (2003) made a statistical
study of fatalities and vehicle ownership including variables related to economic
growth. The study was conducted between 1963-99 and involved data from 88
countries.
Navin (1994) also included more variables, such as annual seat-belt percentage and per
capita consumption of alcohol, to the 3D function shown in (2.5). The countries chosen
are Canada, the United States and the United Kingdom.
A considerable study made by Page (2001), who has compared safety situations and
trends in the OECD countries from 1980 to 1994. The study developed a statistical time
series model. The model gives a rough estimate of the safety performance of a country
regarding variables other than accident death rates such as: population levels, vehicle
fleet per capita, percentage of young people, and alcohol consumption. Based on this
model, countries showing the best levels are Sweden, the Netherlands and Norway.
Bester (2001) has used a stepwise regression analysis to indicate the variables that
should be added or removed in the model. The study used collected data from different
international sources and many variables used in the model such as: national
infrastructure, paved roads and socio-economic factors (e.g. GDP per capita).
A large number of studies have focused on the effectiveness of road safety measures
that countries have taken. Comparing practices of seat belt use, helmet use, and
enforcement levels are such examples. Several studies tried to link these practices to the
trend of accidents or fatalities. Elvik & Vaa (2004), for instance, have evaluated the
effectiveness of various road safety measures in different countries.

28

2.1.4

The fourth generation: Linking product, practices and strategic benchmarking

The last generation of road safety benchmarking has realised the necessity of having a
systematic way to add up all the potential indicators that are related to product, practices
and strategies into one model or one equation or an aggregated index. This will give a
broad picture of benchmarking and not focus on one or few particular aspects. This will
also be useful to identify the success areas from failure in any country.
Very little research has been done on this direction of integrated benchmarking in road
safety. In one of the few studies in this area, Asp & Rumar (2001) have developed an
integrated model called Road Safety Profile (RSP). This model included a list of
quantitative and qualitative indicators that can describe, explain and compare road
safety situations in different countries. The quantitative data were obtained from
international sources and databases (e.g. Globesafe database), while the qualitative
indicators were derived from a survey of questionnaires to experts in each country. The
countries were divided into three different groups of motorisation (low, medium and
high). The RSP technique includes more than 20 direct and indirect road safety
indicators. Each indicator is normalised on a scale from +2 to 2. Then the results are
illustrated as a profile (see Figure 2.3). This illustration can make the comparison
between countries simpler and easier. It can also illustrate the development in a country
over time in a quick and easy illustration. The results of applying RSP to different
countries proved that RSP is also a useful tool for identifying the problems in the
country where actions are needed.
Level of Motorisation: Low

Country 1

Direct safety measures

-2

Country 2
2

-2

Country 3
-2

Personal Risk

-2

Traffic Risk
Indirect safety measures

-2

Road safety statistics

-1

Road safety trend

Road safety R&D

-2

Road safety organisation

-2

Road safety program

-1

Road safety legislation

-1

Traffic police

Driver education

-1

Alcohol in traffic

-2

-2

-1

Speed

-2

Seat belts

-2

Road standard

N.A

N.A

Paved roads

-1

-2

Road expenditure per total

Etc.......

-2

0
2

Figure 2.3: Illustration of Road Safety Profiles (Asp & Rumar, 2001)

29

2.1.5

Summary from literature review (part 1):

Even though some remarkable research has been made over decades in this topic of
international benchmarking for road safety (started by Smeed 1949), many studies
(countries) still rely mainly on traditional performance measurements in terms of traffic
road fatality rates. Several studies and investigations have shown that death rates are not
sufficient to explain the real situation of road safety. Different models, cross sectional
and time series, have been developed to add more indicators to accident death rates and
make the derived model better in describing and comparing road safety between
countries. The possible reasons why death rates are still widely used, even they are not
widely accepted, are the complexity of other models and the absence of any better
alternative.
Few variables are generally included in the traditional modelling process of road safety
benchmarking. However, road safety development in a country is not only measured by
the variables just given, it is more effective where a large number of factors are
involved in the same model. Therefore this dissertation will allow more indicators to be
grouped together. Today computers are developing rapidly, which simplifies the work
and analysis of cross- sectional data and time series data, which was not available in the
past (e.g. to Smeed in 1949).

Summary 1:
There is a need for an integrated benchmarking model in road safety.

2.2

Multidimensional index

One popular approach to reaching a successful integrated benchmarking is to gather all

comparable and meaningful macro-measurements into some sort of multidimensional
index (Nardo et al., 2005, EFQM, 2006). The following part of literature review
explores previous research into aggregating performance indicators into one single
model for the purpose of benchmarking. Due to the absence of earlier models and
methods in road safety in this area of research, the study has to borrow knowledge and
theories from other sectors of science and to apply it to road safety field.
This part is based mainly on a collection of two literature reviews. The first collection
deals with the composite indices for giving a better understanding of the methods
needed for the model building. The second review looks to the business management
models for understanding the concepts product, process and system and how they join

30

together into the business excellence model. However, readers have to note that the
emphasis here in this part of review is on the approach itself rather than on its
application, analysis and results.
2.2.1

Based on composite indices

This philosophy of combining performance outcomes into one single quantitative value,
so called composite index, is not a new one. Many composite indices have been
developed internationally and used in different aspects of life to indicate a progress or
achievements between countries. They cover environmental issues, sustainable
development, globalisation issues, agriculture, economy, information technology,
innovation, and more. A large number of these indices are being developed and
presented with the cooperation of international organisations and bodies. For instance
Human Development Index (HDI), which was developed by the United Nations;
Environmental Sustainability Index by the World Economic Forum; Composite of
Leading Indicators by OECD; European Innovation Scoreboard Index by EU, and
Overall Health System Index by the World Health Organisation. Some other
international indices were developed at universities and research institutes, for example
The Growth Competitiveness Index by Harvard University in the United States. A
review of these studies is reported by John et al., (2001), Saisana and Tarantola (2002),
Freudenberg (2003) and Andrew (2004).
Multi-national discipline and experts were also involved in constructing and assessing
the international composite indices. A.K. Sen, for instance, is a Nobel Prize holder and
is one of the key developers of the Human Development Index (HDI) in terms of his
theory on poverty and welfare on a global scale (Andrew, 2004). Since 1990 the
Human Development Reports (UNDP, 2004) have measured national achievements
annually by using HDI, according to: life expectancy index, standard living index,
and education index. Since then several other supplementary indices to HDI have been
developed such as the Human Poverty Index (HPI), Gender-Related Development Index
(GDI) and Gender Empowerment Measure (GEM).
Until now, no similar composite index has been developed and used in road safety
issues. Most previous work in international road safety assessment has focused on one
or a few indicators as outlined in the previous section. Road safety represents a
complex phenomena, and because it concerns everyone and society as a whole, there is
a need for a systematic approach that can capture all (or most) of the performance
outcomes together.

31

In general, there is a common set of processes for constructing composite indices, which
involve five steps (see e.g. Saisana and Tarantola, 2002): selection of indicators,
developing of a theoretical framework, normalisation, weighting and aggregation, and
validation.
A number of studies (for example Nardo et al., 2005; Andrew, 2004; Saisana and
Tarantola, 2002) have assessed and evaluated different composite indices of
development in terms of their methods and techniques. The evaluation has divided the
literature into two types (schools), between those studies, which have been in favour of
the idea of aggregating performance indicators into some sort of overall index, and
those studies, which have opposed or criticised it. A brief summary of these literature
reviews can be made regarding the strengths and weaknesses of composite indexes as
follows:
Proponents of composite indices argue that there are several strengths in developing and
using such indices because:

They are better for making simple comparisons.

They are better for summarising complex issues into a single value.
They are better for providing the big picture.
They are better for sending summary tailored information to policy makers
where they are too busy to look into details.
They are better for easily attracting public.
They are better for simply and efficiently communicating with everyone.

Conversely, many opponents of indices argue that there are several weaknesses in
developing and using composite indices:

They are far from reaching a single and standard methodology. As a result,
much conceptual framework is now available with different calculation
techniques and treatment of indicators.
They are far from giving correct and accurate conclusions.
There is a danger of overload information (indicators).
There is a danger of missing some important information (indicators).
There is a danger of relying on subjective selection of indicators.
There is a danger of relying on subjective assumptions of weightings.

Basically, this debate is based on the amount of information (indicators) that is missed
in the index, methods and weights. In this sense, Nardo et al., (2005) has addressed the
debate surrounding composite indices by saying that:

32

It is hard to imagine that debate on the use of composite indicators will ever be
settled (Nardo et al., 2005)
Also Nigel et al. (2003) has commented on this debate saying:
The two views regarding aggregate indices are not as black and white as may
appear. In fact, they are necessarily complementary (Nigel et al., 2003).
To give an example, Frederik (2002) has criticised (HDI) index because it excludes
several essential indicators on the quality of life, which drives national decision makers
to concentrate their efforts on education and health only. However, despite these
shortcomings, HDI gets widespread media coverage because of its advantages, is now
being widely accepted and is the most successful index worldwide.
Consequently, from road safety perspective, a high level of indicator aggregation is
necessary in order to increase the awareness of road safety problems among the public
and policy makers. However, building a new concept is a challenging task. Even taking
the advantages of aggregated indices; one has to consider carefully their limitations. In
doing such, one has to inform clearly the sources of data, how the indictors were
aggregated, and how the weights are measured or assumed. It is also necessary to have
insight into the methodological issues and options that have already been used by other
indices, when calculating our road safety aggregated index. The obtained composite
index in road safety may not possibly show the whole picture in a country (answering
all questions regarding the overall performance in a country), but at least this index
should be able to show most of the picture (answering most of the questions).
In general, this brief review also suggests that there is a need for internationally
accepted methodology for building a composite index. This methodology can then be
applied to any (or most) composite indices for any (or most) fields of science. It seems
that little work has been done in this direction, though it may not be an impossible task
in my view.

Summary 2:

There is large and growing number of composite indices everywhere for

international benchmarking. This is because of the advantages they offer.
Despite this, there is no single composite index used for benchmarking
road safety.
Even taking the disadvantages of composite indices, they can serve as a
suitable tool for building RSDI model.

33

2.2.2

Business excellence models

The success in any company/organisations performance is no longer linked only to the

quality of a product but also to the core quality of services, delivery, practices, human
resources, competence, etc. Business, in general, recognises the importance of having a
multi-dimensional index for managing, assessing and controlling the performance in the
company. This has led to the development of quality benchmarking models, quality
improvement models and excellence models, which are today quite popular and are
widely accepted and used in modern economic and business research. Reviews of these
models are reported by Ghobadian and Hong (1996), Kanji (1998), Dahlgaard et al.
(1998) and Vokurka et al. (2000).
In the earliest stages of quality development, the focus was on product, profit and
market share (Cooper 1990). Since then quality concept has improved in different
phases (e.g. control, process, inspection, etc.) resulting in the modern concept of Total
Quality Management (TQM)1 which has put together most dimensions of the overall
company performance and capabilities into the same model (Zhang, 2001).
Furthermore, there are today many quality award models that have been inspired by
TQM principles, which are widely used. The most popular ones are the Malcolm
Baldrige National Quality Award (MBNQA)2 in the United States and the European
Foundation for Quality Management (EFQM)3 in Europe. Also Sweden has developed a
quality award model for the most distinguished organisation of TQM-work of the year
in Sweden, which is organised by the Swedish Institute for Quality (SIQ), see (Eriksson,
2004). There are some similarities in the framework of these models, regarding, for
instance the consideration of operations and results of an organisation as a total.
According to the EFQM model (EFQM, 2006), there are two main criteria, Enablers
and Results. The Enablers criteria cover what an organisation does. The Results
criteria cover what an organisation achieves. Factors such as leadership, people
management, resources and product management are called the Enablers, while
customer satisfaction, employee satisfaction and product quality are called the
Results. The enablers deliver the results, however the success in business
performance is achieved when both Enablers and Results are used.

1 Total Quality Management (TQM) started to be used in the mid 1980s in order to understand the
Japanese quality management methods.
2 Malcolm Baldrige National Quality Award (MBNQA) was introduced in 1987 as a model for
excellence in total quality management in the United States.
3 The EFQM Excellence Model was introduced in 1992 as a tool for assessing organisations for the
European Quality Award.

34

Many studies have shown the benefits of the adoption of business excellence models
into other aspects of life. For instance, Dunn and Mathews (2001) and Nabitz and
Klazinga (1999) have applied EFQM excellence within the public sector (healthcare
sector).
Most of the above quality models give greater weighting points to the end business
results and human dimension (leadership, customer satisfaction and employee
satisfaction). This understanding might be useful when we apply these principles to
RSDI to judge the weights of indicators.
Economic performance principles can serve as an inspiration to road safety. However,
applying economic thinking to road safety field is not an easy task and requires new
approaches of analysis because of the complex nature of road safety as a process, as
practices or product. To give an example, there are different kinds of products available
on the market (e.g. goods, services, ideas, software, models, tools, etc.). However, the
safety product in terms of number of deaths, injuries and cost is not a physical object to
be found on the market or be bought or sold by a business sector. The safety product is
much more complex. The safety product is road accidents which equals loss, while the
reduction in the total number of deaths and injuries and associated costs equal profit.
This profit is an immediate profit while the total quality of outcomes (e.g. RSDI) is the
long-term profit.
Improved total quality (operations and results) has led to sustainable improvements in
profits and customer satisfaction in the field of business. Similarly in road safety, the
improvement in operational quality (good practices) and quality results (accident death
rates) will lead to sustainable progress in road safety. The understanding of this
principle will lead to practices performance. Any successful integrated benchmarking
has to be carried out by linking practices to the end results. This is a sort of deep-level
learning in an organisation or a country.
By understanding how an organisation improves its performance and excellence, it may
be able to translate its principles into the conceptual framework of RSDI.
Consequently, in order to become successful in road safety as well as in business, each
part of the society or organisation must work properly together towards the same goal
with so much understanding that each person and each activity affects the overall
outcome. The reduction in road accidents in a country (or municipality or big city) is a
sign of efficiency and development throughout that society.

35

Summary 3:

2.3

Any successful integrated benchmarking has to be carried out by

linking the key practices to the end results.
Economic performance principles in an organisation can serve as
an inspiration to road safety performance in a country in terms of
product-practices and system.

Conclusions

There are three main summaries that can be learnt from the literature review made in
this chapter:

36

Summary 1: There is a need for an integrated benchmarking model in road

safety.
Summary 2: There is large and growing number of composite indices
everywhere for international benchmarking. This because of the advantages they
offer. Despite this, there is no single composite index used for benchmarking
road safety. Even taking the disadvantages of composite indices, they can serve
as a suitable tool for building RSDI model.
Summary 3: Any successful integrated benchmarking has to be carried out by
linking the key practices to the end results. Economic performance philosophy
and principles in an organisation can be served as an inspiration to road safety
performance in a country in terms of product-practices and system.

Chapter 3: The Theoretical Framework of Macro- Performance

Indicators in Road Safety in Road Safety

Chapter 3

The Theoretical Framework of MacroPerformance Indicators in Road Safety

This chapter attempts to identify the most measurable and comparable
national performance indicators (master-list) that could be used as
international benchmarks in road safety.

The indicator is a measurement that quantifies something that affects the road safety
level and it can be measured in some common terms such as a percentage (e.g.
percentage of front-seat belt usage), a rate (e.g. number of fatalities per population), or
as qualitative information (e.g. level of national data collection and reporting: Very
Good, Good, Fair, Unsatisfactory). The focus of the dissertation research is on the
quantitative indicators.
Additionally, the indicator depends on the type of data being collected, what we want to
measure, and for whom this indicator will be used (target group). Indicators can be used
at the highest national levels to measure or assess the progress being made towards a
certain goal, such as reducing the number of accidents and injuries in the whole society.
At the local level, indicators can be used to measure the daily/monthly activities of road
users through which measures can reduce their involvement in an accident (probability
of accident) at a certain place and time (e.g. young pedestrians involved in an accident
at a particular intersection). The focus of this dissertation research is on the national
(macroscopic) performance indicators.
There is a growing interest internationally in macroscopic quantitative performance
indicators in road safety (e.g. IRTAD, 2006; SafetyNet, 2005a, among others), due to
the importance they can play in describing the road safety situation in a country,
comparing the development and assessing the effect of different measures.

37

This chapter will investigate the second research questions (specified in Section 1.2) for
the purpose of building a multi-dimensional index, which is:
What are the most commonly used performance indicators for benchmarking
road safety internationally?
To answer this research questions; we have first to find out where macro-performance
indicators stand between other types of indicators. Figure 3.1, for example, shows
hierarchical development of the data and availability from local to national level. The
pyramids base level represents the primary accident data and indicators at local level.
This is where most of the detailed and specific data is available. The second level of the
pyramid describes data by more aggregated indicators at national level. It is known that
not all indicators and data used at a local and provincial level can be used nor are
available at a national level. The third level of the pyramid refers to all macroscopic
road safety performance indicators that could indicate and monitor the countrys
progress over time in road safety, which will also allow international comparisons. The
fourth level of the pyramid refers to the key macroscopic road safety performance
indicators that can be integrated into a single index. At the top, the national
multidimensional index integrates and summarises much (or all) of the key macroperformance indicators into a single value RSDI. As can be seen from the figure, the
more movement towards RSDI level, the higher decision-making and less detail of data
will be reached. The decision makers in a country will be able to make appropriate
decisions and take appropriate actions for a strategy of a future development.

RSDI
High level of
policy and
decision-making

Aggregation level 4
Key
macro-performance outcomes
Aggregation level 3
Master-list
of macro-performance outcomes
Aggregation level 2
National road safety specific data
Aggregation level 1
Accident primary database, local indicators
and secondary indicators

Data details

Figure. 3.1. The pyramid of road safety indicators and

levels of aggregation (adapted from WRI, 1995)

38

This chapter aims to identify a master-list of performance indicators (the third level of
the pyramid), which can be useful for international benchmarking purposes in road
safety as well as for RSDI building. This work will be the starting point for finding a
key list of applicable performance indicators that can be used for most countries
worldwide (the fourth level of the pyramid). The key short-list of indicators will be
examined next chapter in terms of data availability among the selected countries and the
conceptual design of RSDI.
The approach followed here, for finding the master-list, is divided into three stages.
First stage attempts to find the most important indicators in relation to safety (in terms
of risk mainly, exposure and consequences) with support from the research literature.
In the second stage, indicators are grouped into different dimensions and two main
categories: direct measures in how they (indicators and dimensions) illustrate road
safety level and development in a country, and indirect measures on how they offer a
significant affect on the overall road safety performance. The third stage investigates the
criteria that have to be used during the selection process of indicators and dimensions
towards the final master-list.

3.1

Stage one: Identifying the macro-indicators in relation to risk, exposure and

consequences

As young road users represent a high risk in accidents all over the world, we might say
that a developing country, which has a high proportion of young people in the total
population, should have more casualties than those of the developed countries where
they have low birth rates as well as an older age group. Similarly, countries with a high
proportion of motorcycles or vulnerable road users on roads should have more
casualties than those countries where most road users are well protected inside vehicles.
In the same context, as rich people are healthier than poor people because they have
better access to hospitals and health services, we might say that rich countries are
healthier (safer) than poor countries. In the same way, one can try to assess the
differences between countries in relation to other types of factors such as: weather
differences between countries, age of vehicles in a country, the expenditure rate on road
safety measures, etc.
Although many attempts have been made to describe road accident problems, it may be
able to describe the problem as a sign of illness in society. To diagnose the problem,
there are direct and indirect symptoms. The direct symptoms in road safety can be easily
seen in society from simple observation and data such as: speed problems, alcohol and
driving, vehicle conditions, road user behaviour, etc. The indirect symptoms in road

39

safety are not simply obvious and they need more checking and examination such as:
the traffic management, education, traffic police enforcement, legislations, etc.
There are many descriptive models in road safety that have attempted to formulate a
theory or model that can explain why accidents happen. One widely used method is to
illustrate the road safety problem as a function of three dimensions: exposure, accident
risk for a certain exposure and injury consequences (Rumar, 1999).

3.1.1

What are the Exposure, Risk and Consequences?

Exposure is an important dimension in road traffic that refers to the amount of travel in
which accidents may occur. The more we travel on roads, the higher the probability that
an accident will occur. Without traffic or mobility, there will be no accidents and no
road safety problems. Many studies show that there is a correlation between vehicle
traffic volume and the total number of accidents. Moreover, traffic volume is one of key
pillars together with environment and safety, which is needed to promote the
sustainability of transport in any country.
At present, there are many ways of measuring exposure. In international comparisons as
well as for national statistics, the population size, urban population, the number of
vehicles, length of road network and number of driving licenses are possible and
available indicators of exposure. However, these indicators, as found in many studies,
do not always work as good measures of exposure. This is due to the differences in
socio-economic conditions between countries, population density, vehicles per citizen,
and transport mode split.
The number of kilometres vehicles travel is generally considered a suitable exposure
measure and is defined as the sum of the distance travelled by all motor vehicles in the
country over a year (the distance travelled per vehicle multiplied by the number of
vehicles). Unfortunately this measure is not often available in most countries. Many
countries do not have a national system for counting travel volume on the basis of this
measure, where they estimate the kilometres driven by either the total fuel sales in the
country or they give an estimated equal average distance for each car driven (amount of
kilometres per year). In fact, these assumptions are not fully accurate because of the
differences in fuel consumption, type of fuel (Gasoline, diesel, etc) and the distances
travelled by each type of car and vehicle. One should also remember that the amount of
kilometres driven is greater on urban roads, which are busier than those in rural areas.
The lack of detailed and quality exposure data is a challenging issue and therefore,
international comparisons are often conducted on the basis of a per-capita population or

40

per-vehicle. There is a need for detailed exposure information on a specific population

(e.g. children or elderly road users), non-motorised modes (e.g. cyclists and
pedestrians), motorcycles, and driver information. This information is often less known
and available in many developing countries than in highly developed countries. The
classifications of roads also differ in their standards from one country to another (e.g.
national/regional/motorways). Additionally, there is a need for special counts of traffic
volume by specific type of vehicle (e.g. lorries and vans, heavy trucks, etc).
Calculation and gathering of annual exposure traffic data is not a simple issue and it
needs to be conducted in a regular and systematic manner for each group of road user,
type of roads, time, etc. Surveys are expensive in many countries especially in
developing countries. Fortunately, more advanced technologies have now become more
available in most countries. They are cheap and effective in calculating exposure of
travel (e.g. widespread use of mobile telephones, telecommunications technologies
inside vehicles and along roads) that it is hoped will provide better and more accurate
exposure variables that can be used in international comparisons in the near future.

Risk: is determined as the probability of an accident happening per units of exposure or

it is evaluated as the size of severity of this accident. The higher the accident risk, the
higher the probability of an accident occurring for a given road user in one particular
place and time. At micro level, the accident risk in any location can be observed from
actual accidents data (after accidents happen) or from observed conflicts (before
accidents happen) using estimations of speed, distance and time to accident (Hydn,
1987). However, at macro level the term risk of accident is observed directly from
accident death rate. When comparing different countries, the indicators of fatality per
population and fatality per vehicle are widely used and they vary from country to
country and over time. Neither measures take into consideration the characteristics of
the type of transport modes or the road users. The risk per exposure unit has generally
shown a clear decrease over time in most countries. This is an indication of an
improvement in the overall accident situation in most countries where they produce
lower accident death rates for each unit of exposure. However, the exposure unit itself
(i.e. number of vehicle or vehicle kilometres travelled) has continued to increase and
more casualties occur on roads in many countries. It is internationally found that the
risks for accidents are higher in the countries with low motorisation level and in
countries that have experienced very rapid and quick motorisation.
Consequences: The third dimension in describing the road safety situation is the risk of
injury severity in an accident, which refers to the outcome of accidents in terms of
injuries. The severity of the consequences of an accident ranges from fatalities and
serious injuries down to slightest and damage only.

41

By multiplying the three dimensions (Exposure, Risk, and Consequences) we get the
total number of killed or injured persons or accidents in road traffic. Rumar (1999)
described the road safety problem as a function of three dimensions, which are
exposure (E), accident risk (A/E) for a certain exposure and injury risk (I/A). This
is illustrated in Figure 3.2 where the volume of the cube determines the size of the road
safety problem. Any change in any one of these three dimensions will change the whole
safety situation in a country.
Accident Risk (A/E)

Exposure (E)

Injury severity (I/A)

The shaded area indicates the number of accidents (A).

Figure 3.2: Road safety problem described by
three-dimensional cube (Rumar, 1999)

The total number of deaths or injured persons (I) = Exposure (E) Accident risk (A/E)
Injury severity (I/A).
This formula can be transformed to the fatality rate. An example of this expression is:
fatalaties
exposure accidents fatalaties
=

inhabitants inhabitants exposure accidents

If we take the number of vehicles as an exposure measure, we find:
Personal Risk = Motorisation Traffic Risk
These formulas would allow us to compare and illustrate the road safety situation
among countries. Thulin & Nilsson (1994) have shown how exposure, risk and
consequences vary for different transport modes and age groups in Sweden.

42

In principle there are three main ways for reducing this size of safety problem in
response to any change in the three dimensions:

Reducing exposure factors: by reducing the amount of travel per person or vehicle
and the total reduction in traffic volume.
Reducing risk factors: by for instance improving driver skills, road user education,
vehicle performance, road standards, legislation and enforcement.
Reducing accident severity: by protecting people better in vehicles from injury
severity. Protecting pedestrians and other vulnerable road users by vehicle design,
and protecting two wheelers by using appropriate helmets.

Shifting travel from means of transport with high exposure and risk (e.g. motorcyclists)
to means that have a low level such as public transport can influence the level of the
safety situation in the country.

3.1.2

Correlation between the quantified macro factors and road accidents

In road safety, there are a large number of causal factors contributing to road accidents
and their severity. There is no single cause of accidents and it is hard to pick up one
factor or even a few of them as being more important than the others. From this,
removing any factor that contributes to or causes the accidents will not provide a quick
nor fully satisfactory solution to road accident problems. In other words, these factors
are not fully independent of each other and each one is influenced by other factors. To
be able to measure the effect of any given factor to accident frequency, one has to keep
all other relevant factors equal, which is a complicated task.
The correlation between any factor and road accidents means that they both increase
and decrease simultaneously. For example, if an increase in speed causes an increase in
road accidents it means both are correlated. The degree of such a correlation differs
from factor to factor and is usually measured by using different statistical techniques
and it ranges from zero to one. If one, it means the factor is highly correlated to road
accidents and if zero there is no correlation at all. This section is not intended to discuss
these techniques in detail, but instead to focus on results from several sources (literature
survey and meta-analysis, e.g. the Handbook of Road Safety Measures by Elvik & Vaa,
2004) that determine which most macro factors have contributed to accident occurrence
and consequences. In all approaches, the chosen factors should be relevant to the endoutcome of road safety that one is seeking to measure.

43

I. Risk and Road User Behaviour: Different studies indicate that the human factor is
the major contributory factor to accidents (for example violation of speed limits). At the
same time, any error in the system and on roads will lead to unsafe road user behaviour:
Speed and risk of crash involvement: Speed has been identified as a highly important
influencing factor concerning road safety risk and consequences. An increase in average
speed results in a higher risk of involvement in an accident and greater severity. In
many countries, speed contributes to a significant percentage of all deaths on the roads.
Leaf & Preusser (1999), for example, concluded that reducing vehicle speeds could
have a highly significant influence on pedestrian accidents and injuries. Garber &
Gadiraju (1988) determined that accident rates increased with increasing variance of
speed.
Alcohol and risk of crash involvement: Drivers with high BAC (Blood Alcohol Content)
in their blood run a greater risk of being killed than those with zero BAC (sober
drivers). Hakkert & Braimaister (2002) provided a review of many studies and reported
that the risk in traffic will increase rapidly with BAC. Such results have given the basis
for setting BAC limits in many countries (e.g. .08 g/dl). Thoresen et al. (1992) have
shown a positive correlation between the total number of fatalities in Victoria State in
Australia with alcohol sales and inverse relationship with random BAC breath testing.
Age of drivers and risk of crash involvement: Road accidents are the leading cause of
death for young drivers and motorbike riders. The risk by age group per kilometre
travelled and per hour exposed to traffic is higher among young people (15-24) and old
(65+). However the exposure to traffic for young road users is higher than old people.
Evans (1991) reported that young male drivers are overrepresented in accidents in the
US. Page (2001) concluded from a survey in OECD countries that the higher the
proportion of young people in the population, the higher the number of road accident
fatalities.
Use of helmets: Motorcycle helmets have been shown to have a clear impact on
reducing fatal and serious injuries by between 20% and 45% (WHO, 2004). The same
study shows that bicycle helmets reduce the risk of head and brain injuries by between
60% and 80%. Wearing helmets reduces the probability of being injured by around 25%
(Elvik & Vaa, 2004). Many countries have legislated mandatory helmet use, which has
been effective in preventing, or reducing the severity of two-wheeler riders
(motorcyclists and cyclists). The use of helmets varies from country to country. In highincome countries, the usage rate tends to be high.

44

Use of seat belts: Road accident research has found that seat belts reduce fatal injuries
significantly and can reduce the risk of fatal injury to front-seat passengers. The use of
seatbelts varies from country to country. In high-income countries, the usage rate tends
to be high. In Sweden for instance, seatbelt usage exceeds 90% (Koornstra et al., 2002).
The use of seat belts reduces the probability of being killed by 40-50% for drivers and
front-seat passengers and by 25% for passengers in the back seats as shown in (Elvik &
Vaa, 2004). Regarding the use of safety seats for children and infants, studies (e.g.
WHO, 2004) have shown that infant deaths in cars are reduced by 70% and for small
children by 50%. Mandatory seat belt use has been proven to provide strong protection
against fatalities in accidents in different countries according to various studies.

II. Risk and Road Conditions: Fewer accidents resulting in fewer injuries happen on
motorways than on other types of roads because of the separation between vehicle
movements according to their speed (no high speed variance). Elvik & Vaa (2004) show
that the rate of injury accidents per million vehicle kilometres of travel on motorways is
about 25% of the average for all the public roads. Road surface conditions, poor road
surface, defects in road design and maintenance contribute to increased risk of
accidents. Bester (2001) reported that paved roads lead to lower fatality rates. Besides,
many studies have assessed the safety performance of similar roads between countries
by producing some sort of map or star rating (risk index) for roads (e.g. EuroRAP,
2006).
III. Risk and vehicle related factors: New cars tend to have more safety and protection
features, such as air bags, anti-brake system (ABS), etc. There is relation between
vehicle age and risk of a car crash. One study (in WHO, 2004) showed that occupants in
cars manufactured before 1984 run almost three times the risk of new cars. Many
developed countries have improved vehicle crashworthiness and safety, which means
better passenger protection (and for VRUs) in the event of a crash. Many countries in
the European Union (EU) as well as USA have set out legislation for safety standards in
motor vehicles, for instance the New Car Assessment Programme (NCAP), where
vehicle crash performance is evaluated by rating the vehicles models according to their
safety level for occupant protection, child protection and pedestrian protection. Vehicle
defects increase the risk of accident. The size of vehicle is crucial; the greater the mass
of the vehicle (e.g. heavy trucks), the more protection people have inside the vehicle
(their occupants) and the more others are involved in fatal accidents to. It is known that
poor vehicle maintenance and technical conditions can also contribute to accidents. In
terms of periodic vehicle inspection, different research shows different results. (Elvik &
Vaa, 2004) concluded in the review of macro-studies that there is no clear evidence that
periodic vehicle inspection has an effect on the number of accidents, while (Hakim et al.

45

1991) presented in another review of macro-studies that the periodic inspection of motor
vehicles reduces the number of road fatalities.

IV. Risk and post-crash injury outcome: Different studies have shown that fatality
rates are correlated with the level of medical facilities available in the country expressed
in terms of population per physician and population per hospital bed, see (Jacobs &
Fouracre, 1977) and (Mekky, 1985). A review of a European study, (WHO, 2004),
showed that about half of deaths from road accidents occurred at the accident spot or on
the way to the hospital. Noland (2003) concludes that medical care has led to reductions
in traffic-related fatalities in developed countries over time (1970-1996). The variables
used are: infant mortality rates, physicians per capita, and average acute care days in
hospital.
V. Risk by different transport modes: The ETSC report (1999) has compared the risk of
different transport modes in EU countries and it shows that the risk for cars (expressed
in terms of fatality per hundred million passenger hours or kilometres travelled) is 10
times greater than buses. The risk for the vulnerable modes (foot, cycle and motorcycle/moped) is 80-200 times more than buses. Motorised two-wheelers are the highest
risk among all other modes. The study gave a rough assumption of travel speed by the
transport modes.
VI. Socio-Economic Factors and Risk: There are many socioeconomic factors that
contribute to the causes of accidents. Some of the major factors are the following:
Gross National Product GNP: It is widely known that the motorisation rate (vehicle per
population) increases with income (GNP per capita). This may affect both exposure and
the risk of fatal accidents. Many studies (e.g. World Bank, 2003) have shown that the
fatalities per vehicle appear to decline rapidly with income. Maybe this reflects the shift
from vehicles with high risk (motorbikes, foot) to safer and protected vehicles (e.g.
four-wheelers) or it may show more funds and expenditure being spent by the country
on its road safety measures. There is a negative relationship between income growth and
the number of road accidents in the long term (Hakim et al, 1991). The increase in
income leads to safer vehicles and more investment in road infrastructure, which leads
to fewer road accidents and casualties. However, it should be clear that the
improvement of income could also increase the travel distance (higher exposure) and
alcohol consumption (higher risk).
Unemployment: Few studies have used the unemployment factor as a risk factor for
accidents. It appears to be negatively related to accidents and casualties. Hakim et al.
(1991) has shown in the literature review he made that an increased unemployment rate

46

in country might reflect on the ability to pay for a single journey and a reduced exposure
to the whole journey. Page (2001) included employment (percentage of population in
employment) into his model in the study conducted for the OECD countries (19801994). The higher employment figures showed an increase in the number of fatalities.

Urban population: Urban roads will have more accidents and fewer fatalities or severity
per kilometre travelled than rural roads, because of the density of vehicles and the lower
speeds of travel. Hakkert and Braimaister (2002) have shown in one macro-study that
countries with a high level of urbanisation will have higher population densities and
they may experience lower levels of fatalities and serious injuries. Page (2001) has
found that the population who live in urban areas have fewer road accident fatalities
than other places. Bester (2001) also reported similar results, that countries with higher
road densities will have fewer fatality rates. Shorter distances to medical services can
explain this.
Illiteracy: Bester (2001) has analysed socio-economic factors in different countries and
he found that the illiteracy percentage has a statistically significant effect on the national
fatality rate. He explained that a country that can read and write is expected to influence
the ability of road users to understand the rules of the road and road signs.

VII. Risk and other factors: Different macro studies have shown that the risk of
crashing will increase by other factors such as: poor visibility, using hand-held mobile
telephones, dark conditions, wet roads and roads that are covered with snow or ice
(Elvik & Vaa, 2004), (Evans, 1991). There is an inverse relationship between accidents
and technology level in a country as well as and the enforcement effectiveness level:
Technology level: Few studies have described the decline in the number of fatalities in
all industrialised countries as a result of the increase of technology use in vehicle and
road infrastructure accompanied with a better communication between road-vehicledriver-environment (i.e. Evans, 1991; OECD, 2003). We are waiting for better
measurements in this area at macro-level, for instance by assessing the level of
development and availability of Intelligent Transportation Systems (ITS) applications in
a country: such as intelligent speed adaptation in cars, electronic driving licence, speed
enforcement cameras, etc.
The enforcement effectiveness: the traffic police enforcement effectiveness has been
shown to have a clear impact on the number of accidents. It increases the perceptions of
drivers to the risk of accidents. Several studies highlighted the strong relationship
between the increased police enforcement and the reduction in number of accidents
(see, e.g. Zaidel, 2000; Escape, 2003; Cameron and Diamantopoulou, 2000) can be

47

measured by factors related to the proportion of drivers who had been checked by traffic
police for speed, alcohol or seatbelt.
Furthermore, it seems that an increase in the average price of gasoline reduces the
number of trips and the exposure (Hakim et al., 1991). Similarly, there is an inverse
relationship between number of accidents and the number of driving licenses delivered
(Van and Wets, 2003). Moreover, the road safety audit process is shown to have a clear
impact on the number of accidents (Proctor et al., 2001). However, there is lack of data
concerning all these factors and they are not available in many countries.
To sum up, the following table lists the general macro factors, grouped into three
categories of exposure, risk and consequences:
Table 3.1: Macro-factors influencing exposure, risk and consequences
Factors influencing exposure to traffic
- Economic factors and GNP per capita
- Urban population density, and other demographic factors
- Type of travel mode choice
Factors influencing risk of accident
- To driver: speed, alcohol and driving, being young, etc.
- To groups of road users: unprotected road users
- To vehicles: motorcyclists, heavy trucks, non-motorised traffic, vehicle inspection, age of car, etc.
- To roads: unpaved, defects in road design, poor maintenance, etc.
- To environment: darkness, fog, ice, etc.
Factors influencing accident severity
- Human factors: speed, helmets worn by users of motorcyclists and cyclists
- Vehicle: passive safety (e.g. seatbelts, airbags, child safety seat, vehicle safety and protection
standards)
- Crash-protective roadsides, guardrails, barriers
- Poor rescue and pre-hospital emergency care
- Poor country health care system

It is, perhaps, desirable to emphasise again that this research analyses indicators and
dimensions at national level, so it is irrelevant to benchmark country performance in
road safety in terms of any micro indicator, even though such indicators can influence
accident risk, exposure and consequence. The following are some examples of micro
indicators: time of accident, travel route, length of trip, mix of traffic, location (i.e.
urban/rural areas, intersections), type of crash (i.e. side, front, or rear crash), gender, age
distribution.

48

As one can see, from the summary table that the three categories exposure-riskconsequences are highly interrelated and many indicators are included in more than one
category (e.g. speed can influence both risk and consequences). Therefore, it needs to
gather the indicators into practical dimensions of which each represents a single topic
area of progress in road safety, as will be discussed in the next section.

3.2

Stage two: Finding a list of macro-dimensions in road safety

The safety situation can be quantified by a combination of risks in traffic as

probabilities (e.g. speed risk, alcohol and driving risk, etc.). For instance this can be
indicated by the following formula (Koornstra, 1996):
SAFETY = ( w1.R1 + w2 .R2 + ..... + wn .Rn ) kilometres
where

(3.1)

R is the risk of particular factor

Kilometres is the exposure in traffic, and
w is the weight of the risk factor in a particular country

Additionally, (i.e. Navin, 1994; Haight, 1983) accidents are caused by a combination of
five main dimensions (human-vehicle-environment-road-system). These dimensions are
not fully independent of each other and each dimension is influenced by many factors.
This function can be expressed as:
Y=F (X)
where

(3.2)

Y is the number of accidents or the accident rate

X is a vector matrix of the explanatory variables in each of (human-vehicleenvironment-road-system).

Currently, many studies (i.e. European studies) also recognise the importance of having
a comprehensive set of performance themes in road safety. The two studies ETSC
(2001, 2003) have identified four themes which are: first in behaviour: speed, alcohol,
seat belts; second vehicles: passive safety; third roads: percentage of roads meeting
design standards; and fourth on trauma management: arrival time and quality of medical
treatment.
SafetyNet (2005a) has been directed at identifying uniform comparable safety
performance indicators and dimensions between all EU Member States. It focuses on
seven performance areas which are: the use of alcohol and drugs, speeds, use of
seatbelts and crash helmets, use of daytime running lights, vehicles, roads, and trauma
management. The study also focuses on the causes of accidents and the number of

49

deaths by group of road user (i.e. pedestrians, children, mopeds & motorcycles, cyclists)
or by contributory factor (i.e. speed, alcohol and driving).
A research work by Ecorys & SWOV (2005) presented possible operational and output
indicators in road safety. In such an effort, the study identified some several key areas
such as: increased enforcement, drivers licenses, passive and active safety,
infrastructure, commercial drivers, and trauma management.
Based on the brief review and discussion presented above and in previous sections, the
study has identified nine dimensions (see Figure 3.3) of which each corresponds to a
special area of road safety. The dimensions listed are:

Traffic risk: the risk of a person being killed in road accidents per vehicle or per
vehicle-km.
Personal risk: the risk of a person being killed in road accident per number of
inhabitants.
Changing trend: the percentage change of death trend over time. There is an indication
of an improvement in the overall traffic risk and personal risk in most countries
including LMCs, while the number of deaths has been reduced mostly in HMCs alone
over time.
Safer vehicles: this dimension assesses the safety characteristics of vehicles in a country
that affect the number of vehicle-related crashes (e.g. type of vehicle, new cars in a
country, inspection of vehicles, index of national crashworthiness).
Safer roads: this dimension measures different aspects of the quality and conditions of
roads in a country in relation to road safety (e.g. motorway level, paved roads and
national expenditure on roads).
Safer people (road user behaviour): this dimension assesses human behaviour and
traffic safety with respect to speeding, drink-driving, helmet use and seatbelt use.
The socioeconomic factors: this category considers a countrys development and
investments in relation to health level, education level per capita, urban population and
income level.
Traffic police and enforcement: measures traffic police and enforcement effectiveness
levels in the country.
Road safety organisational structure: measures the efficiency of the national road safety
programme, action plans, legislations and how much cooperation is made between the
key bodies responsible for road safety actions in the country.
The groups measure road safety development in terms of direct or indirect classes. The
direct indicators are derived measures (outcomes or direct symptoms), for example
death rates (traffic risk and personal risk) are considered as direct measures for

50

explaining national road safety development. The indirect dimensions are individual
means (practices or indirect symptoms) in the way they could describe the development
in a particular theme to road safety. The indirect dimensions can be also seen from topdown or bottom-up approach according to their contributions to the national road safety
development. For instance, policymakers can focus on human-vehicle-road dimensions
to make changes at top level, or they can focus on organisational structure (e.g.
education, legislations, and traffic safety management), socio-economic system and
enforcement to make changes at local level. However, both of these approaches
contribute toward the same goal of national road safety development.

Indicators related to Socio-economic

factors

Direct
dimensions

Traffic Risk
deaths per
vehicles

Indicators related to Road safety

organisational structure

Personal Risk
deaths per
population

Indicators related to Traffic police

and enforcement

Percentage
change of
deaths trend

Indirect
dimensions

Indicators related to Safer vehicles

Indicators related to Safer roads
Indicators related to Road user
behaviour

Figure 3.3: The selected macro dimensions in road safety

Furthermore, special care should be taken to avoid double counting. The indirect
dimensions of performance should focus mainly on indicators that are related to
practices. For instance, the indicator of percentage of seat belt use is more suitable
than the indicator of traffic deaths due to not using seatbelts, and similarly the
percentage of new cars is more suitable than traffic deaths due to poor vehicle
conditions and so on. The indicators of traffic deaths by group (pedestrians, cyclists,
etc) or by contributory factor (speeding, alcohol and driving, not using helmets, due to
poor road conditions, due to poor vehicle conditions, etc.) are all represented by traffic
risk and personal risk (direct dimensions).

51

3.3

Stage three: Criteria for selecting macro-performance indicators and

dimensions in road safety

To assess the usefulness and applicability of the indicators and dimensions, a criteria
has to judge the selection and quality of indicators in the master-list.
The value of each chosen indicator is expected to correlate with the road safety
outcome, which means it is related to the probability of accidents to occur under
different conditions. However, It is possible to find multiple indicators relating to a
single dimension, which needs a clear selection especially of which ones to choose and
why. For instance, by taking the health level in a country as an example (road safety is a
health matter), possible indicators could measure this: the proportion of the total health
care expenditure as a percentage of GDP, life expectancy at birth (years), physicians per
population, hospital beds per population, etc. However, a higher number of hospital
beds in a country does not necessarily lead to better health. In addition Life
expectancy at birth which is being used by HDI, is not enough to measure the health
level in a country. The focus should be first on more comprehensive and efficient
indicators and second on the outcomes (outputs)4 not on the mere numbers. In terms of
road safety, it will be useful to add up the severity index (proportion of deaths per total
accident casualties) as a good indicator of health, which severity index decreases with
better medical facilities and rescue services. However the number of injuries is still
widely underestimated in a large number of countries. According to a global study by
Jacobs et al., (2000), only 50 percent of road injuries are reported from road accidents.
Consider, for instance, the indicators that can be used to measure the change of the
number of drivers above the legal BAC limits or driving exceeding speeds limits
checked by police in one particular country over a year. This change could indicate a
higher real traffic violation rate, but at the same time it can be a sign of an increased
level of reporting and checking by police.
On the way to develop a complete set of performance indicators that can be used for
international comparisons, one needs to remember that such a choice is restricted by
certain conditions and requirements for both the indicators and data as following:

First, the indicators should have effect on road safety if any change in indicator
occurred. They should represent the improvement in the situation and be reasonably
accepted from different studies and literature surveys.
4 An example of the outcome indicators (in health sector) that have been developed in recent years are:
the number of specific surgical operations made in the country annually, and the level and spread of
infectious diseases in the country.

52

Second, one should care about the quality of each indicator and data. There are usually
several data sources available from which we can find data to measure. But the data
should come from one or more reliable sources (national and international).
Third, the indicators chosen should be clear and with a precise definition. For example,
an indicator of safer vehicles or urban roads standards without a clear and precise
definition of what we mean by the words safe or standards could easily lead to
unclear collection of data. This might very well lead to a misunderstanding of what
results are being achieved. In addition, the number of deaths in accidents cannot be
directly compared internationally since the individual definitions of road accident deaths
differ from country to country. Most European countries use the standard of dead within
30 days of the accident occurred, which many LMCs use the definition of deaths as on
the spot or within 24 hours.
Fourth, the reliability of any indicator means that there is no real, major and sudden
change in the indicator for a country being measured between different sources and over
time.
Fifth, indicators should be simplified, to various degrees, in order to make it possible to
measure and to be easily understood by the widest possible audience. Some indicators
are simple, have a relatively direct meaning and can be expressed in units which most
people are comfortable with (e.g. motorcycles as a percentage of total fleet vehicles).
Other indicators are more complicated but have a long experience in the field and are
supported by research (e.g. deaths per vehicle kilometres travelled).
Sixth, we should always use a group of indicators relating to the desired objective we
want to describe. But at the same time, we should not allow the set of indicators to
become too many because that will take too much time to interpret and analyse (also it
is a matter of cost). There is no exact number of indicators; rather the number should at
least capture the results sufficiently for what we want to obtain. The chosen indicators
should be as minimal as possible. For instance, road user behaviour in the country may
require many indicators to capture the major aspects. However, we should remember
that if we have identified a large number of indicators for a single aspect, this might
mean that the aspect is too complex or too important or more data is being collected
than necessary. In brief, it needs as well a balanced set of indicators as possible.
Last, as is known, the data collection process should be available year-to-year. This will
make the indicators available and accessible whenever data is needed. The data and
indicators should be updated more frequently.

53

3.3.1

Sample of survey and the multidimensional index

It is not possible to collect data from all countries to examine the chosen indicators. This
needs work from international organisations. A sample of countries might well be
enough to select for the purpose of this dissertation.

3.3.2

Quantitative versus qualitative indicators

Not everything in road safety that is known and is important can be counted. Several
indicators of development involve subjective judgments. For example, the degree of
development of the National Road Safety Programme (NRSP) in a country cannot be
easily quantified into numbers. Often these judgments can be measured, using
questionnaires or opinions of expert panels, and they can be translated into quantitative
rating systems. But here we need to assume that the expert panels are recognised as full
experts in this area of interest. Also we will assume that the combined information they
have is good enough to judge on the whole issue we seek for. Moreover, we have to
note that it would be difficult to obtain the opinions of experts regularly and whenever
they are desired.

3.3.3

IT supports the macro-performance indicators

Computer databases can facilitate the accessibility to a large road safety data, indicators
and other information from different countries. This will allow a quick analysis of the
data with regular updating. Then the databases could provide this information to the
country policy makers and to the public, which will help them in drawing attention to
these phenomena. However, such database must be regularly maintained to ensure the
accuracy of information. The Globesafe database (Globesafe, 2006) in Sweden is one
example, among others, of such international database.

3.4

Building a master-list of macro-performance indicators in road safety

Based on the criteria and discussion in previous sections, a master list of macro
indicators and dimensions is gathered in Table 3.2. The indicators are classified into
three classes according to their data availability in different countries, quality and type.
A high availability of data means that enough data is always available to access.
Medium availability means that data is under development and there is a call for data
collection in these indicators. A low rating means that data is currently not available and
they require further development and collection of data in the future for a large number
of countries, especially from developing countries. For instance, one acceptable

54

indicator is deaths per vehicle kilometre and it is an important element in road safety,
but this measure is still not available in most countries. In fact, there are several macro
indicators that may play a role in road safety development in a country, but
unfortunately some indicators are hard to measure for now and they will be kept for
further and future development.
A high quality of data means that data comes from reliable sources with minimum
errors. Acceptable data means that it is fairly reliable and accurate. Poor quality means
that we have a data quality problem and the results are questionable and should not be
fully trusted. Not rated means indicators are not identified for now and there is no
indication of what the quality could be in future. For instance, the indicator percentage
of motorways per road network has poor quality of data and there is difference of
definitions of road standards across countries.
The size of quality and availability of the indicators have been estimated based on a
sample of countries from different parts of the world. The Globesafe database
(Globesafe, 2005) was used as a prime source of data. However, any future attempt to
estimate the quality and availability of data would require special techniques done
directly with computer.
The indicators, as mentioned earlier, may either be quantitative/objective or
qualitative/subjective. In some indicators, there is lack of quantitative information and it
will be better if subjective indicators are included. Subjective indicators can be
considered as complementary indicators if they are translated to a quantities rating
system. Both organisational structure and enforcement measures are based mainly
on subjective assessments and at this stage it is a major problem to gather experts
assessment regarding both dimensions and to measure subjective indicators. However,
both dimensions can be developed and become more concrete indicators for the future.

55

Table 3.2: The master list of macro road safety indicators and dimensions
Core Dimensions and Indicators
Traffic Risk:
Death rate (per vehicles-km)
Death rate (per person-km)
Deaths per vehicles
Personal Risk:
Death rate per population
Death rate per population age
Changing trend in the number of deaths (%)
Road User Behaviour:
Percentage of seat belt use
Percentage of crash helmet use
Percentage of drivers above the legal BAC limit in police checks
Percentage of all drivers exceeding speed limits in police checks
Consumption of alcohol per capita (litres)
Minimum age for driving
Vehicle Safety:
Index of national crashworthiness (vehicle crash performance)
Distribution of vehicles by age: Percentage of new cars
Mass classes of car fleet (%)
Percentage of buses and coaches in total vehicle fleet
Percentage of vehicles excluding motorcycles
Roads Safety:
Percentage of roads paved
Total paved roads (km) per capita/vehicles
Total motorways/freeways (km) per capita/vehicles
Km of motorway per km of paved road
National expenditure on roads (engineering/maintenance) % GDP
National expenditure on roads (safety measures) % of GDP
National expenditure on roads per total vehicles
Socioeconomic indicators:
Percentage of urban population
Percentage of 15-24 years old in the total population
Population density (people per km2).
Life expectancy (years)
Population per physician
Average acute care days related road accidents in hospital
Severity index (number of deaths per total casualties)
Illiteracy: percentage of persons over 15 years unable to read
Gross National Product (GNP) per capita
Percentage of people in unemployment
Traffic Police and Enforcement:
The annual number of random breath tests (per vehicles)
The annual number of speed-violation notices (per vehicles)
The annual number of seat belt violation notices (per vehicles)
The technical means/equipment available to the traffic police
Organisational structure:
The development of the National Road Safety Council
How far the cooperation between the key stakeholders and NGOs
The funds level are spent on road safety measures
Legislations level, data collection level and statistics
Inspection of vehicles
Number of national campaigns in the last three years
Number of driving licenses delivered per total vehicles fleet

Availability

Quality

Type

Low
Low
High

Good
Acceptable
Good

Objective
Objective
Objective

High
Medium
Medium

Good
Acceptable
Acceptable

Objective
Objective
Objective

Medium
Medium
Low
Low
Medium
Medium

Acceptable
Acceptable
Poor
Poor
Acceptable
Acceptable

Objective
Objective
Objective
Objective
Objective
Objective

Low
Low
Low
High
High

Acceptable
Acceptable
Acceptable
Acceptable
Acceptable

Objective
Objective
Objective
Objective
Objective

High
High
Low
Low
Low
Low
Low

Acceptable
Acceptable
Poor
Poor
Poor
Poor
Poor

Objective
Objective
Objective
Objective
Objective
Objective
Objective

High
Medium
Medium
Medium
Medium
Low
Medium
Medium
High
Medium

Acceptable
Acceptable
Acceptable
Acceptable
Poor
Poor
Acceptable
Acceptable
Acceptable
Poor

Objective
Objective
Objective
Objective
Objective
Objective
Objective
Objective
Objective
Objective

Low
Low
Low
Low

Acceptable
Acceptable
Acceptable
Not Rated

Objective
Objective
Objective
Subjective

Low
Low
Low
Low
Medium
Low
Low

Not Rated
Not Rated
Not Rated
Not Rated
Not Rated
Acceptable
Acceptable

Subjective
Subjective
Subjective
Subjective
Subjective
Objective
Objective

The increase in availability of data will improve the quality of data. In other words, the
quality of indicators is good when data is available. At the same time, the importance of
each indicator depends on its type, availability and quality. Thus, for increasing the

56

usefulness and applicability of the indicators and dimensions, efforts have to be made to
enable a high quality data delivery, especially to those indicators are considered
important and vice versa, see Figure 3.4.

Quality of
indicators
(Accuracy)

Indicator C

Indicator A
Indicator B
Importance
of indicator

Figure 3.4: Criteria to improve indicator quality and importance

To conclude, the selection of the macro-indicators and dimensions is not final. The
indicators will continue to be developed and refined. This selected master-list is useful
for different benchmarking purposes. The selected indicator should be clearly defined,
measured, regularly available and be comparable over time. In next chapter, the most
appropriate indicators from (Table 3.2) with high data availability and acceptable
quality will be selected and linked together into a short key list, which will be useful for
the construction of RSDI.

57

58

Chapter 4: The conceptual framework of RSDI

Chapter 4

The conceptual framework of RSDI

The aim of this chapter is to describe the conceptual framework of RSDI
and how it can be developed and applied. Also it shows the possible
benefits and challenges in developing such multi-dimensional index. The
RSDI prioritises the key national outcomes in road safety. Three pillars
and nine key dimensions are involved in RSDI framework. This chapter
identifies short-term list of key performance indicators for road safety that
can be applied uniformly for most countries in each of HMCs and LMCs.

4.1

Introduction

This chapter assesses the fourth and fifth research questions (stated in Chapter 1) i.e.:
What are the knowledge and criteria that must be addressed when aggregating the
indicators into one single index?
Do death rates have to be replaced with the new index, or be a supplementary part, or
be part of the new index?
One of the major challenges to assess road safety in a country is that the traditional
measurements used are not powerful enough to model the complexity of road safety
situation in a country (as already discussed in Chapter 2). Most previous work in road
safety benchmarking has focused on the product of road safety in terms of accident
death rates. However, road safety performance in a country needs to be placed and
understood in a broader context in terms of key practices and outcomes. This has
brought me to develop RSDI in earlier studies (Al-Haji, 2003) and (Al-Haji, 2005).

4.2

What is RSDI?

The RSDI can be defined according to its conceptual criteria, methodological

approaches, process or the relation between its components. These definitions are
interrelated and complement to each other as shown as follows:

59

The RSDI is an international benchmarking model that indicates and communicates in a

comprehensive and easy way the severity of the road safety situation in a specific
country and/or in comparison to other countries in time.
This definition is based on the overall goal stated in Chapter 1.
The RSDI is an integrated benchmarking model for road safety performance and
development.
This definition is based on the discussion in Chapter 2-part 1 (see Figure 2.1).
The RSDI provides a framework for developing a basket of road safety performance
indicators that can be used to indicate how well the overall performance of road safety
in a country is carried out.
This definition is based mainly on the discussion in Chapter 3 (see Figure 3.1).
The RSDI translates the outcomes about road safety in a country into a single
quantitative value (composite index).
This definition is based on the discussion in Section 2.2.1 (see composite
indices).
The RSDI links the key practices that affect the overall outcome to each other and to the
end-results (accident death rates). This will ensure that countries are benchmarked on
the effectiveness of their outcomes and experiences in road safety.
This definition is based mainly on the discussion in Sections 2.2.1 and 2.2.2 (see
e.g. business excellence models).
This chapter will further discuss and analyse these definitions with more focus on the
theory and principles behind RSDI.
At the beginning, let us first simplify further the framework of RSDI in relation to its
intentions. The RSDI is intended to answer:

Which country performs better than another?

Why is a specific country more successful than others?
How and what measures a successful country has used to improve its road safety
work.
What actions have to be taken to improve road safety performance in a country
in future?

The RSDI is also intended to rank the performance of any country in terms of its
practices and product of road safety against:

60

The performance of other similar countries.

A reference country.
A target value.
Past performance (e.g. previous year) for the same country.

The RSDI results ranged from 0 till 100. The higher values indicate a higher level of
safety in the country. The lower values indicate the worst performance in country in
terms of road safety level/work. The target value of RSDI is 100 and it indicates how
close a country is to provide safer roads and safety infrastructure (safety programmes,
law enforcement etc). An RSDI-value will stimulate and influence countries to improve
their road safety work and will help them to determine the performance gap between
their current results and the targets set up.

4.3

The expected benefits of using RSDI

As we move into the 21st century, we also move into this increasingly improved
technology in communications such as the Internet and processing speed via broadband
capacity etc. Today computers are developing rapidly, which simplifies the work and
analysis of a large amount of international road safety data that was not available
previously. This development has made the work on a multidimensional index easier
and brought it closer to reality. Briefly, the main purposes of RSDI are:

To be useful in describing the contributions of various indicators to the overall

level of road safety.
To provide a broader picture of the whole road safety situation in a country
compared to the traditional models.
To be a simple, quick and more easily understandable form. This can increase
the awareness of the problem among both the general public and policy makers.
Policy makers will be able to take appropriate decisions, setting targets and
priorities for the future.
To serve as a tool for Self-Assessment, allowing countries to identify their areas
of strengths and weaknesses in road safety.
To tell if a country is going into the right direction to prevent injuries and reduce
the risk of injury in road traffic.
To make the key stakeholders in road safety and NGOs become more actively
involved in the process of promoting road safety in the country.

61

4.4

RSDI quality criteria

The quality of RSDI or any composite index depends on the quality of the indicators
and the data used in the index. The major obstacles in constructing any international
index are the differences in definitions, non-collection of data, non-reliability of data,
non-validity of data, and under-reporting. These problems are major problems in many
countries and it leads to less reliability on indicators and as a result create less reliability
on the derived RSDI index. The size of the problem of quality data differs from country
to country and it seems to have more severe consequences in many LMCs.
A number of issues can be addressed here concerning the criteria of RSDI:

RSDI must be drawn from annual and national statistics that come from reliable
sources.
RSDI should be valid and regularly updated based on comparable and available
indicators/data.
RSDI should provide a clear description of the selected indicators and the
theories behind them.
RSDI should provide a clear selection of methodologies in the construction and
analysis.
The total number of indicators used in RSDI should be a relatively small and
well-balanced set (as possible) that is able to capture the most of desired results.
RSDI should be as relevant as possible for most countries worldwide.

This set of criteria aims to improve the transparency, validity and accuracy of the RSDI
index.

4.5

The principles and philosophy behind RSDI

In RSDI, the success in national road safety reflects the whole quality-related outcomes.
It is a multiple dimensions of outcomes. Managing people (e.g. road users behaviour),
system quality (key practices) and product quality (accident death rates) will make it
possible to assess success in the road safety work carried out in a country.
There are key practices in a country that are more likely to be considered good and
useful outcomes for international benchmarking such as seat belt use, helmets use,
motorways level, etc. It can be assumed that the change for road users behaviour (e.g.
increase of seat belt use) or the changes of road and vehicle safety quality are also
outcomes/success factors as a result of a particular policy and practice in a country.

62

RSDI can serve as an award for development of the country, which provided good
outcomes in the whole road safety system.
According to Figure 2.1 (Chapter 2), RSDI was described as an integrated
benchmarking model of all the three types of benchmarking: Product, Practices and
Strategic Benchmarking. However, different studies (i.e. Rumar, 1999) indicate that the
human factor (road users) is the major contributory factor to accidents. The Human
outcomes (activities) play the key role in the development and understanding of road
safety in a country. Thus, the human dimension has the potential to become one of the
main pillars of RSDI.
There are three pillars of outcomes involved in the framework of RSDI, which are
illustrated in Figure 4.1. Each pillar comprises a number of dimensions (which will be
explained later):

People focus: Examines how successful the country has been in promoting
safety among its people and making people behave safer on roads. Enhancing
the awareness of the people first so that the country knows what, how and why
to improve its road safety?
Service/system focus: Examines how effective a country has been in improving
its services and system results in terms of safer cars, safer roads, higher
investment on safety measures, education, better health and rescue services,
better organisational structure, etc. It examines how the country addresses its
objectives and targets as well as actions to achieve these targets.
Product focus: Examines a countrys performance in accident death rates.

The product of safety (in terms of accident death rates) may answer partly which
country performs better than another, but it will not answer why one country is better
and what actions/measures another country has to take to improve its performance?
Traditional benchmarking efforts focus only on the superficial level (comparisons and
rankings of accident death rates) and not on the overall system (internal level) and
human assessments (core level). Superficial outcome in road safety can be easily seen in
society from simple observation and data such as: number of deaths, injuries and cost,
whereas the internal and core outcomes in road safety are not simply obvious and they
need more checking and analysis about for instance traffic management, education,
human behaviour, traffic police enforcement, legislations, etc.

63

Human
performance

Safer road users

"behaviour"

Socioeconomic
performance

RSDI

Organisational
performance

Product
performance

Percentage
change of
deaths trend

Personal risk
"deaths per
population"

Traffic risk
"deaths per
vehicles"

System
performance

Safer roads

Enforcement
performance

Safer vehicles

Figure 4.1: RSDI conceptual framework (overall road safety performance)

It must be emphasised that there is no guarantee that countries with good scores only in
the quality of the product (accident death rates) will not have trouble in the future in
terms of number of deaths and injuries. Maybe there is a low risk of death rates (traffic
risk and personal risk) for some particular reasons but there is probably, at the same
time, an increasing risk somewhere else (e.g. enforcement, speed, alcohol and driving,
pedestrians, etc.). Accident death rates reflect the exposure measures in terms of number
of population and number of vehicles, but they do not reflect the overall continuous
improvement in road safety. A national failure in a given year is not necessarily due to
poor performance in this year but it is probably due to weaknesses over a period of time
in different aspects and activities. This can explain why the decrease in the number of
deaths (or accident death rates) has stopped or fluctuated or has started to show an
upward trend in some countries (municipalities). This can be illustrated with two
examples from HMCs. The first example comes from Sweden, which has a successful
record over decades against drunk driving and its consequences and it has shown a
continuing decline in the total number of road traffic deaths. However, recent figures
(between 2000 and 2002) have shown an increase of the number of alcohol-related
traffic deaths in the country. This was explained as a result of the reduction of alcohol
taxes, which lead to an increase in total consumption per capita (Laurell, 2003). This
unacceptable performance in this area had led Sweden to increase its efforts, make
modifications and take countermeasures (e.g. enforcement and media campaigns) to
mitigate this problem. The second example from Canada and the United States where

64

despite the decrease in the overall number of road user deaths and related risks between
1997 and 2000, there was a large increase in motorcycle deaths occurring during the
same period (OECD, 2002b). These figures gave a sign of failure. Countries have to
reduce such problems early, before it is too late. To do this there is a need for effective
analysing tools.
The system quality and human performance are the missing links in describing and
comparing the complexity of road safety situations between countries. If there is a
failure in any area, the success in that country will fail as well. Thus RSDI may provide
an early, quick and direct warning signal. If RSDI shows poor performance in any area,
changes can be made to prevent a worsening situation.
The combination of product-system-human measurements into RSDI will be also useful
to determine how similar countries in transport patterns can be different in road safety
performance, especially for those countries that have very close accident death rates
(e.g. United Kingdom, Sweden and Netherlands).
Figure 4.1 shows that the product of road safety is produced by human and system
activities. At the same time, there is a relationship between both human and system
pillars. This can be explained by the fact that any error in the system (e.g. road, vehicle,
enforcement, etc.) will lead to unsafe road user behaviour. On the contrary, any human
error (e.g. decision-making error) will lead to an unsafe system. The RSDI provides a
framework for identifying a set of human and system indicators simultaneously, which
influence each other and ultimately they influence the product results.
According to the discussion in the previous chapter (Figure 3.3), nine key dimensions
have been selected in the construction of RSDI. Each dimension has been broken up
into indicators or sub-indices. The dimensions have been linked together by using the
approach direct-indirect outcomes. The direct outcomes refer to the product of road
safety in terms of accident death rates, while indirect outcomes refer to the system and
human performance. The indirect outcomes can be seen either as individual means in
the way that they can describe the development of a particular aspect relevant to road
safety or that they can offer a significant effect on the death rates. There are many
popular composite indices, which have included both direct and indirect measures into
the same index. For instance, the European Innovation Scoreboard Index (The European
Commission, 2004) includes both R&D expenditure (indirect) and the number of new
products and services (direct) in order to measure the innovative activity in a country.
Another example, the Composite Health System Performance Index (Lauer et al., 2004)
includes health level as a direct indicator and health distribution as an indirect
indicator.

65

The choice of indicators, that can be useful for making a meaningful international
benchmarking, depends on what is available at national-level data among countries. The
data collected at national level should indicate the countrys performance and should
summarise the large amount of detailed data that is available at the local and provincial
level.
In this context, there is a serious need to understand and distinguish between input
factors, process factors, output factors and outcome factors, which represent together the
overall road safety performance (see Figure 4.2). At the local level, input factors refer to
the traffic and country resources that produce the outputs (e.g. number of accidents,
deaths, injuries). These may be measured in terms of items such as number of vehicle,
population, length of road network, environmental factors, etc. The process (activities)
reflects the best use of road safety resources (inputs) that will produce measurable
outputs. These processes are based for instance on a combination of the "three Es" of
road safety: Education, Engineering and Enforcement. The output indicators are
expected to lead to the performance outcomes, but by themselves they do not explain
anything for international benchmarking (e.g. number of accidents, deaths, injuries,
material damage). At the top level, the RSDI gather the national key-performance
outcomes into a single value, which represents the best view of road safety performance
in a country.
The advantage of using RSDI is that results can be linked to strategies and action plans
which can start from the RSDI outcome to make changes in the input and vice versa. As
can be seen from the figure, the more movement towards RSDI level, the higher
responsibility, decision-making and less detail of data will be reached. The decision
makers in a country will be able to make appropriate decisions and take appropriate
actions for a strategy of a future development.

66

etc.

Local

Number of driving
licences holders

Quantitative targets

Urban vs. Rural data

Environment, fog, ice,

darkness

Economic factors (e.g.

GNP per capita)

Number of road users

(e.g. pedestrians, cyclists,
etc.)

Local-provincial-national

etc.

Research and development

etc.

Breath tests per vehicles

No. of drivers use seat belt

No. of cyclists use helmets
etc.

National

% of seat belt use

% of helmets use
etc.

Human performance

etc.

Good practices
Technology transfer

No. of random breath tests

% of motorways

Percentage of new cars

Age of cars
Km of motorway

HDI

Education, health, GNP

Services/system quality

Traffic Risk
Personal Risk

Product quality

Performance indicators

Outcomes of performance

Number of accidents, fatalities,

injuries, cost, etc.

Indicators/outputs

Action plans

Institutional framework and

coordination

Traffic safety management

Safer vehicles

Safer roads

Safer people

The three Es: Engineering,

Education, Enforcement

Length of roads by type of

roads

Number of population

Investment in road safety

Process/activities

Number of vehicles by
type of travel mode

Inputs

Figure 4.2: The input-output-outcome-RSDI framework

International

RSDI
performance
in a country

RSDI performance/aggregation

Key
Performance
Indicators

Gap of
performance
"success"

RSDI
target

Success

67

4.6

The RSDI model from an economic perspective

As mentioned in the literature review the success of any company/organisations

performance is no longer linked only to the quality of a product but also to the core
quality of services, delivery, practices, human resources, competence, etc. Business, in
general, recognised the importance of having a multi-dimensional index for managing,
assessing and controlling the performance in the company as EFQM model (EFQM,
2006).
Economic performance principles can be served as an inspiration to RSDI. However,
applying economic thinking to road safety field is not an easy task and requires new
approaches of analysis because of the complex nature of road safety as a process, as
practices or product. The safety product is road accidents which equal loss, while the
reduction in the total number of deaths and injuries and associated costs equal profit.
This profit is an immediate profit while the total quality of outcomes (e.g. RSDI) is the
long-term profit.
By understanding how an organisation improves its performance and excellence, it may
be able to translate their economic performance factors into a more similar form to
RSDI pillars and principles. In order to do this, an attempt (see Table 4.1) was made to
interpret economic terms with the RSDI concepts (human, service and product) and
other general terminological sets of road safety.
The focus on road users (as employees and customers) will encourage the involvement
of road users in the road safety process, which will make them understand their
contribution to the overall development of road safety. The focus on policy makers
(employers) will encourage them to make roads, vehicles and system safer and as a
result this will ultimately make the product of safety profitable. The focus on product
will bring action plans and professional know-how to the product testing in terms of
best practices, and to the product marketing in terms of road safety campaigns.
The human dimension in road safety is interrelated and complex, and thus it is often
difficult to identify. People can make process and product at the same time. People can
reduce their risk in traffic and increase safety at the same time. To give an example,
road safety policy makers in a country are employers in terms of their efforts to improve
road safety, but they are also customers, at the same time, when they are exposed to
traffic risk as drivers, pedestrians, passengers, etc. Understanding the human dimension
in road safety or any business is the best approach to making everything around work
effectively and safely.

68

Several empirical studies in economy have shown that training and education of
employees are necessary to make a successful TQM in an organisation. Similarly, the
education of road users and training of traffic safety personnel (e.g. health, police,
engineers, etc.) are essential for road safety improvement.
Table 4.1: Applying economic performance principles to RSDI performance

Human
Performance

Economic performance factors

for an organisation/company

Road safety performance factors in a

country

Employer leadership, management

and organisation

Market and information analysis, e.g.

market share, demand and supply
Organisation management and
planning

Policy makers (e.g. national road safety

council "NRSC")
Road users and traffic safety personnel (e.g.
health, police, engineers, etc.) education,
training, behaviour, etc.
Public and society satisfactions of road
safety levels and risks
Country diagnosis problems, causes and
priorities
Traffic safety management (e.g. National
Road Safety Programme "NRSP")

Product design

Action plans

Sources and supply quality

Sources and investments in road safety

Countermeasures: the four Es: Education,
Engineering, Enforcement and
Encouragement
Good practices and lessons learned

Employee education, skills, training

and development
Customer satisfaction

Service
Performance

Manufacturing and engineering

Testing during the production
Marketing, publicity, advertising and
media
Increase in productivity

Product
Performance

Continual improvements
Increase in profit

Campaigns, media, engaging stakeholders

and NGOs
Reduction in number of accidents, deaths,
injuries and risk
The reduction percentage of accident death
rates over time
Reduction in the total cost of road accidents

In order to become successful in road safety as well as in business, each part of the
society or organisation must work properly together towards the same goal with so
much understanding that each person and each activity affects the overall outcome. The
reduction in road accidents is a sign of efficiency and development throughout the
society.
Most of the quality economic models give greater weighting points to the end business
results and human dimension (leadership, customer satisfaction and employee
satisfaction). This approach can inspire RSDI methods to judge the weightings of
indicators (discussed in the next chapter).

69

Improved total quality in operations and results has led to sustainable improvements in
profits and customer satisfaction. Similarly in road safety science, the improvement in
operational quality (good practices) and quality results (accident death rates) will lead to
sustainable progress in road safety.
The following points summarise the main philosophy behind RSDI:
1- No single measure explains all national success in road safety. Everyone in the
country is responsible for the overall outcome. Everyone is part of the problem,
but everyone is also part of the solution.
2- The RSDI attempts to identify key practices that affect the overall national
outcome and to link these practices to each other.
3- Any successful international benchmarking in road safety has to link practices to
the end results (accident death rates).
4- RSDI focuses on results of human-system-product.
5- Human performance is considered to be one of the main pillars of RSDI.
6- RSDI is a tool for continuous improvements in national road safety.

4.7

The process of RSDI building

There are four-stages in the process of RSDI building. They are carried out according to
the research method mentioned in Section 1.3 (see Table 1.1):

1. Selecting the right quality indicators and dimensions (contributes to Theory

Building).
2. Using the right methods to combine the chosen indicators together with
weightings into RSDI (contributes to Model Building).
3. Applying RSDI for the right sample of countries that have similar patterns of
motorisation and transport characteristics (Empirical Studies).
4. Implementing and reviewing RSDI results (recommendations) theoretically,
empirically (Model Assessment). The results can also be assessed from each
country by involving the right stakeholders in the process of development.
Stage one was explained in previous chapters, however further discussion is made in the
next section. The following chapter describes the methodological approaches for
aggregating the normalised indicators with weightings into RSDI. Empirical studies
about the usefulness of having RSDI and how it can be used in practice are discussed in
Chapter 6 and Chapter 7. Stage four is assessed in Chapter 8, further discussion
according to the RSDI results and its recommendations are left to every country/region
to make/take their review, responsibility and necessary actions.

70

4.8

Selecting the right indicators to be added into RSDI

The selection process of indicators is the first and most important stage of RSDI
building since it will determine the significance of the obtained results at a later stage of
study. Much attention and time should be paid in testing and developing good indicators
because this is the key step to any successful RSDI in the future.
Selecting key macro-performance indicators for a multidimensional index such RSDI is
a complex issue, especially in which ones to choose and why. This selection depends on
what is available from international sources and what can be useful for meaningful
comparisons. The selected performance indicators should make a valuable contribution
to the overall development in road safety in a country, whether the indicator can show
any performance progress or not. Furthermore, RSDI will measure what is achieved in a
country not what a country is doing regarding road safety.
In the previous chapter, some kind of handbook/master-list of performance indicators
and dimensions was set up that will possibly be useful in assessing overall country
performances in road safety (see Table 3.2). This long list is made to ensure the longterm vision of RSDI. Additionally, in the same chapter, a detailed information was
compiled on the correlation between each indicator and safety with support from the
literature either in how they illustrate road safety level and development in a country
(direct measures) or how significantly they affect death rates (indirect measures).
As seen from the Table 3.2, the list of performance indicators is relatively long and it is
not possible to gather all into the RSDI index, not only because they are many and it
will take too much time to interpret and analyse, but also because some of the indicators
require further development, which means availability internationally, quality and
definitions. For instance, one widely acceptable indicator is deaths per vehicle
kilometre, but unfortunately the number of vehicle kilometres is not often available in
most countries, and it varies considerably in HMCs. Hence this indicator will be
excluded from RSDI at this stage of the analysis. Furthermore, the indicator percentage
of motorways per road network is a good indicator since motorways have the lowest
risk of injury accidents compared to other types of roads because of the separation
between vehicle movements according to their speed. This indicator has a low quality
and reliability of data as there is a difference in definition of road standards across the
countries. So this indicator again will be hard to apply to a large number of countries,
especially in LMCs. Severity index is another factor of interest in international
comparisons. It is measured as the percentage of deaths per total number of casualties
(death, serious and slightly injured). The severity index decreases with better medical

71

facilities and rescue services in the country. Again, this indicator cannot be included in
RSDI because the injury figures are still under-estimated significantly in many LMCs.
Simple indicators have been used and developed by a number of international
institutions and databases such International Road Federation (IRF), International Road
Traffic and Accidents Database (IRTAD), World Bank, etc. But they are, unfortunately,
not detailed enough. Besides, there are many important indicators and data missing in
these international databases such as percentage of seat belt use, percentage of helmet
use (cyclists and/or motorcyclists), average age of the vehicle fleet in the country,
national expenditure in road safety measures, percentage of motorways, number of
alcohol breath tests, number of speed checks, etc. This data is not available for a large
number of countries including some HMCs and there should be a call for data collection
and further development for these indicators.
An earlier study (Al-Haji, 2005) has shown that RSDI correlates well with the Human
Development Index (HDI). This was explained in terms of the relationship between
road accidents and each of the HDI components, these are the level of income of the
country, health level and education. The increase in income leads to safer vehicles and
more investment in road infrastructure and safety countermeasures. The death rates are
correlated with the level of medical facilities available in the country. Education can
influence the ability of road users to understand the rules of the road and road signs. In
addition, HDI is widely accepted and valid worldwide. Hence it would be possible to
include this dimension in RSDI (as a socio-economic development index) for both
LMCs and HMCs.
The selection of indicators shown in Table 3.2 is the first and most important source but
not the final one. The indicators will continue to be developed and refined before RSDI
will be fully conceptualised. The selection and development of indicators should
correspond to the four-stage model of RSDI mentioned earlier in Section 4.7.
The size of the safety problem differs from country to country and from region to
region. The countries face different challenges and priorities in road safety, thus their
key performance indicators can be different as well. In the short-term, RSDI will
include indicators with high availability and acceptable quality, while the indicators that
require further development and collection of data will be used in long-term RSDI
development. This long-term goal should be done with a comprehensive list of key
standard performance indicators that can be accomplished in three possible ways:
1. By asking each country to set up some kind of performance information centre at
national level regarding RSDI key indicators. It is known that collecting annual

72

detailed data (exposure and risk) in a country has to be conducted in a regular and
systematic manner, which is quite expensive and it needs knowledgeable and trained
personnel as well. The performance information centre of RSDI will collect data and
supplementary information for each indicator and will develop methods to estimate
the necessary indicators. This centre has to work closely with a National Road
Safety Council (NRSC) in each country as well as with national traffic accident
statistic databases. Most countries have their national database for accident data
collection and analysis such as the Swedish Traffic Accident Data Acquisition
(STRADA), National Traffic Accident Statistics Database in United Kingdom
(STATS19), and the Traffic Accident Registration (FARS) in the United States.
Additionally, several LMCs have used a computerised system called the
Microcomputer Accident Analysis Package (MAAP), which was developed by the
UK Transport Research Laboratory (TRL).
2. By asking regional/local consultants to assist RSDI in finding and/or estimating the
required data and information from each country. It might be useful if a leading
international institution or agency is involved in the development of RSDI because
this will increase the transparency, accuracy and validity of data. There are many
international traffic accident databases which have been developed by a number of
regional institutions. These can support RSDI in one way or another. Such
international institutions are: EU, WB, WHO, ADB, UN, GRSP, etc.
3. By developing different research studies and projects in which they can complement
each other and contribute separately to the overall long-term goal of RSDI, each
research project will address one aspect of RSDI and ultimately contribute to RSDI
vision. This will allow RSDI to use the projects results together and enable a
complete interpretation of road safety situation in a country. To accomplish this, a
first standard and relationship must be defined and identified between the selected
projects, where each project will develop one sub-index ranges from 0 (worst) to
100 (best) to each dimension (see Figure 4.3). Fortunately, in many regions, there is
a number of interesting benchmarking projects available (e.g. European projects),
which would make significant support to the RSDI profile. For instance the EuroRap5 project focuses on safer roads; Euro-NCAP6 project focuses on safer vehicles;
the SARTRE7 project focuses on safer people between EU countries; and the
Escape8 Project focuses on traffic enforcement in Europe. A further interesting
5 Euro-RAP project: European Road Assessment Programme, which assesses the safety performance of
similar roads between European countries.
6 Euro-NCAP project: European New car Star Assessment Programme, which assesses the safety
performance of vehicles European countries.
7 SARTRE project: Behaviour of European car drivers in relation to various topics of traffic safety such
as drinking and driving, speeding and seat belt wearing.
8 Escape project: It assesses the potential of enforcement tools and measures to improve safety on
European roads.

73

project from Australia (Cameron & Diamantopoulou, 2000) can support RSDI in
terms of its enforcement performance index, which has developed a combination of
outcomes to measure an enforcement index for the State of Victoria in Australia.

Traffic Risk

Safer roads
index

Safer vehicles
index

Personal
Risk

Safer people
index

Enforcement
Index

Organisational
Index

HDI

Figure 4.3: Sharing the long-term vision of RSDI

(The size and proportion of each dimension is given as an example only)

By reaching such long-term goals, RSDI will be able to produce an annual international
report as well as regional and individual country reports.

4.9

Identifying a short-term key list of indicators for road safety performance

There is no definitive list of performance indicators in road safety that are

internationally accepted and suitable for all countries. RSDI must be tailored to most
countries worldwide according to the availability and reliability of data. Therefore, the
first step is to prioritise the indicators and then to define a shortlist of key performance
indicators that would be applied uniformly for most countries. In doing this, it is
suggested two RSDI indices be developed; one index is relevant for most countries in
LMCs and one index for most countries in HMCs. RSDI will clearly include additional
set of core dimensions and indicators for HMCs as they have a more developed data
collection procedure than is the case in the LMCs. There is a number of composite
indicators that have been applied to countries according to their development. An
example is the Human Poverty Index (HPI), which is produced by the United Nations,
measured two indices HPI-1 and HPI-2 for developing and developed countries
respectively.

74

At this stage of the study and based on the previous discussion, seven indicators in six
dimensions for LMCs and sixteen indicators in eight dimensions for HMCs are
suggested (see Table 4.2). These indicators will be tested in the two empirical studies in
following chapters. They do not fully address the development, but still they give a
broader picture of road safety rather than just focusing on individual aspects. It might be
better if a selection and test of a few key indicators were made from the beginning and
then gradually adding/modifying more and more over time towards the key standard
list.
The first chosen indicator is the level of traffic risk in a country, which deals with the
death rates. The death rate per vehicle is currently used, while death rate per vehicle-km
is not used due to lack of data. Unfortunately, motor vehicle kilometre data is often
estimated and not also available in many HMCs and for every year (see IRTAD, 2006).
The second indicator is the personal risk, which is defined as deaths per total
population.
The third indicator is the percentage change in the number of deaths over a period of
time.
The dimension of road user behaviour is considered to be the heart of RSDI. It is based
on the percentage of seat belt use in front seats for LMCs and in rear seats for HMCs.
Using this indicator will enable RSDI to distinguish between HMCs whose
achievements in seat belt use in front seats are very similar. Additional indicators used
in this dimension are the percentage of two wheelers (motorcyclists and cyclists)
helmet use, the percentage of drivers above BAC (Blood Alcohol Concentration) and
the percentage of all drivers exceeding the speed limits.
The vehicle safety dimension is based on the percentage of the vehicles (not
motorcycles) in the total vehicle fleet. Additional indicators have been included into
RSDI for HMCs including the average age of the vehicle fleet where new cars tend to
have more safety and protection features, such as air bags, anti-brake system (ABS), etc.
The safety level of roads for HMCs is based on the percentage of the motorways per
total road network in country in addition to the national expenditure in road
maintenance as a percentage of the total investment, while RSDI for LMCs uses the
percentage of paved roads.
In the socioeconomic index, the HDI has been included into RSDI for both LMCs and
HMCs. HDI describes the level of health, education and income in a country.

75

The enforcement dimension measures traffic enforcement effectiveness levels in the

country for HMCs. Unfortunately; most LMCs have too little data regarding this
dimension. Three indicators are included in the enforcement dimension, which are the
proportion of drivers who had been checked for alcohol, proportion of drivers who had
been checked for seat belt, and the proportion of drivers who had been checked for
speed.
The last proposed dimension of RSDI is the road safety organisational index that
measures the level of cooperation between the key bodies responsible for road safety
actions in the country, how much funding is spent on road safety measures, how far the
national action plan in a country is from the national/region target (i.e. EU target of
halving road deaths by 2010), number of national campaigns, etc. This dimension is not
addressed in this study due to a serious shortage of objective indicators and data for
both HMCs and LMCs.
In table 4.2, the maximum and minimum values for each of the selected indicators based
on a wide-ranging sample of countries have been identified (estimated in few cases).
The sample was taken from different countries in different years with different levels of
development and motorisation. It can be seen from the table that huge differences exist
between the countries in all indicators (distance between maximum and minimum
values). In addition, the range of values is different from one indicator to another, which
means that the change percentage (increase or decrease) between indicators is different.
For example an increase 10% in the motorway network of a country is probably more
significant than the same increase of paved roads. In similar discussions, the percentage
change of a value at a low level might be different from the same change at a high level
of the same indicator. For example, a decrease of 10% in traffic risk at a low level as
many developed countries seek to achieve this goal (e.g. from 1.2 to 1.08) is perhaps
more significant than a same decrease at a high level (e.g. from 300 to 270).
The indicators contribute differently to the RSDI. The low values of traffic risk and
personal risk show a good safety level in a country, while in contrast, the safety issue
increases continuously with the increase of the values of other indicators.

76

Organisational
performance

Traffic police and

enforcement

Socio-economic
level

Safer roads

Safer vehicles

Road user
behaviour

Percentage of crash helmet use

Percentage of crash helmet use

10
10
Not Rated
Not Rated

Proportion of drivers who had been checked for seat belt

The funds are spent on road safety measures (% of GNP)
Number of national road safety campaigns (2000-2003)

10

0.281

Proportion of drivers who had been checked for speed

Proportion of drivers who had been checked for alcohol

Human Development Index (HDI)

Percentage of paved roads

Human Development Index (HDI)

Investment in road maintenance as % of total

% of motorway length per total road network

Percentage of paved roads

Average age of the vehicle fleet (number of years)

10

% Drivers that never drive faster than speed limits

Percentage of vehicles not motorcycles

10

% Drivers that never drink before driving

Percentage of vehicles not motorcycles

0.0
0.0

-25
Percentage of seat belt use (front seat)

1.2

Minimum

Percentage of seat belt use (rear seat)

Changing trend

Deaths 30 days per 100,000 population

Deaths 30 days per 10,000 vehicles

Indicators for LMCs

Percentage change of deaths (2000-2003)

Deaths 30 days per 100,000 population

Deaths 30 days per 10,000 vehicles

Traffic Risk

Personal Risk

Indicators for HMCs

Dimension

Sources: IRF, IRTAD, World Bank WDI, United Nations, Globesafe database.
Maximum and minimum values have been estimated from sample of several countries with different levels of development
Not rated means that there is insufficient data available in many countries and it is uncertain what the range of results could be
The choice of targets is estimated on the idea that there is potential progress ahead for all countries in road safety development.

Safer
System

Safer
People

Safer
Product

Pillar

Not Rated

Not Rated

40

40

40

0.963

100

40

95

30

80

80

78

78

+25

70.0

600

Maximum

Not Rated

Not Rated

45

45

45

0.98

100

65

97

90

90

80

80

-25

Target

Table 4.2: The suggested list of key performance indicators and dimensions, which can be included into RSDI for LMCs and HMCs

77

At the same time, it might be better if a target value was set up for the minimum and
maximum values that are identified, according to a special performance scale for each
indicator. This can be estimated on the idea that there is potential progress ahead for all
countries in road safety development. This target is not simply a static end-goal but it
has to be adjusted and be exposed as a challengeable target over time as much as
possible. This target depends on the characteristics of the studied countries or regions
and how much progress in road safety that has been made. For instance, RSDI can
assess the development in road safety in EU countries to reach its ambitious target for
reducing the number of road traffic deaths by 50% by 2010. In this case, the sub-targets
can be set up for different type of roads (urban, rural, motorways), for individual groups
of road users (i.e. 50% reduction of the number of pedestrian deaths by 2010) and for
different contributory factor (i.e. 50% reduction of the alcohol-related traffic deaths).
Each sub target should contribute to the overall national target.

4.10 Conclusion
This chapter has described the conceptual framework of the RSDI and how it can be
developed and applied. Also, it has shown the possible benefits and challenges in using
this composite index. The RSDI prioritises the key national outcomes in road safety and
aggregates them into a summarised index. Furthermore, this chapter argues that death
rates focus only on the superficial level (product) and ignore the overall system (internal
level) and human assessments (core level). Thus any successful international
benchmarking in road safety has to link practices to the end results (death rates).
It has also been emphasised that the number and type of indicators that will be used in
RSDI depend on the availability and quality of data in the country. A standard list of
macro-performance indicators and dimensions that should be widely useful, available
and ready for future data needs, must be set up to build an index that can be useful, valid
as a widely accepted framework.

78

Chapter 5: Methodological Approaches

Chapter 5

Methodological Approaches
Four main approaches (objective and subjective) for building RSDI are
proposed and described. The strengths and weaknesses of each approach
are summarised by the end of the chapter. The next step will put RSDI into
practice.

5.1

Introduction

The main question of this chapter is how to combine the performance indicators of road
safety that have been identified in previous chapters into a composite measure.
To answer this question, let us first simplify things by looking at the road safety
performance from another perspective. Imagine that RSDI is a university degree
diploma, where countries represent the students, the world equals the university and
each dimension of road safety is an educational course, for example:
- Safer people (road user behaviour) is the first course - represented by Physics.
- A safer road is the second - represented by Mathematics.
- Safer vehicle is the third- represented by Computer science.
Generally in education, a single course for a student is evaluated by the sum of credit
points for the single tasks (written exams) within the course, while the overall degree
diploma is evaluated by the sum/average of points from all courses that student has
taken. If students (countries) really perform well in several or all courses (road safety
dimensions) that will put them up the class (world) as they have the most points.
Students can perform better in one course than another and thus they can acquire a
certificate for a single course, while the degree diploma (RSDI) needs good

79

performance in all courses9. When students (countries) perform lower than the average
level of the class (average level of RSDI), they are often required to submit
complementary essays or assignments (good practices in terms of road safety). Students
(countries) who have obtained a certain amount of points by completing the required
courses will be passed or admitted to the second level of challenge.
There are several point grading systems in use in different countries and it differs from
country to another (e.g. GPA10 Grade Point Average, Swedish system11, common
system12). Additionally, in general, there are often two common ways of combining the
course credits into a final record ranking of a student:
The first method is the unweighted grade point method where all courses are given
equal importance, which makes the calculation of the final record simpler and easier
when combining the course credits together.
The second method gives different weightings to the courses on the basis of the
workload/importance of each course (essays, assignments, research, etc.). The
advantage of this method is to encourage students to continue studying more
challenging, high-level courses.
In making weighted grade point system, countries (students) should be awarded
different points for each activity received to reflect the fact, for instance, that high
seatbelt use in front and rear seats (Advanced Mathematics) is better weighting than
high seatbelt use in front seats only (Introduction to Mathematics). Again this
depends on the target of the evaluation and type of countries (students). Thus, we cannot
ask students (e.g. developing countries) to pass advanced mathematics if they have
failed in the first level of mathematics.
Subsequently, another question arises, which is how to combine these courses to a form
of a standard scale for all students in all countries. An interesting approach to this
problem is translating or using a 100-point scale for all courses. Rather than summing
up the total points from all courses (with weights), students scores should be by the end
9 The RSDI system is taken under credit system, where each country receives a credit out of 100 points.
The Pass/Fail point scale is not used by RSDI as it is not appropriate to blame countries with Pass/Fail
system.
10 A GPA is a number calculated using two factors: credits earned and grade points.
11 In Sweden: the grades used in high schools are: MVG - Mycket vl godknd (Passed with special
distinction >90%), VG - Vl godknd (Passed with distinction >75%), G - Godknd (Passed >50%), IG Icke godknd (Fail<50%); while in universities, with some exceptions, the grades used: VG - Vl
godknd (Passed with distinction), G - Godknd (Passed), U - Underknd (Fail). Source:
http://www.wikipedia.org/ (accessed 2006-03-31).
12 A Excellent, B Above Average, C Average, D Below Average, F Failure

80

expressed as a percentage of the total obtained points. For example, a mathematics

course would get 20/100 of the total weighting of the degree.
Finally, the third question arises, which is important: Are all students (countries) taking
part (or interested) in the same course or degree? Or are they ready to acquire the same
tasks and levels or not?
The performance of students (countries) may vary from place to place and from person
to person, therefore the courses (road safety dimensions) are not really comparable
everywhere. Thus there are two options, either to lower the level of the courses (quality)
so that would certainly involve more students (countries) which seems good as more
students become involved, or to divide students into different groups according to their
levels and competence (LMCs and HMCs in terms of road safety), which would make it
more sensible.
As one can see from the above example there are different approaches and stages one
has to take into account before examining and aggregating the performance into one
value. In fact there is no internationally accepted standard method yet that is available to
weight indicators before combining them together for any type of composite index (see
for example Nardo et al., 2005).

5.2

Construction of RSDI

The major steps used in the process of constructing RSDI are the following and perhaps
in some cases, the steps are interrelated:
1.
2.
3.
4.

Finding the key indicators and dimension (discussed in previous chapter).

Normalising (standardising) the indicators.
Weighting the indicators.
Combining the chosen indicators into (RSDI) by using different approaches.

Most composite indices follow the linear additive regression form. This means that we
can measure each indicator (dimension) individually and then aggregate all normalised
dimensions into one index. In other words, in order to aggregate apples, oranges and
pears into one index (RSDI), one basket (dimension) for each type of fruits has to
match, then we normalise each basket alone with the same scale, then we can measure
each basket individually and finally aggregate them into one index.
The general form of the overall measure of road safety is simply:

81

RSDI = f (H, S, P)
where

(5.1)

H is the human performance (basket of apples).

S is road safety system (basket of oranges).
P is product quality (basket of pears).
Each of these baskets is the sum of sub-dimensions and indicators.

The general and simplest composite index of RSDI will take the form:
n

RSDI = wi Di

(5.2)

i =1

where:

Di are normalised dimensions;

wi are the weights for Di

The weights of each dimension in RSDI range between 0 and 1; the sum of weights is
one. In all approaches for calculating RSDI, we classify the countries into three groups
by achievements in RSDI ranged from 0 till 100 with high road safety development
(RSDI of 70 or above), medium (40 till 70) and low (less than 40). The higher values
indicate a higher level of safety in the country. The lower values indicate the worst
performance in country in terms of road safety level and vice versa.
One should remember that any change in the process of constructing RSDI (e.g. change
in number of indicators, weights, methodologies, etc.) would change the final results of
the countrys ranking as well.

5.3

Normalising the indicators

Before integrating the selected indicators together into RSDI, we have to normalise the
values of indicators. The indicators have different scales (i.e. tens to millions) and units
(i.e. percentages to rates). The idea of normalisation makes all indicators acquire the
same magnitude before mixing them together (like comparing oranges with apples). We
need to ensure that no value will dominate others in the final RSDI scores.
There are many techniques used to normalise the indicators and each technique has its
own advantages and disadvantages. These techniques depend on data availability,
sample size of countries, type of aggregation methods and the expected influence of
variables in the final index (e.g. more interest in new data). The most common methods
used in normalising data of international indices are the general linear transformations
techniques that measure the distance between the actual value and the best and worst
values of data. Other methods for normalising data include distance to the mean,
distance to the median, distance to the standard deviation, distance from year to

82

previous year, distance to target, etc. The normalisation methods used in this study will
be discussed further under aggregation approaches in Section 5.5.

5.4

Weighting the variables

The choice of weights for multidimensional indices is crucial. If any change in weight
values has occurred, this may bring different results and change the countrys ranking.
For instance, if Personal Risk gets a higher weight, this may benefit the position of
countries with higher population density. Again if Alcohol and driving gets a higher
weight, this may benefit the position of countries with low or zero alcohol consumption.
In addition any change of approach (methodology, system) in calculating the index will
change the weights and the final results of the index.
The weights can be measured or estimated empirically from data collected subjectively
from experts opinions or based on statistical and theoretical models. A number of
studies (for example Nardo et al., 2005; Andrew, 2004; Saisana and Tarantola, 2002)
have explained, assessed and evaluated the advantages and disadvantages of different
weightings methods used in several research studies. Yet, there is no single and standard
methodology used for weighting the variables. As a result, several conceptual
frameworks are now available with different calculation techniques, weightings and
treatment of indicators. Frederik (2002) has commented on that the weighing remarks
are always questionable: No weighting system is above criticism.
The weights of indicators have to be given first to the sub-index (dimension) and then
weighting all dimensions together. The weights are valued on their importance and
impact on the whole situation (e.g. road safety). This could be shown in the correlation
between the indicator and road safety improvement.
Missing of data is another issue that should be taken into account. It leads to less
reliability on indicators and creating less reliability on the derived index. The weighting
according to missing data is not simple, maybe we can give less-weight to the indicators
with more missing data, but here we will approach the question of data collection and
this can skew the good scores of RSDI towards countries with more availability of data.
The problem of missing data differs from one country to country and it seems to have
more severe affects in developing countries. In fact, there are some possibilities to solve
or reduce this problem, for instance by estimating the values from nearest data/year with
similar characteristics; by taking an average value (before and after if available); by
using time series regression models to predict the missing value, or by using another
year or mixing of years in case of cross-sectional data. In some countries where there is
lack of data for one or more indicators, the RSDI will be hard to calculate or estimate.

83

In this case, it would be better if we ignore the value of the country for this particular
year (Not Available).

5.5

Combining the chosen indicators (together with weights) into RSDI by using
different approaches

Four methods for determining the weights are proposed for RSDI building through two
objective methods that are: simple equal average and principal components analysis
(PCA); and two other subjective methods that are assessment technique from experts
opinions and assessment technique from literature and theory review. The PCA
method is a statistically based option. The simple equal average method uses equal
weights for all indicators and countries, whereas the other methods give different
weights dependent on the importance of indicators.
The methods chosen are widely used and accepted in constructing different international
composite indices. However, each of these methods has its advantages and
disadvantages. The idea of using several methods as in other indices is to experiment
with different approaches and weights and to compare the obtained results afterwards,
before deciding on one or more techniques to be used in future. The following sections
describe each of the approaches in more detail.

5.5.1

Approach 1: Using Simple Average

In many international composite indices (i.e. HDI), all the variables included are given
the same weight. This means that all indicators in the composite index have the same
importance. In the calculation of RSDI by using "simple average method, each pillar of
RSDI is given equal weight. At the same time, all the sub-dimensions and indicators
included in each pillar are given equal weights as well. This means that all indicators in
the RSDI have same importance. This makes the determination of weights much
simpler and less subjective. It is also good for the sake of transparency.
Before combining the selected indicators together with weights into RSDI, the values of
indicators have to be normalised. The idea of normalisation is to ensure that all
indicators and dimensions are given the same scale and unit before mixing them
together into RSDI so that no value should dominate others in the final RSDI scores.
The normalisation method distance to a target seems the most suitable method for
RSDI construction. This method takes into consideration the development of indicators
over time. Besides, it gives the possibility to apply RSDI to individual samples of
countries separately and then to combine the final RSDI scores altogether (e.g. in one
table). The choice of targets is estimated on the idea that there is potential progress
ahead for all countries in road safety development. However, it should be noted that any

84

change in the target value will lead to change in the performance value of a country but
will not affect the final ranking of countries.
Each indicator/dimension will come up with a value of between 0-100 as a percentage.
In doing this normalisation, the distance to target method is used as:
Normalisation =

(Target value)
*100
(Actual value)

(5.3)

Normalisation =

(Actual value)
* 100
(Targetl value)

(5.4)

The selection of either of the formulas (5.3) and (5.4) depends on the safety level if it
has an increasing (much is better) or decreasing rate (less is better) in change with
respect to the value of indicator. For instance, the low values of traffic risk, personal
risk and average age of the vehicle fleet show a good safety level in a country, while
in contrast the safety issue increases continuously with an increase in the values of other
indicators such as percentage of motorways and percentage of seat belt use.
As the indicator percentage change of road accident deaths has positive and negative
values, it is suitable to normalise it by the (minimum, maximum) normalisation formula
as follows:
Normalisation =

(Maximum value - Actual value)

* 100 (5.5)
(Maximum value - Minimum value)

Furthermore, adding all indicators (with equal weights) together in RSDI will give more
score to the pillars that have included many indicators (e.g. the system pillar). Hence, it
is necessary to consider the three overall pillars at the final stage of aggregation into
RSDI. Each pillar (Pj) is calculated as the simple average of the indicators included. The
standard formula will be:
n

Pj = wi I i

(5.6)

i =1

where:

Ii are the normalised indicators;

j = 1,2, 3;
wi are the weights for indicator I (the sum of all weights should be 1).

The RSDI is measured as the simple average of the three pillars where each pillar is
given equal weight (one-third):
1 3
RSDI = Pi
(5.7)
3 i =1

85

However, one has to note that there is a wide-range of values between countries in all
indicators. For instance, the estimated range of values for traffic risk (killed per vehicle)
is 1.2 to 600 (see Table 4.2).
According to Julia (2003), the main attractions to this approach (simple average) among
many composite indices (i.e. HDI) are its simplicity, its transparency and it is less
subjective than other methods. On the other hand, many studies (i.e. Michael, 2003)
have criticised this method for making all combined indicators equally important, which
is not true in real-world applications.
The following approaches give different weights to the indicators, so each dimension
will receive an amount of weights depending on the importance of the indicators in each
dimension.
5.5.2

Approach 2: Expert Judgements

The weightings in this approach are determined according to experts judgment. The
principle of this approach is to submit checklists to a group of experts in road safety.
They will be asked to mark the weights for different indicators and to assess the
necessity of each indicator subjectively. The responses will be transformed to the rating
system of RSDI, the larger weight is the more important and the total of weights should
equal one.
This method has been proven to be a useful tool in some studies, for instance in the
Growth Competitiveness Index (GCI) developed by Harvard University (John et al.
2001) where this index combines objective and subjective sub-indices. The subjective
variables in GCI are derived from a survey of business managers from around the globe.
There are many important areas of road safety that are hard to measure in quantitative
forms. It would be possible for RSDI to include subjective indicators in the long run.
Both dimensions of organisation and enforcement can be supplied by subjective
assessments, for example the assessment of the technical means that are available to the
traffic police in the country and the assessment of national road safety programme.
There are different techniques that can be applied in assessing the survey results and
opinions of experts such as the Multi-Criteria Analysis (Mendoza et al. 1999) and
Delphi Technique (Linstone and Turoff, 2002). For now, and because of the absence of
our survey data among experts, these methods are not discussed further here.

86

One example of how to convert the subjective results into the numerical scale of RSDI
is shown in Table 5.1. A question that can be put to experts is: "would you mark the
overall weight of each indicator as excellent, good, average, poor or inadequate?" The
subjective assessment will be graded as the RSDI scoring system from 0 to 100. The
final sum of weights will be one.
Table 5.1: The subjective weighting results in a numerical scale of RSDI
Inadequate

Poor

Average

Good

Excellent

RSDI range

Less than 20%

20-40%

40-60%

60-80%

More than 80%

RSDI Rating

10%

30%

50%

70%

90%

However, this approach depends on the expert panel and it requires a criterion assuming
that the expert panels are recognised as fully qualified experts in the area of road safety
to judge on the issue we are dealing with. Besides, it should be clear that it would be
difficult to obtain the opinions of experts regularly and frequently.
5.5.3

Approach 3: Subjective weights based on previous experience

In the calculation of RSDI by using this method, the chosen indicators will have
different importance (weights) in the RSDI based on their impact on the whole situation
of road safety. The sum of weights should equal one (or 100 points). We use the same
normalisation formulas to the target (5.3 and 5.4).
These weights will not be the same in all countries and regions due to the differences in
their priorities in terms, for instance, standards of roads, vehicles types, road user
behaviour and traffic management. For instance, the Northern European countries see
drink-driving as the one main problem, therefore a low performance can be noticed in
this area (i.e. in terms of alcohol-related traffic deaths). At the same time, any national
actions by Northern Europeans to overcome the drink-driving problem and reduce the
end results in terms of number of deaths will also result a better performance in the
process and practices such as enforcement indicators, national campaigns, etc. On the
other hand, many Southern European countries are attempting to reduce speeding
problem and increase safety-belt use. The usage of seatbelts tends to be lower in South
Europe compared to Northern EU. Thus, the Southern European countries have focused
their actions that relate to their priorities (in this example: Speeding). Moreover,
countries where type of road users is a priority, such as motorcyclists in the Southeast
Asia, could focus their efforts on reducing the risk to motorcyclists and improving their
safety by increasing the use of crash helmets. Thus, in respect to RSDI, the selection
weights and performance indicators may take into account the national/regional

87

priorities. Every region can select the appropriate RSDI indicators and their weights for
making special regional benchmarking, which will relate to its priorities. For example,
the safety of motorcyclists could become an important indicator of RSDI in countries
(regions) where it is considered a priority problem. However, one has to note that the
international RSDI should ultimately include/summarise the indicators that reflect the
overall priorities and problems of most countries and regions in the world.
In this respect, this weighting method takes into account the earlier case studies,
accident investigations, statistical analysis and meta-analysis studies from the chosen
countries/region that can indicate the weight (importance) of each indicator in relation
to accidents. Such studies have been developed by international and regional
institutions, for instance: EU, WB, WHO, ADB, UN, Sida and GRSP.
In addition, we have to judge several alternatives of weightings based on earlier
discussion (in Chapter 4) where RSDI gives greater weighting points to the end results
and human behaviour. For instance, one of the major contributory factors in road
accidents is human error resulting from the violation of traffic laws. Rumar (1999) has
shown that the human factor is the major contributory factor in 94/95% of accidents
based on two in-depth studies carried out in the UK and USA. However, this result will
not be as the same in LMCs due to the differences in the standards of roads, vehicles
and traffic management. Thus, we have to select from among several alternatives of
weights, making a choice and evaluating the results afterwards. This alternative will not
be the one finally chosen, we should review the weighting and the obtained results and
if we disagree with this choice, we will make the necessary changes till we reach the
optimal choice of weights for the selected countries or globally.
Based on this discussion, an example of the alternative choices of weights is shown in
Table 5.2. Both product (direct measure) and behaviour index (indirect measure) receive
larger weights than the system performance.
Table 5.2: Alternative choices of weighting the pillars of RSDI
RSDI Pillars
Aggregate Product performance
Aggregate Human performance
Aggregate System performance
Total aggregate

Alternative 1

Alternative 2

Alternative 3

Percentage score
40
40
20

Percentage score
35
35
30

Percentage score
30
50
20

100

100

100

For instance by using Alternative 1, we give (40%) weight for each of the product and
human behaviour and (20%) to the system index (Alternative 1), the RSDI is calculated
as follows:

88

RSDI= 0,4PI+ 0,4HI+ 0,2SI

where

(5.8)

PI is the Product index,

HI is Human behaviour index, and
SI is the System index.

The following figure illustrates RSDI model following these assumptions of weights.
INPUTS & PROCESS

PRODUCT
RESULTS
INPUTS

PROCESS

PEOPLE
RESULTS
SYSTEM
RESULTS

LOCAL

SUCCESS

OUTCOMES OF PERFORMANCE
(i.e.40%)

(i.e.40%)

KEY
MACROPERFORMANCE
INDICATORS

RSDI
(100%)

(i.e.20%)

NATIONAL

(100%)

INTERNATIONAL

Figure 5.1: An illustration of the combined aggregated results

with weights into RSDI Model (Based on the original RSDI model in Figure 4.2)

However, there is no dominant choice of weights and it is necessary that further indepth studies be carried out to investigate the accidents data, causes and priorities in the
selected countries (or internationally if data from most countries is available) before
making any better assessment of weights.
5.5.4

Approach 4: Principal Components Analysis

Principal Component Analysis (PCA) is a multivariate technique used to find a

combination of indicators. It describes the variation of the original data by means of a
smaller set of dimensions (termed principal components). PCA is a very promising
technique that is widely used in many fields of science and has also been used in the
construction of some of the composite indices for instance the Internal Market Index for
EU countries (Tarantoal, 2002). The advantage of this technique is that the weighting of
various indicators is based on statistical methods rather than subjective judgements. It
lets simply the data itself to decide on the weighting issue, which is good from
transparency point of view. The PCA technique requires all the indicators to be analysed
together not only the dimensions as the previous methods did. The data of indicators is
first normalised to zero mean and unit variance by using the formula:
Normalisation =

(Actual value - Mean value)

(Standard deviation)

(5.9)

89

The adjustment data is transformed into a new set of uncorrelated principal components.
Each component contributes to a proportion of the variance. The sum of variance in the
PCs is equal to the sum of the variance in the original values. Each Principal
Component is a linear combination of the original indicators where the first component
PC1 explains the maximum amount of variance, while the next principal component
PC2 explains the next largest proportion of variance and so on. Together, all the
principal components account for 100% of the variation. The general form for the
created principal components PCj in the linear combination of the indicators Xi is:
p

PC j = ai X i

(5.10)

i =1

where:

ai is the regression coefficients:

p number of indicators
p

a
i =1

2
i

=1

(5.11)

In the calculation of RSDI we compute all data and we consider the number of principal
components that could explain most of the variation. It is usually between 70% and
90%. This technique allows us to pick out a minimal number of PCs that summarise the
original data.
It must be stressed that PCA may produce more than one index with different scales.
Thus there is a need to convert the PCA indices into one single value (RSDI) with scale
range from 0 to 100. In doing this, we first normalise the values of the components
(indices) obtained by using the simple average method and then we attach their weights
together in the additive regression model.

5.6

Possible applications and illustration of RSDI

In order to put RSDI into practice and to check its applicability, we have to measure and
test RSDI along with the weighted values of indicators/pillars for a sample of countries
and from different parts of the world. Such an empirical study is also needed to more
fully assess the usefulness of the performance indicators included in RSDI.
When measuring RSDI, one has to remember that many indicators are composed
together into a single number and a lot of work went into producing the results (e.g. data
collection, selection, weightings). Moreover, an understanding of this RSDI score/rank
requires a closer look at the data behind each indicator. An example of how to illustrate
RSDI results is shown in Figure 5.2. It shows the difference between the dimensions

90

and countries at a three-level of performance of RSDI. This will be useful for

identifying the problems in the country where the necessary actions should be taken in
the future.
Traffic risk
100
80

Organisational index

Personal risk

60
40
20

Enforcement index

Safer vehicles

Socioeconomic level

Safer roads

Safer people (behaviour)

Countries with high RSDI (70 or above)

Countries with medium RSDI (40 to 70)
Countries with low RSDI (less than 40)

Figure. 5.2: An example of illustration of the three RSDI levels and dimensions

The following figure shows another example of how to illustrate RSDI results. This
illustration gives quick and clear assessment of RSDI results for a large number of
countries with relatively high number of dimensions.
RSDI
System
Human
Product
Countries/Regions

Speed

Alcohol

Seatbelt

Enforcement Management

HDI

Crash
helmets

Traffic Personal
Risk
Risk

Country (Region) A
Country (Region) B
Country (Region) C
Country (Region) D
Country (Region) E
etc.
Countries (regions) with high RSDI (70 or above)
Countries (regions) with medium RSDI (40 to 70)
Countries (regions) with low RSDI (less than 40)

Figure. 5.3: An example of illustration of the RSDI levels and dimensions

for different countries and regions (not actual dimensions)

91

5.7

Summary

This chapter has described the methodological approaches. There was choice of possible
approaches; however none of them is obviously the right one. In principle, each
approach has its advantages and disadvantages that can complement each other. This
will require further investigation and applications and this will be examined in next
chapters (moving from RSDI philosophy and methods to practice).

92

hapter 6: empirical study 1: Applying RSDI to HMCs in EU

Chapter 6

Empirical application 1:
Applying RSDI to eight HMCs in EU
The aim of this chapter is to apply RSDI for HMCs where a sample of
eight European countries is chosen and three methods of RSDI are used
for assessing and classifying these countries according to their overall
road safety performance. The chapter starts with an overview background
of the road safety situation in the EU region, in particular, on the main
studies that have been made regarding performance indicators and
benchmarking in the EU.

6.1

Background- road safety in the EU

No doubt that HMCs Highly Motorised Countries have experienced a successful

record of reducing traffic accident deaths and injuries in the last few decades. This
development is a result of a long process of investments and actions in the overall road
safety system accompanied with high-level of commitment at national and regional
level (i.e. EU countries). Despite this development, the current number of accidents and
casualties is still considered too high and unacceptable throughout HMCs.
For policy makers, one main challenging approach to solving the road safety problem is
setting clear targets. Such targets are basically necessary to ensure that there is an
integrated follow-up action plan in the country. However, setting efficiency targets in
any country needs a systematic benchmarking system that allows the country to
compare its outcomes (progress) over time, or with similar outcomes with other
countries.

93

The European Commission, for example, has set up an ambitious target to halve the
number of traffic deaths by 2010 in the EU (i.e. 25,000 deaths for in the EU-2513 and
20,000 deaths for the EU-15), see e.g. European Commission (2000) and ETSC (2004).
According to some recent reports e.g. (ASSESS, 2005), (OECD, 2006), the present
levels of road deaths seem too high as compared to the stated target. This suggests that
additional actions and efforts have to be made in the EU region, if the target has to be
achieved by 2010.
There has always been an interest in benchmarking road safety achievements and
progress between European countries. Several large-scale benchmarking studies have
been carried out either within EU Union vision (i.e. SafetyNet, 2005a) or within OECD
vision, mostly of Western Europe (i.e. Page, 2001; IRTAD, 2006). At the same time,
several small-scale benchmarking studies have been conducted across the EU, for
example, the SUNFlower study (Koornstra et al., 2003) has focused on the good
performing countries in road safety (Sweden, the United Kingdom and Netherlands),
and the SECBelt study (ETSC, 2005) focused on road safety causes and problems in the
Southern, Eastern and Central European countries.
Collecting simple safety performance indicators for benchmarking purposes has been
used and developed by a number of European institutions and databases such as the
European Road Federation (ERF), Eurostat, International Road Traffic and Accidents
Database (IRTAD), Community database on accidents on the roads in Europe (CARE).
Additionally, several European initiatives and research studies have assessed the
performance across the EU in particular areas of road safety. Examples of such
benchmarking studies include the EuroRAP study (focuses on safer roads in the EU),
EuroNCAP study (focuses on safer vehicles in the EU), the ESCAPE study (focuses on
police enforcement in the EU), and the SARTRE study (focuses on the behaviours of
road users in the EU). This has already been discussed in brief in Chapter 4 (Figure 4.4).
Furthermore, there is an ongoing debate in EU literature on the development of key
performance themes in road safety across EU, in particular, on the work on uniform
comparable performance indicators. This development has been mainly reported by
ETSC14 (2001), ETSC (2003), SafetyNet (2005b) and Ecorys & SWOV (2005).

13 EU-25: After the enlargement in May 2004, the EU becomes twenty-five member countries. The
EU-15: Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, the
Netherlands, Portugal, Spain, Sweden and the UK. The EU-10 new members are: Cyprus, the Czech
Republic, Estonia, Hungary, Latvia, Lithuania, Malta, Poland, Slovakia and Slovenia.
14 The European Transport Safety Council (ETSC) is an active organisation in road safety in EU.
Available at: http://www.etsc.be (last visited 2006-04-11)

94

To summarise, many answers have been given to questions related to benchmarking and
performance in the EU. Many European projects, initiatives and studies have
strengthened the work in the area of benchmarking and performance indicators in terms,
for instance, of availability, quality and standards of data. However, there are two main
issues that have not attracted much attention yet, first is the need for aggregated tools
for benchmarking and setting targets, and the second issue is the necessity to have a
clear distinction between output-operational-outcomes of performance indicators for
benchmarking purposes.

6.2

Data and indicators included in RSDI

Eight European countries have been chosen for this empirical study15: Belgium,
Germany, Spain, France, Italy, The Netherlands, Sweden, and the United Kingdom.
According to IRTAD database, the selected countries tend to have a high level of
motorisation (over 500 motor-vehicles/1,000 inhabitants). Additionally, the selected
countries have shown high achievement in human development index with a HDI of
0.928 or above (UNDP, 2005).
The selected indicators and dimensions included into RSDI are presented in Table 6.1,
whose data set was collected from year 2003 through different sources. This selection
reflects mostly the availability and reliability of data for all the selected countries.
6.2.1

Limitation and quality of data

One major limitation, with regard to data sources, was the difficulty in collecting all of
the required data from one source or database. In this study, data comes from several
sources, primarily from the studies mentioned in previous section. However, in general,
care must be taken when we use more than one source because sources may use
different definitions for the same indicator.
Unfortunately data of two wheelers (motorcyclists and cyclists) helmet use in most EU
countries are not available and in some Eastern and Southern Europe countries are not
collected at all (ETSC, 2005). Furthermore, the selected EU countries have very close
results in the percentage of paved roads and in some cases the quality of this indicator
varies from one source to another. Hence both indicators are not represented in the
current study.

15 Standard country abbreviations: B: Belgium; D: Germany; E: Spain; F: France; I: Italy; NL: The
Netherlands; S: Sweden; the UK: the United Kingdom.

95

Due to the lack of accurate data relating to road user behaviour and police enforcement
from all the selected countries, this study has relied on SARTRE report (SARTRE,
2005) for finding the indicators: X6, X12, X13 (see Table 6.1). The SARTRE results are
based on a survey of opinions (questionnaires and interviews) about road safety
measures (e.g. behaviour and enforcement) between European car drivers.
However, this type of indicator, based on opinion survey, has several shortcomings with
regard to its accuracy of the results and the methods used in assessing the road users
opinions. For instance, it is uncertain whether the right sample of road users has been
asked, whether the road users really understood and answered all questions as intended
(see Sven, 1994), and whether the surveys were carried out annually (in case of
SARTRE project, the survey were carried out every 5th year or so, which is rather
long). However, despite these limitations, this type of indicator has advantages and
much to recommend because:

The data can be collected relatively easily.

The data collection is usually relatively low-cost (it depends on the sample size).
The enforcement and behaviour performance are essentially based on the idea
that any increase of the drivers perception (expectations, opinions) to the risk of
accident will increase their safety.
This type of indicator is widely developed and used and in different fields of
science. For instance, in a lot of quality management research, several indicators
are collected through questionnaires to find information about customers and
employee satisfaction or the image of the business in society (i.e. Hallowell,
1996). As discussed earlier (Chapter 4), high business results are related to high
customer satisfaction. Additionally, in many composite indices, data has been
collected and weighted based on public opinions. For instance, Daniels et al.
(200) measured the perception for health problems in developing countries.

In Table 6.1, target performance value for each indicator has been established and
estimated on the idea that there is progress ahead for all countries in road safety
development. In order to meet the whole target of the EU (i.e. 50% reduction of the
number of traffic deaths by 2010), sub-targets must be achieved in all road safety
sectors (indicators). However, it is not easy to decide how much the expected sub-target
of each indicator would contribute to the total EU target and that will be left for future
work.

96

25

X3: Percentage change of deaths (2000-2003)

X4: Percentage of seat belt use (rear seat)

Changing trend

Enforcement

Socio-economic
level

Safer
roads

Safer
vehicles

Road user
behaviour

1,1

X2: Deaths 30 days per 100,000 population

Personal risk

X13: Proportion of drivers in each country who had

been checked for speed in last 3 years

X12: Proportion of drivers in each country who had

been checked for alcohol in last 3 years

13

23

0.945

18,00

X10: Investment in road maintenance as % of total

X11: Human Development Index (HDI)

1,20

X9: % of motorway length per total road network

5,6

X7: Average age of the vehicle fleet (years)

94,94

86

X6: % drivers reporting they do not drive faster than

average in last 3 years

X8: Vehicles not motorcycles

26

X5: % drivers that never drink before driving

13,1

2,55

X1: Deaths 30 days per 10,000 vehicles

Traffic risk

Indicators

Dimensions

20

24

0.930

12,00

5,20

90,99

6,6

81

28

75

-11,9

8,0

1,23

17

32

0.928

15,00

0,40

87,31

8,5

83

23

20

-6,5

12,8

2,15

13

32

0.938

17,00

0,91

93,68

7,3

80

24

45

-25,0

10,2

1,68

0.934

8,00

2,10

82,06

8,1

78

16

10

5,6

10,5

1,59

27

37

0.943

58,00

2,00

94,44

6,9

66

44

47

-5,0

6,4

1,23

NL

41

0,949

62,00

0,60

93,97

8,9

76

78

74

-10,5

5,9

1,06

Sources: X1, X2, X3, X8 (IRTAD, 2006); X4 (ETSC, 2003); X5 (Vereeck and Deben, 2003); X6, X12, X13 (SARTRE 3, 2005); X7 (Eurostat, 2003);
X9 (SafetyNet, 2005a); X10 (ERF, 2005); X11 (UNDP, 2005).
Standard country abbreviations: B: Belgium; D: Germany; E: Spain; F: France; I: Italy; NL: The Netherlands; S: Sweden; UK: United Kingdom.
The choice of targets is estimated on the idea that there is potential progress ahead for all countries in road safety development.

System

Human

Product

Pillars

Table 6.1: Road safety data and key indicators for the selected HMCs in European Union

38

0.939

47,00

0,84

96,49

6,3

86

52

75

2,2

6,1

1,14

UK

45

45

0,98

65

97

90

80

80

-25

Targets

97

2003

2003

2003

2003

2003

2003

2000

2003

1998

2002

20002003

2003

2003

Year of
data

6.3

Calculation of RSDI

Following the methods of RSDI mentioned in Chapter 5, we measure the RSDI for the
selected countries by combining all their indicators together. By using the method
Simple Equal Average, the values of the indicators are first normalised to the target
values by using the formulas (5.3 and 5.4). As the indicator percentage change of road
accident deaths (2000-2003) has positive and negative values, it is suitable to
normalise it by the (minimum, maximum) method (formula 5.5). The normalised
values of indicators are presented in Table 6.3.
The RSDI is measured as the simple average of the three pillars where each pillar is
given equal weight (one-third):

RSDI =
where

1 3
Pi
3 i =1

(6.1)

Pi is the value of the normalised pillars.

Within each pillar the included indicators are given equal weights as follows (see also
Table 6.2):

100
100
100
X1 +
X2 +
X3
9
9
9
100
100
100
P2=
X4 +
X5 +
X6
9
9
9
100
100
100
100
100
100
100
X7 +
X8 +
X9 +
X 10 +
X 11 +
X 12 +
X 13
P3=
21
21
21
21
21
21
21
P1=

Alternatively, we can give subjective weights to the indicators based on previous

experience. The study judges the selection of weightings among several alternatives
based on earlier discussion (see Chapter 5) where RSDI should give greater weighting
points to the end results and human behaviour. At the same time, the judgement takes
into account the earlier case studies, accident investigations and meta-analysis studies
from the chosen countries/region that can indicate the importance of each indicator in
relation to accidents (e.g. European Commission, 2000). An example of the subjective
choice of weights that can be taken in this study is (40%) for each of the product and
human behaviour and (20%) for the System index (see Table 6.2):
RSDIj= 0,4PI+ 0,4HI+ 0,2SI
where

98

PI is the Product index,

HI is Human behaviour index, and
SI is the System index.

(6.2)

Within the System Index, we can give equal weights to the four dimensions: safer
vehicles (X7 and X8), safer roads (X9 and X10), HDI (X11) and enforcement (X12 and
X13), each of which has a weighting of (5%). The distribution of weights between the
indicators can be expressed:
40
40
40
40
40
40
20
20
20
X1 + X 2 + X 3 + X 4 + X5 + X 6 + X 7 + X8 + X9
3
3
3
3
3
3
8
8
8
(6.3)
20
20
20
20
+ X10 + X11 + X12 + X13
8
4
8
8
RSDI =

Table 6.2: Selecting the weights of the indicators of RSDI (objectively and subjectively)

Pillar

Indicators

Product

X1: Deaths 30 days per 10,000 vehicles

100/9

40/3

X2: Deaths 30 days per 100,000 population

100/9

40/3

X3: Percentage change of deaths (2000-2003)

Total weighting scores (Product)
X4: Percentage of seat belt use (rear seat)
Human
X5: % drivers that never drink before driving
Behaviour X6: % drivers reporting they do not drive faster
than average in last 3 years
Total weighting scores (Human)

System

Weighting scores
Weighting scores
Method
Method
(Simple Average) (Subjective weights)*

100/9

40/3

100/3 points

40 points

100/9

40/3

100/9

40/3

100/9

40/3

100/3 points

40 points

X7: Average age of the vehicle fleet

100/21

20/8

X8: Vehicles not motorcycles

100/21

20/8

X9: % of motorway length per total road network

100/21

20/8

X10: Investment in road maintenance as % of total

100/21

20/8

X11: Human Development Index (HDI)

X12: Proportion of drivers in each country who
had been checked for alcohol in previous three
years
X13: Proportion of drivers in each country who had

100/21

20/4

100/21

20/8

100/21

20/8

been checked for speed in last 3 years

Total weighting scores (System)

The overall weighting scores (RSDI)

100/3 points

20 points

100 points

* Greater weights are given to indicators that are related to Product and Human pillars.

However, it is necessary that further in-depth studies be carried out to investigate the
accident data, causes and priorities in the selected countries before making any better
subjective assessment of weights.
The results obtained by these two methods are shown in the Table 6.3.

99

49,75
73,14
83,36

46,51 39,06 63,05

59,52 49,02 100,03

62,89 47,62 38,74

87,72 81,97 45,64

France

Italy

Netherlands 81,30 78,13 59,98

94,34 84,75 70,98

Spain

Sweden

UK

X5

X6

93,75 65,00 95,56

92,50 97,50 84,44

58,75 55,00 73,33

12,50 20,00 86,67

56,25 30,00 88,89

25,00 28,75 92,22

93,75 35,00 90,00

31,25 32,50 95,56

X4

Human

84,77

91,48

62,36

39,72

58,38

48,66

72,92

53,10

Human
index
X8

X9

X10

X11

X12

X13

6,67

23,08 94,69 71,11 37,78

8,89

15,56

6,67

79,37 99,47 14,00 72,31 95,82 20,00 84,44

56,18 96,87 10,00 95,38 96,84 91,11

72,46 97,36 33,33 89,23 96,22 82,22 60,00

61,73 84,59 35,00 12,31 95,31

68,49 96,58 15,17 26,15 95,71 71,11 28,89

58,82 90,01

75,76 93,81 86,67 18,46 94,90 53,33 44,44

89,29 97,87 20,00 27,69 96,43 51,11 28,89

X7

System

66,49

64,72

75,83

44,77

57,44

54,59

66,77

58,75

System
index

100

Notes: (X1, X2, X3,.., X13) are the normalised indicators that are used to calculate RSDI.
The values are normalised by using the distance to target (except X3 which is measured by using the minimum-maximum normalisation method).

71,78

69,52

49,54

72,51

81,30 62,50 73,72

41,74

Product
index

Germany

X3

39,22 38,17 47,82

X2

Belgium

X1

Product

Table 6.3: RSDI scores using simple average technique and subjective weights

74,34

79,85

70,44

44,75

61,78

50,93

70,73

51,20

76,65

83,68

69,88

46,01

63,61

51,20

72,23

50,63

RSDI
RSDI
(Simple (Subjective
Average) weights)

By using the Principal Component Analysis (PCA), we normalise the data of indicators
to zero mean and unit variance by using the formula (5.9). We then compute all
indicators and consider the number of principal components that could explain most of
the original data (variation). The advantage of this technique is that the weighting of
various indicators is based on statistical methods rather than subjective judgements. It
lets simply the data itself to decide on the weighting issue, which is good from
transparency point of view.
In table 6.4 we present the eigenvalues derived from the PCA. By examining the
eigenvalues, we see that the eigenvalues of the first five components are more (or near)
than unity (1.00). So we have to consider the first five principal components (PC1, PC2,
PC3, PC4 and PC5) that together explain 92.2 % of the data, which is quite good.
The number of Principal Components obtained for this study is relatively high (5
components), which means that the total information of the indicators could not be
summarised in few components (sub-indices). This means that a low correlation exists
between the original indicators.
Table 6.4: The Eigenvalue analysis of the normalised indicators
Eigenvalue 5,152 2,430 1,951 1,526 0,922 0,720 0,2989 0,000 0,000 0,000 -0,000 -0,000 -0,000
Proportion 0,396 0,187 0,150 0,117 0,071 0,055 0,023 0,000 0,000 0,000 -0,000 -0,000 -0,000
Cumulative 0,396 0,583 0,733 0,851 0,922 0,977 1,000 1,000 1,000 1,000 1,000 1,000 1,000
PC1

-0,331 -0,395 -0,100 0,356 0,400 -0,205 0,022 0,303 -0,035 0,408 0,281 0,207 0,114

PC2

-0,095 -0,131 0,228 0,220 -0,095 0,249 -0,508 0,192 0,284 -0,061 -0,136 -0,390 0,503

PC3

-0,422 -0,245 -0,115 0,127 -0,093 -0,280 0,282 -0,394 0,481 -0,125 -0,393 -0,080 -0,046

PC4

-0,124 -0,133 0,676 -0,221 0,162 -0,100 0,204 -0,293 -0,241 0,216 0,142 -0,415 0,010

PC5

-0,128 -0,032 -0,152 0,317 0,240 0,681 0,371 -0,038 -0,312 -0,090 -0,244 -0,180 -0,045

The PC1 explained 39.6% of total variance. The other 4 Principal Components
explained 18.7%, 15.0%, 11.7%, and 7.1% of total variance, respectively. The formulas
of the principal components are given in details below, where X1 to X13 are the
normalised values of the original indicators. The obtained principal components are
uncorrelated with each other and each explains on part of the total variation of the
original data.
PC1=-0,331X1-0,395X2-0,100X3+0,356X4+0,400X5-0,205X6+0,022X7+
0,303X8-0,035X9+0,408X10+0,281X11+0,207X12+0,114X13

101

PC2=-0,095X1-0,131X2+0,228X3+0,220X4+-0,095X5+0,249X6-0,508X7+
0,192X8+ 0,284X9- 0,061X10-0,136X11-0,390X12+0,503X13
PC3=-0,422X1-0,245X2-0,115X3+0,127X4-0,093X5-0,280X6+0,282X70,394X8+ 0,481X9-0,125X10-0,393X11-0,080X12-0,046X13
PC4=-0,124X1-0,133X2+0,676X3-0,221X4+0,162X5-0,100X6+0,204X70,293X8-0,241X9+0,216X10+0,142X11-0,415X12+0,010X13
PC5=-0,128X1-0,032X2-0,152X3+0,317X4+0,240X5+0,681X6+0,371X70,038X8-0,312X9-0,090X10-0,244X11-0,180X12-0,045X13
By replacing the values of the normalised indicators for each country in the formulas
above: PC1, PC2, PC3, PC4 and PC5, we obtain the results shown in Table 6.5. It must
be noted that the PCA method has produced five components (sub-indices) with
different scales than RSDI. Thus there is a need to convert these PCs into one single
value (RSDI) with scale range from 0 to 100. In doing this, we first normalise the values
of PCs by using the simple average normalisation method. Then we calculate RSDI by
attaching the normalised PCs with their derived weights into RSDI (see also Table 6.5).
Table 6.5: The adjustment procedures of PCA weights to match the scale of the RSDI
Normalised Normalised Normalised Normalised Normalised
PC1
PC2
PC3
PC4
PC5

PC1

PC2

PC3

PC4

PC5

-1,76

0,51

-2,60

-0,24

-0,69

16,20

55,70

0,00

41,70

40,79

26,44

-0,17

1,59

2,07

-1,50

0,02

42,04

78,12

100,00

8,23

64,67

55,72

-2,36

-1,23

-0,11

-0,13

0,71

6,55

19,76

53,42

44,59

88,01

27,69

-0,33

-0,86

-0,24

-1,81

0,27

39,46

27,32

50,46

0,00

73,25

36,05

-2,76

-0,41

1,37

1,96

-0,22

0,00

36,59

85,05

100,00

56,85

37,99

NL

2,30

-0,08

0,30

0,27

-1,91

82,09

43,57

62,19

55,27

0,00

60,73

3,40

-2,18

-0,08

0,45

0,75

100,00

0,00

53,95

60,00

89,21

65,68

UK

1,68

2,65

-0,72

0,99

1,07

72,06

100,00

40,31

74,21

100,00

74,28

PC5

PC1

PC2

PC3

PC4

Eigenvalue (proportion)

39,6

18,7

15

11,7

7,1

Weights out of 100

42,53

20,09

16,11

12,57

7,63

RSDI

For a snapshot of the performance in the chosen countries, we Plot PC1 and PC2 in
Figure 6.1, this shows that both components are not correlated. It also shows the
similarities and dissimilarities between different groups of countries regarding the data
used in the study. According to the performance of PC1, that represents the most

102

variation in the original data, countries can be classified into three groups. It seems that
the countries to the right of the vertical axis show good results. The countries to the left
show worse results, while other countries are around the average.

U.K.

Second Component

Germany

Belgium
Netherlands
Italy
France

-1

Spain

Sweden

-2
-3

-2

-1

0
1
First Component

Figure 6.1: Score Plot of the first two Principal Components

As discussed in Chapter 5, although PCA has a number of advantages, it has also some
limitations. The best results are usually obtained from PCA when the original indicators
are highly correlated and when a large sample size of countries being used. Perhaps we
see that there is a justification for selecting only one of the five components, i.e. PC1,
but this will lead to a loss of too much information.
6.4

Summary of results

In the following table, the final RSDI scores and ranks are summarised from this
empirical study in accordance to the methods used in calculation. A quick look at the
results indicates a remarkable difference between the selected European countries. The
first two methods have shown near results to RSDI for all countries, while a
considerable percentage of reduction for RSDI scores occurred to all countries (except
UK) by using PCA method. These results, together with those results from the second
empirical study, will be discussed and analysed in more detail in the Chapter 8. This
would subsequently support the findings and the recommendations of the thesis.

103

Table 6.6: The final RSDI scores and ranks from the empirical analysis and approaches

Sweden
U.K.
Germany
Netherlands
France
Belgium
Spain
Italy

104

RSDI
Simple
average

Ranks

RSDI
Subjective
weights

Ranks

RSDI
PCA

Ranks

79,85

83,68

65,68

74,34

76,65

74,28

70,73

72,23

55,72

70,44

69,88

60,73

61,78

63,61

36,05

51,20

50,63

26,44

50,93

51,20

27,69

44,75

46,01

37,99

7: Empirical application 2: Applying RSDI to LMCs in

7

Southeast Asia

Chapter 7

Empirical application 2
Applying RSDI to LMCs in Southeast Asia
The aim of this chapter is to apply the RSDI tool to five LMCs for
benchmarking their road safety performance. This benchmarking is
carried out against reference targets. The countries chosen for this
empirical study are neighbouring countries in Southeast Asia: Cambodia,
Laos, Malaysia, Thailand and Vietnam.

7.1

Background- the road safety situation in Southeast Asia

Most of the LMCs Less Motorised Countries range between the lower and medium
level of the motorisation and a few are now starting to enter the phase of rapid
motorisation as seen especially in emerging and reformed economies. An example of
emerging economies is some countries of Southeast Asia (e.g. Malaysia) where
economic growth is transferred into increased car ownership. The reform economies are
mainly situated in East European and the Baltic countries where the transition to a
market economy has brought dramatic changes to cities as most of the road users have
shifted from public to private transport use.
The majority of LMCs are in a period of a rising number of deaths; however there is no
guarantee that the development of road safety in LMCs will follow the same
development in HMCs. One main factor that may lead to these differences is that most
HMCs have never experienced the same road traffic conditions as LMCs (e.g. a high
proportion of motorcycles as in many of South and Southeast Asia countries).
Most published data from international sources exclude the category of motorcycles
when they report the number of registered motor vehicles in different countries. This

105

will actually make little difference for HMCs where motorcycles account for only a
small percent of all vehicles on the roads, but in countries such as Southeast Asia where
there is a high rate of motorcycles (more than 50% in many of these countries), this
exclusion makes the international comparisons less accurate (Al-Haji & Lindskog,
2005).
In order to put RSDI into practice and to check its applicability to LMCs, five
neighbouring countries in Southeast Asia have been chosen: Cambodia, Laos, Malaysia,
Thailand and Vietnam. These countries are part of ASEAN16 region and share common
transport patterns and characteristics.
According to ASNet17 and Globesafe18 databases, in 2003, the selected countries tend to
have low to medium levels of motorisation. It is highest in Malaysia and Thailand (over
400 motor-vehicles/1,000 inhabitants), and lowest in Vietnam, Cambodia and Laos (less
than 200 motor-vehicles/1,000 inhabitants). The level of motorcycle ownership is high
in the chosen countries, where it accounted for 95% of the total number of vehicles in
Vietnam, 80% in Laos, 75% in Cambodia, 71% in Thailand and 48% in Malaysia. As a
result, the casualties in respect of motorcycles constitute the majority of road users in
accidents, nearly 90% in Cambodia, 70% in Vietnam and 60% in Malaysia.
According to the Human Development Report of the United Nations (UNDP, 2004), the
selected countries have shown medium level of the human development index (HDI)
out of a field of 175 countries worldwide. Malaysia is ranked (59), Thailand (76),
Vietnam (112), Cambodia (130) and Laos (135).
7.2

Data and limitations

As a matter of fact, dealing with a data collection and quality in LMCs we often face a
dilemma. The availability, reliability, normalisation, transparency and quality of data
are challenges and problems in most LMCs, which have to be taken into account before
making meaningful international comparisons (see the criteria of RSDI in Section 4.4
and discussion earlier in Section 4.9). In this study, seven indicators are chosen in
RSDI, which have been categorised in six dimensions (see Table 7.1). The data set was
16 ASEAN refers to the countries: Brunei, Cambodia, Indonesia, Laos, Malaysia, Myanmar, the
Philippines, Singapore, Thailand and Vietnam.
17 ASNet ASEAN Road Safety Network is an Internet platform for road safety between ASEAN
countries. ASNet platform is hosted by Linkping University, Sweden. http://www.asnet.org (last visited:
2006-01-04).
18 Globesafe is an Internet-based tool that collects, harmonises and analyses the road safety data for the
purpose of global comparisons, Globesafe tool is hosted by Linkping University, Sweden.
http://www.globesafe.org/ (last visited 2006-01-04).

106

collected mainly from year 2003 through different sources. Due to the lack of accurate
data relating to road user behaviour from the selected countries, this study has relied on
estimated figures, made by consultants from Asian Development Bank (ADB) through
the ASNet database, for finding the indicators: X3 (seat belts usage) and X4 (Crash
helmets usage). The use of helmets and seatbelts differs from country to country and it
tends to be higher in major cities, urban areas or specific roads (e.g. motorways) than
other places. Therefore, the ADBs estimated figures were taken across the whole
country in each case.
However, the selected data in this study would be the starting point toward the
development of a set of key performance indicators for LMCs. Any better collection and
quality of indicators in future work will lead to better and accurate assessment of RSDI
for these countries.
The reference target has been established subjectively for each indicator according to
the earlier discussion made in Section 4.9. The targets are estimated based on the idea
that there is potential progress ahead for all countries worldwide in road safety
development.
Table 7.1: The key performance indicators and dimensions in the selected countries
(year 2002-2003)
Safer Product

Safer People

Traffic
Risk
(Deaths per
10,000
vehicles)

Personal
Risk
(Deaths per
100,000
person)

D1
X1

D2
X2

X3

Cambodia

22,73

7,54

Laos

20,87

Malaysia

Country

Safer System
Safer vehicles Safer roads

Seat belt
use %
(estimated)

Helmet use
%
(estimated)

Vehicles not
motorcycles
(%)

Paved
Roads
(%)

HDI

X4

D4
X5

D5
X6

D6
X7

10

10

24,8

16,0

0,57

9,81

10

10

19,9

14,0

0,53

4,88

25,64

70

70

51,8

76,0

0,79

Thailand

5,23

20,41

30

30

29,1

94,0

0,77

Vietnam

10,94

16,15

10

10

5,6

25,0

0,69

80

80

97

95

0,98

Reference
targets

D3

Notes: 1. X1, X2, X3.X7 are the selected performance indicators that used to calculate RSDI. 2. D1, D2, D3.D6 are
the selected dimensions. 3. The indicators X1, X2, and X5 are derived from the ASNet database for ASEAN countries.
4. The source of X6 is the World Development Indicators (WDI). For details, see: http://www.worldbank.org/data/.
Data of paved roads from Thailand is questionable. 5. HDI is taken from the Human Development Report 2004. For
details, see: http://hdr.undp.org/2004/. 6. The consultants from Asian Development Bank (ADB) have estimated the
data of X3 and X4 through the ASNet database.

107

7.3

Calculation of RSDI

Following the methods of RSDI mentioned in Chapter 5, we calculate the RSDI for the
selected countries by combining all their indicators together.
By using the Simple Equal Average method, the values of the indicators are first
normalised to the target values by using the formulas (5.3 and 5.4), see Table 7.3.
Second, each pillar of RSDI is given equal weight (one-third):

RSDI =
where

1 3
Pi
3 i =1

Pi are the normalised pillars

Within each pillar the included indicators are given equal weights as follows (see also
table 7.2):
100
100
X1 +
X2
6
6
100
100
P2 =
X3 +
X4
6
6
100
100
100
P3 =
X5 +
X6 +
X7
9
9
9
P1 =

Alternatively, we can give subjective weights to the indicators based on previous

experience. We rely first on earlier case studies from the region that indicate the
importance of each dimension/indicator in relation to accidents. Examples of such
studies include studies from Asian Development Bank (ADB) and Asia-Pacific
Regional Road Accident Database (APRAD); see (ADB, 2005), (ADB, 1996). As a
result, it is possible to judge that both motorcycles (and helmets use) can play an
important role in calculating RSDI in these countries, so these indicators can be given
larger weights in RSDI. Secondly, the selection of weightings can be judged, among
several alternatives, based on earlier discussion in Chapter 4 that the indicators related
to end results and human behaviour should be given greater weighting points in RSDI.
Furthermore, it may be advisable to give less weight to the indicators with low
availability (or estimated figures such as seatbelt use or helmet use). However, here we
will approach the question of data collection in other countries regarding the same
indicators, which can skew the good scores of RSDI towards countries with more
availability of data.

108

An example of subjective choice of weights that can be taken in this study is 40% for
the product, 30% for the human behaviour and 30% for the System index (see again
Table 7.2):
RSDIj= 0,4PI+ 0,3HI+ 0,3SI
where

PI is the Product index,

HI is Human behaviour index, and
SI is the System index.

The same function is clarified below by the distribution of weights between the
indicators:
RSDI= 0,20X1+0,20X2+0,15X3+0,15X4+0,15X5+0,5X6+0,10X7
Table 7.2: Selecting the weights of the indicators of RSDI (objectively and subjectively)

Pillar

Indicators

Weighting scores Weighting scores

Method
Method
(Simple Average) (Subjective weights)*

X1: Deaths 30 days per 10,000 vehicles

100/6

20

X2: Deaths 30 days per 100,000 population

100/6

20

Product
Total weighting scores (Product)
Human
Behaviour

100/3 points

40 points

X3: Percentage of seat belt use (front seat)

100/6

15

X4: Percentage of helmet use (motorcyclists)

100/6

15

Total weighting scores (Human)

100/3 points

X5: Vehicles not motorcycles

System
X6: % of paved roads
Performance

100/9

15

100/9

100/9

10

X7: Human Development Index (HDI)

Total weighting scores (System)
The overall weighting scores (RSDI)

100/3 points

30 points

30 points

100 points

* Greater weights are given to indicators that are related to product, human behaviour and vehicle
ownership.

However, this subjective selection of weights is not final and further in-depth studies
would be needed to investigate the accidents causes and priorities in these countries
before making any better assessment of weights.

109

The results obtained by these two methods (simple average and subjective weights) are
shown in the Table 7.3.
Table 7.3: RSDI scores using simple average technique and subjective theories
RSDI

RSDI

X1

X2

X3

X4

X5

X6

X7

(Simple
Average)

(Subjective
weights)

Cambodia

4,40

66,31

12,50

12,50

25,57

16,84

58,16

27,13

28,39

Laos

4,79

50,97

12,50

12,50

20,52

14,74

54,08

23,39

24,12

Malaysia

20,49

19,50

87,50

87,50

53,40

80,00

80,61

59,61

54,32

Thailand

19,12

24,50

37,50

37,50

30,00

98,95

78,57

42,83

37,28

Vietnam

9,14

30,96

12,50

12,50

5,77

26,32

70,41

22,24

20,99

100,00 100,00 100,00 100,00 100,00 100,00 100,00

100,00

100,00

Reference
targets

In the calculation of RSDI by using Principal Component Analysis method, we

compute all indicators and we consider the number of principal components that could
explain most of the original data. By examining the eigenvalues in Table 7.4, we see
that the eigenvalues of the first two components are more than unity (1.00). So we
consider the first two principal components (PC1 and PC2), which together explain 93.6
% of the data, which is highly good.
The formula of the first principal component is:
PC1=0,401X1-0,077X2-0,419X3-0,419X4-0,383X5-0,4X6-0,418X7
where X1 to X7 are the normalised values of the original indicators. This component
accounts for (5,2873/7)=75.5% of total variance. There appears to be a contrast between
X1 (traffic risk) on one hand, and the other indicators on the other hand. As is shown,
the weight of X2 (personal risk) is very small, which means that X2 has no affect on
PC1, so X2 can be neglected in this component.
In a similar way, the second principal component is:
PC2=0,24X1-0,858X2+0,071X3+0,071X4+0,403X5-0,179X6+0,047X7
This component accounts for (1,264/7)=18.1% of the total variance. It is contrast with
X2 (personal risk) and X6 (paved roads) on one hand, and the other indicators on the
other hand. In this case, the weight of X2 (personal risk) is very high, which means that
X2 has a considerable affect on this component.

110

Table 7.4: Eigenvalue analysis of the covariance matrix of normalised values

Eigenvalue

5,2873

1,264

0,0065

0,0000

0,3292

0,113

-0,0000

Proportion

0,755

0,181

0,001

0,000

0,047

0,016

-0,0000

0,755

0,936

1,000

1,000

0,983

0,999

1,000

0,401

-0,077

-0,419

-0,419

-0,383

-0,400

-0,418

0,240

-0,858

0,071

0,071

0,403

-0,179

0,047

Cumulative
First principal
component PC1
First principal
component PC2

We can substitute the values of the normalised indicators for each country in the
formulas, PC1 and PC2. The results are shown in Table 7.5. As be seen, the results of
the normalised PC1 indicates that RSDI level is high for Malaysia followed by
Thailand. Not surprisingly, Cambodia and Laos scored higher by using PC2 formula as
they have better records in personal risk, because, as discussed earlier, a country at an
early level of motorisation tends to have a growing number of traffic-related deaths, but
with low growth in the number of population-related deaths (personal risk). However,
one has to remember that PC2 accounts only 18,1% of the total variance.
For a snapshot of the performance in these chosen countries, we plot PC1 and PC2 and
we see both components are not correlated (Figure 7.1). This figure provides a map of
the similarities and differences between the countries regarding the values of PC1 and
PC2. It shows that Malaysia and Thailand are situated to the lower-left of the vertical
axis, while Cambodia and Laos are situated in the upper-right corner of the figure. The
reference targets remains so far ahead of all countries for both values of PC1 and PC2.
This means that there is still hard work ahead for these countries in road safety
development.

1,5

Reference targets
Cambodia

Second Component

1,0

Laos

0,5

0,0

Vietnam

-0,5

-1,0

Malaysia
-4

-3

-2

Thailand
-1
0
First Component

Figure 7.1: Score Plot of the new scores of PC1 and PC2 in the sample of countries

111

It must be stressed that principal components analysis has produced two indices with
different scales. Thus there is a need to convert the two PCs into one single value
(RSDI) with scale range from 0 to 100. In doing this, we first normalise the values of
PC1 and PC2 by using the simple average method. Then we calculate RSDI by
attaching the normalised PC1 and PC2 with their weights (proportions) into RSDI (see
Table 7.5).
Table 7.5: The adjustment procedures of PCA weights to match the scale of the RSDI
Countries

PC1

PC2

Cambodia

2,164

0,868

1,38

75,64

15,74

Laos

2,240

0,513

0,00

62,36

12,06

Malaysia

-1,863

-1,156

75,38

0,00

60,80

Thailand

-0,618

-1,133

52,53

0,85

42,53

Vietnam

1,385

-0,612

15,71

20,34

16,61

Reference targets

-3,308

1,520

100,00

100,00

100,00

Eigenvalue (proportion)
Weights out of 100

75.5
80,66

18,1
19,33

7.4

Normalised PC1 Normalised PC2

RSDI

Summary of results

In the following table, the final RSDI scores and ranks from the empirical analysis are
summarised in accordance with the three methods used in calculating RSDI. The results
from this study indicate a remarkable difference between the selected countries. In the
next chapter, these achieved results together with those results that obtained from the
empirical study regarding HMCs will be discussed, analysed and compared.
Table 7.6: The RSDI scores and ranks from the empirical analysis and approaches

Malaysia
Thailand
Cambodia
Laos
Vietnam

112

RSDI
Simple
average

Ranks

RSDI
Subjective
weights

Ranks

RSDI
PCA

Ranks

59,61

54,32

60,81

42,83

37,28

42,53

27,13

28,39

15,74

23,39

24,12

12,06

22,24

20,99

16,61

Chapter 8: Analysis and Discussion of Results

Chapter 8

Analysis and Discussion of Results

The aim of this chapter is to present and discuss the research results
obtained from the two empirical studies and the theoretical
discussion. The aim of the empirical studies was to put RSDI theory,
philosophy and methods into practice.

The goal of this dissertation is to develop, present and test a new international integrated
tool for benchmarking road safety in order first to bridge a gap in the literature in this
area of research and second to offer a simple and comprehensive tool to policy makers
and public to use. To accomplish this goal, two empirical studies were carried out to
determine the applicability of this tool in real world applications and from different
parts of the world. This chapter will investigate the results and explore mainly the
seventh (last) research question (specified in Section 1.2), which is:
How can the applicability (usefulness) of this index be checked and evaluated?
To answer this research question, let us pose it in a simpler version by dividing it into
two general questions from theoretical and empirical perspectives:
Fist question: How to evaluate the tool from the results of empirical studies and
the methods used in calculation?
Second question: How to evaluate the results of this tool from the theoretical
discussion?
For this purpose, this chapter is divided into two sections. The first section assesses the
empirical results, while the second section will use an analytical method called SWOT
to assess theoretically the overall Strengths, Weaknesses, Opportunities, and Threats of
RSDI.

113

8.1

An Empirical Assessment: results and discussion

The results of RSDI from the two empirical applications are presented in Table 8.1.
Three methods have been used in calculating RSDI, one method used equal weights for
all indicators and countries, whereas the other methods give different weights depending
on the importance of indicators either subjectively or statistically.
Table 8.1: The RSDI scores and ranks from the two empirical applications and RSDI
approaches
HMCs

Sweden
U.K.
Germany
Netherlands
France
Belgium
Spain
Italy
LMCs

Malaysia
Thailand
Cambodia
Laos
Vietnam

RSDI
Simple average

Ranks

RSDI
Subjective
weights

Ranks

RSDI
PCA

Ranks

79,85
74,34
70,73
70,44
61,78
51,20
50,93
44,75

1
2
3
4
5
6
7
8

83,68
76,65
72,23
69,88
63,61
50,63
51,20
46,01

1
2
3
4
5
7
6
8

65,68
74,28
55,72
60,73
36,05
26,44
27,69
37,99

2
1
4
3
6
8
7
5

RSDI
Simple average

Ranks

RSDI
Subjective
weights

Ranks

RSDI
PCA

Ranks

59,61
42,83
27,13
23,39
22,24

1
2
3
4
5

54,32
37,28
28,39
24,12
20,99

1
2
3
4
5

60,81
42,53
15,74
12,06
16,61

1
2
4
5
3

From the above table, it can be seen that there is a wide variation in RSDI among the
selected countries. The change of method in calculating the index has changed the
weights and as a result the final results of the index as discussed further below.
The RSDI results from HMCs indicate that the RSDI level is high for Sweden, the UK
and Germany. The opposite is true for the Belgium, Spain and Italy as they obtained a
lower level of RSDI. The Spain and Belgium have almost the same RSDI values. The
countries can be classified into two groups of RSDI according to the Simple average
and Subjective weights based on previous experience methods. The countries Sweden,
the UK, Germany and the Netherlands have achieved high levels of RSDI (over than
70), while France, Belgium, Spain and Italy are grouped at the medium level of RSDI
(40 till 70). No country has obtained a lower road safety development (with RSDI of 40
or less) by using these two mentioned methods. The distances between the RSDI groups
are rather high. From the Table 8.1 we also see the results of top RSDI countries are

114

rather far ahead of the countries at the low level (e.g. between Sweden (83,36) and Italy
(46,01)).
The RSDI results from LMCs put Malaysia and Thailand at the top level of the sample
list of countries, and are followed by Cambodia, Laos and Vietnam. Laos and Vietnam
have shown almost the same RSDI values. The countries can be classified into two
groups by using RSDI ranged from 0 till 100 where Malaysia and Thailand ranked at
the medium level (40 till 70), while Cambodia, Laos and Vietnam are at the low level of
RSDI (less than 40). On the other hand no country has obtained a high road safety
development (with RSDI of 70 or above). As in HMCs, the distances between the RSDI
groups in LMCs are rather high. From the previous table we see the results of Malaysia
and Thailand are far ahead of the rest of countries at the low level.
It should be noted that there are several factors that can affect the country performance
include: the selection of dimensions/indicators, normalisation techniques, weighting
system, etc. There are several statistical tools can be used to measure the uncertainty
and sensitivity of the results in relation to the input data, for example the uncertainties
analysis method (UA) and sensitivity analysis (SA); a review of these methods is
available in Saisana et al. (2005). However, these methods are not applied in this study
because of small sample size of both empirical studies. Besides that the thesis has made
in advance careful assumptions and discussions regarding data selection and
normalisation methods. Therefore, a simple uncertainty analysis for the RSDI results
can be made regarding weighting techniques as follows.
As we see in Figure 8.1 the chosen countries have obtained very similar RSDI
performance scores by using the approaches Simple average and Subjective weights.
In contrast, these countries are highly influenced by the choice of the weights of PCA
method. The uncertainty change range in some countries is rather high about 25% (e.g.
Spain, Belgium and France), which is taken as the difference between the maximum and
minimum RSDI values for the country. Therefore, we may say that RSDI is quite
sensitive to the change of weights and methods of calculations, especially by using the
PCA method.
As can also be seen, the RSDI scores, in most countries, are positively impacted through
this choice of the Subjective weights method where the scores of RSDI have slightly
increased in most countries. In contrast, the RSDI scores in most countries are heavily
impacted (negatively) through the choice of PCA method. Besides, the PCA method has
led to different ranking of RSDI than the other two methods. For instance, the UK,
Netherlands, Italy and Vietnam scored higher by using the PCA method.

115

The results from this analysis indicate also a remarkable difference between the selected
countries even at the same level of motorisation (e.g. France and United Kingdom) and
even at the same level of accident death rates (e.g. Sweden and Netherlands).

90
80
70
60

Figure 8.4: Comparing the RSDI scores between the selected

EU countries according to the change of weights/approaches

50
40
30
20
10

Ita
ly

Sp
ai
n

Be
lg
iu
m

Fr
an
ce

an
y

et
he
rla
nd
s
N

er
m

.K
.
U

Sw

ed
en

RSDI- Simple Average

RSDI- Based T heories

RSDI- Principal Components

70
60
50
40
30
20
10
0
Malaysia
RSDI- Simple Average

Thailand

Cambodia

RSDI- Based T heories

Laos

Vietnam

RSDI- Principal Components

Figure 8.1: Comparing the RSDI scores between the selected

Samples of countries according to the change of weights/approaches

8.1.1

Why does RSDI differ between countries?

The results of RSDI differ considerably between the chosen countries. To understand
these obtained outcomes further, one has to note that there are many indicators
composed together into a single number and a lot of work behind the final RSDI
outcomes (e.g. data collection, selection, weightings, etc.). One needs to investigate the
differences of the normalised dimensions. In addition, there is a need to take a closer
look at the data behind each indicator, how they are collected, trusted and measured.

116

There is no single measurement that explains all the performance in country according
to the data collected as well as the philosophy of RSDI. There is always a need for
continuous progress in all sectors of road safety performance and in all countries. As an
example, France is ranked first in terms of its remarkable reduction in number of deaths.
However, when it comes to actual and overall performance (RSDI), France does not
hold such top position.
Furthermore, as it is shown in Table 6.3, there is a wide-range of performance values
between EU countries in all indicators. No country is the best at every practice they do.
As an example, Sweden is the best performing country in terms of the indicators X1, X2,
X5, X10, X11 and X12. United Kingdom has higher levels with respect to the indicators
X4, X6, X8 and X13. Germany achieved the best score in two indicators, namely: X4 and
X9. Similar observations can be observed for Belgium, which has higher levels with
respect to X6 and X7. France has achieved the most progress within the period from
2000 to 2003, cutting the number of road deaths by 25%. This remarkable reduction was
explained in (FIA, 2004) as a result of high political will and consideration to road
safety in France (it was one of the Governments key priorities in 2002). The reason for
the poor performance in UK in terms of alcohol enforcement (indicator: X12) is due
probably to the reason that still no random breath testing has been introduced and
applied in the U.K (ETSC, 2006). Between 2000 and 2002, there was an increase in the
number of alcohol-related traffic deaths in Sweden (Laurell, 2003). This unacceptable
performance has led Sweden to take extra countermeasures (e.g. enforcement among
others), which may explain the high performance in this area in Sweden (X12).
However, the enforcement of speed (X13) seems to be low in Sweden, which means (if
results of opinions are correct) there is a sign of failure in sustainable development in
road safety and this should be considered. Furthermore, motorcycles are a major
component of the traffic fleet in Italy, while these are a minority in North Europe. This
can explain the performance scores of the indicator X8.
Consequently, there is a considerable difference between the performance in each pillar
of RSDI and countries. As can be seen in Figure 8.2, taking simple average as a method,
Sweden is the best performing country in terms of product performance and road
users behaviour, while Netherlands has higher levels with respect to System
performance. This illustration will be useful for each country to see in which areas it is
performing well and what do they need to keep progressing, and in which areas still
need more addressing in the future. As can also be seen from Figure 8.2 South European
countries generally perform lower than Northern European countries in relation to RSDI
performance pillars.

117

Product performance
100

75

50

25

Human behaviour
performance

System performance

Sweden

Netherlands

Germany

UK

Product performance
100

75

50

25

Human behaviour
performance

System performance

France

Italy

Spain

Belgium

Figure 8.2: An illustration of the differences between various pillars

of road safety development in some HMCs (in 2003).

In a similar discussion, an example of how to illustrate the dimensions of RSDI results

(simple average method) in LMCs is shown in Figure 8.3. As can be seen, Malaysia is
the best performing country in terms of traffic risk, road users behaviour, HDI and
safer vehicles. Thailand has higher levels with respect to road situation. Not
surprisingly, Cambodia scored number one in personal risk level because (as was
discussed earlier in Figure 2.2) a country at an early level of motorisation tends to have
a growing number of traffic-related deaths (traffic risk), but with low growth in the
number of population-related deaths (personal risk). Again, this illustration will be
useful for country to see in which areas it is performing well, and which areas still need
actions to be taken.

118

Traffic risk
100
75

HDI

50

Personal risk

25
0

Safer people

Safer vehicles

Safer roads

Cambodia

Laos

Malaysia

Thailand

Vietnam

Figure 8.3: An illustration of the differences between various dimensions

of road safety in some LMCs (in 2003).

8.1.2

Can RSDI indicators and results be generalised to all countries?

The research to date has been generated in small sample size of countries (five LMCs
and eight HMCs) and it is still questionable whether it is possible to apply the same set
of performance indicators used in this study to other LMCs (e.g. in Africa or Latin
America) or to other HMCs (e.g. USA, Japan, Australia). As discussed earlier in Section
4.9, there is no definitive list of performance indicators suitable for all countries. RSDI
must be tailored to most countries in each LMCs and HMCs according to the
availability and reliability of data. However, further applications and research is
therefore needed in this respect.
Fortunately, many researches and initiatives in different countries have moved from
national level to initiatives (databases, applications and research projects) at regional
level (EU, ASEAN, OECD, Latin America). This would simplify the work of RSDI,
with regard to data sources and accessibility, by looking to regional databases instead of
national ones. As one might know that the regional sources use similar definitions for
the same indicators and for all countries in the region. However, to accomplish the
international vision and ambition of RSDI, all regions must provide the required and
same set of information/data to RSDI.
Additionally, care must be taken when generalising the RSDI results obtained in this
study to other countries because this study represents a small sample of countries and
perhaps the results will not be the same for a large sample of countries and from

119

different regions. For instance, we cannot generalise the same results (weights) of RSDI
in Southeast Asia to the whole LMCs, and similarly when we talk about North Europe
and South European countries.
8.1.3

Can we combine the results from the two empirical studies together?

There is a number of composite indices that have been applied differently to countries
according to their development. As an example Human Poverty Index (HPI), which is
produced by the United Nations, measured two indices HPI-1 and HPI-2 for developing
and developed countries respectively.
The RSDI was calculated differently for countries that are considered LMCs and
countries considered being HMCs. The RSDI (for LMCs) is comprised of 7 weighted
indicators, while the RSDI (for HMCs) is comprised of 13 weighted indicators.
Therefore, we have to look to RSDI scores differently between LMCs and HMCs. For
instance, Malaysia has scored higher than Italy, Spain and Belgium. However, this
result should not be taken seriously because the indicators used were not the same. For
better judgement, countries have to take the same exam (given the same indicators and
weights).
However, on a global scale, the RSDI between these two groups of countries (LMCs,
HMCs) is likely to overlap. This can happen, for instance, because many LMCs have
low personal risk (higher population density) and low or zero alcohol drinking and
driving. Besides, some LMCs have experienced an increased level in motorisation with
growing investment and development in road safety (especially in rich LMCs). If this is
the case of overlapping between the results of RSDI in both LMCs and HMCs, Figure
8.4 shows one possible way to illustrate this.

RSDI
100
Country D

HMCs

world class in
road safety

Country C
average

40-50
Country B

LMCs

Country A
0
Time
Figure 8.4. RSDI performance development

120

8.1.4

Which approach (method) of RSDI is the best?

In fact, each method used for RSDI calculation has its advantages and disadvantages
that can complement each other in terms of their complexity, weightings,
subjective/objective method and transparency. A summary of the general strengths and
weaknesses of the four approaches is given in Table 8.2. The summary is based on the
general methodological principles that have been discussed in earlier sections as well as
in Chapter 5.
As discussed earlier, the choice of any method is crucial and any change of method in
calculating the index may change the weights and the final results of the index. It is not
easy to say which method has given the best combinations of the indicators and weights
or to say that the objective approaches were in some way better than subjective
approaches. However, by looking to the obtained RSDI results from the two empirical
studies, while paying attention to the limitations of the methods shown in Table 8.2, we
may conclude:
The PCA method makes the determination of weights less subjective and with better
transparency. However it did not work reasonably well in constructing RSDI for this
study and data. Of course, it makes the determination of weights less subjective and
with better transparency. However the results achieved in the end did not tell the actual
performance of indicators towards the target values between countries. This method
deals appropriately with the correlation between indicators, however, in this study; the
inter-correlations between the indicators were not high with each other. Therefore the
number of Principal Components obtained for the study regarding HMCs was relatively
high (5 components). Another observation is that the PCA method has widened the
scores of the total RSDI in countries compared to other methods. Therefore, we may say
that RSDI is quite sensitive to the change of weights and methods of calculations by
using the PCA method. However, one cannot assess fully the usefulness of the PCA
method because it is sensitive to small sample of countries, which maybe it gives better
results if we apply it to bigger sample of countries. Therefore, we must be cautious in
interpreting the final results. Another disadvantage to the PCA method is that it requires
special skills to calculate and understand the obtained results. This can perhaps
contradict with the main aim of RSDI for being a simple and easily understandable tool
to everyone.
The simple average weighting method might have been more efficient for constructing
RSDI. The results seem satisfactory and they represented the actual performance
between indicators (towards target). It made the determination of weights less subjective

121

and it was also good for transparency. However, making all combined indicators equally
important is not true in real-world applications of road safety.
The choice of weights by using the Subjective weights was subjective in the
beginning (based mainly on RSDI philosophy and previous experience), however, this
method has given satisfactory results and better representation of the actual performance
of indicators.
To sum up, there was choice of possible approaches; however none of them is
definitively the right one. This will require further application. However, attaching
different weights to different indicators subjectively can be one suitable solution for
RSDI building and this method has much to recommend it
Table 8.2: Summary of the strengths and weaknesses in the four approaches
Approaches

Using simple
average

Strengths

Limitations

Relatively objective method.

Widely used (i.e. HDI).
Tests are already developed
internationally.
Simple and quick method.
Good for transparency.
Good for validity.

Different weights.
Represent the opinions of experts in
the field.
Relatively simple.

Expert
judgements

Subjective
weights
based on
previous
experience

Principal
Components
Analysis
(PCA)

Different weights.
Inexpensive method.
Relatively good for transparency.

Different weights.
Relatively objective method.
Good for transparency.
Deals well when the original indicators
are highly correlated.
Deals well with double counting.

122

All indicators have the same importance

(equal weights).
Dominated by dimensions those have
included the most indicators.

Subjective method.
Lack of transparency.
The scale of weighting is not precise
enough.
Difficult to obtain the opinions of experts
regularly and frequently.
Subjective method.
Choice of weights is up to authors.
Finding weights to some indicators is
limited (based on literature).
Relatively complex method (in terms of
finding the most priorities/problems in
the countries/regions).
Sensitive to missing data.
Relatively complex method. It requires
special skills.
More than one principal component is
often identified. We have to take further
step to convert them into RSDI.
The chosen PCs explain most but not all
of the variation in the original data.
Some indicators maybe perform better in
one PC than another.
Sensitive to small sample of countries.
Lack of interpretive meaning in some
cases.

8.2

Theoretical Assessment: What are the RSDIs Strengths, Weaknesses,

Opportunities and Threats?

To summarise, assess and provide an overall picture of the prospects for a success when
using RSDI, a more detailed analysis has to be carried out about its positive and
negative sides. A useful tool for such an analysis is the so-called SWOT model, which
is an acronym for Strengths, Weaknesses, Opportunities, and Threats. This model is a
good framework used for assessing the position and direction of a company or any
particular project or idea that is challenging an organisation in terms of its internal
dimensions (Strengths and Weaknesses) and the external dimensions (Opportunities and
Threats). A review of SWOT method can be found in Manktelow (2007).
Four main questions were developed to guide the SWOT study regarding RSDI:
1.
2.
3.
4.

What are the RSDIs advantages?

Which weaknesses must be minimised or improved to reach our long-term goal?
What are the good chances (opportunities) that might move the RSDI forward?
What are the obstacles (threats) that might hold the RSDI back?

Carrying out an analysis using the SWOT framework will enable us to maximise RSDI
advantages from strengths and opportunities, and to understand how to avoid or
minimise weaknesses and the expected threats in future.
What are the RSDIs strengths?

Facilitate international benchmarking.

Increase the awareness of the problem among both public and policy makers.
Lower details of data as RSDI summarises a lot of information about road safety
into a single value.
Broader picture of the whole road safety situation in a country compared to the
traditional models in road safety.
Simple and quick format.
Easier to understand and communicate with public.
A guide to identifying areas for development.
Monitor improvements and change.
Can engage the key stakeholders and NGOs in the process of promoting road safety
in the country.
Expertise available to develop RSDI.
Ongoing demand in road safety benchmarking internationally.

123

What are the RSDIs weaknesses?

No final selection of key standard list of indicators.

Insufficient data available for every key performance indicator.
Definitions vary from one country to another.
Time series data is often short.
Lack of understanding the linkage between key practices and product.
Large amount of information hidden behind the RSDI record, which should be
supported with illustrations when necessary.
Subjective selection of indicators.
It is hard to develop quantifiable indicators for some areas and it may need
qualitative judgement (e.g. organisational performance).
Can be subject to misuse or misleading if policy makers rely only on the indicators
given in RSDI and ignore some other important information.
Methodologies shortcomings (the advantages and disadvantages of methodologies
were given in previous Section).
Transparency question regarding the selection of indicators and weights.

What are the RSDIs opportunities?

International interest in road safety will be growing as road accidents will become
the worlds third leading cause of death by the year 2020.
Data collection is rapidly growing and there are opportunities to obtain more data
and better information.
Fast improving in computer databases which can facilitate the accessibility and
analysis to a large road safety data from different countries.
Lessons learned from other research (e.g. development of other composite indices)
Good collaboration with international bodies, NGOs and other partners.
Increased expansion of projects that can support each dimension of RSDI (e.g. EuroRap, Euro NCAP).
Planning with international projects. Not currently linked to any but the opportunity
is there.
Ongoing interest in RSDI, which has been shown by international projects such for
instance (SafetyNet, 2005b)19 in EU and (Planzer, 2005)20 in Latin America and the
Caribbean (more examples were given in Section 1.6).

19 The SafetyNet project is a large research project of the EU, funded by the European Commission.
The study referred to RSDI, see (SafetyNet, 2005).
20 This is a large comparative study in road safety project in Latin America and the Caribbean, which is
sponsored by the United Nations Economic Commission for Latin America and the Caribbean. The study
referred to RSDI and described its methodology (translated to Spanish), see (Planzer, 2005).

124

What are the RSDIs threats?

Lack of data.
Limited coverage and marketing of RSDI.
Lack of lead international agency to run.
New competitors (similar benchmarking models or tools).

After completing the SWOT analysis, an action plan in future work has to be set up for
each point to overcome the identified weaknesses and threats.
8.3

Summary

The empirical assessment can conclude that the varieties of road safety in many
indicators between countries could be explained by differences in one single RSDI
scores.
There was no final choice of possible approaches for RSDI building. This will require
further applications. However, the assessment of this study suggests the choice of
weights by using the subjective weights method as it has given satisfactory results and
better representation of the actual performance of indicators by the end.
From the theoretical assessment it can be concluded that there are currently challenges
and limitations to build such a multi-dimensional index, but in spite of this it can be
assumed that most of these problems can be solved over time. The threats, in general are
negligible compared to the opportunities of having such a model for both national and
international road safety purposes.

125

126

Chapter 9: Conclusions and future work

Chapter 9

Conclusions and future work

This chapter starts with a short summary of the research approach used
to investigate the main overall goal of the thesis and research questions
that was stated in Chapter 1. This is followed by discussion and
conclusions whether the dissertation has successfully answered these
research questions. Based on this discussion, overall conclusions of this
research are outlined. Lastly, the dissertation suggests some directions
for further research in relation to the topic of this research.

There is a considerable amount of literature that has highlighted the importance and
usefulness of comparing achievements and practices in road safety between countries.
The progress in road safety in any country needs a reliable and standard measurement to
rely on its comparisons and assessments. Two main problems have directed an interest
in undertaking this dissertation research:
1. The lack of good and comprehensive measurement in road safety for
benchmarking nationally or internationally, which can be of interest for a wide
range of research purposes in road safety. The current traditional measurements
used in benchmarking are not powerful enough to model the complexity of road
safety situation in a country, where they focused on few indicators (e.g. the
product of road safety in terms of accident death rates). However, road safety
performance in a country needs to be placed and understood in a broader context
in terms of key practices and outcomes (e.g. health, education, road user
behaviour, enforcement, engineering, etc.).
2. The lack of simple, easy and clear understandable measurement that can be
used as a policymakers analytic tool for showing road safety level in their
countries and enable them to take appropriate decisions, setting targets and
priorities for the future.

127

Therefore, the main overall goal of this dissertation was (as stated in Chapter 1),
specifically:
To create an international benchmarking model that indicates and communicates in
a comprehensive and easy way the severity of the road safety situation in a specific
country and/or in comparison to other countries in time.
However, to accomplish this goal, there was found a second target, which is less
apparent but also of great importance. This target concerns the type and criteria
selection of macro performance indicators that will be included in this benchmarking
model.
Therefore this dissertation has attempted to research road safety performance at the
national scale, factors that influence it (performance), and to gather all comparable and
meaningful macro-measurements into a multidimensional index, called Road Safety
Development Index (RSDI). In Chapter 4, several definitions to RSDI were given
according to its conceptual criteria, methodological approaches, process or to the
relation between its components. These definitions can complement to each other. For
instance, RSDI was defined according to its process as follows:
The RSDI links the key practices that affect the overall outcome to each other and to
the end-results (accident death rates). This will ensure that countries are
benchmarked on the effectiveness of their outcomes and experiences in road safety.
To narrow the research further, this dissertation specifically addressed seven research
questions, as follows:
1. What are the most commonly used benchmarking models for road safety?
2. What are the most commonly used performance indicators for benchmarking
road safety internationally?
3. What are the key performance indicators in road safety that can be applied
uniformly for most countries?
4. What are the knowledge, criteria and methodologies that must be addressed
when aggregating the indicators into one single index?
5. Do death rates have to be replaced with the new index, or should it be a
supplementary part, or be part of the new index?
6. How can the new index be applied internationally for a sample or more of
countries?
7. How can the applicability (usefulness) of this index be checked and evaluated?

128

The results described and discussed within the dissertation are summarised in the
following sections.
9.1

Summary of research approach

To accomplish these goals and preliminary research questions, a research approach was
conducted into four stages: theory development, model building, empirical studies and
model assessment. Due to the absence of earlier models and methods in road safety in
the area of international aggregated tools, the thesis has borrowed and translated
knowledge and theories from other sectors of science (e.g. international composite
indices, business management models) for the purpose of this research and RSDI
building.
9.2

Conclusions about each research question

The following findings of this dissertation were reached and are presented in terms of
the research questions.
The first research question is:
What are the most commonly used benchmarking models for road safety?
Many answers, in literature reviews, have been given to questions related to
benchmarking and performance in road safety where they have strengthened the work in
the area of benchmarking and performance indicators in terms, for instance, of
availability, quality and standards of data. However, there are two main issues that have
not attracted much attention yet, first is the need of an aggregated tool for benchmarking
in road safety, and second the need of a clear selection of key performance indicators
that can be used internationally. This dissertation has divided the reviewed
benchmarking models into four types:
1. Product Benchmarking is used to compare death rates and motorisation to either
cross-sectional data or time-series data.
2. Practices Benchmarking is used to compare activities related to human-vehicleroad performance (e.g. seat belts use, crash helmets use, motorways level, etc.)
3. Strategic Benchmarking is used to compare National Road Safety Programme
(NRSP), management and organisational framework.
4. Integrated Benchmarking is used to compare countries in terms of the three
previous types of benchmarking altogether.

129

The results obtained in this study have proved the need to have an integrated
measurement for benchmarking and assessing the overall performance in a country and
against other countries.
The second research question is:
What are the most commonly used performance indicators for
benchmarking road safety internationally?
Attempting to explore this research question was a more difficult task. The research
approach, for the sake of simplicity and transparency, was to let the literature results
speak for the selection of macro-performance indicators. In this way, the approach was
divided into three stages. First stage has attempted to find the most important indicators
in relation to safety (in terms of risk mainly, exposure and consequences) with
support from the literature. In the second stage, indicators were grouped into different
dimensions and two main categories: direct measures in how they (indicators and
dimensions) illustrate road safety level and development in a country, and indirect
measures in how they offer a significant affect on the overall road safety performance.
The third stage investigated the criteria that have to be used during the selection process
of indicators and dimensions. It was clear that the chosen indicators must be easy,
meaningful, available, measurable, and comparable worldwide and be able to cover
future data. This has allowed this research to determine and classify the key macro
performance indicators in road safety into some sort of a master-list. However, the
selection of the key macro-indicators from this work was not final and they will
continue to be developed and refined.
The findings of this research have not only suggested which indicators of performance
in road safety might be important, but also suggested how one might collect the type of
interesting data, based on special criteria, and use it. The results have also shown that
the obtained master-list (long-list) of performance indicators can be useful for different
purposes of international benchmarking in road safety as well as for RSDI building and
analysis.
The third research question is:
What are the key performance indicators in road safety that can be applied
uniformly for most countries?
The most appropriate indicators from the master-list with high data availability and
acceptable quality were selected and linked together into a short key list, which was
useful for the construction of RSDI as well. The master list was mainly based on
theoretical discussion and literature review, while the key list is a refined table, which

130

built on the basis of criteria (high availability and acceptable quality in the country), the
empirical case studies, and the conceptual framework of RSDI. Therefore the key list
contributed more to the model building of RSDI.
As shown in this dissertation, the key-list has distinguished between what indicators we
wish to have in RSDI and what we have actually been able to collect/include. The study
argued that measuring the right indicators is much better than measuring the wrong ones
even they have better data-availability and quality. In this respect, RSDI would become
available for possible real-world applications, valid and accessible whenever data is
needed.
One possible way to allow the key list be used internationally, as examined in the thesis,
is to identify two short-term lists of key performance indicators. One list could be
applied uniformly for most countries in LMCs and a second one, much more detailed,
for most countries in HMCs.
The fourth research question is:
What are the knowledge, criteria and methodologies that must be addressed
when aggregating the indicators into one single index?
Three different frameworks of RSDI were presented, which support and complement
each other. The first model is a preliminary framework built on the basis of the
integrated benchmarking between all the three types of benchmarking: Product,
Practices and Strategic Benchmarking (presented in Figure 2.1). The second model is a
modified and updated framework built on the argument that the human outcomes (road
user behaviour) play the key role in the development and understanding of road safety
in a country. Thus, there were three pillars of outcomes involved in the second
framework of RSDI (presented in Figure 4.1):

People focus: Examines how successful the country has been in promoting
safety among its people and making people behave safer on roads.
Service/system focus: Examines how effective a country has been in improving
its services and system results in terms of safer cars, safer roads, higher
investment on safety measures, education, better health and rescue services,
better organisational structure, etc.
Product focus: Examines a countrys performance in accident death rates.

The third RSDI framework looked at the differences between performance indicators in
road safety in terms of input factors, process factors, output factors and outcome factors,

131

which represent together the overall road safety performance in any country (presented
in Figure 4.2).
Additionally, the dissertation explored the existing knowledge, theories and method in
other sectors of science for building a multidimensional index from two directions:
The first direction explored the knowledge of composite indices for giving a better
understanding of the methods needed for RSDI building. Four approaches of composite
indices (normalisation, weighting and aggregation) were considered in this study. Two
objective methods are the simple equal average method and principal components
analysis (PCA), and two subjective methods are assessment technique from experts
opinion and assessment technique from literature and theory review. The PCA
method is a statistically based option, the simple equal average method used equal
weights for all indicators and countries, whereas the other methods gave different
weights depending on the importance of indicators.
The second direction looked to the business management (excellence) models to
understand further the principles/concepts (pillars of the model: e.g. product-processsystem) and how these concepts join together into the multidimensional model.
Business, in general, recognised the importance of having a multi-dimensional index for
managing, assessing and controlling the performance in the organisation. Improved total
quality in operations and results has led to sustainable improvements in profits and
customer satisfaction in the organisation. Similarly in road safety science, the study
believed that any improvement in good practices and quality results (accident death
rates) would lead to a sustainable progress in road safety as well.
This research has not only provided a strong evidence of the necessity to have such a
tool (RSDI), but also demonstrated the benefits, criteria of RSDI and the required
mechanism in how to design such tool and use it.
The fifth research question is:
Do death rates have to be replaced with the new index, or should it be a
supplementary part, or be part of the overall performance of RSDI?
The dissertation emphasised that the product of safety in terms of accident death rates
may answer partly which country performs better than another, but it will not answer
why one country is better and what actions/measures another country has to take to
improve its performance. Accident death rates focus only on the superficial level
(product) and not on the overall system (internal level) and human assessments (core
level).

132

Furthermore, there is no guarantee that countries with good scores only in the quality of
the product (accident death rates) will not have trouble in the future in terms of number
of deaths and injuries. Maybe there is a low level of death rates (traffic risk and personal
risk) for some particular reasons but there is probably, at the same time, an increasing
risk somewhere else (e.g. enforcement, speed, alcohol and driving, pedestrians, etc.).
Accident death rates reflect only the exposure measures in terms of number of
population and number of vehicles, but they do not reflect the overall continuous
improvement in road safety.
The study has shown that the system quality and human performance are the missing
links in describing and comparing the complexity of road safety situations between
countries. If there is a failure in any area of product-human-system, the development in
a country will fail as well. Thus any successful international benchmarking in road
safety has to link practices to the end results (accident death rates). If RSDI shows poor
performance in any areas, changes can be made before a worsening situation will
actually start. Therefore, RSDI may provide an early, quick and direct warning signal.
The results from the empirical studies have indicated a remarkable difference between
the selected countries even at the same level of motorisation (e.g. France and United
Kingdom) and even at the same level of accident death rates (e.g. Sweden and
Netherlands).
Based on the discussion above, accident death rates have become part of the overall
performance of RSDI.
The sixth research question is:
How can the new index be applied internationally for a sample or more of
countries?
Apparently, it was not possible to collect data from all countries worldwide to examine
the chosen indicators. This needs an enormous amount of data collection and time. A
sample of countries was enough to select and examine for the purpose of this
dissertation.
In order to achieve adequate results of RSDI, international comparisons have to be
carried out between similar countries (regions), as much as possible, at the same level of
development, similar type of transport system and characteristics (causes of accidents
and priorities), and similar levels of motorisation. However, such task was not possible
because, among other things, of the lack of a clear definition of development between
countries. Thus, the study has addressed vehicle ownership (motorisation) as the most

133

appropriate criterion to classify countries internationally. By referring to the key list of

performance indicators, RSDI has included different indicators for countries that are
considered LMCs Less Motorised Countries and countries considered to be HMCs
Highly Motorised Countries.
As such, data set was collected from year 2003 through different sources. This selection
reflected mostly the availability and reliability of data for all the selected countries. The
first empirical study came from eight European countries, while the second empirical
study came from five Southeast Asian countries.
The seventh and final research question is:
How can the applicability (usefulness) of this index be checked and
evaluated?
The goal of this dissertation was to develop, present and test a new international
integrated tool for benchmarking road safety in order first to bridge a gap in the
literature in this area of research and second to offer a simple and comprehensive tool to
policy makers and public to use and communicate. To accomplish this goal, two
empirical studies were carried out to determine the applicability of this tool in real
world applications and from different parts of the world.
In answering this research question, two main conclusions can be drawn, whether RSDI
made a comprehensive tool, on the one hand, and whether RSDI became a simple and
easy tool, on the other hand.
From the empirical assessment of results that made in Chapter 8, it was concluded that
the varieties of road safety in several indicators between countries could be explained
by differences in the final RSDI scores.
From the theoretical assessment (SWOT) also made in Chapter 8, it was concluded that
there are currently challenges and limitations to build RSDI, but in spite of this it was
assumed that most of these problems could be solved over time.
As regards the question whether RSDI became a comprehensive tool, the RSDI results
obtained in this study gave a broader picture of the whole road safety situation in a
country compared to the traditional models, which have shown that RSDI has achieved
this target.
As regards the question whether RSDI became a simple and easy understandable tool,
the results obtained in this study were scored on a scale of 0 to 100, where higher values

134

indicate a higher level of safety (and sustainability as well) in the country. The lower
values indicate the worst performance. Again, the results of the study have shown that
RSDI have achieved this target as well.
9.3

Overall conclusions and future work

This dissertation has started with several ambitious goals. It has met most of its initial
targets and answered the initial research questions.
This dissertation work has provided four major contributions. The first contribution is
the design and development of the RSDI, which can assess and compare the road safety
achievements in different countries in a comprehensive and easily accessible form. The
second contribution is the development of a master list of macro performance indicators
that can be used for different purposes of international benchmarking in road safety. The
third contribution of the thesis is the development of a key list of applicable
performance indicators, categorised into three pillars and nine dimensions. This key list
can be used for assessing road safety in most countries worldwide based on data
availability, quality and importance. The fourth main contribution is an in-depth
analysis of the knowledge and criteria that are required in the selection and aggregation
of performance indicators in road safety, supported by a literature review.
However, despite this progress, there are some future research directions that could
strengthen the obtained results.
The RSDI needs to be examined with bigger samples of countries and from different
parts of the world in order to determine whether the results obtained in this study may
be generalised.
As shown in this dissertation, there was no final choice of aggregation methodology for
RSDI building. Each methodology has its advantages and disadvantages. This will
require further investigations and more empirical applications. However, attaching
different weights to different indicators of RSDI subjectively, based on RSDI
philosophy and previous experience, can be one suitable solution for RSDI building and
this method has much to recommend it.
Today, the problems of missing data and quality exist in different countries and it is
probably that the work on solving these problems will continue in near future.
There are number of interesting benchmarking studies available in different regions,
which have assessed the achievements between different countries regarding one or

135

more dimensions of road safety (e.g. safer roads, safer vehicles, enforcement
effectiveness levels). The results from each study can be translated to the scale of RSDI
(as a sub-index) and contribute separately and directly to the overall goal of RSDI. We
are not yet at that point, but the opportunity is there.
Further work, RSDI can collaborate with international indices for measuring
sustainability and development. For instance, RSDI can work closely with HDI in terms
of the relationship between road accidents and each of the HDI components; these are
the level of income of the country, health level and education. In addition, RSDI can
collaborate with special environmental composite indices towards an overall sustainable
transportation in safety, environment and mobility. Other interesting features of RSDI
include a development of disaggregated indicators across municipalities in the country.
This provides a useful basis for local policy makers to address the problems between
different municipalities, big cities and counties inside the country. As might be known,
municipality (big city) data often has homogenous definitions and similar socioeconomic conditions and levels of mobility, which can simplify the calculation of RSDI.
The prospects using the RSDI approach for big cities were discussed in (Al-Haji & Asp,
2004 and 2006a).
In conclusion, this research hopes to be useful for the national policy makers and the
general public since it will help them to understand the magnitude of the road safety
problem more easily and to draw their attention to this phenomenon. It will also help the
researchers to understand the relationship between different factors at national level and
the overall road safety level in a country. It also hopes the discussion, findings and
results of this research can contribute to the efforts of enhancing the collaboration,
cooperation and technology transfer of road safety between countries and regions across
the globe.

136

References
Adams, J., (1987). Smeeds law: some further thoughts. Journal of Traffic Engineering and Control, 10
(7), pp. 70-73.
ADB Asian Development Bank, (2005). Arrive alive, ASEAN Commits to Cutting Road Deaths
Association of Southeast Asian Nations Regional Road Safety Strategy and Action Plan (20052010).
ADB, Manila.
ADB Asian Development Bank, (1996). Road Safety Guidelines for Asia & Pacific Region, Regional
Technical Assistance Project Report, Asian Development Bank.
Ahmed, P. K., and Rafiq, M. (1998). Integrated benchmarking: A holistic examination of select
techniques for benchmarking analysis. Benchmarking: An International Journal, 5(3), 225242.
Al-Haji, G., Asp, K., (2007). Traffic accidents reduction strategy, best practices from European States.
The International Conference in Traffic & Its Contemporary Issues, Kuwait, May 12-14, 2007.
Al-Haji, G., Asp, K., (2006a). New Tools for Assessing and Monitoring National Road Safety
Development. The 2nd International Road Safety Conference, Dubai - United Arab Emirates, pp.31.
Al-Haji, G., Asp, K., (2006b). The Evolution of International Road Safety Benchmarking Models:
Towards a Road Safety Development Index (RSDI). The International Journal Science & Technology
for Highways, Vol.3, pp.3-9.
Al-Haji, G. (2005). Towards a Road Safety Development Index (RSDI)- Development of an International
Index to Measure Road Safety Performance. Licentiate Thesis No. 1174, Linkping University,
Sweden.
Al-Haji, G., Lindskog, P., (2005). Road Safety in Southeast Asia- Factors Affecting Motorcycle Safety,
The ICTCT international workshop on measures to assess risk in traffic, Campo Grande, Brazil.
Al-Haji, G., Asp, K., (2004). Applying Road Safety Development Index (RSDI) for Big Cities. The 6th
International Conference: Traffic Safety Management for Big Cities, St. Petersburg, Russia, pp. 218222.
Al-Haji, G. (2003). Developing Road Safety Development Index, The ICTCT international workshop on
improving safety by linking research with safety policies and management, Soesterberg, the
Netherlands.
Al-Haji, G., Asp, K., (2002). Road Safety Perspective in Arab Countries- Comparative Study and
Analysis of Progress, SORIC 02 Conference (Safety on Roads), Bahrain, pp. 116-121.
Al-Haji, G., (2001). Traffic Safety in Developing Countries- New Approaches in Technology Transfer by
Using Distance Education Technique. Masters thesis, LITH-ITN-EX-2001:156-SE, Linkping
University, Sweden, pp. 7-23.
Andreassen, D.C., (1985). Linking deaths with vehicles and population. Traffic Engineering and Control,
Vol. 26, pp. 547549.
Andrew, S., (2004). Literature Review of Frameworks for Macro-indicators. Centre for the Study of
Living Standards. Ottawa, Canada.
ASNet database (2005). ASEAN Road Safety
http://www.asnet.org (Last visited: 2006-04-18).

Network,

Linkping

University,

Sweden.

Asp, K., Rumar K., (2001). The Road Safety Profile. The RetsNet News Regional Traffic Safety
Network. No. 1.
Asp, K., Rumar, K., Andersson, J., (1998). Urban Transport in a Road Safety Perspective. VTIUtveckling AB, Sida-Publications, The Swedish International Development Cooperation Agency
(Sida), Sweden.

137

ASSESS (2005). Assessment of the contribution of the ten and other transport policy measures to the
midterm implementation of the White Paper on the European Transport Policy for 2010. European
Commission. Belgium.
Bester, C.J., (2001). Explaining national road fatalities. Accident Analysis and Prevention. Vol. 33, pp.
663672.
Broughton, J., (1991). Forecasting road accident casualties in Great Britain. Accident Analysis and
Prevention, 23 (5), pp. 353362.
Broughton, J., (1988). Predictive Models of Road Accident Fatalities. Traffic Engineering and Control,
May 1988, ISSN:0041-0683, pp. 296-300.
Cameron, M., Diamantopoulou, K., (2000). Index to measure traffic enforcement effectiveness in each
Police Region of Victoria. Monash University Accident Research Centre, Report No. 172.
CARE
Community
Road
Accident
Database,
http://europa.eu.int/comm/transport/home/care/index_en.htm. (Last visited: 2006-04-18).

(2006):

CHEMS Commonwealth Higher Education Management, (1998). Benchmarking in Higher Education:

An International Review. Twente, CHEMS.
Cooper, R. G. (1990). Stage-Gate System: A New Tool for Managing New Products, Business Horizon,
May-June, pp 44-54.
Dahlgaard, J., Kristensen, K., Kanji, G. (1998). Fundamentals of Total Quality Management, Chapman &
Hall, London.
Daniels N, Light D, Caplan R. 2000. Benchmarks of fairness for health care reform: a policy tool for
developing countries, Bulletin of the World Health Organization, vol.78, no.6, pp. 740-750.
Dorsch, J. and Yasin, M. (1998). A framework for benchmarking in the public sector. Literature review
and directions for future research, International Journal of Public Sector Management 11(2/3) 91-115.
Dunn, B., Mathews, S. (2001). The pursuit of excellence is not optional in the voluntary sector, it is
essential, International Journal of Health Care Quality Assurance, Vol. 14 No.3, pp.121-5.
Ecorys and SWOV, (2005). Impact Assessment Road Safety Action Programme- Assessment for mid term
review. European Commission.
Elvik, R., Vaa, T., (2004). The Handbook of Road Safety Measures. Elsevier Amsterdam, ISBN: 0-08044091-6, pp.66 & pp. 676-803.
Eriksson, H. (2004). Organisational value of participating in a quality award process: a Swedish study.
The TQM Magazine, Volume 16, Number 2, pp. 78-92(15).
ESCAPE, (2003). Traffic enforcement in Europe: effects, measures, needs and future. Final report of the
ESCAPE consortium.
ETSC European Transport Safety Council, (2006). Enforcement Monitor, newsletter on traffic law
enforcement in EU, http://www.etsc.be/documents/Enforcement%20Monitor%2005.pdf (Last visited:
2006-04-18).
ETSC, (2005). SECBelt Report: The Safety of Vulnerable Road Users in the Southern, Eastern and
Central European Countries, Brussels.
ETSC, (2004). Road Safety in an enlarged Europe: challenges and opportunities for the 25 EU Member
States. The 6th European Transport Safety Lecture.
ETSC, (2003). Transport Safety Performance in the EU. A Statistical Overview, Brussels.
ETSC, (2001). Transport safety performance indicators. European Transport Safety Council ETSC,
Brussels, pp. 15-42.
ETSC, (1999). Exposure data for travel risk assessment: current practice and future needs in the EU.
European Transport Safety Council, Brussels, Belgium, pp.12.
EFQM European Foundation for Quality Management, (2006), EFQM Excellence Model, EFQM,
Brussels. http://www.efqm.org (Last visited: 2006-04-18).
ERF, (2005), The European Union Road Federation (ERF)- Road Statistics.

138

European Commission, (2001). European transport policy for 2010: time to decide. COM(2001) 370.
White paper of the Commission of the European Communities. Brussels, Belgium.
http://europa.eu.int/comm/energy_transport/library/lb_texte_complet_en.pdf (Last visited: 2006-04-18).
European Commission, (2000). Priorities in EU road Safety; Progress Report and ranking of actions.
EC-COM(2000)125.
EuroNCAP, (2006). European New Car Assessment Programme. Available at: http://www.euroncap.com
(Last visited: 2006-04-18).
EuroRAP, (2006). European Road Assessment Programme. Available at: www.eurorap.org (Last visited:
2006-04-18).
Eurostat, (2005). European Union statistical database- Transport. The Statistical Office of the European
Communities.
Evans, L., (1991). Traffic Safety and the Driver, New York: Van Nostrand Reinhold, P25-43, pp. 60-95.
FIA Foundation for the Automobile and Society, (2004). FIA Press Release, European Road Safety
Charter,
Available
at:
http://www.fia.com/resources/documents/989353331__06_04_2004_Road_Safety_2.pdf (Last visited:
2006-04-18).
Fieldwick, R. and Brown R.J. (1987). The effect of speed limits on road casualties. Traffic Engineering
and Control, Vol. 28, pp 635-640.
Fred, W., (2001). A road safety information system: from concept to implementation. Leidschendam,
SWOV Institute for Road Safety Research, The Netherlands, pp. 7-9.
Frederik, B., (2002). An overview and evaluation of composite indices of development,. Social Indicators
Research 59: 115-151. Netherlands, 115-151.
Frits, B., Jacques, C., Siem, J., Kees, Montfort., (2001). Nonlinear time series aspects of road safety
research. SWOV Institute for Road Safety Research, Leidschendam, Netherlands.
Garber N. J., Gadiraju, R., (1988). Speed Variance and its Influence on Accidents. Foundation for Traffic
Safety, Washington, DC.
Ghobadian, A., Hong, S.W. (1996). Characteristics, benefits and shortcomings of four major quality
awards, International Journal of Quality & Reliability Management, Vol. 13 No.2, pp.10-44.
Globsafe database, (2005). Global road safety database. Some parts of this database are available on-line:
http://www.globesafe.org/ (Last visited: 2006-04-18).
Haight, F., (1983). Traffic Safety in Developing Countries, Journal of Safety Research, Vol. 14, No. 1,
pp. 1-12.
Hakim, S., Shefer, D., Hakkert, A.S., Hocherman, I. (1991). A Critical Review of Macro Models for Road
Accidents. Accident Analysis & Prevention, Vol. 23, No. 5, pp. 379-400.
Hakkert, A.S., and Braimaister, L., (2002). The uses of exposure and risk in road safety studies. SWOV
Institute for Road Safety Research, the Netherlands.
Hakkert, S. & McGann, A. (1996). A Comparative Study of Road Safety in Australian States. Research
Report ARR 278, ARRB Transport Research: Vermont South, Victoria.
Hallowell, R., (1996). The Relationships of Customer Satisfaction, Customer Loyalty, and Profitability:
An Empirical Study. International Journal of Service Industry Management, 7 (4), 27-42.
Hydn, C. (1987). The development of a method for traffic safety evaluation: The Swedish Traffic
Conflicts Technique. Bulletin 70. Institute fr Trafikteknik, LTH, Lund University, Sweden
IRE Innovating Regions in Europe, (2006): available http://www.innovating-regions.org/ (Last visited:
2006-04-18).
IRF International Road Federation, World Road Statistics, different editions and years, Geneva,
Switzerland, and Washington DC, United States.
IRTAD (2006). IRTAD International Road and Traffic Accident Database, different editions and years,
BAST, Germany.
Jacobs G, Aaron-Thomas A, Astrop A., (2000). Estimating global road fatalities. TRL Report 445,
Transport Research Laboratory, 2000, London.

139

Jacobs, G.D., Fouracre, P.R., (1977). Further research on road accident rate in developing countries.
TRRL report LR 270. Transport and Road Research Laboratory, Crowthorne, Berkshire.
Jacobs, G.D., (1982). The potential for road accident reduction in developing countries. Transport
Reviews 2, pp. 213224.
Jacobs, G.D., Hutchinson, P., (1973). A study of accident rates in developing countries. TRRL report LR
546. Transport and Road Research Laboratory, Crowthorne, Berkshire.
John, W., Mcarthur., Jeffery D., (2001). The Growth Competitiveness Index: Measuring Technological
Advancement and the Stages of Development. Global Competitiveness Report 2001-2002, Centre for
International Development at Harvard University, pp. 28-51.
Julia, Salzman., (2003). Methodological Choices Encountered in the Construction of Composite Indices
of Economic and Social Well-Being. Centre for the Study of Living Standards, Ottawa, Canada.
KANJI, G. K., (1998). Measurement of business excellence, Total Quality Management, Vol. 9, No. 7,pp.
633-643.
Koornstra, M.J., (1992). The evolution of road safety and mobility. IATSS (International Association of
Traffic and Safety Sciences), Research, Vol.16, No.2, pp. 129-148.
Koornstra, M.J., Oppe, S., (1992). Predictions of road safety in industrialised countries and Eastern
Europe; an analysis based on models for time series of fatality rates and motorised vehicle kilometres
or amounts of passenger cars. Proceedings of International Conference Road Safety in Europe, VTI,
Linkping, Sweden, pp.49-70.
Koornstra M.J., (1996). The quantifying of road safety developments, Proceedings of the Conference
Road Safety in Europe, VTI conference, Sweden, pp167-186.
Koornstra, M., Lynam, D., Nilsson, G., Noordzij, P., Petterson, H-E., Wegman, F., Wouters, P., (2002).
SUNflower: a comparative study of the development of road safety in Sweden, the United Kingdom, and
the Netherlands. SWOV Institute for Road Safety Research, Leidschendam, the Netherlands, pp 67-127.
Kopits, E. & Cropper, M., (2003). Traffic Fatalities and Economic Growth. The World Bank, Policy
Research Working Paper No. 3035, Washington, DC, World Bank.
Lauer, J. A., Christopher, C. A. K. and Evans, D. B., (2004) World health system performance: the impact
of varying the relative importance of health system goals. Technical report, Health Services Research.
BMC Health Services Research, 4:19.
Laurell, H., (2003). Towards alcohol free roads in Europe. Swedish National Road Administration.
Accessed in 27th February 2006 from: http://www.eurocare.org/btg/conf0604/papers/laurell.pdf (Last
visited: 2006-04-18).
Leaf, W.A., and Preusser, D.F, (1999). Literature Review on Vehicle Travel Speeds and Pedestrian
Injuries.
US
National
Highway,
Traffic
Safety
Administration,
http://www.nhtsa.gov/people/injury/research/pub/hs809012.html (Last visited March 15, 2005).
Linstone, H., Turoff, M (2002). The Delphi Method Techniques and Applications, University of Southern
California. http://www.is.njit.edu/pubs/delphibook/delphibook.pdf (Last visited: 2006-04-18).
Manktelow, J. (2007). SWOT Analysis: Understanding Strengths, Weaknesses, Opportunities and
Threats, Mind Tools Ltd. http://www.mindtools.com/swot.html (Last visited Mrach 10, 2007).
Mekky, A., (1985). Effect of road increase in road motorisation levels on road fatality rates in some rich
developing countries. Accident Analysis and Prevention, 17 2, pp. 101109.
Mendoza, G.A., Macoun, P., Prabhu, R., Sukadri, D., Purnomo, H. Hartanto, H., (1999). Guidelines for
applying multi-criteria analysis to the assessment of criteria and indicators. The Criteria & Indicators
Toolbox Series No. 9. Centre for International Forestry Research. pp. 34-37.
Michael, F., (2003). Composite indicators of country performance: a critical assessment, OECD, Paris.
Minter, A. L., (1987). Road casualties- Improvements by Learning Processes. Journal of the Traffic
Engineering and Control, Feb. 1987, pp. 74-79.
Nabitz, U.W., Klazinga, N.S., (1999). EFQM approach and the Dutch quality award, International
Journal of Health Care Quality Assurance, Vol. 12 No.2, pp.65-70.

140

Nardo, M. M. Saisana, A. Saltelli and S. Tarantola, A. Hoffman and E. Giovannini., (2005). Handbook
On Constructing Composite Indicators: Methodology And User Guide, OECD Statistics Working
Paper.
Navin, F., Bergan, A., Qi , J., (1994). A Fundamental Relationship for Roadway Safety: A Model for
Global Comparisons. Transportation Research Board, Transportation Research Record, 1441,
Washington D.C., pp. 53-60.
Nigel, J., Jonathan, L., and Murray. P., (2003). The usefulness of aggregate indicators in policy making
and evaluation: a discussion with application to eco-efficiency indicators in New Zealand. pp39-42.
NHS National Health Service, (1999). The NHS Performance Assessment Framework, London,
Department of Health.
Noland, R.B., (2003). Medical Treatment and Traffic Fatality Reductions in Industrialised Countries.
Accident Analysis and Prevention, 35 (6), pp.877-883.
OECD Organisation For Economic Cooperation and Development, (2006). Achieving Ambitious Road
Safety Targets. Paris. Organisation for Economic Co-operation and Development OECD. Paris.
OECD, (2003). Road safety: Impact of new technologies. Organisation for Economic Co-operation and
Development OECD. Paris.
OECD, (2002a). Road Safety Performance- Trends and Comparative Analysis. Last accessed in 27th
February 2006 from http://www.oecd.org/dataoecd/56/32/2487308.pdf
OECD, (2002b). Aggregated Environmental Indices: Review of Aggregation Methodologies in Use. Final
Report. Paris: OECD.
OECD, (1997). Road Safety Principles and Models: Review Of Descriptive, Predictive, Risk and Accident
Consequence Models. OCDE/GD(97)153, Paris.
Oppe S., (2001a). International comparisons of road safety using Singular Value Decomposition.
Leidschendam, SWOV Institute for Road Safety Research, The Netherlands, pp. 8-18.
Oppe, S., (2001b). Traffic Safety Development in Poland. SWOV, Leidschendam, Netherlands.
Oppe, S., (1991a). The Development of Traffic and Traffic Safety in Six Developed Countries. Accident
Analysis and Prevention, Vol.23, No.5, pp. 401-412.
Oppe S., (1991b). Development of traffic and traffic safety: global trends and incidental fluctuations.
Accident Analysis and Prevention; Vol.23: pp. 41322.
Oppe S., (1989). Macroscopic models for traffic and traffic safety. Accident Analysis and Prevention,
Vol.21, pp. 225-232.
Page,Y., (2001). A statistical model to compare road mortality in OECD countries. Accident Analysis &
Prevention, 33, pp. 371385.
Peden M, Scurfield R, Sleet D, Mohan D, Hyder AA, Jarawan E, Mathers C, (edi). (2004). The world
report on road traffic injury prevention. WHO, Geneva.
Planzer, R., (2005). Traffic Safety in Latin America and the Caribbean, Actual Situation and Challenges,
UN-ECLAC The United Nations Economic Commission for Latin America and the Caribbean. pp 2728. http://www.eclac.cl/publicaciones/RecursosNaturales/2/LCL2402PE/lcl2402e.pdf (Last visited:
2006-04-18):
Pollitt, C., Martin, C., Joss, R., (1994). Performance management in government: issues and illustrations,
OECD Public Management Occasional Papers, No. 5, OECD, Paris.
Proctor, S., M. Belcher, P. Cook., (2001).
Telford Publishing, London. United Kingdom.

Practical

Road.

Safety

Auditing,

Thomas

Rumar K. (1999). Transport safety visions, targets and strategies: beyond 2000. First European Transport
Safety lecture, Brussels, European Transport Safety Council.
Routio, P. (2006). Arteology or the Science of Artefacts, University of Art and Design Helsinki.
http://www.uiah.fi/projects/metodi/123.htm (Last visited: 2006-04-18).
SARTRE (2005), European Drivers and Road Risk, INRETS: Arceuil, France. http://sartre.inrets.fr (Last
visited: 2006-04-18).

141

SafetyNet, (2005a). Annual Statistical Report 2004, Building the European Road Safety Observatory
Workpackage 1 Task 3 Deliverable No 1. European Commission, SWOV, Netherlands.
SafetyNet, (2005b). State of the art Report on Road Safety Performance Indicators, European
Commission, SWOV, Netherlands, pp 16 and pp80:
http://safetynet.swov.nl/inhoud/Fixed/WP3/sn_wp3_d3p1_stateoftheartreport_v2p0.pdf (Last visited:
2006-04-18).
Saisana, M. and Tarantola, S., (2002). State-of-the-art report on current methodologies and practices for
composite indicator development. European Commission Joint Research Centre. Italy.
Saisana, M., Saltelli, A., Tarantola, S., (2005). Uncertainty and Sensitivity Techniques as Tools for the
Analysis and Validation of Composite Indicators. Journal of the Royal Statistical Society A. 168 (2):117.
Scott, P., (1986). Modelling time series of British road accidents data. Accident Analysis and Prevention,
18: pp 109-117.
Silvak, M., (1983). Society's aggression level as a predictor of traffic fatality rate. Journal of Safety
Research 14, pp. 9399.
Smeed, R.J., (1949). Some statistical aspects of road safety research. Journal of Royal Statistical Society,
Series A 112, pp. 134.
Spendolini, M.J., (1992). The Benchmarking Book, AMACOM, New York, NY.
Sven, D., (1994). The SARTRE 2-tables. Another set of opinions about traffic and traffic. safety of some
European drivers, VTI rapport 403/403A, Sweden.
Tangen, S. (2005). Demystifying Performance and Productivity, International Journal of Productivity
and Performance Management, Vol.54 No. 1.
Tessmer, J., (1999). Comparing International Crash Statistics. Journal of Transportation and Statistics,
pp. 159166.
The European Commission (2004), European Innovation Scoreboard 2004: Comparative analysis of
innovation performance, Commission Staff Working Paper, Brussels, SEC(2004) 1475.
Thoresen, T., Fry, T., Heiman, L. and Cameron, M., (1992). Linking economic activity, road safety
countermeasures and other factors with the Victorian road toll. MUARC Report No. 29. Victoria.
Australia.
Thulin, H., and Nilsson, G., (1994). Road Traffic, exposure, injury risks and injury consequences for
different travel modes and age groups, VTI Report, 390A, Swedish Road and Transport Research
Institute, Linkoping, Sweden.
Timo, E., (1998). Risks, exposures and accident data, VTI Conference 9A, part 6, nr 9A:6, s 85-96,
Linkping, Sweden.
Turner C, McClure R & Pirozzo S. (2004) Injury and risk-taking behaviour a systematic revie,
Accident Analysis and Prevention, 2004 36:93-101.
UNDP, (2005). Human Development Reports. Available at: http://hdr.undp.org/reports/global/2005/ (Last
visited: 2006-02-22).
UNDP, (2004). Human Development Reports. Available at: http://hdr.undp.org/reports/global/2004/ (Last
visited: 2005-05-18).
Van den Bossche, F. and Wets, G., (2003). Macro Models in Traffic Safety and the DRAG Family:
Literature Review. Report RA-2003-08, Diepenbeek, Belgium.
Van Vliet, P. and Schermers, G., (2000). Sustainable Safety: A new approach for road safety in the
Netherlands. Ministry
of
Transport,
Rotterdam,
The
Netherlands. http://www.rwsavv.nl/pls/portal30/docs/1771.PDF (Last visited 2007-03-12).
Vereeck, L. and L. Deben (2003). An International Comparison of the Effectiveness of Traffic Safety
Enforcement Policies, Limburg University, Belgium.
Vokurka, R., Stading, G.L., Brazeal, J.A (2000). Comparative analysis of national and regional quality
awards, Journal of Quality Progress, Vol. 33 No.8, pp.41-9.
Watson, G. (1993). How process benchmarking supports corporate strategy, Planning Review, pp.12-5.

142

WHO, World Health Organisation, (2004). World Report on Road Traffic Injury Prevention, Chapter 3
Risk Factors, Geneva. http://www.who.int/world-health-day/2004/infomaterials/world_report/en/
(Last visited March 10, 2005).
World
Bank,
(2005).
The
World
Development
http://www.worldbank.org/data/ (Last visited: 2005-05-18).

Indicators

(WDI).

For

details:

World Bank, (2003). Traffic Fatalities and Economic Growth. Policy Research Working Paper Series
3035.
WRI., (1995). Environmental Indicators: A Systematic Approach to Measuring and Reporting on
Environmental Policy Performance in the Context of Sustainable Development. World Resources
Institute, Washington D.C.
Zaidel, D. (2000). Traffic Law Enforcement Practices in Sixteen European Countries, WP5, VTT,
Finland.
Zhang, Q., (2001). Quality Dimensions, Perspectives and Practices: A Mapping Analysis, International
Journal of Quality and Reliability Management, Vol. 18, No. 7, pp. 708-721.

143

144

The End

145