Академический Документы
Профессиональный Документы
Культура Документы
Principal Supervisor:
Professor Martin Skitmore
Associate Supervisor:
Professor Tony Sidwell
Research Centre:
School of Urban Development,
Faculty of Built Environment and Engineering
2006
Risk factors leading to cost overrun in the delivery of highway construction projects.
DECLARATION
The work contained in this thesis has not been previously submitted for a degree or diploma
at any other higher education institution. To the best of my knowledge and belief, the thesis
contains no material previously published or written by another person except where due
reference is made.
Signed: ______________________________
Date: ________________________
II
This research is dedicated to my wife, MT, who believed in me as she wrote .....
This is a dream that you thought you would never get the chance to experience, and
here it has been presented to you in a very short time. Maybe, so you cant think about
your decision too long!!! Grab it and know that I am here to support you and know
that I will not only be behind you but around you. I love you and you deserve this great
opportunity to fulfil l your dream (MT, 2002).
III
Risk factors leading to cost overrun in the delivery of highway construction projects.
ACKNOWLEDGMENTS
I would like to express my sincere thank you to the following people, without whose help,
this thesis would not have been possible.
Many thanks go to my principle advisor and supervisor, Professor Martin Skitmore for his
unwavering support and guidance throughout this research. Especially for his ability and
skill in maintaining a focus on client and industry needs in highway construction. I thank
Martin for re-introducing me to statistical theory and its practical applications in research. I
am truly lucky to have him as my advisor and supervisor who guided me through my
research project from the start to the end. Also, I would like to express gratitude to my
associate advisor, Professor Tony Sidwell, for his comments and discussions. As well, I
would like to thank all the staff in the school of Urban Design for their support.
I would like to give a very special thanks to Alan McLennan, Adjunct Professor in
Engineering at the Queensland University of Technology. Allan contributed much of his
valuable personal time in guiding me in the practical directions of the research. I am
particularly thankful for his expert facilitation of the elicitation process that was so important
to my research.
To my mentor, Dr. Mel Silverman of New Jersey, I thank you for the confidence and
determination you have instilled in me during our long friendship. In particular, thank you
for the endless email support and direction you provided during this long journey you
picked just the right times throughout the journey to 'rattle my cage' and so get me refocused
once more.
Most importantly, I would like to thank my wife and best friend MT, for her love and
support. She inspired me to keep on going, even when it became physically and emotionally
tough thank you, my Linda. To my family, I thank you for all your patience, love and
support. To my three special daughters, Michelle, Suzanne and Rebecca, thank you for all
you have done for me.
IV
ABSTRACT
Accurate client budget estimates are critical to the initial decision-to-build process for the
highway construction projects. This decision-to-build point in a project's development is
seen as the international standard for measuring any subsequent cost estimate inaccuracies
involved (National Audit Office/Department of Transport, 1992; World Bank, 1994;
Nijkamp and Ubbels, 1999), with accuracy being defined as the difference between the
initial project estimate at the decision-to-build stage and the real, accounted project cost
determined at the time of project completion. Expressed as a percentage of estimated cost,
this is often termed cost escalation, cost overrun or cost growth, and occurs as a result of
many factors, some of which are related to each other, but all are associated with forms of
risks. The analysis of these risks is often a necessary step for the improvement of any given
estimating system and can be used to diagnose trouble spots and to pinpoint areas where
project estimating accuracy improvement might be obtained.
In this research, highway projects in Queensland, Australia that have suffered significant
cost overrun are analysed. The research seeks to address the gap in the knowledgebase as to
why highway projects overrun their costs. It focuses on understanding how client projects
budgets go wrong, when dealing with project risk.
The foundation for this research is drawn from the post-mortem analysis of highway
projects, each costing in excess of A$1m and whose final total expenditure exceeded budget
by 10% or greater. The research identifies client risk variables which have contributed to
significant cost overrun and then uses factor analysis and also expert elicitation, using
nominal group technique, to establish groups of importance ranked client risks. Stepwise
multivariate regression analysis is then used to investigate any correlation of these risks,
along with project attributes such as highway project type, indexed project cost, geographic
location and project delivery method to the percentage of cost overrun.
The research results indicates a correlation between the reciprocal of project budget size and
percentage cost overrun that can be useful in clients determining more realistic decision-to-
build highway budget estimates when taking into account project size in relation to economy
of scale.
V
Risk factors leading to cost overrun in the delivery of highway construction projects.
CONTENTS
STATEMENT OF ORIGINAL AUTHORSHIP............................................................................. II
ACKNOWLEDGEMENTS .......................................................................................................... IV
ABSTRACT ...............................................................................................................................V
LIST OF FIGURES....................................................................................................................... XI
LIST OF TABLES ....................................................................................................................... XII
LIST APPENDICES ....................................................................................................................XV
CHAPTER 1
Introduction to research project........................................................................................................1
1.1 Background to the Research................................................................................................1
1.2 Research problem ................................................................................................................3
1.3 Research purpose.................................................................................................................7
1.4 Research objectives .............................................................................................................7
1.5 Justification for the research................................................................................................7
1.6 Methodology .......................................................................................................................8
1.7 Outline of the research thesis organisation.......................................................................9
1.8 Delimitations of scope and key assumptions.....................................................................10
1.9 Conclusion.........................................................................................................................11
CHAPTER 2
Literature review into project risks and cost overrun .....................................................................13
2.1 Preamble............................................................................................................................13
2.2 Introduction .......................................................................................................................13
2.3 Definitions .........................................................................................................................14
2.3.1 The Nature of Risk ...............................................................................................17
2.3.2 Risk and uncertainty .............................................................................................17
2.3.3 Risk management .................................................................................................18
2.3.4 Risk evaluation and acceptance............................................................................19
2.3.5 Risk assessment....................................................................................................20
2.3.6 Risk assessment models .......................................................................................20
2.3.7 Risk control processes ..........................................................................................21
2.4 Risk engineering................................................................................................................22
2.5 Qualitative risk assessment................................................................................................23
2.6 Quantitative risk assessment..............................................................................................23
2.6.1 Simulation ............................................................................................................24
2.6.2 Artificial logic applications ..................................................................................25
2.7 Assessment of probabilities and consequences .................................................................25
2.8 Elicitation of risk ...............................................................................................................27
VI
2.8.1 Expert elicitation ..................................................................................................27
2.8.2 Expert elicitation using the Delphi method..........................................................28
2.8.3 Elicitation using semi-structured interviews ........................................................28
2.8.4 Nominal group technique.....................................................................................29
2.9 Organisational risk culture ................................................................................................ 29
2.10 Project risk management ...................................................................................................30
2.11 Project site condition risks ................................................................................................33
2.12 Managing risks through procurement practices ................................................................35
2.13 Risk distribution in project procurement...........................................................................36
2.13.1 Risk allocation through standard-form contracts .................................................39
2.14 Payment methods for contract risks ..................................................................................39
2.15 Delivery processes for projects .........................................................................................40
2.16 The traditional method of project procurement................................................................42
2.17 Improving project delivery methods for risk.....................................................................44
2.17.1 Build-Operate-Turnover project delivery ............................................................45
2.17.2 Relationship/Alliance/Partnering contracting ......................................................46
2.18 Project contracting ............................................................................................................48
2.18.1 Procurement auctions...........................................................................................49
2.18.2 The competitive tendering process.......................................................................49
2.19 Pre-qualification of contractors.........................................................................................50
2.19.1 Research into client pre-qualification practices ...................................................52
2.19.2 Choosing contractors on low bids ........................................................................54
2.20 Bid evaluation ...................................................................................................................56
2.20.1 Research into bidding...........................................................................................57
2.20.2 Sub-contractor selection.......................................................................................58
2.21 Project budget estimating..................................................................................................58
2.21.1 Estimating processes ............................................................................................59
2.21.2 Early project estimates .........................................................................................60
2.21.3 Project cost overrun .............................................................................................60
2.21.4 Research into project estimating ..........................................................................61
2.21.5 Factors influencing costs......................................................................................63
2.21.6 The relation between cost accuracy and scope.....................................................63
2.21.7 Scope change........................................................................................................64
2.22 Cost forecasting models ....................................................................................................65
2.23 Project cost contingency ...................................................................................................68
2.24 Estimating methods for cost contingency .........................................................................70
2.25 Literature summary ...........................................................................................................74
VII
Risk factors leading to cost overrun in the delivery of highway construction projects.
CHAPTER 3
Research methodology ...................................................................................................................77
3.1 Introduction .......................................................................................................................77
3.2 Research strategy...............................................................................................................77
3.3 The research procedure......................................................................................................80
3.4 Research steps ...................................................................................................................81
3.5 Stage 1: Review literature .................................................................................................81
3.6 Stage 2: Establish data source of highway construction projects ......................................82
3.6.1 Indexing of historical road project prices .............................................................84
3.7 Stage 3: (a) Determine cost overrun factors from historic project data.............................85
3.8 Stage 3: (b) Factor analysis using principal component analysis and factor
rotation on cost overrun factors to consolidate data ..........................................................85
3.8.1 Factoring methods ................................................................................................85
3.8.2 Method of factor extraction ..................................................................................86
3.8.3 Sample size for factor analysis .............................................................................86
3.8.4 Type of factor rotation employed .........................................................................87
3.8.5 Number of factors in analysis...............................................................................87
3.8.6 Characteristics of samples in the factor analysis ..................................................87
3.9 Stage 4: Use nominal group technique (NGT) to elicit, review and prioritize
cost overrun risk groupings and highway project types ....................................................88
3.9.1 Focus group ..........................................................................................................88
3.9.2 Delphi technique...................................................................................................88
3.9.3 Nominal group technique .....................................................................................89
3.9.4 Selection of experts for NGT process ..................................................................90
3.9.5 Group composition ...............................................................................................91
3.9.6 Ranking process ...................................................................................................91
3.9.7 NGT workshop evaluation ...................................................................................92
3.10 Stage 5: Undertake data analysis and statistical modelling ...............................................92
3.10.1 Regression analysis basis assumptions.................................................................95
3.10.2 Correlation analysis..............................................................................................96
3.10.3 Statistical regression.............................................................................................96
3.10.4 Sample size of data...............................................................................................97
3.10.5 p-value ..................................................................................................................97
3.10.6 t-test ......................................................................................................................98
3.10.7 Other considerations.............................................................................................99
3.11 Ethical considerations........................................................................................................99
3.12 Conclusion.........................................................................................................................99
VIII
CHAPTER 4
Data collection, analysis techniques and statistics .......................................................................101
4.1 Preamble..........................................................................................................................101
4.2 Stage 2: Establish data sources of highway construction projects ..................................101
4.2.1 Project number and location ..............................................................................102
4.2.2 Description of works..........................................................................................102
4.2.3 Method of project delivery.................................................................................103
4.2.4 Programmed cost................................................................................................105
4.2.5 Actual cost..........................................................................................................105
4.2.6 Percentage (%) ...................................................................................................106
4.2.7 Indexing of project costs to 200203 out-turn prices.........................................106
4.3 Stage 3 (a): Determine cost overrun factors from historic project data .........................107
4.3.1 Exclusion of some initial historical project data ................................................109
4.4 Stage 3 (b): Use factor analysis (principal component analysis) and factor rotation on cost
overrun factors to consolidate data..................................................................................109
4.4.1 Factor analysis test parameters...........................................................................110
4.4.2 Principal component analysis.............................................................................111
4.4.3 Scree plot............................................................................................................113
4.4.4 Factor rotation ....................................................................................................114
4.5 Step 4: Use nominal group technique (NGT) to elicit, review and prioritise principal
cost overrun risk groupings and highway project types .................................................117
4.5.1 Identification and selection of experts ...............................................................117
4.5.2 Agenda Item 1: Research project background ...................................................118
4.5.3 Agenda item 2: Workshop aims/desired outcomes and explanation of
NGT for workshop .............................................................................................118
4.5.4 Agenda Item 3: Principal cost overrun grouping exercise .................................118
4.5.5. Agenda Item 4: Ranking of risk groupings ........................................................121
4.5.6. Agenda Item 5 of workshop: Development of highway project types...............125
4.5.7 NGT Workshop Agenda Item 6: Workshop feedback and assessment..............127
4.6 Application of NGT workshop findings to project data..................................................129
4.7 Step 4: Undertake data analysis and statistical modelling ..............................................131
4.7.1 Model dependent variable ...............................................................................131
4.7.2 Model prediction variables..............................................................................131
4.7.3 Geographic project type .....................................................................................132
4.7.4 Geographic data and model coding....................................................................132
4.7.5 Highway project construction type reference number .......................................135
4.7.6 Construction type data and model coding ..........................................................135
4.7.7 Project delivery types.........................................................................................136
4.7.8 Project delivery data and model coding .............................................................136
4.7.9 Indexed highway project programmed cost continuous variable.......................137
4.7.10 Project high level risks .......................................................................................138
4.7.11 Project high level risk data and model coding ...................................................138
IX
Risk factors leading to cost overrun in the delivery of highway construction projects.
X
List of Figures
XI
Risk factors leading to cost overrun in the delivery of highway construction projects.
List of Tables
Table 1.1: Highway projects over A$1m that exceeded estimate by >10%.................................5
Table 1.2: Research procedures ...................................................................................................9
Table 2.1: Methods for determining risk acceptance .................................................................19
Table 2.2: Quantitative risk assessment methods.......................................................................21
Table 2.3: Qualitative risk assessment methods.........................................................................21
Table 2.4: Sources of common project construction risks .........................................................32
Table 2.5: Target accuracy of highway design estimates...........................................................62
Table 2.6: Classes of project estimates ......................................................................................69
Table 2.7: Contingency ranges in highway estimates ................................................................70
Table 3.1: Situations for differing research strategies................................................................79
Table 3.2: Research procedures .................................................................................................81
Table 3.4: Model techniques ......................................................................................................92
Table 3.5: Interpretation of p-values ..........................................................................................98
Table 4.1: Number of projects analysed...................................................................................101
Table 4.2: Sample of data of completed highway ....................................................................102
Table 4.3: Refined listing of highway project work types .......................................................103
Table 4.4: Composition of project delivery methods ...............................................................104
Table 4.5 Project delivery code................................................................................................104
Table 4.6: Programmed costs of projects .................................................................................105
Table 4.7: Actual costs of projects ...........................................................................................106
Table 4.8: Percentage of cost overrun for analysis periods......................................................106
Table 4.9: RICI indices and factors applied to project costs....................................................107
Table 4.10: Project cost overrun factors derived from historic highway data..........................109
Table 4.11: Sample display of project and cost overrun matrix data used in analysis.............110
Table 4.12: KMO and Bartlet tests...........................................................................................110
Table 4.13: Initial and extraction communalities for cost overrun variables ...........................111
Table 4.14: Total variance explained unrotated ....................................................................112
Table 4.15: Rotated and ordered component matrix ................................................................115
Table 4.16: Principal factor groupings from rotated component matrix ..................................116
Table 4.17: Expert group composition .....................................................................................117
Table 4.18: Experience profile of expert group membership...................................................118
Table 4.19: Cost overrun risk factor grouping worksheet........................................................120
XII
Table 4.20: Agreed mapping of cost overrun factors to groupings..........................................121
Table 4.21: Ranked groupings derived from importance index...............................................122
Table 4.22: Final NGT highway project construction types....................................................125
Table 4.23: Desired workshop project construction types for future grouping .......................126
Table 4.24: Workshop evaluation for Question 1 ....................................................................127
Table 4.25: Workshop evaluation for Question 2 ....................................................................127
Table 4.26: Workshop evaluation for Question 3 ....................................................................128
Table 4.27: Workshop evaluation for Question 4 ....................................................................128
Table 4.28: Sample of low level to high level risk group mapping to projects .......................129
Table 4.29: Incidences across projects for the HL/ risk groups...............................................130
Table 4.30: Split of rural and urban geographic types.............................................................134
Table 4.31: Data coding for geographic types .........................................................................134
Table 4.32: Construction project type reference number.........................................................135
Table 4.33: Sample data coding for only project types 1 to 6 .................................................136
Table 4.34: Project delivery codes...........................................................................................136
Table 4.35: Sample data coding for project delivery codes .....................................................137
Table 4.36: Sample of indexed programmed cost $m continuous variables............................138
Table 4.37: High level risk groupings codes............................................................................139
Table 4.38 Sample of high level risk grouping codes against projects....................................139
Table 4.39: Case-wise diagnostics...........................................................................................140
Table 4.40: Project outlier details ............................................................................................140
Table 4.41: Variables entered/removed in forward selection mode ........................................142
Table 4.42: Forward selection mode summary using dependent variable of % over cost .......143
Table 4.43: Coefficients for forward selection mode ..............................................................143
Table 4.44: Excluded variables in forward selection mode .....................................................145
Table 4.45: Variables entered/removed in stepwise selection mode .......................................146
Table 4.46 Summary of multivariate linear regression for three modes..................................147
Table 4.47: Sensitivity testing of coefficients for secondary stepwise selection mode ...........148
Table 4.48: Sensitivity testing for stepwise selection mode summary ....................................149
Table 4.49: Coefficients for stepwise selection mode .............................................................149
Table 4.50: Excluded variables in stepwise selection mode using revised data ......................150
Table 4.51: Residual statistics for stepwise selection mode for revised project data ..............150
Table 4.52: Variables entered/removed in stepwise selection mode .......................................150
Table 4.53: Sample of data using reciprocal transformation ...................................................154
Table 4.54: Model summary using reciprocal indexed prog. cost $m transformed data.........155
XIII
Risk factors leading to cost overrun in the delivery of highway construction projects.
XIV
List of Appendices
Appendix A: Road Input Cost Indices for the period 1995-96 to 2002-03 .......................207
Appendix B: Programmed and actual project costs indexed to year 2003 ........................211
Appendix C: Questionnaire format to establish NGT membership...................................217
Appendix D: Questionnaire containing summarisied NGT responses ..............................223
Appendix E: Client risks in highway project delivery.......................................................233
XV
Risk factors leading to cost overrun in the delivery of highway construction projects.
XVI
CHAPTER 1
Introduction to research project
CHAPTER 1
The delivery of projects is performed using mainly traditional processes that have evolved
from history and the industrial revolution, where specialisation of professional organisations
was the key trend (Pakkala, 2002). This means that architects, engineers, specialty
contractors, and the industry have adopted a segmented rather than an integrated type of
process. In developed countries, the majority of projects have used the project delivery
method of design-bid-build (DBB) (Gould, 1997), alternatively known as the tradition
method, and typically consisting of discrete feasibility/preliminary design, full design,
construction and operation phases (Fig 1.1).
Feasibility/
preliminary Design Construction Operation
design
The traditional procurement process has been developed to focus on clarity, separation of
phases and provide a transparent and independent bidding stage. The process can be
inefficient and take a long time and often sets up opposing stances between participants,
however the clarity of this process is particularly attractive to clients who need to
demonstrate probity (Sidwell et al., 2002).
Although lack of cost control in all of the project development phases can contribute to cost
control problems, of particular interest is the time the client makes the decision to build
(Hester et al., 1991). In the traditional method, this is often made towards the end of the
design phase. Accurate budget estimates are critical to the initial decision-to-build process
1
Risk factors leading to cost overrun in the delivery of highway construction projects.
for the construction of capital projects (Ward, 1999). These are usually based on a number of
factors such as the complexity of the project, the speed of its construction, the location of the
project and degree of unfamiliarity (Baker et al., 1999). Project cost overrun can be caused
by rising costs from inflation and inadequate analysis of information (Kayode, 1979) and by
costing methods (Akpan and Igwe, 2001).
Engineering designs have a high level of influence on project costs and sometimes
unsatisfactory design performance can lead to cost overrun (Barrie and Paulson, 1992).
There have been few instances where an engineering design is so complete that a project
could be built to the exact specifications contained in the original design documents (Chang
(2002). Many construction problems are due to design defects and can be traced back to the
design process (Bramble and Cipollini, 1998). Design and project specific factors such as
vagueness in scope, design complexity and project size affect the cost estimate of a project
(Akinci and Fischer 1998).
For capital projects, cost estimates are first prepared to enable clients to make reliable
decisions regarding economic feasibility and justification. Early project estimates are often
prepared on limited scope definition and little information regarding the specific parameters
that are needed in the completed facilities (Zeitoun and Oberlender, 1993). As well, they are
often prepared under severe time constraints (Chang (2002). Estimates, even when grossly
inaccurate, often become the basis upon which all future estimates are judged and for the
project team, their performance and overall project success are often measured by the client
by how well the final project cost compares to the initial cost estimate (Hester et al., 1991).
For the project client, accurate cost estimates are vital for business decisions on strategies for
asset development, potential project screening, and resource commitments for existing and
proposed project developments. Accurate estimates are critical to the initial decision-to-build
process for the construction of capital projects (Flyvbjerg et al., 2002).
There are generally four stages involved in developing project cost estimates (Queensland
Department of Main Roads, 2000:
1. Strategic stage: estimate of the cost of undertaking the planning functions required to
produce the options analysis and business case. Usually applies to non-routine
projects where significant cost will be incurred before the proposal is officially
recognised as a project.
2. Concept stage: This provides comparative costs of options prepared as part of the
options analysis. No project cost is produced at this stage. Concept approval gives
authority for the development of the preferred option.
3. Preliminary Design: This develops the total project cost estimate based on the final
design solution but usually before commencement of design, detailing and
documentation.
4. Detailed design (Decision-to-Build): This is prepared on completion of the detailed
design when final plans, specifications and bill of quantities are available. The
detailed design generally forms part of the scheme prototype documents needed for
approval to call tenders and forms the benchmark against which the final cost of
completion is measured.
2
CHAPTER 1
Introduction to research project
While cost estimates become more accurate over time, it is the cost estimate at the time of
making the decision-to-build the project that is of primary interest (Flyvbjerg et al., 2002).
This point in the projects development is seen as the international standard for measuring
the inaccuracy of project cost estimates (National Audit Office Department of Transport,
1992; World Bank, 1994; Nijkamp and Ubbels, 1999). Accuracy is defined as the difference
between the initial project estimate at the decision-to-build stage and the real, accounted
project cost determined at the time of project completion. Expressed as a percentage of
estimated cost, this is often termed cost-escalation, cost overrun or cost growth.
Total project cost estimates include the costs of all component activities from the initiation
of the project proposal to finalisation. These include the cost of developing the concept,
investigations, developing the design, acquiring land, altering public utility plant,
construction, project administration and handover (Queensland Department of Main Roads,
2000).
Estimating methods and the accuracy of project cost estimates are major reason for project
cost changes (Hester et al., 1991). When an inaccurate original estimate is prepared for a
project and is used to compare the actual cost of that project, then there can be a noticeable
difference, referred to as a cost change (Flyvbjerg et al., 2002). Clients assume that routine
changes in projects project will only affect work in the change areas, whereas, in reality, the
effects can extend well beyond specific change areas (Hester et al., 1991). Project cost
accuracy is very important to clients as it enables them to have better cost control over
projects. However, construction projects are notorious for running over budgets (Hester et
al., 1991).
Clients estimating policies usually focus on the preparation of unlikely to be exceeded but
not excessively conservative estimates (Flyvbjerg et al., 2002). In the case of highway client
organisations, this usually means that the estimate prepared at any stage of a project has a
90% confidence factor of not being exceeded at the cost-at-completion (Queensland
Department of Main Roads, 2000).
3
Risk factors leading to cost overrun in the delivery of highway construction projects.
Figure 1.2 shows the required estimate ranges that are normally required by clients in the
stages of project estimates.
High profile highway projects in the US, such as Bostons Central Artery/Tunnel, known as
the Big Dig and Virginias Springfield Interchange have made engineers, contractors and
public taxpayers acutely aware of the problem of cost overrun. For example, the Big Dig was
estimated at a cost of US$2.6 billion (1982 dollars) but was expected to be completed at a
cost of US$14.6 billion (2002 dollars) with completion then anticipated in 2005 (NAS,
2003).
Cost overrun of highway projects in Queensland, Australia, has been significant over a
relatively long period of time. The Queensland Governments Roads Implementation
Program 200405 (Queensland Department of Main Roads, 2005) reports on the proportion
of projects costing more than $1m that have had significant cost overrun by exceeded their
programmed estimate by more than 10%. As can be seen from Table 1.1, cost overrun in
highway projects have had a serious impact on program budgeting from the view of the
client. On average, 1 in 10 highway projects of => $1m has significantly exceeding their
decision-to-build budget (Queensland Department of Main Roads, 2005).
4
CHAPTER 1
Introduction to research project
Table 1.1: Highway projects over A$1m that exceeded estimate by >10%
Planning and programming future highway construction projects are vitally important tasks
in highway organisations (Wang and Chou, 2003). A construction program outlines how
highway funds are to be spent over time and any deviation from the stated program often
brings a quick response from the public, the press and politicians. When this occurs, the
highway organisation loses creditability and time is often taken defending deviation from the
published program (Flyvbjerg et al., 2002). On the other hand, if a highway organisation can
produce realistic program estimates, especially at the decision-to-build stage that it is able to
abide by, then the agency's image can be enhanced.
Project owners, such as highway agencies, are usually engaged in specific types of
construction projects with unique features. For example, highway construction projects are
characterised by their linear complexity, with their greatest risk lying below ground level due
5
Risk factors leading to cost overrun in the delivery of highway construction projects.
to the relatively larger footprint, as compared with say building structures (Halligan et al.,
1987). Project risks can be derived by reviewing historical data and thus ensuring
consideration is given to potential cost overrun (Touran, 2003). Historical data may be used
as a guide; however estimators and project managers also use their experience and
professional judgement to weigh the competing factors to arrive at the most likely value
(Yeo, 1990).
The analysis of project risks is a necessary step for the improvement of any given estimating
system and can be used to diagnose trouble spots and to pinpoint areas where greater
improvement can be obtained (Touran, 2003). Formal and informal databases usually exist
for specific domains, such as in assessing risks or cost overruns in construction projects.
Many of these databases are deemed commercial-in-confidence, particularly in highly
competitive industries (Al-Bahar and Crandell, 1990). In-house historical databases are often
inadequate or disjointed, unavailable, or supplemented with personal knowledge. These
databases are company and/or project specific and are not usually able to be uniformly
applied across new projects (Al-Bahar and Crandell, 1990).
Models have been established showing cost influencing factors derived from past records of
construction costs (Wilmot and Cheng, 2003). Extrapolation of past trends has been used to
forecast future overall construction costs (Koppula, 1981; Hartgen et al., 1997), however
such models are usually only used for short-term forecasting because of their reliance on the
notion that past conditions and specifications are not always retained into the
future.Completed construction project data from the building industry has been used with
regression analyses to forecast the actual cost of projects (Skitmore and Ng, 2003). Models
of cost overrun across processing projects have been derived from quantitative data from
completed projects using factor analysis and multivariate regression (Trost and Oberlender,
2003).
The level of project risk contingency in estimates has a major impact on their financial
outcomes for clients. If contingency is too high it might encourage poor cost management,
cause the project to be uneconomic and aborted, or lock up funds that is not available for
other projects (Dey et al., 1996). On the other hand, if the contingency allocation is too low,
then it may be too rigid and set an unrealistic financial environment, resulting in
unsatisfactory performance outcomes (Touran, 2003). In some areas of the public sector,
there is a tendency to remove contingency provisions in budget submission, as contingencies
are often seen as fats leaving no allowance to express anticipation of any project risk
(Yeo, 1990).
The most common method of allowing for uncertainty in estimating has been the addition of
around 10% contingency to the most likely estimate of the known works (Burger, 2003).
Hartman (2000) argues that this is an unscientific approach and a reason why so many
projects finish over budget. The common method of allocating 10% contingency is regarded
as overly simplistic and heavily dependent on estimators' faith in their own experiences
(Yeo, 1990). The Department of Main Roads Queensland adopts a 10% contingency for all
detailed design estimates unless there are other reasons why they should be adjusted
(Queensland Department of Main Roads, 2000) and as shown in previous Table 1.1 could be
seen as being significantly inadequate to cover the client risks in highway construction.
6
CHAPTER 1
Introduction to research project
The purpose of this research is to establish the nature and extent of cost overruns of
Queensland highway projects and to develop a method for improving budget estimating
practices. The research purpose is also to develop a more definitive risk contingency
allocation regime for overall highway projects that can supersede the arbitrary models
present in most highway projects.
2. How does the amount of highway cost overrun in such highway projects correlate
with their project types, size, delivery processes and client project risks when
historical project data are analysed?
Numerous factors affect project construction costs and most construction cost models
developed in the past have used only a few of the many possible influential factors identified
to date.
7
Risk factors leading to cost overrun in the delivery of highway construction projects.
Research of this type has also been hampered in the past because adequate data has not been
available. However, more data sources of completed projects now appear to be becoming
available, particularly in the public sector. Once identified, post mortem analyses of these for
project risks can ultimately provide the client with more confident final project budgets that
are unlikely to be exceeded.
The interrogation of in-house historical databases is probably the best source of data to
assess risk occurrences or consequences of risk events and in many cases these databases are
inadequate or disjointed, unavailable or supplemented with personal information bias (Al-
Bahar and Crandell, 1990). These databases are company and project specific and may not
necessarily be able to be uniformly applied to new projects.
Research such as that for the Washington State Department of Transportation has identified
risk factors that have strong association with project construction costs, indicating that cost
overruns, expressed as a percentage of the original contract amount, has tended to increase
with the size of the project (Hinze and Walsh, 1997). Williams (2003), in his research into
highway contract overruns, identified the need to study further correlations between different
highway project types and constructed project cost overruns. There is little evidence in the
research to date that has identified such correlations in highway projects in Australia. As
well, many research projects to date consider only the final outcome of contracts within the
project and have not considered the clients risks associated with the full project budget.
That is, the failures that lead to cost overrun in the overall project from the time of the
client's decision-to-build is made until its completion.
Clients require different contingencies for different elements of projects (Eden et al., 2005).
The establishment of a range of contingencies can require a considerable amount of work by
estimators and so a simple contingency across the board is included in order to acknowledge
the difficulty of pinning down project uncertainty (Baccarini, 2004). This research aims at
addressing this issue by providing clients with a cost overrun model which correlates risk
contingency with highway project attributes.
1.6 Methodology
This research adopts an historical analysis as the foundation to the methodology and is an
important approach for presenting information (Kirszner and Mandell, 1992). As well, the
use of historical data assists in providing an insight into current problems relating to cost
overruns in highway project estimates through the examination of what has happened in the
past.
8
CHAPTER 1
Introduction to research project
Five stages form the basis of the research as outlined in Table 1.2 and are designed to
provide answers to the research questions.
4 Use nominal group technique (NGT) to elicit, review and prioritise principal
cost overrun risk groupings and highway project types.
This Chapter 1 begins the background phase of the research by providing the objectives,
proposed methodology overview, scope and project organisation.
Chapter 2 is a literature review from professional journals, books, internet searches and from
interviews with highway construction delivery experts. Chapter 2 essentially provides a
review of the current state of the art in construction risk, contract delivery methods and
project cost estimating and cost control processes used in and its management and cost
estimation of risk. Brief definitions of aspects of construction risks, procurement and the
project management of projects are also contained within Chapter 2.
Chapter 3 discusses the research methodology necessary to achieve the research objectives.
The research method adopted is that of analysing historical data collected on highway
construction projects. The research techniques of factor analysis (FA) using principal
9
Risk factors leading to cost overrun in the delivery of highway construction projects.
component analysis (PCA) and rotation and then nominal group technique (NGT) are
proposed that will consolidate data on the types of construction risks and the types of
highway projects that are associated with client budget cost overruns. Multivariate regression
analysis is then used to identify potential correlations between budget cost overruns and the
various attributes and risks present in projects with excessive cost overrun.
Chapter 4 contains the various data analyses of the historical project information. Factor
analysis and then nominal group technique are used to initially break down the bulk of the
client risks identified from a post mortem of historical project data. The data analysis uses
multivariate regression analysis to investigate the statistical significance of the available data
between documented client project construction risks, highway project types, project size,
location, project delivery methods. These are then correlated against the size of completed
project budget cost overruns.
Chapter 5 discusses the findings from the data analysis carried out in Chapter 4 and develops
conclusions and recommendations that are derived from the research and how the research
objectives align with the findings. Both specific and general recommendations are provided,
as well as recommendations for future research.
Government contracts like the QDMR are typically awarded either through procurement
auctions or are negotiated price contracts with local authorities of internal workforce units.
All the highway projects chosen in this research have contractor pre-qualification so as to
ensure that all project constructors have demonstrated their ability to do similar highway
projects in the past. As well, where appropriate, contractor surety bonding has formed part of
the project delivery process.
Cost growth, cost changes and cost overrun are considered to have the same meaning for the
purpose of this research, and they can be defined as the difference between the final project
cost and the cost estimate at the time of the decision-to-build after design for a particular
project.
The total project cost estimate for the client includes the estimated costs of all component
activities from the initiation of the project proposal to finalisation. These include the cost of:
developing the concept design and business case
conducting investigations and developing the design
detailing the design
acquiring land
altering public utility plant
construction
project administration and project handover.
10
CHAPTER 1
Introduction to research project
The definition adopted for a client or project owner is an individual or organisation for
whom something is built or delivered under some form of contract agreement. The
Queensland Department of Main Roads (QDMR) is the client/owner and is the organisation
for which the highway projects are constructed. It should be noted that some historic project
data utilised in this research may be additionally influenced by long term economic
conditions and contracting climates that may vary from time to time. The variation in such
external project conditions over the seven year analysis period representing the project data
may provide some influence over any correlation of results.
Only highway construction projects are considered in the research. Other transport
infrastructure and building construction are not included. This research study uses QDMR
data for projects completed for the financial years from 1997 through to 2003. The data used
is unique due to the project variables such as project estimate and final costs, project
location, project delivery type, highway project type and reasons for project cost overruns.
All project data used came from published sources within Queensland and all relate to
projects that were initially estimated to cost A$1m or greater and whose final completed cost
substantially exceeded their initial budget (by more that 10%).
1.9 Conclusion
A focus of research effort in the area of project cost overrun will provide a more basic
understanding of the linkages of risk to highway project delivery. The purpose of the
research project is to carry out an empirical analysis in order to ascertain whether the data
support the hypothesis that there is a statistical correlation between highway project cost
overruns and project attributes.
The research concentrates effort on the study of aggregate data about highway construction
projects in order to identify common trends that occur. While the research finds a history of
underestimation of costs and the over-optimistic assumptions about performance of a
substantial number of project budget estimates, the research also specifically contributes to
the body of knowledge in project construction management and cost estimating in the
following ways. It has:
combined risk assessment and expert elicitation techniques in investigating project
management estimating and cost control issues when estimating highway projects
provided a quantitative and qualitative project assessment that will help highway
project decision makers define unforeseeable disturbances more reliably ahead of
time, so that corrective measures can be better taken into account in project design
and estimating
provided an in-depth post mortem analysis of client risks that have led to significant
project cost overrun and thus has strengthened the understanding of why highway
projects can overrun substantially above project decision-to-build budgets.
identified client risk factors that are aimed at simply providing a better understanding
of the risk parameters and management contingency requirements in project, thus
validating better risk contingency weightings in budget estimates
utilised an established expert elicitation technique in the research project in a way
that required minimal impact on the time of construction experts, but at the same
time providing an acceptable outcome to a difficult elicitation requirement
accounted for economy of scale in highway projects by proposing ranges of over and
above % contingencies for varying sizes of highway projects.
11
Risk factors leading to cost overrun in the delivery of highway construction projects.
12
CHAPTER 2
Literature review into project risks and cost overrun
CHAPTER 2
2.1 Preamble
The change in project cost, or cost growth, occurs as a result of many related factors all of
which are associated with some form of risk (Flyvbjerg et al., 2003). Analysis of the reasons
for project cost overrun of construction projects is a necessary step for the improvement of
any given cost estimating system and can be used to pinpoint areas where the greatest
improvement can be obtained. As part of this process, Chapter 2 provides a literature review
of aspects of project delivery risks which often contribute to the potential for cost overrun.
It begins with some basic terms and definitions, followed by a discussion on the nature of
risk its identification and its management, particularly in the construction industry. The
review then looks at the concept of risk engineering, followed by aspects of expert elicitation
of risk probabilities. It then investigates the significance of risk assessment in tendering and
contract administration by using varying forms of project procurement used within the
construction industry. It then goes on to look at the various components that can influence
estimate cost accuracy of projects during project construction.
This literature review then focuses on the various project delivery mechanisms used in
project construction and the influence that they have on minimising and managing project
uncertainty, particularly for clients. Reviewed also are project cost estimating and
procedures that are used in minimising client risk and in reducing project cost overrun.
Project cost estimating and also project contingency, with emphasis on highway construction
projects are also considered.
2.2 Introduction
The procurement of constructed facilities is a process which involves many complex and
interrelated steps incorporating questions of value for money, probity and fitness for
purpose. Over more than three centuries, a procurement process has been developed that
focuses on clarity, separation of phases and a transparent independent bidding stage. This is
often referred to as the traditional project delivery process (Flyvbjerg et al., 2002).
Unfortunately, it has some reported serious drawbacks. For example, the process
management can be inefficient and take an extended time and so often sets up opposing
stances between the project participants which can eventually compromise the measures of
success of a project in terms of time, budget and technical performance (Sidwell et al.,
2002).
On the other hand, the main barriers to achieving project success are the changes in the
project environment (El-Choum, 1994). The problem multiplies with the size of the project
as uncertainty in project outcomes increase (El-Choum, 1994). Large-scale construction
projects are exposed to uncertain environments because of such factors as planning, design
13
Risk factors leading to cost overrun in the delivery of highway construction projects.
and construction complexity. In addition, the presence of various interest groups (such as the
project owner, consultants and contractors) as well as resources (such as materials,
equipment, project funding, climatic, economic and political environment and statutory
regulations) all add to project uncertainty. Other factors contributing to uncertainty include
the complexity of the project, the speed of its construction, the location of the project, and its
degree of unfamiliarity (Ahmed et al., 1999).
One of the major causes of business failures is related to the client. Client-generated risk
factors can be stated as a client's financial ability to meet the cost of the work, its claims
record, changing needs, and the construction sophistication. In turn, these risks can put a
strain on the contractor's cash flow and can increase the actual cost of a project during
construction (Kometa et al., 1996).
2.3 Definitions
The Project Management Institute (Project Management Institute, 2000:21) defines a project
as:
a temporary endeavour undertaken to create a unique product or service: Temporary
meaning that every project has a definite beginning and a definite end: Unique
meaning that the product or service is different in some distinguishing way from all
similar products or services.
Projects generally involve large, expensive, unique or high-risk undertakings that have to be
completed by a certain date, for a certain amount of money, and deliver some expected or
anticipated level of performance. These three criteria of success have become widely used. It
captures the essential task of the project manager, and the essential trade-offs that they can
make. Kohrs and Welngarten (1986:87) report seeing a sign: Good! Fast! Cheap! Pick any
two. This analogy can also be true in the project construction sense, where there is usually a
trade-off of some sort on at least one of these parameters by the project owner.
McCoy (1986) has tried to develop an integrated success criteria based on this three-fold
criterion. At its most simplistic, Avots (1984) suggests that schedule is most important early
in the project, but during the project cost becomes most important and after the project only
technical performance is remembered. Salapatas and Sawle (1986) define success to have
been achieved only when three groups perceive success: the client (based on performance,
budget and reputation), the contractor (based on profitability, reputation, client and public
satisfaction) and the customer/public (based on environment, reliability and cost). Potter
(1987) has found from experience that success and failure can in fact be very close and
Sykes (1982) supports this by pointing out that many large projects have been saved from
disaster only because of fortuitous circumstances.
14
CHAPTER 2
Literature review into project risks and cost overrun
In project development, the project client plays a very important role. The term client refers
to that person or organisation investing in the construction of built facilities (Mak and
Picken, 2000).
Dake (1992) examines the concept of risk from an historical perspective. In the seventeenth
century, risk was defined as the probability of an event occurring, with a focus on either the
losses or gains that the event would represent if it occurred. The interesting aspect of this
early definition is that there was as much attention paid to benefits as to losses. This is a
perspective that has been lost in the twentieth and now twenty-first centuries. In the popular
sense, the term risk carries largely negative connotations of loss or harm that generally has
implications of negative or adverse results from an uncertain event. For example, Fishburn
(1984) calls a certain bad event 'risky' and Statman and Tyebjee (1984) see risk as being a
high probability of failure.
Jaafari (1990) on the other hand, sees risk as being the presence of potential or actual
constraints that could stand in the way of project performance by causing partial or complete
failure during construction and commissioning, or at the time of using the project.
Risk has been defined by Chapman and Ward (1997: 58) as being:
the exposure to the possibility of an economic and financial loss or gain, physical
damage or injury, or delay as a consequence of uncertainty.
By comparison, Young (1996) sees risk in the project management environment as being any
event that could prevent the project realising the expectations of the stakeholders as stated in
the agreed project brief or agreed definition.
15
Risk factors leading to cost overrun in the delivery of highway construction projects.
Dias and Ioannou (1995) conclude that there are two types of risk:
1. Pure risk that exists when there is the possibility of financial loss but no possibility of
financial gain (e.g. physical damages)
2. Speculative risk that involves the possibility of both gains and losses (i.e. financial
and production risk).
Hillson (2002) argues that the common usage of the word risk only centres on the negative
outcomes. Ward and Chapman (2003) argue that risk is often associated with adversity,
things that may go wrong, and threats to projects.
However, Royer also points out that having such a focus of risk on adversity means project
risk management tends to focus more on processes and methods that reduce the effect of
threats.
The general conceptualisation of procurement has been expanded by McDermott (1999) into
the context of project construction. Rowlinson (1999) furthers this systems view of
procurement by including elements such as contract strategy, culture and finance.
Kumaraswamy and Dissanayaka (1998) trace and link the definition of procurement to the
action or process of acquiring or obtaining material, property or services at the operational
level. They also define building procurement as being the amalgam of activities undertaken
by a client to obtain a building.
Construction procurement, on the other hand, is defined as being the framework within
which construction is brought about, acquired or obtained (Kumaraswamy and Dissanayaka,
1998). A high degree of specialisation has evolved in the procurement of the various goods,
materials and services in the construction industry. This has led to a network of supply
chains that include multiple layers of sub-contractors and interlinked suppliers.
The next section looks at risk in both the clinical and project sense, as well as at the various
processes for identify and managing risks within projects.
16
CHAPTER 2
Literature review into project risks and cost overrun
While risk is fairly well documented in the literature, the terminology is not consistently
applied across construction, project management, engineering, health and safety,
environment, business and other industries (del Cano and de la Cruz, 2002). Risk can be
classified as voluntary or involuntary, depending on whether or not the events leading to the
risk are under the control of the persons at risk or not (del Cano and de la Cruz, 2002). In the
theoretical sense, Cvethovich and Earle (1992) view risk not as an inherent quality of the
physical world but as a representation of the interaction between physical and psychosocial
characteristics with the assessment of risk involving judgements about what is valued.
Kumamoto and Henley (1996) identify five attributes of risk. These are:
1. Likelihood
2. outcome
3. significance
4. causal scenario
5. population.
Uher (1994) identifies 34 individual risks and categorises them into a single model, referring
to some as activity risks that may affect individual activities, while others were global risks
that were common to all activities. The majority of risks Uher identifies are global risks.
Rutgers and Haley (1997) developed a model that identifies four distinct phases of risks in a
project:
developmental risks technical, commercial/financial feasibility
project economics, permits/authorisation, third-party intervention and political
change
construction risks schedule, cost, performance, design changes, interest rate
escalation, consequential damage, force majeure/country risk, currency changes
operational risks market changes, statutory changes, unrest/strikes, acts of God,
third-party liability etc.
Decisions are concerned with variables which are normally classified as risks or
uncertainties. Risks are unknowns, the probability of the occurrence of which can be
assessed by statistical means (risks are usually insurable). Uncertainties are unknowns, the
probability of the occurrence of which cannot be assessed (uncertainties are uninsurable)
(Chapman and Ward, 1997). It is possible, however, for a decision-maker to assign a
subjective probability to an uncertainty. As knowledge increases in conjunction with the
amount and detail of statistical data, areas of uncertainty are progressively converted to areas
of risk (del Cano and de la Cruz, 2002). The evolution of weather data and associated
forecasting techniques is such an example (Fellows et al., 2002).
Risk links strongly with the uncertainty of the probability and consequence of a risk event.
Ayyub and McCuen (1997) point out that uncertainty has two types of origins non-
cognitive and cognitive. Non-cognitive uncertainty results from physical randomness. This
type of uncertainty is normally dealt with by employing current statistical and probabilistic
17
Risk factors leading to cost overrun in the delivery of highway construction projects.
science (Chapman and Ward, 1997). Cognitive types of uncertainty result from humans
expressing subjective judgements. Blair (1999) uses a fuzzy set theory approach for the
development of costs and schedules for complex engineering systems. Uncertainty always
exists in the modelling and project management of complex construction projects and this
uncertainty is due to the model representing real systems and is also attributed to humans
who express risk in subjective terms.
Raftery (1994) points out that risk and uncertainty characterise situations where the actual
outcome for a particular event or activity is likely to deviate from the estimate or forecast
value. As well, risks exist in projects because of their uniqueness and temporary nature and
can impact on the project contractor and sub-contractors, stakeholders and project owner in a
variety of ways. Leu et al. (2001) point out that during project implementation, many
uncertain variables dynamically affect the project duration and the costs can thus change
accordingly (del Cano and de la Cruz, 2002).
Risk management is the process by which clients and their project managers make decisions
based on data generated in risk assessments. Risk management involves making educated
decisions about different configurations, construction scenarios and operational parameters.
The Australian/New Zealand Standard 4360:1999 defines risk management as a generic
framework for establishing the context, identification, analysis, evaluation, treatment,
monitoring and communication of risk. However, the standard is not prescriptive, but rather
describes the systems and processes required for risk management. Tuohey (2002) points out
that the use of this standard is an integral element of good management practice across an
enterprise and makes the Standard an ideal choice as the framework for project risk
management. It provides a uniform, structured system that can be used across an
organisation and by which clients can better understand risk issues on individual projects.
The management of risk is seen as an essential part of the management of projects. Strategies
for mitigating risks on projects include(Turner, 1999; OGC, 2002).:
reducing the uncertainty associated with the project
avoiding the risk by finding a different way of doing the project
abandoning the project
reducing the likelihood of the risk occurring or the impact on the project
transferring the risk to other parties such as contractors or insurance companies
accepting the risk and creating a contingency plan
Risk management is also dependent on numerous factors such as industry sector, the size of
the project, and the stage of the project life cycle (Baker et al., 1999). To include some of
these ideas for the management of risk, the risk register is often been seen as the starting
point (Williams, 1993a). Also, Williams (1993b) provides a discussion of how the risk
register can assist in the allocation of risk and the preparation of risk management plans. The
use of project risk registers is often seen as an important step in the reuse of historical project
information (del Cano and de la Cruz, 2002). They can be seen as repositories of personal
knowledge or organisational memories where experiences about risks and responses are
continuously recorded (Tah and Carr, 2001).
18
CHAPTER 2
Literature review into project risks and cost overrun
However, Williams et al., (1997) point out that the project risk register fails to capture the
inter-relationships between risks and the systemic structure within the risks. This makes it an
inadequate tool for the capture and representation of risks, and therefore they question the
basis for analysis and decision making.
Dey (2001) has identified the following as general benefits that can be achieved from the
application of risk management in any type of project:
issues of the project are clarified and allowed for right from the start
decisions are supported by thorough analysis of available data
structure and definition of the project are continually and objectively monitored
contingency planning allows controlled and pre-evaluated responses to risks that
materialise
clearer definitions of the specific risk associated with a project
encourages problem-solving innovative solutions to problems within a project
provides a basis for project organisation structure and appropriate responsibility
matrices
builds up a statistical profile of historical risk for modelling future projects.
Kaplan and Garrick (1984) assert that the purpose of risk analysis and risk quantification is
always to provide input to an underlying decision model which involves not just risks but
19
Risk factors leading to cost overrun in the delivery of highway construction projects.
also other forms of costs and benefits. Risk analysis is a systematic process for evaluating a
risk at the systems level.
Once risks are determined, the next process involves the modelling and quantifying risks
under risk assessment. The following sections deal with these aspects.
Risk assessment provides qualitative and quantitative data to decision makers for later use in
risk management and when project risk assessment is undertaken properly and implemented
correctly, and in sequence (Ayyub and Wilcox, 2000). It enables continual improvement in
project decision making and can increase the likelihood of the successful completion of
projects to cost, time and performance objectives. It is an effective tool in meeting clients
needs in project delivery (Halpin, 1998).
A strong risk management program is essential in all project delivery strategies and is the
central aspect of project management (Shepherd, 1997). On the other hand, some people do
not even consider risk assessment to be a form of science, referring to it as art or even
hocus pocus (Gregory, 1989). Jackson (1989) argues that risk assessments are as full of
unknowns and value-laden assumptions as to lend such processes to be of no real benefit.
A number of systematic models have been proposed for use in the risk-evaluation phase of
the risk-management process. Kangari and Riggs (1989) classify these methods into two
categories classical models (i.e. probability analysis and Monte Carlo simulation) and
conceptual models (i.e. fuzzy-set analysis). They noted that probability models suffer from
two major limitations. Firstly, some models require detailed quantitative information that is
not normally available at the time of planning. Secondly, the applicability of such models to
real project risk analysis is limited, because agencies participating in the project often have
problems with making precise decisions. Such problems are often ill defined and, thus,
require subjective evaluations that classical models cannot handle.
The risk assessment control process attempts to answer the following three questions derived
from Kaplan (1991):
1. What can go wrong?
2. What is the likelihood that it will go wrong?
3. What are the consequences if it does go wrong?
In order to perform risk assessments, several methods have been created to help answer these
questions. Each of these methods is suitable in certain stages of a projects lifecycle. The
characteristics of particular risk assessment methods are shown in tables 2.2 and 2.3.
20
CHAPTER 2
Literature review into project risks and cost overrun
Qualitative risk analysis uses expert opinion to evaluate the probability and consequence of
interaction within a system. Table 2.3 details some qualitative risk assessment methods used
An essential function of the construction project manager is the control of projects and hence
the control of risks. Risk monitoring is required so as to respond to events that occur over the
course of a project. Risk control can be achieved through the updating of risk management
21
Risk factors leading to cost overrun in the delivery of highway construction projects.
plans with new information, identifying alternatives to unplanned risk events, and by
mitigating unplanned risks (del Cano and de la Cruz, 2002). Perry and Hayes (1985) suggest
a checklist of risks that may occur throughout the life span of any project. Bent (1988), in
introducing the concepts of control, asserts that the fundamental elements of control are the
cost element and the project schedule. However, Might (1984), in carrying out a survey on
the effectiveness of project control systems on high-risk US Department of Defence and
NASA projects, reports that such control mechanisms are unclear and hence ineffective.
Williams (1995) describes various risk identification and analysis tools used by researchers
and practitioners in project risk management that can be used in controlling risk. Tummala
and Leung (1999) developed a methodology governing risk identification, measurement,
assessment, evaluation and control and have applied it in managing costs associated with risk
for transmission-line projects.
Most risk analysis reported in the literature so far has centred around analysing the duration
of projects and many authors, such as Farnum and Stanton (1987), present the distribution of
duration of activities as classical beta distributions. For example, Berny (1989) proposes a
distribution for practical simulations. Turner (1999), on the other hand, suggests expert
judgement, plan decomposition, assumption analysis and brainstorming for effective
identification of risk factors in projects. Delphi technique is used by Dey (1999) for
identification of project risk factors.
Cooper et al. (1985) suggests a risk-engineering approach where systematic risk evaluation
could be performed by subdividing a project into its major elements and analysing the risk
and uncertainty associated with each in detail. A classic risk analysis process is shown in the
upper tiers of Figure 2.1. The dotted box of risk engineering is shown to highlight this as a
backdrop to risk assessment and risk management.
22
CHAPTER 2
Literature review into project risks and cost overrun
R
This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
Patton (1986:35) offers a more technical definition in the context of program evaluation by
describing qualitative data as consisting of:
Detailed descriptions of situations, events, people, interactions, and observed
behavior; direct quotations from people about their experiences, attitudes, beliefs, and
thoughts; and excerpts or entire passages from documents, correspondence, records,
and case histories.
23
Risk factors leading to cost overrun in the delivery of highway construction projects.
approach categorises risks into high/medium/low probability and for this purpose can be
defined arbitrarily.
Construction simulation, fault tree analysis, fuzzy stochastic applications, risk premium and
expected net present value are just a few of the many quantitative risk assessment techniques
that can be used. These are explained in detail in the following paragraphs.
2.6.1 Simulation
Ang et al. (1984) define simulation as being the process of replicating the real world on a set
of assumptions and conceived models of reality. Simulation can be applied to construction,
manufacturing, public health, transportation, business process re-engineering and a host of
other industries (Banks et al., 1996). Simulation is one method to quantitatively assess
project construction risks. It is normally used to represent large, complex projects or systems
because it is much less expensive to experiment with models than real systems. Simulation is
particularly useful for studying systems in the design stage, because it is good for predicting
performance or cost of a proposed facility to be built.
Discrete event (DE) simulation is the particular modelling of a system as it evolves over time
in which state variables change instantaneously at separate points in time (Law and Kelton,
1991). As an example, the simulation of the process of bridge concrete components being
manufactured has the discrete steps of:
an order arrives at a concrete precast yard
component moulds are manufactured
multiple step casting processes
inspections
shipping to the bridge site.
At each point in the process the variables change, that is from raw material to finished
components or structure.
A construction simulation that is built on potential construction scenarios can identify risky
elements, resource constraints and potential bottlenecks (Mak and Picken, 2000). As well, it
can define sequences, construction times, facilities, materials, transport, labour etc.
Simulation is also well suited to calculate the consequences of a potential risk event that may
result in a cost escalation. For example, the environmental concern that occurs when
building on former industrial sites through the uncovering ground contaminants. The
subsequent risks of schedule delays and cost escalations can be approximated through the
use of simulations of various rehabilitation treatments (Hegazy and Ayed, 1998).
24
CHAPTER 2
Literature review into project risks and cost overrun
Artificial logic applications can be developed to consistently emulate the way humans think
and judge by following accountability rules. Fuzzy sets and logic are used to capture
qualitative domain expert opinion for developing affordable project schedules and budgets.
Although this technique cannot substitute for deterministic scheduling and costing methods,
it does complement modelling methods in cases where vague and incomplete project
information is present. Blair (1999) explains that expert opinion can be used to model
construction simulations using fuzzy stochastic techniques. Applications include:
estimating project cost (Paek et al., 1993)
developing a risk-based cost and schedule (Blair, 1999)
developing the schedule (Ayyub and Haldar, 1984)
performing risk-based decision making (Blair et al., 2000)
The use of fuzzy theory is particularly suited to construction issues where there is a severe
lack of realistic historical data (Hegazy and Ayed, 1998)
In general, epistemic risk constitutes the biggest problem in project planning and risk
prediction, where there is little historical evidence on which to base predictions (del Cano
and de la Cruz, 2002). The definitions of probability and risk have implications for how
these are perceived by project owners. Key work on the perception of uncertainty has been
carried out some time ago by Tversky and Kahnernan (1981). Around the same time,
MacCrimmon and Wehrung (1986) describe a study of managers' attitudes to taking risks
and how they choose options and risk action/outcome benefits in standardised risk situations.
Fellows et al. (2002) point out that a decision will have a range of possible outcomes
between quite well-defined extremes. They point out that it is not likely that the probability
of the occurrence of each possible outcome is equal everywhere within the range. The
probabilities may be represented as a graph of probability against possible outcomes and is
referred to as a probability density, with the normal distribution curve being the most
common example (Fellows et al., 2002).
25
Risk factors leading to cost overrun in the delivery of highway construction projects.
Elkington and Smallman (2002) make the point that, in approaching risk management,
managers have the luxury when working in the world of financial risk of vast statistical
databases to instruct probability. Whereas project managers outside of finance, who face
operational risks constantly need to decide the chances of an identified risk occurring. Some
use historical project data, and others use unwritten past experience. But in many cases, the
likelihood of a risk occurring is derived by means of an educated guess (Elkington and
Smallman, 2002).
Databases, either formal or informal, usually exist for specific domains, such as in assessing
risk in construction projects. Many of these are commercial-in-confidence, particularly in
highly competitive industries such as the oil industry. Two early examples of established
databases of project risks are Niwa and Okuma (1982). They describe a well-structured
database in use at Hitachi, and the value of the database for knowledge transfer of project
risk. Ashley (1987) describes a number of examples of expert systems that are based on risk
knowledge.
In-house historical databases are probably the best source of data to assess risk event
probability or consequence. However, in many cases these databases are inadequate or
disjointed, unavailable, or supplemented with personal knowledge (Al-Bahar and Crandell,
1990). These databases are company and/or project specific and are not usually able to be
uniformly applied across new projects.
While industryaverage data exists for the some construction operation variables, (say for
excavators), a firms own historic database can provide improved accuracy in forecasting
production risk and hence costs. Such databases can be developed by project owners, as well
as contactors, and can be used for future projects and comparison with industry averages.
However, many organisations do not maintain such detailed databases. For example, Iseley
and Gokhale (2003) provide an example where the equipment selected for the land clearing
and grubbing activities for a project was based on only the estimators own experience.
Production estimating is extremely effective when experienced estimators are responsible for
making the necessary productivity risk decisions. However, more detailed databases would
permit decisions to be made at a lower level by less experienced people without sacrificing
accuracy (Iseley and Gokhale 2003). This would allow the more experienced estimators to
use their time more effectively. In addition, more detailed databases provide better
information to substantiate the impact if a change in conditions should develop (Iseley and
Gokhale 2003).
When available, government-collated data is generally overly broad and may be used to
gauge or forecast overall trends. For example, the number of accidents per thousand hours of
work can be broken down by specific industry, and yet, this data is difficult to apply to
specific construction project risk probabilities as the data usually considers all types of
construction, from residential, road-building, heavy-civil, to large commercial developments.
26
CHAPTER 2
Literature review into project risks and cost overrun
Databases for construction projects may be available from in-house governmental agencies,
such as the Australian Bureau of Statistics (ABS), from industry sources or from published
academic literature.
Deriving probability values for different cost elements of an estimate was first suggested by
Hertz (1964) as a means of reducing uncertainty. It was recommended as an integral part of
project appraisal by the World Bank, over 30 years ago (Reutlinger, 1970). With this method
it is possible to quantify the range of a particular cost estimate. Probability analysis, in the
context of cost estimating, entails assigning estimated values of each cost element and
analysing the risk associated. The individual probability distribution of each cost component
is then aggregated so as to derive the overall probability distribution of the total cost.
27
Risk factors leading to cost overrun in the delivery of highway construction projects.
Expert elicitation is a valid method to develop risk assessments and is the formal process of
obtaining assessment probabilities and consequences when the information is subjective. The
expert elicitation process needs to be systematically structured and can use a participative
method such as Delphi or modified Delphi techniques to quantify probabilities and
consequences (del Cano and de la Cruz, 2002). Personnel with a risk-analysis background
that are familiar with the construction, design, operation, and maintenance of projects need
to define these issues in the form of specific questions.
The Delphi technique (Linstone and Turoff, 1975) can be used to anonymously elicit the
opinions of experts concerning risks. It provides feedback to experts in the form of
distributions of their opinions and reasons. They are then asked to revise their opinions in
light of the information contained in the feedback. This sequence of questionnaire and
revision is repeated until no further significant opinion changes are expected. The technique
is designed to protect anonymity of the experts opinions and reasoning. Sackman (1972)
points out some shortcomings of the Delphi and these can be characterised as:
The information and questions provided to experts need to be carefully reviewed to
ensure objectivity
It is difficult to summarise and present to the group a common evaluation scale that is
interpreted uniformly by the experts
The benefit of experts participating in active dialogue is missed
It can be difficult and time consuming to explore disagreements between experts.
For the rapidly paced project management world, any expert elicitation method needs to be
efficient because project managers are continually confronted with not having enough time
for all the activities required of them (Laufer, 1996).
The purpose of semi-structured interviews is to extend the scope of a study to a wider range
of exponents and to test the conclusions drawn from exploratory work. This can be achieved
by developing an outline list of questions developed for consideration by the interviewee
with key questions about the issues being researched. The technique of personal construct
28
CHAPTER 2
Literature review into project risks and cost overrun
elicitation, as described by Stewart and Stewart (1981), has been used in project
management technique assessment. Bannister and Fransella (1989) define a personal
construct as being a bipolar distinction, or scale, which is used when contrasting different
people, objects or situations.
Semi-structured interviews can provide opportunity to evaluate tacit knowledge based on the
experience of individual practitioners. Koskinen, Pihlanto and Vanharanta (2002) explain
tacit knowledge as that which expresses itself in human actions in the form of evaluations,
attitudes, points of view, commitments and motivation
When considering risk in terms of monetary outcome, there are three general attitudes
toward risk (Kumamoto and Henley 1996). These are:
risk-aversive
risk-neutral
risk-seeking.
29
Risk factors leading to cost overrun in the delivery of highway construction projects.
An example of this can be seen in say an individuals investment practice. Some investors
prefer to buy stock of small companies in the high technology sector. Typically this is
considered a risk-seeking stock choice because there is a chance of losing an initial
investment but the potential monetary gain can be quite substantial. Economists agree that
those who take over or bear the risk have to be rewarded for doing so. Begg et al., (2000)
comment that many economic activities consist of:
the more-risk-averse bribing the less-risk-averse to take over their risks.
A very early study by Reutlinger (1970) reports on the results of the attitude to risk taking of
US executives in relation to corporate investment decisions. It showed that, although there
were widely different attitudes towards risk taking, even within the same corporation, there
was a tendency for most of those surveyed to be conservative when assessing risk.
In the project management context, it is generally reported that large established companies
are risk-aversive, governmental ventures are risk-neutral to risk-aversive, and smaller newer
companies are risk-seekers. Therefore, risk analyses will account for the acceptability of risk
according to an organisation's established risk attitudes (Kumamoto and Henley, 1996).
Paul and Gutierrez (2005) point out that when a contractor is risk neutral, then the risk
premium will not vary across contract type. They assert that, in general, a contractor will
tend to be risk neutral if the contract is small relative to their total volume of business. In
such cases, risk neutral decisions will maximise the contractors profit. However, if a
contractor is risk averse, then he or she will tend to factor in a higher risk premium in a fixed
price contract since more financial risk needs to be borne than in say cost-plus contracts
(Paul and Gutierrez, 2005).
30
CHAPTER 2
Literature review into project risks and cost overrun
contributors to risks in construction projects (Ahmed et al. 1999). Baker et al. (1999) report
that in a survey of over 100 companies in both the oil and construction industries, companies
commonly respond to assessed risk, however the construction industry concentrates almost
exclusively on the reduction of financial risk and less on managing technical risk.
Construction is an industry associated with many risks. Sampling of the most prevalent risk
will convince any causal observer that the construction process is risky. The literature has
several case studies of construction risk management from general building construction
(Baker et al., 1995), offshore construction (Curole, 1995), or highway construction (Wang
and Chou, 2003). In the field of project management, risk management is a recognised
practice to help clients deliver projects on schedule and within cost (Project Management
Institute, 2000). However, the risk management performed in the construction industry has
traditionally been through the use of gut feel or a series of rules-of-thumb (Al-Bahar and
Crandall, 1990). This may be due to a lack of understanding of the benefits and cost, the
perceived difficulties, or cumbersome processes in developing risk management (McKim,
1992; Ward, 1999). In discussing the US construction industry, Kangari (1991:46) supports
this notion by pointing out that:
Most managers rely primarily on their judgements, rules of thumb, and expertise and
that the conventional algorithmic models developed for risk analysis are not generally
accepted by management for decision making under uncertainty.
Since the early 1990s, a variety of risk management processes have been written about.
These include Al-Bahar and Crandall (1990), Del Cano (1992), BSI (1999), NASA
(Rosenberg et al., 1999), and the US Department of Transportation (FDOT, 2003). De Cano
and de la Cruz (2002) have described the most noteworthy, comprehensive, and sound
project risk management processes. These include:
PRAM this was the first highly comprehensive process developed by a large team,
including both practitioners and academics (Simon et al., 1997; Chapman and Ward,
1997)
RAMP this risk management process has characteristics similar to those of the
PRAM process in scope, structure, and conception but importantly has been
conceived for the construction environment. RAMP has been developed jointly by
the Institute of Actuaries and the Institution of Civil Engineers to assess all kinds of
risks and uncertainties so that they can be identified, evaluated, reduced and
controlled (ICE, 1998).
The Project Management Institute (2000) points out that both PRAM and RAMP tend to
reflect more a British European way of performing project risk management.
PUMA (Project Uncertainty Management) This is an integrated methodology
based on a hierarchically structured, flexible, and generic project risk management
31
Risk factors leading to cost overrun in the delivery of highway construction projects.
process for construction projects from the point of view of the owner for individual
projects (del Cano and de la Cruz, 2002).
Project construction risk can have physical or capability related aspects (Zack, 1997).
Physical risks are those events that prevent one from completing the project or increase the
costs and schedule such as acts of God, weather, impracticability, or other things that are
beyond the control of the project team. Capability related risks are those that interfere with
performing the work but management has a choice in minimising the risk such as poor
quality, safety and equipment selection. Project Management Institute (2000) classifies risks
as:
external (uncontrollable)
internal (controllable).
Project size can be a major cause of risk, as can other factors such as:
the complexity of the project
the speed of construction
location of the project
degree of project unfamiliarity.
Edwards (1995), Smith and Bohn (1999), and Kim and Bajaja (2000) list potential
construction risks. Table 2.4 details some of the most common risk sources in construction
projects.
32
CHAPTER 2
Literature review into project risks and cost overrun
One of the main sources of construction risk derives from the site conditions in which
projects are planned to be constructed. The next section looks at project site conditions as
sources of risks.
Studies from around the world have consistently highlighted the recurring frequency of
claims for unforeseen ground conditions. In Canada, Revay (1992) reports that inadequate
site and/or soil investigations and unexpected ground conditions were among the six most
frequent causes of claims. A study in Hong Kong by Kumaraswamy (1997) aimed at
identifying root causes of claims for extension of time and extra payments on construction
projects, found that unforeseen ground conditions was ranked fourth in the top 10 common
categories of construction claims. In the US, studies by Halligan et al. (1987) on Federal
Highway Administration funded state highway construction projects indicate that claims for
ground conditions accounted for approximately 35% of the total dollar amount paid to
contractors for claims.
The traditional position on site information has been that there is no obligation on the owner
to provide the contractor with any information on the nature of the site and sub-soil
conditions (Mak and Picken, 2000). It is therefore the responsibility of the contractor to
investigate these matters and to be satisfied with the feasibility of constructing the works on
the designated site. Even if there is relevant information in the owners possession, there is
no obligation in English law to disclose it to a contractor. There are a number of civil law
judgements which establish precedence in this area:
1. Thorn v. London Corporation (1876) 1 App Case 120
2. Dillingham Construction v. Downs (1972) NSWLR 49
3. Morrison-Knudsen International v. The Commonwealth of Australia (1972)
13 BLR 114.
On the other hand, some jurisdictions impose such an obligation only in special
circumstances. For example, US public project owners have an obligation to supply site
information to the contractor where the circumstances are such that the contractor cannot
otherwise obtain it (Glower, 1990; Smith, 1995). If the owner decides to provide such
information and the information is incorrect, then the contractor may have remedies against
the owner for breach of warranty, misrepresentation or negligent misstatement.
33
Risk factors leading to cost overrun in the delivery of highway construction projects.
In Great Britain, legislation gives the contractor some protection against exclusion of
liability for inaccurate information even if it was not provided fraudulently such as that of
Section 3 of the Misrepresentations Act 1967 (as amended by Section 3 of the Unfair
Contract Terms Act 1977). However, because of the uncertainties of geotechnical
engineering it has been suggested that such riders would be effective in excluding the normal
consequences of misrepresentation, provided that the owner has not been careless or
fraudulent (Wallace, 1995).
In judging the physical site conditions, the contractor is assumed to have taken into account:
the site information
publicly available information referred to in the site information
information obtainable from a visual inspection of the site
other information which an experienced contractor could reasonably be expected to
have or to obtain.
Ndekugri and McDonnell (1999) point out that it is commercially impracticable for tenderers
to carry out full geotechnical investigations even if there are no time constraints in the
particular bid. In general, a contractors entitlement operates on the basis of the test of
foresee ability. If the physical conditions or obstructions are not foreseeable by an
experienced contractor then the contractor may claim. Jones (1996) notes that there has been
no widespread acceptance of foresee-ability as a general philosophy of risk allocation, but
there has been a strong international acceptance of alternative philosophies such as those put
forward by Abrahamson (1984). By contrast, Eggleston (1996) suggests that a contractor can
be compensated for time and money if the physical conditions are such that an experienced
contractor would have decided at the time of entering into the contract that they had such a
small chance of occurring that it would have been unreasonable to allow for them. Shadbolt
(1990) believes that this is as it should be because it is reasonable for an owner to expect a
contractor to exercise their skills and judgement and for the owner to rely on them in that
respect.
ASCE (1989), Smith (1995) and OReilly (1995) all propose that risk allocation should be
motivational and risks should be allocated to the party with the greatest opportunity to
influence the magnitude of the risk. With regard to ground condition, this is with the project
owner. It is their site and they can commission a thorough site investigation that will, within
limits, sensibly define the ground conditions and safeguard against the risks.
Attempts have been made in contracts to limit the contractors entitlements to make claims
on account of climatic/weather conditions. Climatic conditions on site have been removed
from some prominent contract provisions because it has always been difficult to assess
whether such conditions could or could not have been foreseen by an experienced contractor
(Seppalla, 1991). However, while the contractor may not claim for costs in such cases, a
claim for an extension of time for exceptionally adverse climatic conditions may be made
under the appropriate clause. In some standard contracts like the NEC, a contractor can claim
for weather conditions that occur, on average, less frequently than once in 10 years, but their
entitlement includes both costs and time.
34
CHAPTER 2
Literature review into project risks and cost overrun
Construction industry studies, such as by Latham (1994) and Egan (1998), echo an
underlying lament that can be traced back many decades to the Simon Report in 1944, the
Emerson Report in 1962, the Banwell Report in 1964 and the Tavistock Report in 1966.
These all point out many problems arising from poorly structured project procurement
systems.
The procurement of constructed facilities is a process which involves many complex and
interrelated steps. Along the way there are questions of value for money, probity and fitness
for purpose. Over more than three centuries, the traditional procurement process has been
developed that focuses on clarity, separation of phases and a transparent independent bidding
stage. Unfortunately, there are some serious drawbacks. For example, the process can be
inefficient and take a long time and often sets up opposing stances between participants
(Sidwell et al., 2002). However, the traditional process does have some notable
characteristics which continue to hold relevance in todays climate. These characteristics are
illustrated in the RIBA Plan for Work (RIBA, 1984) which emphasises the step by step
functional contribution and progressive definition, the output from each stage being the input
into the subsequent stage. The clarity of this process is particularly attractive to clients who
need to demonstrate probity.
Project procurement processes in the Australian construction industry are largely derived
from the United Kingdom where the traditional process which separates design and
construction, and awards the construction to the lowest bidder, has been developed over the
centuries, particularly since the industrial revolution. The traditional approach to project
management is shown in Figure 2.3 below.
35
Risk factors leading to cost overrun in the delivery of highway construction projects.
Wang and Chou (2003) assert that risk management needs to be made more efficient and
effective so all parties can understand:
their respective responsibilities
risk event conditions
risk preferences
risk management capabilities.
They also point out that if a project contractor has different perceptions of risk allocation to
those of the client or even a lack of clear understanding of risk management, then the
contractor will inappropriately manage the risks in construction projects by assuming that the
risk events or consequences are not the contractors responsibilities.
On the other hand, Carmichael (2000) argues that there may be an overall saving to the
project client by accepting some risk. Generally client expect that contractors bear some risk,
36
CHAPTER 2
Literature review into project risks and cost overrun
though some take this to an extreme and, through contract condition mechanisms, require
contractors to bear essentially all risks.
There is a basic conflict between the risk and motive provisions in most contracts. If a risk is
placed on the contractor, they have a strong motive to minimise that risk, whereas the client
has no such motive.
Many standard forms of contract hardly mention risk and frequently contracts are not clear
on who bears the risk. Robinson (1987) claims this as the strength of design-and-build
contracts
A survey by the Victorian Government, Australia (Victorian Government, 1999) provides
some interesting insights into the strategic decision making by various parties in the
allocation of risk in public projects. A factor that tends to defeat the efficient allocation of
risk in construction contracts is that project owners may have significantly more power at the
contract formation stage than does the contractor. This power can be derived from a number
of sources, such as:
the competitive bidding process, with unreasonable time pressures and often
involving a large number of bidders
in-house risk management and legal expertise which greatly exceeds that of the
contractor
a general reluctance by some contractors to price adequately for the risks that they
are being required to accept.
The joint use of formalised risk management processes can provide all parties with a way out
of this bind by providing a revised contract form of potentially greater benefit to both the
contractor and owner. However, Tuohey (2002) points out that most contracts are drafted
unilaterally by the owner and risks are allocated without the benefit of a risk management
process. Tuohey (2002) also reports that, in some Australian Government Agencies, it is now
a common requirement for the tenderer to include draft risk management plans for the entire
project. He adds that it is easy to believe that in the near future, risk management will
become a necessary and formal process in the execution of all projects. More projects will be
likely to succeed as a result.
Construction contracts are the written agreements signed by the contracting parties which
bind them and also define relationships and obligations. In any particular contract, the
project owners goals can best be achieved by selecting the contract type that will most
effectively motivate the contractor to the desired end and is also dependent on completeness
of information for the bidder(s) at tender time and the extent the owner wishes to take
specific risk (Zaghloul and Hartman, 2003). Given the opportunity, an owner should favour
37
Risk factors leading to cost overrun in the delivery of highway construction projects.
efficient allocation of risk between parties to a project that simultaneously reduces risk and
improves performance.
The contract terms define how such risk is distributed between the owner and the contractor.
In establishing the context of a contract which is composed of specifications and clauses,
both owners and contractors strive to understand each others goals and capabilities, and
risks should be allocated accordingly (Akinci and Fischer, 1998). However, a comprehensive
empirical study about the allocation of risk in contract clauses conducted by Ibbs and Ashley
(1987) concluded that on an aggregate basis, owners are at risk 11% of the time, contractors
are at risk 46% of the time, and both parties are at risk 43% of the time. These results
confirm that owners use contracts to minimise their risk exposure.
Liabilities and responsibilities of each contracting party are allocated through the conditions
of contract. It has been postulated that inappropriate and unclear risk allocation among the
contracting parties generates avoidable construction claims and disputes (McGowan, 1992;
Fisk, 1997; Kumaraswamy, 1997). One of the main areas where risk management can be
applied is in developing clearer and more appropriate conditions of contract that provides
clearer definition of the risks and hence lead to their proper allocation and management
during the construction process. However, contract language alone is not sufficient to clearly
specify risk apportionment between the contracting parties as contract clauses can be
interpreted in different ways by different parties (Hartman et al., 1997).
Zaghloul and Hartman (2003) report that appropriate risk allocation is a significant
contributor to low transaction costs in specific projects and is a vital issue in the success of
the contracting process. On the other hand, Gransberg and Ellicot (1997) point out that, in an
ownercontractor relationship at least, a common aim of owners appears to be to avoid risk
as far as possible by allocating as many risks as it can to the contractor. Carmichael (2000)
notes that a general belief is that the owner should, wherever possible, define the work as
carefully as possible beforehand, irrespective of the type of contract used so as to reduce the
risks to all parties.
The appropriate contracting method and the contract documents for any construction project
depend on the nature of the project, but an appropriate contracting method coupled with
clear and equitable contract documents do not by themselves ensure project success. This is
especially the case where parties work together in the face of uncertainty and complexity,
whilst having diverse interests and conflicting agendas. The attitudes of the contracting
parties and the co-operative relationships among the project participants are important for
successful project delivery.
The theme of risk allocation in procurement methods is seen as a critical issue in the
selection of construction contracts. Latham (1994) points out that clients often prematurely
select procurement methods without consideration of the appropriate division of risks. The
risk involved could be physical in nature but is more likely to be mercantile (Langford et al.,
2003). This supports Taylors (1993) argument that consideration of risks at the procurement
stage could save money and time and also improve quality. As well, Smith (1995a) has
confirmed that inappropriate risk management can also lead to higher costs for the client.
According to Hibbert et al. (1990), procurement choice advice is often made to clients even
without hard factual evidence of judging the successes of previous procurement decisions.
38
CHAPTER 2
Literature review into project risks and cost overrun
Any construction project involves risk and there is no possibility of eliminating that risk. All
that can be done is to regulate the risk allocated and then to properly manage the risk. This is
usually done through the language of the construction contract.
Not only are procurement issues a factor influencing risk, but, as Uff and Odams (1995)
noted, contractual risks are seldom analysed and quantified. Hence it is important that clients
select appropriate procurement routes and contract strategies which fit the clients ambitions
for a project. Invariably, this will include factors which will deliver value for money
(Langford et al., 2003).
Kozek and Hebberd (1998) point out that decisions regarding risk sharing or risk shifting are
made within the context of a project owners contracting policy. Hughes (1997) asserts that
the purpose behind using standard-form contracts is to allocate risks fairly between the
parties
Zaghloul and Hartman (2003) define one way in which the contracting parties can attempt to
address the right responsibilities for risk by dealing directly with the issue of legal liability
arising from certain contract clauses such as disclaimer (exculpatory) clauses. Such clauses
attempt to transfer one partys risk (which may be legal liability) to another by contractual
terms (Hartman, 2000). The use of disclaimer clauses to allocate risk has been identified by
studies and industry practice as a key reason for increasing the overall cost of a project
(Khan, 1998; Zack, 1996).
When a risk is shifted to a contractor and the contractor has no means by which to control
the occurrence or outcome of the risk, the contractor either insures against it or adds a
contingency to the bid price (Jergeas and Hartman, 1994). A study conducted by de
Neufville and King (1991) in the US found that contractors add significant premiums, in the
order of 3%, to their risks when they have a low need for work or projects that have high
risks. Zaghloul and Hartman (2003) state that recent studies of Canadian contracts indicate
that the use of disclaimer clauses can carry a premium of between 8 and 20%, depending on
whether business conditions were favourable low need for work, fair contract
administration, suitable contract type, and completeness of design works; or adverse high
need for work, high technical complexity, unfair contract administration, and incomplete
design works.
39
Risk factors leading to cost overrun in the delivery of highway construction projects.
A number of possible payment schemes can be used within a contract to manage risk. A
good overall summary of the main types of payment scheme is given in In't Veld and Peeters
(1989). They describe cost-reimbursable schemes such as:
cost plus percentage
cost plus fixed
cost plus incentive
schemes that contain both fixed price and fixed price with incentive provisions.
Carmichael (2000) describes two contract groups that are generally classified according to
the form the consideration from the project owner takes:
1. Where the consideration is either a stipulated sum of money covering all the work or
is a set of monetary rates covering the components of the work, the contracts are
referred to as fixed-price contracts
2. Where the contractor is paid the cost of the work together with an additional amount
for the use of the contractors services, the contracts are referred to as prime-cost
contracts.
The polarisation of construction from design in the construction industry may have arisen
from expected efficiencies in specialisation and the perceived need for independent design
and oversight. But the resulting fragmentation and adversarial contractual cultures have now
been seen by many to be an unfortunate departure from the single-point procurement
solutions provided by master builders in previous centuries.
Research into procurement best practice has identified a need for wider consideration of
what constitutes project deliverables (Rowlinson and McDermott, 1999; Walker and
Hampson, 2002). The construction industry is usually characterised by its complexities,
40
CHAPTER 2
Literature review into project risks and cost overrun
Not until more recent years has there been interest in seeking alternative methods, even
though these so-called alternative methods were used in the past procurements of a number
of public sector projects. Below are terms applied to procurement types from Pakkala
(2002).
Design-Bid-Build (DBB Traditional Method): This system was developed during the
industrial revolution and resulted in the creation of specialised professional movements of
architects, contractors and engineers. This approach has been the standard choice of project
delivery systems for many years. In this model, the owner/client procures the services of a
design consultant to develop the scope of the project and complete design documents which
are then considered as legal documents for use in selecting a contractor who builds in
accordance to the specifications developed by the design team. The owner maintains most of
the risks. Typically in a public organisation, the proposal is in an open competition for a low
price. The contractor that wins the award is legally bound to produce the project at a certain
price, schedule, and minimum level of standard care. After completion of the project, the
owner is responsible for operations and maintenance of the project. The owner is also
responsible for all the financing needs.
Design-Build (DB): This systems heritage dates back to the construction of the pyramids,
when it was referred to as the master builder. Design-build is simply a project delivery
method in which the owner/client selects an organisation that will complete both the design
and construction under one agreement. Upon completion, the owner is then responsible for
all the financing aspects.
41
Risk factors leading to cost overrun in the delivery of highway construction projects.
Construction Management (CM At Fee Agency): This is the process similar to DDB, the
traditional model, in which the owner/client is responsible for the design, bidding, and
construction of the project. However, the construction management organisation takes on the
responsibility for the administration and management, the constructability issues, day-to-day
activities, and also assumes an advisory role to the owner/client. The CM organisation has no
contractual obligations to the design and construction entities. Again, the owner is
responsible for operations and maintenance of the project as well as the financing aspects.
Lump Sum: This is considered as a fixed-price agreement for the total work and products of a
given project. Sometimes this is also referred to as a fixed-price contract. Any changes to the
contract should be agreed upon by both parties, and they are usually described under change
orders.
Unit Price or Schedule of Rates: This refers to price considerations for specific aspects of
construction and materials. For example, the cost per unit is for physical work or products,
such as the price base for so many metres of guardrail etc.
Pure O & M (Operations & Maintenance): This is a delivery method in which the
owner/client secures both operations and maintenance for a project under one agreement
from a single provider, usually called the contractor. The owner is also responsible for
financing aspects.
42
CHAPTER 2
Literature review into project risks and cost overrun
Pakkala (2002) also identifies that Great Britain is one country which deviates substantially
from this practice and uses the Design-Build (DB) models and the Design-Build-Finance-
Operate (DBFO) model (sometimes referred to as Public Private Partnerships or PPP), which
integrates the philosophy of integration and external financing for these processes. He also
notes that the DB model is beginning to be used more frequently for many types of
infrastructure projects, particularly in power plants, water works, bridges, rail, road and also
in some local authority infrastructure projects.
Langford et al. (2003) investigated construction costs in 11 similar motorway project parcels
in the UK between 1990 and 1995. Five of these projects were carried out by a traditional
design, tender, construct method. Five were undertaken by a procurement system by which
the contractor bid a lump sum for the work. One project was done as a DB project. The
results of their analysis indicate that, in road works, the construction costs per kilometre was
some 11% less expensive when lump sum contracts were used. Langford et al. (2003) also
reports that their research shows that lump sum projects are much more likely to be
completed within budget and require less management by the client organisation. They also
report that lump sum also delivers a more harmonious working relationship between the
client and contractor.
In the traditional contract, the majority of cost overrun is borne by the owner, whereas in
lump sum bids, the performance risk is shifted to the contractor. Langford et al. (2003) point
out that the pattern of risk sharing due to that of just chance (aleatoric factors) is unlikely to
be changed as both types of contract arrangements permit the usual range of claims for
unforeseen events. However, sharp differences may be seen in risk situations which may be
defined as epistemic (which relates to the knowledge base of those taking the risks). For
example, a contractor who perceives a particular item of work to be particularly high (or
low) risk will price that item accordingly since the contractor sees its knowledge base and
hence their ability to manage the risk as different from that of their competitors. For
example, a contractor considering the risk of delay and additional cost due to a highway
excavation through rock would include in their assessment any special knowledge of the
geology of the area; and consequently the risk of encountering rock.
Carmichael (2000) states that there is an increasing transfer of financial risk from the
contractor to the owner as the contract type moves from fixed-price to prime-cost contracts.
As the owner takes more and more risk, it is expected that the contract price would decrease.
De Neufville and King (1991) identify two ways of compensating for risk when developing a
bid. One is to develop a standard cost estimate not considering risk and varying the mark-up
depending on the risk. The second method is to develop a cost estimate that adjusts
productivity factors or adds contingency based on the risk of each item being estimated and
then apply a standard mark-up to this risk-compensated estimate. They report that, in
practice, most contractors used the latter method. With regard to contract type, a contractor
43
Risk factors leading to cost overrun in the delivery of highway construction projects.
experienced in undertaking contracts of a similar type would seem to require less risk
premium, because the contractor is likely to have greater confidence in being able to
complete the contract in accordance with the clients brief.
In respect of size of contract, Hillebrandt and Cannon (1990) point out that risks, including
errors by the company itself, technical risks, financial risks of the project and onerous
contract conditions, bear little relation to the size of contract and so it is difficult to put a
percentage on the mark-up for them. They also state that, in general, a company that has a
large number of smaller projects is likely to be subject to less risk overall than a company of
similar size with a few large contracts.
Palaneeswaran and Kumaraswamy (2000) report that few public clients venture to adopt
some sort of innovative measures in their project delivery methods, other than limited trials
in pilot or experimental projects.
Zaghloul and Hartman (2003) surveys more than 300 respondents in the Canadian and US
construction industry. They report that, to reach a better risk allocation process, a trust
relationship between contracting parties needs to exist first, and that this can be achieved
through the following stages:
a clear understanding of the risks being borne by each party and who owns or who
can manage the risk
more time and effort in the front end of a project and sufficient expertise to manage
or mitigate the risks and administer the contract
a negotiation phase prior to the start of the contract should exist, this phase is
required to build a trust relationship between the contracting parties (this negotiation
phase can be part of the contract itself)
an adequate risk-sharing or risk-reward system should exist to share the benefits if
the risk does not occur during the project life cycle.
44
CHAPTER 2
Literature review into project risks and cost overrun
Some of the main criticisms of the traditional DBB method are the lack of innovation,
delayed completion periods and cost overrun that are sometimes encountered in projects.
Since the client bears most of the risks of both the design and construction aspects, there
desirably needs to be better practices to ensure the clients needs are being met, quicker
project completion times and cost effective solutions (Pakkala, 2002).
Highway administrators and other public organisations (rail transport, airports and others)
are seeking better practices. Some are using or evaluating alternative models such as DB,
DBOM and BOO. In Pakkala (2002, the main goals of these innovative project delivery
methods are to:
produce projects that have better quality
bring cost savings to the client
transfer risk to the organisation best able to manage the risk
complete projects faster than the traditional methods.
Pakkala (2002) also reports that there are some tools available for predicting which delivery
method might be applicable to projects that are being considered or planned and identifies
the following:
Design-Build Selector (from the University of Colorado)
Project Procurement Selection Model (PPSSM).
The rationale for better risk allocation between owners and contractors ought to be based on
meeting the above conditions as far as possible. Missing one of these criteria is very likely to
trigger an inappropriate risk allocation process for any given project and hence bring
additional cost for the contracting parties (Zaghloul and Hartman, 2003).
An essential part of the agreement between the government and the private contractor is the
allocation of risk between the parties, that is, when an event occurs that influences the cost or
quality of the contracted service, which party should pay to rectify the situation or,
alternatively, which party should gain the resulting benefits (Arndt, 1999). Compared with
conventional delivery methods, there is a higher risk exposure for the BOT sponsors because
of the following:
high front-end development costs
extensive and lengthy negotiations with the host government
multi-party involvement
long-term commitment
equity contribution from sponsors.
45
Risk factors leading to cost overrun in the delivery of highway construction projects.
The identification, analysis, and allocation of various types of risks are important for the
validation of privately promoted infrastructure projects. Dias and Ioannou (1995) have
classified sources of BOT risk in the following 10 risk categories:
1. country (political and regulatory)
2. force majeure
3. physical
4. financial
5. revenue
6. promoting
4. procurement
5. developmental
6. construction
7. operating.
Determining the relative importance of these types of risks is essential for BOT management
decision makers. The decision makers of construction companies should evaluate and rank
BOT projects with respect to their risk (Zayed and Chang, 2002).
Zayed and Chang (2002) propose a risk index for evaluating BOT projects. The index
assesses the risk and rank of BOT projects. To do this, they identify and analyse the main
areas of BOT project risk and a model for calculating a risk index is constructed. The
accuracy and robustness of their model has been verified by the comparison with holistic
evaluations. They also advocate their newly developed approach as being more convenient
than other approaches in dealing with BOT project risk. Nevertheless, this study relies upon
a small academic study group for collecting the evaluation data and they recommend the data
collection zone be increased more widely to practicing professionals in BOT projects in
order to improve the accuracy of the developed risk model.
The construction industry has used many delivery approaches to implement major
infrastructure projects. Over time, dissatisfaction with the adversarial nature of some of the
more traditional approaches, lead to sponsoring organisations adopting various means to
work in more collaborative manners, particularly during the late 1980s and 1990s. Rahman
et al. (2001) suggest that a relational contracting (RC) approach can be mobilised as a
potential route towards improving relationships and team work. The transaction cost
economics approach provides a useful framework for analysing the inevitable differences in
interest between different contracting parties who are members of the project coalition
(Winch, 1989). On the other hand, relational contracting encourages long-term provisions
46
CHAPTER 2
Literature review into project risks and cost overrun
and mutual future planning, and introduces a degree of flexibility into the contract by
considering a contract to be a relationship among the parties (MacNeil, 1980).
Extending the relational contracting approaches to sub-contractors and suppliers through the
supply chain and ensuring their participation at early stages of projects can further increase
benefits (Kumaraswamy and Mathews, 2000). However, relational contracting approaches
are expected to work in almost any environment, if applied properly. This requires
transforming traditional relationships towards a shared culture that transcends organisational
boundaries (CII 1996).
In the US, the partnering model has evolved as a typically non-legal commitment to
teamwork and collaboration. In the United Kingdom, the Institution of Civil Engineers
launched the New Engineering Contract and at the same time project alliances evolved
(initially in the North Sea oil and gas industry). Since then, project alliancing has become
more established in the oil and gas industries in particular and some other major
infrastructure industries in Europe. Incentive contracts are also widely accepted practices
around the world. These differ from alliances in that, under defined incentives, there can be
the traditional scope of work contracts and allocation of risks.
In Australia, project alliances have been taken to a deeper level of sophistication and use,
particularly in major public sector infrastructure projects. A project alliance, to describe it
most simply, is a project/program delivery strategy where the clients and commercial
participants objectives are aligned to maximise performance, proactively manage risk,
reduce cost and achieve outstanding outcomes.
The first two project alliances in Australia were oil and gas industry projects the Ampolex-
Wandoo Alliance and the Western Mining Corporation East Spar Alliance. It was after
completion of these two projects in 1998 when the first three government project alliances
were created. Sydney Water became the first public sector organisation to create a project
alliance on the Northside storage tunnel project, a 26 kilometre tunnel under the northern
suburbs of Sydney with a firm schedule deadline (the Sydney Olympic Games). This was
followed by the National Museum of Australia, known as the Acton Peninsula Alliance
47
Risk factors leading to cost overrun in the delivery of highway construction projects.
(A$150m) (Walker et al., 2000), and the Western Australian Water Corporations Woodman
Point waste water treatment alliance. Since 2002, numerous project alliances have been
undertaken, with the Department of Main Roads in Queensland being the government
agency with the most alliance experience at the end of 2002 in Australia.
Not all projects are suited to an alliance approach. Project alliancing is best suited to those
projects where the traditional risk transfer strategy is not appropriate. In many projects,
outcomes can be enhanced and the project optimised by embracing risk through
collaborative and co-operative contracting against the traditional blind faith transfer or
shifting of risks to others. A fundamental design principle of a project alliance commercial
framework is that if one participant wins, all win; or if one looses, all loose.
The key tangible and intangible benefits of a project alliance delivery approach include:
cost reduction
early completion
improved quality of the asset through enhanced workmanship and finish
improved operability.
Project alliances are not only better equipped to analyse the real risks of the project, but can
also be better equipped to deal with their consequences. On the other hand, Tuohey (2002)
points out that, in practice, there have been a number of issues in the execution of alliance
type contracts. These include:
the risk/return ratio can be such that the rewards to the contractor are low,
particularly if the contractor is required to have some key people tied up in the
alliance for extended periods
the integrated team environment can be quite stressful for people
the nature of long-term alliance contracts are that there is one winner only, with the
consequence that there is the potential for competition in the market to be reduced,
particularly in thin markets like Australia.
The next section of this literature review document focuses on the linkage of risk to project
contracting in project delivery.
48
CHAPTER 2
Literature review into project risks and cost overrun
Various researchers such as Russell et al. (1992), Holt et al. (1994), Kumaraswamy (1996),
Rankine et al. (1996), Russell (1996), Hutush and Skitmore (1997a), Holt (1999),
Kumaraswamy and Walker (1999) and Palaneeswaean and Kumaraswamy (2000) have
studied and reported on various contractor selection practices. All suggest improved
methodologies, but also highlight their strengths and weaknesses in different practices.
Despite the increasing use of alternative forms of project delivery systems in the past two
decades, the methodologies and procedures for bid evaluation, selecting contractors and
awarding contracts have remained relatively unchanged (Hatush, 1998).
Hatush and Skitmore (1998) suggest there are five process elements that are common across
all contract award practices. They are:
1. project packaging
2. invitation
3. pre-qualification
4. short-listing
5. bid evaluation.
The background to the traditional foundation in procurement of bidding for works in the
form of auctioning goes back to ancient times but the earliest academic treatments are
relatively recent, with the contributions of Friedman (1956) from an operations research
perspective and Vickery (1961) from a game theoretic perspective. Auctions have developed
into four basic allocation methods English (first-price open-cry), first-price sealed-bid,
second-price sealed-bid and Dutch (descending). A fundamental implication of the common-
value model that stands in marked contrast with private value auction theory is that, other
things being equal, the bidder with the largest estimate will make the highest bid.
Consequently, even if all bidders make unbiased estimates, the winner will find that he had
overestimated (on average) the value of the rights he has won in the auction (Milgrom and
Weber, 1982). This is commonly known today as the winners curse, due to the winner being
said to be cursed by:
having paid more for the asset than its true value (Marks, 2000:11).
Analytical game theory assumes that rational bidders will anticipate the winners curse and
bid very conservatively to avoid it (Camerer, 2003). However, field study research by Dyer
and Kagel in 1996 in the construction industry found no evidence of winners curse. Neither
does there appear to be any anecdotal evidence of the effects of this (Dyer and Kagel, 1996).
Competitive tendering, particularly for public works, has a long history in places like the
UK. The origins of the practice are reported by Powell (1980) to have come from the early
19th century and basically arose out of dissatisfaction with the then existing system of
measure and value. This was where trades were paid directly by the client for work as it
proceeded. Powell (1980), in quoting Elsam (1826), wrote that:
49
Risk factors leading to cost overrun in the delivery of highway construction projects.
To achieve these objectives, Smith (1995) suggests the involvement of the following:
selection of an appropriate procurement methodology
the preparation of appropriate tender documentation
identification and selection of an appropriate number of suitable contractors who are
willing to tender for the work
choice and acceptance of the most suitable tender.
The nature and form of the competitive arena for the contractor in construction contracting
are largely determined by the project owner or their representative. The choice of bidding
system, coupled with bidder selection practices, has a direct bearing on the degree of
competition, because it affects both the number and the identities of bidders competing for a
particular contract.
Selective tendering systems appear to be more restrictive than open tendering systems,
because the contractor can bid only upon receiving an invitation. For open tendering, the
onus is on the contractor to bid after responding to an advertisement. Private sector clients
appear to have more flexibility in contractor selection, because public sector clients, mainly
for the sake of public accountability, are forced to adopt strict procedures. The tendering
process is an important stage in project delivery. However, the activity involving contactor
selection is an important decision required by any client so as to minimise project risk
exposure.
The next section of this literature review looks closely at research aspects of contractor pre-
qualification.
50
CHAPTER 2
Literature review into project risks and cost overrun
If signs of contractor distress can be detected early on, then it may be possible for a project
owner to either not engage a particular contractor or become more actively involved with the
project before the contractor has suffered irreversible damage which can lead to bankruptcy
and increased losses to all stakeholders. One of the most difficult and important decisions
taken by a project owner for the success of a project is the selection of the contractor. Such a
decision deserves careful attention because of the complexity and sensitivity involved
(Hatush, 1998).
The following four weaknesses are found in contractor selection practices (Holt et al., 1993;
1995):
1. lack of a universal approach
2. long-term confidence attributed to results of pre-qualification
3. reliance on tender sum in decision making
4. inherent subjectivity of the process.
Choosing a contractor based solely on the lowest bid price is seen as one of the major causes
of project delivery problems. Another disadvantage of using the lowest bid as a principal
discriminating criterion is that some contractors who are facing a shortage of work may enter
unrealistically low bid prices, simply to try and maintain cash flow. Therefore, as Hatush and
Skitmore (1998) indicated, financial and technical criteria should be considered in order to
assess the potential of contractors finishing projects on time and to assess whether
contractors have the resources necessary to complete any contract awarded to them.
Disastrous consequences from awarding contracts to the lowest bidders, without due
consideration of their competencies, have led to growing awareness of the needs to
incorporate non-price parameters into contractor selection methodologies. Such fresh
approaches have been documented, for example by Schexnayder and Ohrn (1997), Arditi
and Yasamis (1997), Kumaraswamy and Walker (1999) and Palaneeswaran et al. (1999).
While the risk of contractor default cannot be eliminated by the project owner, it can be
transferred. When the owner signs a contract with a contractor, the risk of contractor default
is normally transferred partly or totally to a surety company. The general contractor may do
the same when signing contracts with sub-contractors. While the end point may be the surety
company, there is still a cost to transfer risk in terms of premiums assessed and the cost of
dispute resolution. The project owner is the one who pays for it in the end (Al-Sobiei, et al.,
2005). For an owner, the costs of default may also include loss of profits due to the project
not being completed on schedule, administrative expenses to analyse the situation, cost to
complete the project beyond the money yet paid to the contractor, fees associated with the
resolution of conflicts, and negative publicity and loss of goodwill within the community.
Given the fact that contractor default depends on environmental, operational and strategic
factors, it is important to use the past history of situations to minimise contractor default. For
any given project, pre-qualification is a highly important decision for the project owner. The
ability to optimise the short listing from a larger number of potential contractors can be as
important as the selection of the right bidder. This is because the quality of the final bidder
can only be as good as those short-listed (Khosrowshahi, 1999).
51
Risk factors leading to cost overrun in the delivery of highway construction projects.
The pre-qualification process basically consists of evaluating the available pool of potential
tenders, identified either by advertisement or invitation, against some given set of criteria in
order to find some group of firms who are all considered competent to carry out the work
and who are expected to provide an acceptable range of prices. The current practice of pre-
qualification is normally made by exercising accumulated experience and judgement in
assessing a given set of input variables such as reputation, past performance, financial
stability, current workload, firms resource capacity, experience records and technical
expertise (Lam et al., 2002).
Different project owners operate their pre-qualification processes under different objectives
and constraints. There are certain characteristics that distinguish public clients from private
clients. Khosrowshahi (1999) asserts that private clients tend to put emphasis on best service,
lowest price, best value-for-money and competitive advantage to their own organisation,
whereas public clients aim to demonstrate accountability for public funds and even-
handedness to suppliers and contractors.
On the other hand, Lam et al. (2000) points out that the criteria for pre-qualification may
differ from each other because the characteristics of the project and contractor are distinct
and dynamic. Masterman (1994) also points out that, in spite of different clients and projects,
there are common characteristics of contractor pre-qualification such as:
reasonable cost
reasonable quality
reasonable security.
In many cases, and particularly in the case of small contracts or much specialised work, a
number of contractors might be chosen simply on the basis of work that they have done for
an owner in the past. Most firms tend to accept such invitations even if they have no
intention of submitting a bona fide tender. Drew and Skitmore (1992) point out that this
method has deficiencies and should not be used on its own except for very small projects as
it is rather unreliable and may result in the selection of bidders that are either not interested
or are unable to provide competitive bids for a particular contract.
52
CHAPTER 2
Literature review into project risks and cost overrun
These ratings can be used to define parameters such as the maximum monetary amount of
work that can be allowed to a contractor or the maximum value of a work that a contractor
can bid for a particular project.
The Department of Public Works Bureau of Hong Kong prepare approved lists according to
relevant expertise, financial status, and technical and managerial capabilities of contractors
as well as their completion of other contracts. Only contractors on the corresponding list can
apply for contracts. Hong Kong Housing Authority maintains a comprehensive Performance
Assessment Scoring System to review the registered contractors performance levels of their
contracting works in the ongoing projects. Scores for contractors are calculated and a
comparative score league is formed. Contractors who fall in the upper section of this league
are invited for bidding in upcoming projects.
The Mass Transit Railway Corporation of Hong Kong, on the other hand, uses a set of pre-
qualification criterion for the evaluation of contractors. Members of a committee score
candidate contractors with respect to these and then use the scores for recommending
contractors pre-qualification. These criteria and sub-criteria are:
Corporate structure
o management
o relations
Experience
o performance on corporations contracts
o construction performance in similar projects
o work experience
Resources and facilities
o staffing
o labour
o construction plant
o planning/programming
o design
o manufacturing/fabrication
o sub-contractors
Workload
o current
o future
Support functions
o safety
o quality management.
53
Risk factors leading to cost overrun in the delivery of highway construction projects.
Although contractors are generally pre-qualified so as to minimise risks and failures, their
performance levels differ widely under different workloads and dissimilar environments.
Palaneeswaran and Kumaraswamy (2000) point out that many Federal states of the US have
recognised this and incorporate well-structured capacity rating assessments in their
contractor selection procedures. For example, the US Washington State Department of
Transportation looks at the total value of uncompleted prime contract work and also
maximum work class type value, such as asphalt/concrete paving, bridges etc., which a
contractor is permitted to have under contract at any one time within the pre-qualification
period.
In Australia, the Queensland Department of Main Roads specifies similar maximum capacity
ratings in their Project Delivery System Guidelines (Queensland Department of Main Roads,
2000). Once these levels are determined, this is used to determine a contractors eligibility to
bid for a new project. The pre-qualification systems of the Queensland Department of Public
Works and Housing, as well as the Queensland Department of Main Roads, use aspects of
technical capacity, management approach, people involvement, business relations and
financial capacity as the pre-qualification criteria. By evaluating applicants with respect to
these criteria, contractors are pre-qualified for two years and are placed at a pre-defined
level. The aim of the system is to ensure proper matching between the size and the
complexity of the projects and the abilities of the contractors (Topcu, 2004).
Hatush and Skitmore (1998) have all published research works on decision criteria used by
clients for choosing a construction contractor. Holt et al. (1994a) carried out a survey of 53
major UK construction client organisations to determine the decision criteria used for
contractor selection and the importance of these criteria in terms of influencing their choice
of contractors. Hatush and Skitmore (1998) report that all clients use a similar set of criteria
for contractor selection, but that the way clients quantify these criteria can be very different
in practice. In these works, a contractors bid amount appears to be the most dominant and
important criterion (Holt et al., 1994; Hatush and Skitmore, 1998).
54
CHAPTER 2
Literature review into project risks and cost overrun
Holt et al. (1994a) developed a method to evaluate contractor pre-qualification criteria and
provided guidelines for practitioners, highlighting areas to address when evaluating a
contractor based on a particular criterion. In Holt et al. (1996), they used cluster analysis as a
means of reducing a large number of potential bidders in order to identify only those suitable
to tender for a particular project. Ng (1996) investigated different decision support systems
(DSS) for contractor pre-qualification. Among the surveyed decision support systems were
database management systems (DBMS), expert systems (ES), fuzzy sets (FS) and case-based
reasoning (CBR).
For the project owner, it is vital to be able to select the best available contractor and
currently a standard pre-qualification model that can be used widely in the construction
industry is not readily available. Most public agencies and companies who perform pre-
qualification have their own specific pre-qualification model. Tran (2002) lists and describes
a number of international pre-qualification models found in literature. These include:
Qualifier-1, Contractor Pre-qualification Model. This employs a dimensional
weighting procedure that produces aggregate weighted ratings of candidate
contractors from questionnaires
Qualifier-2, Knowledge-based system for Contractor Pre-qualification. This is a
knowledge-based system in which the decision of pre-qualification is based on the
user using decision rules, not calculated scores, and where contractors meet the
decision criteria to proceed to the next step in the decision model
Fuzzy Set Pre-qualification Model. This uses fuzzy set theory to include the
uncertainties in the contractor evaluation process and takes into account the
uncertainty associated with the information from contractors, data reliability and the
uncertainty associated with the decision makers
Hypertext Decision Support Pre-qualification Model. This model uses pair-wise
comparison of factors and an aggregate weight for each contractor is developed
Cluster Analysis in Contractor Classification. This involves the evaluation of
candidate contractors using predetermined selection criteria. Cluster analysis is used
on raw data to divide it into a series of sub-sets by which contractors are ranked
Neural Network Pre-qualification Model. Neural networks are used to learn from
historical contractor data
Contractor Pre-qualification Process. This is based on fuzzy logic and is a three-stage
model employing a hierarchical framework
University of Toronto Contractor Pre-qualification Model. This method considers
contractors efficiency measures and calculates the efficiency score of each
contractor.
Once contractor pre-qualification is carried out satisfactorily by the client, the next stage in
managing risks in the delivery of projects is that of bid evaluation. The next section looks at
this process.
55
Risk factors leading to cost overrun in the delivery of highway construction projects.
The intensity of price competition between contractors has been blamed for many industry
ills, particularly in scenarios where price considerations have been deemed paramount. The
apparent, but often misleading, economy and convenience of awarding contracts to the
lowest bidder is particularly comforting in the public sector, where high accountability
regimes require onerous justification when awarded to other than the lowest bidder.
Selecting the best contractor is one of the most important decisions a client has to make.
Conversely, it is equally important for contractors to know why their bids are rejected. In the
case of a public client, the results and reasons for awarding a contractor and/or rejecting
others should be explicit because of public accountability. If the client is a private one,
contractors may also wish to know the reasons for their failure. In either case, the feedback
of contractors weaknesses can only help improve contracting firms to the betterment of the
industry.
Palaneeswaran and Kumaraswamy (2000) point out that all necessary information should be
provided in tender documents and all project procurement procedures and practices followed
by public procurement agencies should be transparent so that contractors can be become
aware of their strengths and weaknesses. This will thus improve their performance levels to
obtain more contracting opportunities while remaining competitive.
A low tender price may seem appealing to the client at tender stage, but the project may
encounter problems if the contractor is, for example, not able to complete the work on time
or even compromises on the construction quality in an attempt to reduce the contractors cost
(Mahdi et al., 2002). Crowley and Hancher (1995) caution that the awarding of a
construction contract to the lowest bidder without considering other factors can result in
problems such as cost overrun, delays and poor performance. They point out that it is very
real that the lowest bidding contractor may tend to adopt a confrontational claims oriented
position once the project is awarded as a means of making-up any financial short fall.
Smith (1995) argues that what is really required from a public sector point of view is not
simply to obtain the lowest price, but to obtain the best value for the money spent. According
to Smith, there is considerable doubt s to whether this can be achieved by taking the lowest
price in a largely unrestricted open competition (Smith, 1995).
Pakkala (2002) points out that, although the low bid process with 100% price criteria is the
easiest and simplest method of choice because it is quantatively measured, objective,
unbiased and not contestable, his study find that this low-bid process is not conducive to the
creation of innovation and long-term savings in the longer term. He also points out that
56
CHAPTER 2
Literature review into project risks and cost overrun
especially for road projects in particular, most project owners usually desire some elements
of quality and technical criteria since road project are used by the general public and there is
demand for quality and reliability. Some of the more common quality criteria identified are:
technical skills
personal skills (resources)
management team
supply chain management
methodology
environmental criteria
relevant experience
past performance.
Some of the quality criteria can be clearly identified as relevant, and important. However,
the difficulty arises when evaluating these aspects because some of them are personal skills
and not totally objective criteria.
Wang (2003) proposed that, before considering bids submitted by competing contractors for
a public procurement project, the owner should determine a project ceiling price or cost
estimate to use as a threshold or reference point for evaluating the bids and points out the
dilemma for the project owner is setting a ceiling price that is sufficiently low to satisfy the
owners interests in cost savings, yet sufficiently high to tender out the project. Wang (2003)
also points out that while some project owners will fix a ceiling price by an average bidding
ratio (winning bid divided by ceiling price) of past projects, most owners merely make a
decision based on gut feeling. Wang (2003) goes further in saying that, despite its popularity,
the historical average bidding ratio is inferior to more systematic evaluation methods. This is
because the major problem with it is that the ratio tends to be unrealistically low, especially
in a slow construction economy when bidders tend to propose unsustainably low bids simply
to get a contract. A further drawback is that the unique characteristics and risks of the project
are ignored.
Research in bidding has focused either on the development of bidding models to assist
bidders in winning contracts (Dozzi et al., 1996; Fayek, 1998) or on the evaluation of
competitive bids (Crowley and Hancher, 1995; Crowley, 1997; Williams, 2003). Statistical
methodologies have been developed which allow the detection of bids which are too low and
are usually considered to be discordant bids because they are in disagreement with the other
bids (Crowley, 1993). Furthermore, Crowley asserts that there is a historical relationship
with data between discordant bids and the magnitude of cost growth experienced within the
57
Risk factors leading to cost overrun in the delivery of highway construction projects.
particular project, with the owner taking the perceived benefit of a lower bid but incurring
the financial risk of having higher project cost growth.
Williams et al. (1999) have shown that there is a systematic relationship across highway
projects between the low bid and the completed project cost for competitively bid highway
projects in New Jersey and Great Britain. They found that as the magnitude of the low bid
increased, the resulting completed project cost tendered to increase absolutely and as a
percentage of the original bid. Williams (2003) suggested that large highway projects tend to
have much higher relative cost increases than small projects. His research also suggested that
methods of minimising the size of competitively bid highway projects, such as breaking a
project into smaller components, may have been useful in minimizing cost overruns.
Williams (2003) also concluded that cost increases for different types of construction will
follow different patterns and that it was not possible to generalise that the low bid and related
project were related in the same way for all types of projects.
Currently in construction, many project activities have been sub-contracted out by the main
contractor. Although sub-contracting has many advantages, it also brings additional risks to
the main contractor and to the client. These risks are uncertainties related to a sub-
contractor's technical qualifications, timeliness, reliability and financial stability (CII,
1989a). A large proportion (up to 90%) of construction activities can be sub-contracted on a
given project and so it is important to probe deeper down the supply chain in order to ensure
that appropriate sub-contractor selection is also a vital element managing risk in construction
projects. Improvements in sub-contractor selection processes have not received the attention
that one would expect from such a significant contribution to the industry.
Price-based selection often squeezes out the more responsible sub-contractors, driving down
both prices and performance levels. A growing appreciation of such anomalies has led to
recent initiatives towards longer term main-sub-contractor relationships (Matthews, 1996).
The evaluation of sub-contractors by main contractors (and vice versa) is a complex
undertaking. An example of a neural network based approach to sub-contractor rating is
proposed by Albino and Garavelli (1998).
The project cost estimate is primarily concerned with the cost of resources needed to
complete the project activities and include all the processes which are employed to maintain
financial control over a project. Common value auctions generally assume that a true cost
exists but it can only be estimated (Friedman 1956). Management theory, on the other hand,
58
CHAPTER 2
Literature review into project risks and cost overrun
tends to assume that the forecast or budget cost is fixed and that production activities are
sufficiently variable to be somehow manipulated to be kept within budget. In project
construction, this fixing of budget costs is often done by reference to the estimators little
black book of figures that are known from experience and thus can provide a reasonable
target cost.
The traditional approach to cost estimating is the derivation of a best estimate from the
knowledge of existing conditions based on current rates and prices in similar situations, but
with adjustments to reflect anticipated differences in say ground conditions, site accessibility
and other factors. A contingency allowance which usually varies according to type of work,
anticipated risk and project stage is added to cover unforeseen expenditure.
Abbasi and Al-Mharmah (2000), Hutchings and Christofferson (2000) and White and
Fortune (2002) examined various management tools and techniques that are used in project
estimating as well as in cost control processes. The most popular areas are detailed as:
analogous
parametric
detailed.
Analogous estimating: This is an approximate estimating method that compares costs with
similar past projects and which often depends on expert judgement. It can be used in the
preparation of the earliest price estimates for the client. The estimator should have the
relevant experience of estimating the cost of similar projects (Ashworth, 1994). Analogous
estimating using the actual cost of a previous, similar project is the basis for estimating the
cost of a current project and is frequently used to estimate the cost when there is a limited
amount of detailed information about the project. Usually estimators retain their own
database of historical project costs from which equivalent or similar cost information may be
drawn (Loftus 1999).
Detailed estimating (Bottom-up): Unit price estimates can be compiled when quantities of
work items may not be precisely determinable but the nature of the work is well defined
(Clough, 1986). This is best suited for works which are relatively simple and repetitive in
nature such as buildings. It involves estimating the cost of individual work items and the
synthesis of cost estimates from resource estimates made at the lowest possible level of
59
Risk factors leading to cost overrun in the delivery of highway construction projects.
A study by Hester et al. (1991) indicates that the estimating method and the accuracy of
project cost estimates could be a major reason for having cost changes. They found that that
most project owners assume that routine changes in the project will only affect the work in
the change area, whereas. In reality, the effects can extend well beyond that specific change
area Hester et al. (1991).
Various cost estimates are made at different stages of the process: project planning, decision
to build, tendering, contracting, and later renegotiations. Cost estimates at each successive
stage typically progress toward a smaller number of options, greater detail of designs, greater
accuracy of quantities, and better information about unit price. Thus, cost estimates become
more accurate over time, and the cost estimate at the time of making the decision to build is
far from final.
Accurate, early cost estimates for engineering and construction projects are extremely
important to the sponsoring organisation and the project team. For the sponsoring
organisation, early cost estimates are vital for business unit decisions that include strategies
for asset development, potential project screening, and resource commitment for further
project development. Inaccurate early estimates can lead to lost opportunities, wasted
development effort, and lower than expected returns.
Early estimates are typically plagued by limited scope definition (and thus high potential for
scope change) and are often prepared under tight time constraints. Furthermore, reliable cost
data are often difficult to obtain during the early stages of a project, particularly if basic
design and geographic issues remain unresolved. Early estimates, even when grossly
inaccurate, often become the basis upon which all future estimates are judged (with future
estimates sometimes being corrected to be consistent with early estimates). However, final
cost often exceeds the initial estimate.
The accuracy of early cost estimates for capital projects has been a major concern and a
subject of much scrutiny over the last 35 years. Hackney (1986) proposed the use of the
definition checklist for applying contingency to capital cost estimates and then validated the
checklist by comparing the definition ratings of 30 projects to their respective levels of cost
overrun. Hackney later revised the checklist in 1986 to specifically address process projects
and developed a separate checklist to apply the definition rating method specifically to
hazardous waste remedial projects (1986).
60
CHAPTER 2
Literature review into project risks and cost overrun
When the Suez Canal was completed in 1869, actual construction costs were 20 times higher
than the earliest estimated costs and 3 times higher than the cost estimate for the year before
construction began. The Panama Canal, which was completed in 1914, had cost escalations
in the range of 70 to 200% (Summers, 1967). More recently are examples such as Denvers
US$5 billion airport that was 200% overspent (Szyliowicz and Goetz, 1995), the 800 million
Danish Kroner Oresund rail/road bridge between Copenhagen and Malmo, Sweden that was
68% overspent, (Flyvbjerg et al., 2003). An example of cost underestimation in Australia is
the Sydney Opera House, with actual costs of approximately 15 times greater than those
projected (Hall, undated: 3).
On the other hand, Yeo (1990) points out that the entire Apollo space program cost was
about US$21 billion, which was only 5% above its initial estimated budget. But few are
aware that there was, in fact, a hefty $8 billion contingency already built into the initial
estimate (Morris and Hough, 1987).
Despite the large number of reported cases, it seems that construction ranging from the
simplest to more complex projects such as transportation systems and oil and gas platforms
have increasingly faced overrun. Raftery (1994) points out that construction projects tend to
have poor reputation for excessive time and cost overrun. Morris and Hough (1987), during a
study of records from different types of projects funded by the World Bank between 1974
and 1988, found that 63% out of 1778 projects had experienced significant cost overrun. In a
study of 204 construction contracts let between 1986 and 1990 in Italy, Tagliaventi (1991)
reported that an average cost overrun of 50%. Research carried out around 1992 by the
Transportation Research Board evaluated construction cost overruns on projects completed
for the Washington State Department of Transportation. The objective was to identify factors
that have the strongest association with construction cost overruns. Results of the analysis,
which examined information from 468 construction projects, indicated that cost overruns,
expressed as a percentage of the original contract amount, tended to increase with the size of
the project. Evidence also suggested that the cost overrun rate increased with the number of
bidders and with the increased dispersion of the various bids submitted per project
(Hinze and Walsh, 1997).
Construction cost estimating on major transport infrastructure projects has not increased in
accuracy over the past 70 years. The underestimation of cost today is in the same order of
magnitude that it was then (Flyvbjerg et al., 2002; Molenaar, 2005).
In the late 1970s, the US Department of Energy recognised the importance of accurate early
cost estimates and contracted with the Rand Corporation to study the capital cost estimation
problems associated with pioneer energy, process plants (Merrow, 1978). During the study,
61
Risk factors leading to cost overrun in the delivery of highway construction projects.
Merrow determined that 74% of cost growth is caused by underestimation (i.e. improper
estimation).
In 1988, the US Federal Construction Council commissioned the Building Research Board to
study construction estimating practices and to recommend techniques for improving the
accuracy of early cost estimates. The Board identified three recommendations for improving
estimating procedures (Morris, 1988):
1. developing standard terminology and formats for budgets and estimates
2. taking steps to ensure that estimators are properly qualified for conceptual estimating
3. expanding the use of parametric and probabilistic estimating techniques for
conceptual estimates.
Ogunlana (1989) carried out research into monitoring cost estimators accuracy of the last
cost forecast prior to tender for projects. Data for the research were collected from seven
County Councils Offices in the United Kingdom. The projects used in this study were all
road projects for new construction or facilities improvements such as roundabouts,
maintenance works, and expansion of existing roads or bypasses. A total of 36 projects were
analysed. The research included a survey to ascertain what level of error was acceptable in
seven design offices. Table 2.5 below shows that all the offices fell within a +/ 15% range.
Later Ogunlana (1991) reported on the accuracy achieved on different stages of road
estimates where the mean accuracy of statutory utility estimates was reported as 35.7% for
12 projects, with the overall value of accuracy for all elements of the projects being 12.8%
with a coefficient of variation of 11%.
Transportation projects have historically experienced significant cost overrun from budget
estimates. A recent study of 258 infrastructure projects spanning a time period of more than
70 years found that project costs are underestimated in approximately 90% of the
cases/projects, and the actual costs averaged 28% higher than estimated on this sample
(Flyvbjerg et al., 2002). Although highway projects fared better than rail and fixed-linked
projects, the sample still displays an increase in project costs of more than 20%.
High-profile highway projects in the US, such as Bostons Central Artery/Tunnel, known as
the Big Dig and Virginias Springfield Interchange have made engineers, contractors, and
the public acutely aware of the problem of cost overrun. For example, the Big Dig was
estimated at a cost of US$2.6 billion (1982 dollars) and was expected to be completed at a
cost of US$14.6 billion (2002 dollars) with completion then, anticipated for 2005 (NAS,
2003).
62
CHAPTER 2
Literature review into project risks and cost overrun
The literature shows that a wide variety of factors influence construction costs. In a study
conducted in Newfoundland on highways, Hegazy and Ayed (1998) found that season,
location, type of project, contract duration and contract size had a significant impact on
individual contract costs. Herbsman (1986) determined that, in addition to input costs of
materials, labour and equipment and the total volume of contracts bid each year (so-called
bid volume) all influenced project costs.
Some factors are intrinsically related to construction organisations which are solely
responsible for managing them, whereas others are closely related to socio-cultural,
economic, technological and political environments within which most organisations
operate.
Cost overrun arises primarily because of four factors (Yeo, 1990; Minato and Ashley, 1998):
external risk due to:
o modifications in the scope of a project
o changes in the legal, economic and technologic environments
technical complexity of the project due to:
o size
o duration
o technical difficulty
inadequate project management due to:
o the control of internal resources
o poor labour relations
o low productivity
unrealistic estimates because of the uncertainties involved.
Project scope describes the work to be performed in a project and so a cost estimate heavily
depends on this scope definition. Lack of proper scope definition has been stated to be a
major source of bad estimates (Cowie, 1987). Vagueness in scope has two implications for
cost control.
1. It decreases the accuracy of cost estimates
2. It creates a potential for changes in scope during the construction stage, which
generally results in an increase in cost to both owners and contractors.
The accuracy of a cost estimate is highly dependent on the level of detail in the project scope
definition. Peurifoy and Oberlender (1989) have divided the accuracy of an estimate into
three levels depending on the preciseness of scope definition. They found that a conceptual
estimate prepared from a project scope definition that does not include design information is
usually accurate within +40 and 10% of the actual cost. A cost estimate prepared upon
completion of the preliminary design is usually accurate within +25 and 5% of the actual
cost. A detailed cost estimate prepared upon completion of the final design is expected to be
accurate within +10 and 3% of the actual cost.
63
Risk factors leading to cost overrun in the delivery of highway construction projects.
Deciding on the likely cost-variation range of a project is subjective. However, the more
experienced the estimator, the greater likelihood that the subjective decision will be based on
objective experience (Bradley et al., 1990). One inherent difficulty in having an estimator
quantify the likely variation of an estimate is that estimators will not always agree on the
magnitude of the variation. It is an established fact that individuals attitude to risk vary.
Some people are by nature risk-seeking, whereas others prefer to avoid risk. These attitudes
are not necessarily based on past experiences in related fields; rather, they reflect the
individuals basic personality. Those individuals with a risk-aversion personality tend to
assume that, regardless of the apparent reliability of the data base, events will arise that
should be covered by an adequate contingency factor added to the best estimate. By
comparison, individuals who are by nature risk-takers will tend to underestimate the
likelihood of the best estimate being too low (Bradley et al., 1990).
There is frequently a substantial contrast when comparing the importance of early estimates
with the amount of information typically available (i.e. scope definition) during the
preparation of an early estimate. Although lack of scope definition often leads to inaccurate
estimates, early estimates (accurate or not) often become cast-in-stone, and future estimates
are unrealistically expected to agree with the early estimate.
In addition to the adverse effect on the accuracy of cost estimates, lack of proper scope
definition creates a potential for change or growth in scope during construction. Cowie
(1987) observes that the majority of cost growth during the construction period derives from
scope growth. Increases in scope, without extending a project delivery date, can only be
executed by increasing the resource intensity over and above the planned level. Hetland
(1994) points out that this often leads to increased unit cost and hence increased total cost of
a project.
Design and project specific factors affect the cost estimate of a project and can include such
other factors as vagueness in scope, design complexity, and project size (Akinci and Fischer
1998). Engineering designs have a high level of influence on project costs and sometimes
unsatisfactory design performance can lead to cost overrun (Barrie and Paulson, 1992).
Anderson and Tucker (1994) reported that their survey found that about one-third of
architectural/engineering projects miss cost and schedule targets. Chang (2002) notes that, as
reported by Smith (1996), there have been few instances where an engineering design is so
complete that a project could be built to the exact specifications contained in the original
design documents. Many construction problems are due to design defects and can be traced
back to the design process (Bramble and Cipollini, 1998).
In their research study to quantify the impact that project changes have on engineering and
construction project performance, Ibbs and Allen (1995) define change as any event which
results in a modification of the original scope, execution time or cost of work. Because
change may occur throughout all phases of a project, their research focuses on the
quantitative impacts that change has on the detailed design and construction phase of
projects. They found that project change has a large effect on the financial performance of a
construction project. Thomas (1985) studied highway construction programs and reported on
selected claims for project changes and cost/schedule overrun on these same projects. The
64
CHAPTER 2
Literature review into project risks and cost overrun
study concludes that project change has a direct effect on costs and schedules of construction
projects, primarily cost/schedule overrun.
Constructability in the highway industry has also gained considerable attention for adding
costs to designed projects. Early research by OConnor et al. (1991) identified that poor
specifications can cause construction rework and delays. Their findings suggest that 22% of
all constructability problems are related to ineffective communication of engineering
information, plans, and specifications, especially inadequacies in project specifications.
Anderson et al. (1999) also confirmed the issue of inadequacies in project specifications in
their research on state highways in the Unites States.
Chang (2002) reported that on four completed case study projects for environmental and
engineering design services for roadway construction projects in California carried out on a
cost-plus-fixed-fee basis, cost increases averaged 25%.
While the reason for cost overrun can be obvious, the problem still remains that an estimate
is a forecast of a cost to be incurred sometime in the future the problem is that the future
is not always predictable. Kayode (1979) reports that project cost overrun are caused by
rising costs largely due to inflation, inadequate analysis and inadequate information. Orji
(1988) is of the opinion that the causes include certain government fiscal/monetary policies,
poor costing of projects, inflation within the economy and some practices of project
participants, especially those involving government projects. A further reason advanced for
the incidence of project cost overrun is attributed to costing methods (Akpan and Igwe,
2001).
The quality of project design documentation has fallen alarmingly over the past 15 to 20
years (Tilley, et al., 2000). It has been reported that many client organisations routinely
tender major projects on sketch plans of partially completed documentation to transfer
design risk to the managing contractor (McLennan and Jorss, 2006). They also reported that
the findings of the task force set up by the Queensland Division of Engineers Australia in
2004 included the following root causes:
inadequate project briefs based on unrealistic expectations of time and cost
lack of integration along the supply chain linking parties, and between project phases
poor understanding and low skills in risk assessment and management
inadequate use of CAD/computer technology in the design process and in the
compilation of specifications.
65
Risk factors leading to cost overrun in the delivery of highway construction projects.
building size
market condition
number of bidders.
The development of cost forecasting models is one of the techniques for assessing project
costs, and hence, ultimately, the projects cost at completion. The development of cost
forecasting models is summarised by Raftery and Ng (1993) as:
First Generation Models: Element cost planning in the UK and by bidding models in
the US
Second Generation Models: using regression analysis
Third Generation Models: probabilistic estimates, frequently based on Monte Carlo
techniques.
Raftery (1994) also suggests that range estimating can be developed to show the form/shape
of project cost distributions. In 1996, Skitmore indicated this might be done parametrically
for tender price forecasts, by fitting one of the standard probability density functions to the
cumulative frequency of the tender price forecast/actual values (Skitmore, 1996).
Oberlender and Trost (2001) and Trost and Oberlender (2003) report on quantitative data
they collected from completed construction projects in the process industry. Their data are
analyzed using factor analysis and multivariate regression analysis. The resulting model,
known as the estimate score procedure, allows a project team to score an estimate and then
predict its accuracy based on the estimate score. They identify six main factors affecting
estimate accuracy and, in order of significance, are:
1. basic process design
2. team experience and cost information
3. time allowed to prepare the estimate
4. site requirements
5. bidding climate
6. labour climate.
Weverbergh (2002) comments that market data without background information on costs,
uncertainly and privileged knowledge only leads to very crude types of cost estimates, but is
likely to be the best that can be provided under the circumstances.
In the past, forecasting future highway construction costs has been achieved in basically
three ways. Firstly, unit rates of construction such as dollars per lane kilometre by highway
type have been used to estimate construction costs in the short term (Hartgen and Talvitie,
1995; Stevens, 1995). However, this method has generally been found to be unreliable,
because site conditions such as topography, in-situ soil, land prices, environment and traffic
loads differ sufficiently from location to location to make the use of average prices
inaccurate for individual projects (Hartgen and Talvitie, 1995).
Secondly, extrapolation of past trends, or time-series analysis, has been used to forecast
future overall construction costs (Koppula, 1981; Hartgen et al., 1997). Typically in these
analyses, construction costs are collapsed down into overall expressions of construction
66
CHAPTER 2
Literature review into project risks and cost overrun
indices. These types of models are usually only used for short-term forecasting because of
their reliance on the notion that past conditions and standards are maintained into the future.
Thirdly, models have been established that describe construction costs as a function of
factors believed to influence construction costs. The relationship between construction costs
and these factors is established from past records of construction costs. Typically, these
models are used to estimate the cost of individual contracts only. These models, with their
relational structure, are the only models expected to provide reliable long-term estimates
(Wilmot and Cheng, 2003).
Recent research has attempted to apply artificial intelligence to the prediction of completed
project costs. These efforts start with the a priori assumption that each project is considered
unique. Attempts have been made to develop extremely complex models that incorporate the
many factors and their interactions that can affect the cost of a construction project. These
efforts have yielded software that requires significant computing power, a sophisticated user,
input of subjective information, and significant required input data. Ashley (1989) describes
a general performance model for analysing individual construction projects that requires the
analysis of many factors such as labour productivity, design, quality, procurement and the
interaction between the factors.
Leu et al. (2001) point out that during project implementation many environmental variables
such as weather, space, congestion and productivity level dynamically affect activity
duration, and hence cost. Leu et al. (2001) also assert that activities need to be executed in
cost effective ways that are based on the principle of construction time/cost trade off. Several
early mathematical and heuristic models have been generated to solve construction time-cost
trade-off problems (Panagiotokopoulos, 1977). By measuring the cost of risk as an input into
a projects estimate, the success or otherwise of initial risk assessments may be seen as a
variance from the projects estimated cost at completion. It has been postulated that this can
provide a measure of the successful outcomes of projects through defined key performance
indicators. Central to any project improvement is the ability to measure aspects of the project
outcomes, including the cost of risk, for benchmarking and performance measurement
purposes.
An early study by Charles and Andrew (1990) identifies factors that influence the
construction change order rates causing cost overrun. These consist of:
size of project
difference between the low bid and the estimate
type of construction
level of competition
quality of the contract documentation
interpersonal relations within the project
policies of the contractor.
It was discovered that a cost overrun rate of 1 to 11% is more likely to occur on larger
projects than smaller ones. Contracts with an award amount less than the estimate are found
to be more likely to have cost overrun rates above 5%.
67
Risk factors leading to cost overrun in the delivery of highway construction projects.
In terms of managing risk on a project, contingency can take many forms. It may be a time
allowance in the program of work for delay such as wet weather, a cost allowance in the
project cost estimate to account for the residual risk accepted by the project manager or a
contingency process in case an event happens. Cost contingency is included within a budget
to represent the total financial commitment for a project client and the quantum of such
contingency is of critical importance to projects.
More recently, the Association for the Advancement of Cost Engineers (AACE 2000: 28) defines
contingency as:
An amount of money or time (or other resources) added to the base estimated amount
to achieve a specific confidence level, or to allow for changes that experience shows
will likely be required.
Contingency can be divided into two categories of risk known unknowns and unknown
unknowns (Project Management Institute, 2000; Hillson, 1999) and can be invoked for
events within the defined project scope that are:
unforeseen (Moselhi, 1997; Yeo, 1990)
unexpected (Mak et al., 1998)
68
CHAPTER 2
Literature review into project risks and cost overrun
Traditionally cost estimates are treated as deterministic, having point estimates for each cost
element that are based on their most likely value (i.e. a single value estimate based on the
most likely values of the cost elements) (Mak et al., 1998). These single values may or may
not accurately indicate the possible value of the estimate, and they certainly do not indicate
the possible range of values an estimate may assume (Toakley, 1995).
The cost performance of construction projects is a key success criterion for project sponsors.
Construction projects are notorious for running over budget (Hester et al., 1991; Zeitoun and
Oberlender, 1993). When cost estimates are being prepared for feasibility, planning and then
design stages, the cost of risk is reflected by the inclusion of a contingency sum. The
contingency sum, usually expressed as a percentage markup on the base estimate is used in
an attempt to allow for the unexpected.
One of the most commonly used methods for contingency allocation is the classes of
estimate method. Almost all major companies in every industry group develop their own
criteria for the classification of cost estimates. As an example, Blok (1982) defines five
classes of estimate as shown in Table 2.6.
Project cost estimates are prepared at different stages over the life cycle of a project. The
estimating techniques used and the associated probable error ranges depend on the
availability of project information and engineering data. For instance, classes IV and V are
basically the study estimates, whereas class III is the budget estimate prepared prior to
budget approval and the final commitment of funds for project execution. Class I and II
estimates are control estimates where the level of accuracy is high (Blok, 1982).
The Department of Main Roads Queensland expects the following contingency allowances
shown in Table 2.7 in their project estimates unless there are other reasons why they should
be adjusted (Queensland Department of Main Roads, 2000).
69
Risk factors leading to cost overrun in the delivery of highway construction projects.
Baccarini (2004) details numerous estimating methods available for project cost contingency
as shown below:
traditional percentage
method of moments
Monte Carlo simulation
factor rating
individual risks expected value
range estimating
regression
artificial neural networks
fuzzy sets
controlled interval memory
influence diagrams
analytical hierarchy process.
When estimating, the most common method of allowing for uncertainty is the addition of a
percentage contingency figure to the most likely estimate of the final cost of the known
works. Although this contingency can be calculated in various ways as detailed previously,
the most common way is to consider around 10% of the estimated project cost (Burger,
2003). Hartman (2000) argues that this is an unscientific approach and thus a reason why so
many projects finish over budget.
As Yeo (1990) also points out, the most common method of contingency allocation could be
regarded as overly simplistic and heavily dependent on an estimators faith in their own
experience.
70
CHAPTER 2
Literature review into project risks and cost overrun
Quantification of contingency allowances for cost estimating items can also be achieved by
applying the risk management processes detailed in AS/NZS 4360 (Standards, 1999).
Historical events may be used as a guide; however estimators and project managers need to
use their experience and professional judgement to weigh the competing factors to arrive at
the most likely value. Where risks are significant and complex, a statistical evaluation such
as the Monte Carlo method can be used. The objective of contingency allocation is to ensure
that the estimated project cost is realistic and sufficient to contain any cost incurred by risks
and uncertainties.
However, Thompson and Perry (1992) outline several weaknesses of using a contingency
amount as follows:
The percentage figure is most likely arbitrarily arrived at and not appropriate for the
specific project
There is a tendency to double count risk because some estimators are inclined to
include contingencies in their best estimate
A percentage addition still results in a single-figure prediction of estimated cost,
implying a degree of certainty that is simply not justified
The percentage added indicates the potential for detrimental or downside risk; it does
not indicate any potential for cost reduction and may therefore hide poor
management of the execution of the project
Because the percentage allows for all risk in terms of a cost contingency, it tends to
direct attention away from time, performance and quality risks
It does not encourage creativity in estimating practice, allowing it to become routine
and mundane, which can propagate oversights.
Eden et al. (2005) point out that it may be important to require different contingencies for
different elements of a project. However, the establishment of a range of contingencies can
require a considerable amount of work by estimators, so they simply add on a 10%
contingency across the board for example in order to acknowledge the difficulty of pinning
down project uncertainty.
When Euro Tunnel (the private company that owns the tunnel under the English Channel)
went public in 1987 to raise funds for the project, investors were told that building the tunnel
would be relatively straightforward. Under Water (1989:37) reported that, in respect to cost
escalation, the prospectus read:
Whilst the undertaking of a tunneling project of this nature necessarily involves certain
construction risks, the techniques to be used are well proven . . . . The Directors,
having consulted the Mitre dOeuvre, believe that 10% would be a reasonable
allowance for the possible impact of unforeseen circumstances on construction costs.
Cooper et al. (1985) points out that, in estimating the cost of a large hydroelectric
development, the cost variability and uncertainty was acknowledged by incorporating a
contingency allowance in the estimate of 10%. This was calculated as a proportion of the
71
Risk factors leading to cost overrun in the delivery of highway construction projects.
total construction cost after subtracting engineering, management and owners costs. In this
case, the contingency proportion reflects past experience industry practice and the feel of the
cost estimating team.
Touran (2003) proposed a probabilistic model for the calculation of project cost contingency
to counter initial project cost estimates by considering the expected number of changes and
the average cost of change. The model assumed a Poisson arrival pattern for change orders
and independent random variables for various change orders. From these elements, Touran
(2003) calculates the probability of cost overrun for a given contingency level. Touran
(2003) also asserts that project owners, such as transportation agencies, who are usually
engaged in specific types of construction projects can, by reviewing historical data of a
specific transit agency, calculate rates of change, size and distribution of changes and
prepare risk profiles or cumulative probability curves for various values of contingency.
Touran (2003) also suggests that such outcomes can be used at the budgeting phase of new
projects to ensure consideration is given to potential cost overrun after projects commence.
Where there is some form of tender documentation provided for bidders, a portion of the
contingency will usually be transferred to the provisional sums section in these documents.
For construction projects that usually use a governments fixed quantities contract, the
magnitude of the final account variations that comprise additions and omissions can be
compared with the contingencies included in estimates. This comparison can be used to
assess the accuracy of the allowance made for the contingencies at the early estimate stage.
With the traditional approach of informed guessing at contingencies, it can then be perceived
that the size and the level of quality of projects will be affected by the contingency amount.
Estimators tend to include an inflated buffer in the contingency estimate. Raftery (1994)
identifies personal bias and differences in personal risk attitude. Kahnaman and Tversky
(1972) refer to this as conservatism. The term conservatism originated from studies of
human information processing in that individuals tended to revise their opinions in the light
of new evidence to a lesser degree than would be expected of an optimal information
processing system. This tendency for humans to be less extreme than optimal in their
judgements has been termed conservatism in probabilistic information processing.
Moreover, due to the effect of negative sanctions (i.e. imposing a penalty for an
underestimate, where tender bids are above the pretender estimate but no reward/penalty for
an overestimate), an over-exaggerated contingency is not uncommon in many project
estimates. For public works projects, this leads to misallocation of resources as more than
sufficient funds are locked up in projects. Musgrave and Musgrave (1984) reported that the
mis-allocation of resources severely disturbs the important fiscal functions of public sector
spending in the areas of:
1. provision of social goods
2. distribution of income and wealth
3. maintenance of a stable economy in the aspects of employment and price levels.
In the case of public expenditure, the task becomes one of distributing the available fund
allocations among competing projects. Having chosen projects for inclusion, the budgeted
capital expenditure is then committed for the duration of the program. Sometimes project
teams can inflate the contingency allowances in an attempt to avoid the need to seek
72
CHAPTER 2
Literature review into project risks and cost overrun
additional funds if budgets become over spent. However, this inflation can become
exaggerated if there is no heavy call on the contingency fund beyond what might reasonably
have been expected for the project. Consequently, project budgets can be seriously under
spent. The magnitude of under-spending can be so large that it is possible to identify
facilities that were previously foregone but could have been included in a construction
program in the first place. Explaining situations such as this can be just as embarrassing for
clients as seeking additional funds because of budgets being inadequate to meet unforeseen
costs.
In some areas of the public sector, there is a tendency to remove contingency provisions in
budget submission, as contingencies are often seen as fats (Yeo, 1990). Consequently there
is no allowance to express the anticipation of risk or for the lack of confidence in project
estimates. As well, the engineering and construction complexities of projects are often
overshadowed by economic, societal and political challenges. In addition to these challenges,
a number of observers suggest that project estimates are purposely misrepresented in
quantum in an effort to secure project approval (Flyvbjerg et al., 2002).
Contingency can have a major impact on project outcome for a project owner. If contingency
is too high it might encourage poor cost management, cause the project to be uneconomical
and so aborted. It may also lock up funds not available for other projects or activities
(Flyvbjerg et al., 2002). On the other hand, if the contingency allocation is too low, then it
may be too rigid and set an unrealistic financial environment, resulting in unsatisfactory
performance outcomes (Dey et al., 1996).
Under many companies risk management strategies, risks are reviewed at intervals
throughout the life cycle of the project and assessments updated to reflect the current level of
uncertainty surrounding the project (Flyvbjerg et al., 2002). Risks for which contingencies
are provided early in a project may some later time be overcome by further investigation or
design modifications. For example, a contingency allowance for rock in cuttings early in the
project may be replaced by specific quantities and costs following geotechnical
investigations to minimise the specific risk exposure. The amount of contingency is
reassessed at project review points to reflect current knowledge and level of uncertainty of
the project with a view to forecasting the most likely outcome (Dey et al., 1996).
HM Treasury (1993) identifies two major categories of contingency that can be incorporated
into construction projects:
Design contingency this allows for changes during the design process for such
factors as incomplete scope definition and inaccuracy of estimating methods and data
(Clark and Lorenzoni, 1985).
Construction contingency this is for changes during the construction process. Under
a traditional procurement arrangement, the project owner engages others to produce
the design before competitively selecting the construction contractor. Subsequently a
contract is signed between the project owner and the contractor, which typically
contains a variations clause to allow for changes and provide a mechanism for
determining and valuing variations. Construction contingency exists to cater for these
variations allowable under the contract between the project owner and contractor
(Staugas, 1995).
73
Risk factors leading to cost overrun in the delivery of highway construction projects.
Mak and Picken (2000) state that contingency can be compared with the total approved value
of contract variations to assess the accuracy of the contingency.
Baccarini (2004) analysed project cost data of 48 road construction projects from an
Australian government road authority for the estimation of construction contingency. He
reports that the organisation uses a traditional percentage approach for estimating
contingency. The main findings of his analysis include:
Construction contingency is on average 5% of award contract value, whilst variations
were 10% of award contract value. This shows a shortfall in contingency of 4.6%
The amount of estimated contingency is significantly inadequate to cater for the total
value of contact variations, by an average shortfall of 47%
There are no significant correlations between project variables and cost contingency
that might be used to predict cost contingency.
The literature supports the notion that accurate, early cost estimates for engineering and
construction projects are extremely important to the sponsoring organisation. Accurate cost
estimates are vital for business unit decisions that include strategies for asset development,
potential project screening, and resource commitments for further project development. The
literature survey has revealed several research studies on highway construction projects
which attempt to predict the amount a construction contract might increase while taking into
account various factors that could be used in such predictions. Research to date has generally
revolved around the cost increase in contracts within projects. Several research studies have
demonstrated that changes initiated during construction projects have a large effect on their
financial performance. Research also demonstrated that estimating methodology and
accuracy of cost estimates can be major reasons for cost increases. Research studies have
also been conducted in order to predict the extent to which the cost of a construction contract
might increase, taking into account various project prediction factors.
Most research has involved considerable effort investigating the construction aspects of
project delivery and the impact of that on the performance and other aspects of projects.
Changes in construction contracts lead not only to increased costs, but also to contract delays
which then affect project delivery. In the history of construction, the nature of construction
risks has led to many cost overruns. The first step in clients managing risks causing cost
overrun is their identification. In the area of highway project construction, the literature
review has unearthed little empirical research that has determined client risks leading to cost
overrun associated certain types of highway projects and their delivery methods.
The literature review has identified many important stages during the delivery of projects
where the management of construction risks are important. The literature review concludes
that, although many of the risk techniques are effective for the particular types of projects
they were applied to, the approaches generally treat projects as independent entities with
little attempt to categorise projects into specific project sub-types from which detailed
74
CHAPTER 2
Literature review into project risks and cost overrun
analyses can be undertaken. In the past, highway cost estimating models have been
established that describe construction risks as a function of factors believed to influence
construction costs. Typically, the models established in this manner have been used to
estimate the cost of individual contracts only, not project budgets.
Empirical research is now required to assess whether certain highway projects properties and
delivery methods indicate higher propensity to cost overrun. The research needs to be
focused on the client, not the contractor, and with a particular focus on overrun relating to
the decision-to-build baseline budget. The outcome from such an analysis would enable
clients to be in a better position to take advantage of these findings when considering
specific types of future highway construction. Data from different highway types should be
studied to identify if there is any correlation between project cost overrun and the type of
highway project constructed.
Chapter 3 follows and describes the methodology adopted in the research to assess client
cost overruns in highway projects.
75
Risk factors leading to cost overrun in the delivery of highway construction projects.
76
CHAPTER 3
Research methodology
CHAPTER 3
Research methodology
3.1 Introduction
Research methodology is the collection of methods to change or create new beliefs and
knowledge. The term methodology means different things to different people and its
meaning continues to evolve (McCuen, 1996). Research methodology deals with the
methods for creating knowledge about the world and the interpretation of this knowledge in
light of the ontological and epistemological positions (Reich, 1994). Ontology deals with the
nature of the things we know about the world or the nature of the world, while epistemology
deals with the relation between humans and their knowledge. Typical epistemological
questions are:
What can we know?
How do we know?
What is truth?
Is there a priori knowledge, and if so, of what?
The following section provides a general description of the research strategy adopted for this
thesis, as well as justification of the methodology.
77
Risk factors leading to cost overrun in the delivery of highway construction projects.
As well, there are a number of different approaches to empirical research as shown in Figure
3.1.
However, this hierarchical view of research strategies is incorrect in that experiments with an
explanatory motive have always existed. Yin (2003) also points out that each research
strategy is not distinguished by this hierarchy, but by conditions such as the:
type of research question posed
extent of control that an investigator has over actual behaviour events
degree of focus on contemporary as opposed to historical events.
The first and most important condition for differentiating among the various research
strategies is the identification of the type of research question being asked. The second
condition is the current state of knowledge of the variables involved in the research (Bennett,
1991). In Chapter 1, the following two research questions were formulated:
1. What client risks are present during the delivery of highway construction projects in
Queensland, Australia that lead to significant project cost overruns?
2. How does the amount of highway cost overrun in such highway projects correlate
with their project types, size, delivery processes and client project risks when
historical project data is analysed?
78
CHAPTER 3
Research methodology
Table 3.1 displays the conditions for various research strategies. The what and how research
questions are likely to favour the use of archival analysis, history or case study research
strategies
On the other hand, although a survey is a strategy that allows a researcher to collect data
directly from sources in a systematic fashion, there are some important disadvantages (Stone,
1978). These include:
decreased willingness or refusal of people to respond to survey probes, because of
suspicion, fear and other form of resistance
most surveys are one-offs and so their capacity to generate data with which to test
causal connections among variables is limited
the sample survey can be an extremely expensive research strategy because of
administrative and personal costs
standardised response formats of many sample surveys may force respondents to
subscribe to statements they dont fully endorse
surveys that are used to collect data may have low response rates.
79
Risk factors leading to cost overrun in the delivery of highway construction projects.
This research adopts the historical analysis approach as the required methodology as it
provides an insight into the current research problem through the examination of what had
happened in the past using analysis, analogy and trend extrapolation of historic data.
The approach entails researching construction delivery practices to identify risk occurrences
as well as risk constraints and processes to minimise client risk exposure during project
delivery that lead to cost overrun. It has also provided a means for the development of a
consensus of risk factors based on expert opinions and trend exploration and model
development of cost overrun based on historical project data and project attributes. The main
activities of this investigation have been presented in Figure 3.2 which provides a detailed
explanation of the research process based on the research plan.
Data / information/
knowledge/
experience Results
Project attributes.
Establish study data. Factor analysis Modelling of cost
Build databases for of cost overrun overrun factors
Conclusions
data collection. data for using multivariate
Recommendation
highways regression
Further research
analysis.
options
Determine cost Develop
overrun factors. principal cost
Determine highway overrun factors
project types. in project types
80
CHAPTER 3
Research methodology
Use nominal group technique (NGT) to elicit, review and prioritise principal cost overrun
4
risk groupings and highway project types.
Undertake data analysis and statistical modelling using multivariate linear regression
5 analysis. Establish correlations between client risks causing cost overrun, project attributes
and project programmed cost.
The following sections explain the justification and conduct of the individual methodology
stages adopted in the research.
This first stage focused on a thorough review and evaluation of the research literature
relating to the research topic. This included topics such as:
nature of project risk and uncertainty
organisation risk culture
project risk management
project procurement practices in managing construction risk
improving project delivery methods so as to manage risk
pre-qualification linkages to contract default
bid evaluation and contract selection practices
project budget estimating procedures
final project cost forecasting
role of contingency in managing cost overrun
expert elicitation of project risks.
81
Risk factors leading to cost overrun in the delivery of highway construction projects.
The research stage required a sample of highway infrastructure projects that was appropriate
for this area of research. The data sample was required to be large enough to allow statistical
analyses of cost overrun factors and project costs. Data on actual costs in transportation
infrastructure projects were relatively difficult to come by. One reason is that such data is
time consuming to generate. For private projects, such data are often classified as
commercial-in-confidence and thus kept from the hands of competitors. Unfortunately, this
also tends to keep private project data from the hands of researchers. For both public and
private projects, data on actual costs have been held back by project owners because more
often than not, actual costs reveal substantial cost escalation that was often considered an
embarrassment to project owners. For public sector projects, funding and accounting
procedures often made it difficult to keep track of the multiple and complex changes that
occur in total project costs over time. For large public projects, reconstructing their actual
total costs would typically entail long and difficult archival work and complex accounting.
The study for this research into cost overrun into various types of highway construction
projects focused on a state highway authority that manages a network of highways and
bridges network within the state of Queensland, Australia. The authority is the Queensland
Department of Main Roads (QDMR) and it is responsible for the 34,000 kilometres of the
states highway network, representing 20% of Queenslands total road network of 174,000
kilometres. It also manages 2740 bridges and 20,000 major culverts. This highway network
represents the states largest single physical asset, with a replacement value in 2005 of
A$26.6 billion.
Queenslands highway network traverses a wide variety of geographic features that make its
highway network unique in Australia. These features include:
freight/tourist passenger routes that traverse coastal topography requiring high
capacity hydraulic designs to accommodate high flood immunity and speedy water
runoff
highway routes that traverse mountainous topography requiring limits on climbing
and descent design grades to accommodate heavy highway freight movements
high traffic density urban highway networks
82
CHAPTER 3
Research methodology
low volume rural main road network interconnecting low rainfall semi-arid regional
centres
high volume national highway and main road networks that are located in coastal
tropical climates with extremely high seasonal rainfalls
medium volume national highway freight corridor that is located predominately in
low rainfall semi-arid and arid rural regions but with expansive soils and unseasonal
extreme weather conditions necessitating specialised pavement designs
limited availability of high quality road-making materials in most rural areas,
necessitating long distance haulage of road making materials
regional transport regulations that allow volume loading of multiple articulated cattle
road trains requiring a network of high speed, high strength, all-weather and wide
highway pavements.
The research was validly limited to the highway network of the state of Queensland. No
other highway network in Australia that has this combination of unique and diversified
geographic and functional constraints and limitations are managed by the one highway
organisation.
The research focused on highway infrastructure projects from QDMR that contained data on
substantial project cost overruns.The highway project construction data was collected from
the published Roads Implementation Program Yearbooks of the Queensland Department of
Main Roads (QDMR) over the period from 1995 to 2003 with a portfolio of projects in the
sample worth approximately A$1 billion. The data were available as published documents.
The project construction cost data was selected on the basis of data availability and all
projects that were known to have information on substantial cost overrun were considered
for inclusion. The construction projects covered 78 different local authority geographic areas
of Queensland out of a total of 125. They represent state highway works in 65% of the
administrative areas of Queensland over the stated time period.
All projects subsequently analysed were those delivered by the traditional design-bid-build
method. The use of project data using this delivery mechanism was justified as there is
considerable support worldwide for delivery of infrastructure projects, such as highways,
using the traditional method. In a global perspective, the delivery of capital infrastructure
investment projects varies in practice from country to country, however, highway
infrastructure project delivery in such countries as Australia, Canada, Great Britain, New
Zealand, Sweden and the US all use the traditional model of design-bid-build in the great
majority of cases (Pakkala, 2002).
It is important to note that the project programmed cost was defined as budgeted project
construction cost at the time of the decision to build the project. This was consistent with
Flyvbjerg et al. (2002) who pointed out that the decision-to-build estimate method focused
on decision making and hence on the accuracy of the information available to decision
makers. It was this cost estimate at the time of making the decision to build that was of
primary interest in the research. This decision-to-build value was also the international
standard for measuring the inaccuracy of cost estimates (National Audit Office and
Department of Transport, 1992; Nijkamp and Ubbels, 1999). For the purpose of the research,
the total project cost estimate included the estimated costs of all component activities from
the initiation of the project design to construction finalisation.
83
Risk factors leading to cost overrun in the delivery of highway construction projects.
The sampling criterion was that of data availability. The highway project construction data
available was only available in a consistent format from 1995 and so did not allow an
assessment of cost overrun prior to 1995. While the project sample was not perfect, it was
considered by the researcher to be the best obtainable, given the limited availability of useful
data in the public domain in this field of research. A data integrity process also required to
ensure that the data presented was true and factual representations of the highway
organisation's historic data. since that data was contained within documents which
represented public statements of the organisations planning, construction and maintenance
activities of the state government, then it was considered that the facts contained within these
documents were true and factual. Each document was authorised by both the state
government Minister and the Director General of the state highway organisation.
Price Index of the Output of the Road and Bridge Construction Industry: This index was first
produced in an experimental form in the March quarter 2000, and first published in the
September 2002 issue of Producer Price Indexes, Australia (ABS catalogue no. 6427.0).
The Road Input Cost Index (RICI): Price indexes for highway projects in Queensland have
been indexed since 1984 and up to 30 June 2004 by the application of a calculated index.
The RICI was used in this research to index up the historical highway costs because it
provided the only continuous highway focused price indices for the financial years
1995/1996 through to 2002/2003 for which the historic project data were available. The
84
CHAPTER 3
Research methodology
Road and Bridge Construction Price Index was not adopted as the index was only time-series
back calculated to an index reference period of 199899 = 100.
The first step in stage three the research required the determination of highway work types
and project cost overrun factors. The available highway data contained individual
descriptions of all the work types as well as the reasons for individual projects having
exceeded the clients programmed budget. These work types and reasons for cost overrun
were recorded in Excel spreadsheets by the researcher. Where common work types or cost
overrun factors occurred across projects, single work types and cost factors were recorded to
cover incidences. All unique work types and reasons for cost overrun were recorded
individually. The research process used the experience in highway construction and the
professional judgement of the researcher to determine the listings of work types and reasons
for cost overrun.
The focus of this analysis was based on the clients exposure to project cost overrun, not that
of contractors delivering the projects. The client focus required that a number of
considerations that were identified in the literature research had to be taken into account
when reporting the cost overrun factors. These included:
the use of design-bid-build contracts that could lead to higher client exposure to
design risks
pre-qualification of contractors that has the potential to limit client risk exposure to
contract default
contract payment types that focused on schedule of rates and bill of quantities
contract clauses that were designed to reduce the clients exposure to certain
construction risks
tender evaluation techniques
contract provisions that limit the client's exposure to adverse physical and latent
conditions and wet weather events.
There are different methods of extracting the factors from a set of data. The method chosen
matters more when the sample is small, the variables are few, and/or the communality
85
Risk factors leading to cost overrun in the delivery of highway construction projects.
estimates of the variables differ. By far the most common form of factor analysis is Principal
components analysis (PCA). PCA seeks a linear combination of variables such that the
maximum variance is extracted from the variables. It then removes this variance and seeks a
second linear combination which explains the maximum proportion of the remaining
variance, and so on. This is called the principal axis method and results in orthogonal
(uncorrelated) factors. PCA analyses total (common and unique) variance.
The PCA technique of factor analysis was considered appropriate because of the limited a
priori knowledge available about the number of different cluster relationships that could be
expected for the data sample. The literature also indicated that PCA provided a deterministic
method to group elements into meaningful subdivisions in order to overcome
multicollinearity problems in the project data. As well, it was a statistical procedure that
could uncover relationships among many variables and in the context of this research the
variables were the specific cost overrun reasons in highway construction projects. In the
factor analysis technique, correlations and interactions among variables are summarised into
a small number of underlying factors. The method aimed at identifying key variables or
groups of variables that controlled the cost overrun system under study.
In adopting PCA, there were four decisions that needed consideration as part of the factor
analysis methodology. These were:
1. method of factor extraction
2. sample size
3. type of factor rotation
4. number of reduced factors (components) to be adopted.
There are different methods of extracting the factors from a set of data. The method chosen
matters more when the sample is small, the variables are few, and/or the communality
estimates of the variables differ. By far the most common form of factor analysis is Principal
components analysis (PCA). PCA seeks a linear combination of variables such that the
maximum variance is extracted from the variables. It then removes this variance and seeks a
second linear combination which explains the maximum proportion of the remaining
variance, and so on. This is called the principal axis method and results in orthogonal
(uncorrelated) factors. PCA analyses total (common and unique) variance. The PCA
technique is appropriate where a limited a priori knowledge is available concerning the
number of different cluster relationships that could be expected for the data sample (Hair et
al., 1998). PCA also provides a deterministic method to group elements into meaningful
subdivisions in order to overcome multicollinearity problems in the project data.
Guidelines for the minimum sample size needed to conduct factor analysis suggested a
minimum sample size of 100 to 200 observations (Guadagnoli and Velicer, 1988). Some
researchers have suggested the ratio of sample size to number of variables as a criterion, with
recommendations ranging from 2:1 through to 20:1. PCA requires a large sample size. It is
based on the correlation of the variables involved, and correlations usually need a large
86
CHAPTER 3
Research methodology
sample size before they stabilise. Tabachnick and Fidell (2001: 588) have advised the
following regarding sample size: 50 observations is very poor, 100 is poor, 200 is fair, 300 is
good, 500 is very good and 1000 or more is excellent. As a rule of thumb, a bare minimum
of 10 observations per variable is desirable to avoid computational difficulties.
In order to facilitate the interpretation of factors, factor analysis requires the rotation of axes.
The rotation procedure does not affect the goodness-of-fit of the factor solutions but serves
to make the output more understandable. Three rotation techniques are in general use:
varimax, equimax and quartermax. Of these, the most popular is Kaiser's varimax algorithm,
which is known to provide the best parsimonious analytical solution (Harman, 1967). This
minimises the number of variables with high loadings on factors, thus causing the factor
loadings of each variable to be more clearly differentiated.
The scree plot of eigenvalues against the number of factors was also used as part of this
process. The plot was used to show the point at which the eigenvalues began to level off and
this plot was used as a cut-off point to support the adoption of the desired number of factors
(Velicer and Jackson, 1990).
Several pre-tests are available to measure the sample characteristics necessary for successful
factor analysis. One is the Kaiser Meyer Olkins test (KMO) for sampling adequacy. KMO
values vary from 0 to 1.0 and values closer to 1 are better. An overall KMO should be 0.60
or higher to develop successful factor analysis (Hutcheson and Sofroniou, 1999). Another
test is the Bartlett Test of Sphericity, which checks if the sample was randomly drawn from a
population in which the correlation matrix was an identity matrix. This uses the determinant
of the correlation matrix to tests the null hypothesis that the correlation matrix is an identity
matrix using a chi-square approximation (Bartlett, 1947) and is particularly relevant when
dealing with a relatively small sample of data (<100) and with a relatively large number of
variables (>10). The Bartlett test sets up a chi-square approximation to determine whether
the developed correlation matrix is an identical matrix in the analysis. Bartletts sphericity
test is particularly relevant when dealing with a relatively small sample of data (<100) and
with a relatively large number of variables (>10). A minimum value of 700 was adopted for
the Bartlett test (Hutcheson and Sofroniou, 1999).
A further test is to examine the Anti-Image correlation matrix. The diagonals on the matrix
should have an overall Measure of Sampling Adequacy (MSA) of 0.5 or above (Hair et al.,
1998). Individual variables can be considered for elimination from the analysis if they are
87
Risk factors leading to cost overrun in the delivery of highway construction projects.
3.9 Stage 4: Use nominal group technique (NGT) to elicit, review and
prioritize cost overrun risk groupings and highway project types
Stage four of the research methodology involved the elicitation of the grouping of cost
overrun factors from experts. Elicitation using experts has been a formal process of obtaining
information or answers to specific questions where the information was highly subjective.
The process has been used by quite a number of disciplines (Ayyub and McCue, 1997).
Koskinen et al. (2002) explain tacit knowledge as expressing itself in human actions in the
form of evaluations, attitudes, points of view, commitments and motivation. The expert
elicitation was adopted in order to obtain a comparison between the factors derived from the
factor analysis technique and those obtained from carrying out an expert elicitation process.
From previous interviews by the researcher, it was important to take note of interviewees
concerns about time management. In the rapidly paced world of construction management, it
was found that an expert elicitation method needed to be efficient because experts in
construction management were continually confronted with insufficient time (Laufer, 1996).
The selection of the elicitation technique was therefore determined to a large extent by the
time that experts were prepared to allocate to the research project.
A number of techniques were available that provided the opportunity to evaluate tacit
knowledge based on the experience of individual practitioners. The three expert elicitation
techniques considered for the research project were:
1. focus group
2. Delphi technique
3. nominal group technique.
Focus group research had long been prominent in marketing studies (Morgan, 1997), in part
because market researchers sought to tap emotional and unconscious motivations not
amenable to the structured questions of conventional survey research. It was reported that the
interaction among focus group participants brought out a range of perspectives through the
language that was used by the members. The reactions of each person spark ideas in others,
and one person may fill in a gap left by others.
The difference between focus group research and group interviewing was important in
considering the appropriate technique. In group interviewing, a standard survey instrument is
administered to respondents simultaneously. In contrast, there is no standard instrument in
focus group studies, only a topic being explored through the exchange of group discussion.
Survey research required a priori theory or at least a list of subtopics as a guide for selection
of items to be included in the survey instrument. In focus group research, there was no a
priori theory.
Linstone and Turoff (1975) identified the Delphi technique as one form of anonymously
eliciting the opinions of experts concerning events and the reasoning behind the opinions. In
88
CHAPTER 3
Research methodology
the process, experts are asked to revise their opinions in the light of the information
contained in the feedback. This sequence of questionnaire and revision is repeated until no
further significant opinion changes are expected. Sackman (1972) identified important
shortcomings in the Delphi which were important in considering the form of elicitation that
was adopted in this research project. These shortcomings were characterised as:
information and questions provided to experts needed to be carefully reviewed to
ensure objectivity
difficulty in summarising and presenting a common evaluation scale to a group that
could be interpreted uniformly by the experts
benefits of experts participating in active dialogue was missed
difficult and time consuming in exploring disagreements between experts.
Delbecq et al. (1986) summarises the NGT decision making process used as:
1. silent generation of ideas
2. round-robin feedback from group members to record ideas
3. discussion of each recorded item for clarification and evaluation
4. individual voting on priority of items with the group decision being mathematically
derived through rank-ordering or rating.
The use of face-to-face meetings to reach agreement sometimes causes problems because:
a senior member of the group (e.g. a boss or person with a dominant personality)
could sway opinions in a manner inconsistent with the information presented
people could be unwilling to change opinions when stated publicly
people could grandstand or posture by sticking to beliefs that may not have been
appropriate, and so show they were actively engaged in the process.
89
Risk factors leading to cost overrun in the delivery of highway construction projects.
All these issues are needed to be managed properly in an elicitation process (Lifson, 1972).
The Nominal group technique was seen as providing the most appropriate method for
eliciting expert opinion as it had to potential to demand less valuable time from the experts
and still provide the appropriate focus and outcome sought.
Possible candidates to the nominal group panel were considered from within several sources:
professional organisations
project management consultants
contracting organisations
client organisations.
All potential group members were contacted via telephone to determine their willingness to
participate in the research and whether they met the following criteria for selection:
over ten years of contract management experience in highway construction projects
had acted in the role of the client representative in at least three infrastructure
construction projects
willing to participate in the entire process
willing to share ideas regarding construction/contract risk assessment and analysis
techniques as part of the research.
A questionnaire was used to obtain relevant information about individuals that profiled
potential members. The quality expert elicitation process involves at least three types of
participants:
1. specialist
2. analyst, and
3. generalist (Keeney and von Winterfeldt, 1991).
1. Specialist: The specialist should be at the forefront of knowledge in their field, and be
recognised as leaders by their peers. They should have the command of knowledge and
flexibility of thought to apply their expertise to the issue at hand. As well, they should be
capable of translating their knowledge into judgements relevant to the problem.
2. Analyst: The analyst conducts the elicitation of the judgement of specialists. Analysts
have knowledge and expertise in statistics, probability theory and decision analysis, as well
as having experience with expert elicitation processes. Their task was seen as assisting the
specialists to formulate the issues, to decompose them, to articulate their judgements, to
check consistency and to help document the specialists reasoning.
3. Generalist: It was important to include the generalist in any elicitation in any complex
technological problems where the application of the specialist knowledge to the overall
problem is not obvious. Generalists are experts with broad knowledge about many or all of
the issues under study. Generalists do not need to be at the forefront of their field, but they
should excel at communicating with the specialists, especially in the translation of project
needs into specialist language. Generalists facilitate the essential bridging between various
issues.
90
CHAPTER 3
Research methodology
Group composition and characteristics are determined by its size and member characteristics.
When considering the size of the group, there were two conflicting tendencies to consider.
An increase in the group size would be accompanied by an increase in diversity of skills,
experience and knowledge. However, the increased group size could also result in a decrease
in individual participation as the larger a group gets the more distorted the issues become in
favour of those members with low thresholds of participation (Chapman, 1998). Also, as the
size of the group increases, the intimacy of interaction decreases, with more members feeling
threatened and inhibited from participation because of the increased impersonality of the
situation. In addition, the size of group membership tends to be related to cohesiveness with
the greater the size of the group, the lower the cohesiveness. Many studies have identified
the conditions under which groups tend to become more cohesive. These include physical
proximity, similarity of work, homogeneity or similarity of attitudes and values, ease of
communication and limiting the size of the group to ideally below 12 (Chapman, 1998).
Research in expert elicitation suggested that five specialists were usually sufficient to cover
most of the expertise and breadth of opinion (Clemen and Winkler, 1985). As well, they
suggested that there may be a need for two or three additional generalists and two or three
analysts. For some issues, the diversity of opinions of the set of experts was important, as
was the individuals expert credentials.
A Likert scale was used in the NGT process to record the comfort level that workshop
participants felt in regard to decisions made by the group. Likert scales and Likert-type
scales (the latter also known as rating scales) are a type of scale that is commonly used to
measure human constructs (Vojir, et al.) The direct application of the Likert scale has often
been adopted in multi-criteria decision making models. The one disadvantage of this scale
method is that it does not allow a fuzzy range of importance to fit through a weighting
assessment.
For ranking the importance of project risk factors to the potential for cost overrun, it was
considered that a collation of the mean and standard deviation of all the participant scores
provided for each grouping would not adequately determine an overall relative ranking as
this would not reflect any relationship between the groupings (Chan and Kumaraswamy,
1997). Therefore the ranking was determined by calculating the Importance Index for each
group, using the following formula:
Where w was the weighting, ranging from 1 to 10, given to each factor; W was the highest
weight, i.e. 10; f was the frequency of the response; and F was the total number of NGT
workshop participants.
91
Risk factors leading to cost overrun in the delivery of highway construction projects.
The final step in the nominal group technique involved the workshop evaluation completed
by the expert group. This sought feedback on the NGT process, the process data provided
and the timing and composition of the workshop.
Results from analysis of models can often provide valuable contributions to knowledge bases
used for decision making (Nilsen and Aven, 2003). However, the development of the model
also implies the trading off of complexity in order to provide satisfactory representation of
the system under study. The ultimate goal of the statistical modelling process was not just to
produce a model but also provide one that was sufficiently credible and acceptable to
decision makers. In addition, a robust statistical technique was needed that would cope with
any undetected data outliers, otherwise the results of the model could have proved
misleading. Within the general classification of a model, there have been a variety of model
techniques that have been applied in the analysis of risk variables. Some are shown in Table
3.4 below.
92
CHAPTER 3
Research methodology
The method of quantification and the analysis techniques that can be used are dependent on
each other, and both are restricted by the information available. Some of the techniques
noted, particularly the network-based solutions, require a large amount of information that is
not always available. Monte Carlo analysis technique has often been used in simulation and
sensitivity analysis to determine the significance of individual sources of project risk and
other project factors such as cost overrun (Flanagan and Norman, 1993). However, given
that a sensitivity analysis is univariate and the effects of project cost overruns are numerous
and vary simultaneously, the use of Monte Carlo analysis in this research was not considered
appropriate.
Data mining was seen as a useful research tool for pre-processing structured information
from available data and using statistical analysis to uncover patterns and relationships hidden
in project related databases. It was reported that data mining compresses even more value
out of repositories of information (Marco, 2000).
Multivariate statistical analysis is a scientific inference used in analytical work where the
problem of inferring from sets of performance measurements on a number of individuals of
objects is constantly faced. The history of science confirms that such inferences can be
successful, and can handle inferred reality as well as a means of reducing the number of
variables. Researchers in sciences such as behavioral, biological or physical have long since
abandoned sole reliance on classical analysis methods (El-Choum, 1994). The construction
industry has been seen as a good example as each project is unique and therefore, separate
treatment is needed to overcome major difficulties that cannot be noticed otherwise.
With multivariate statistical techniques two functions can be derived these correspond to
the distinction between descriptive and inferential statistics. Descriptive statistics requires no
assumptions whatever about the distributions from which observations have been sampled,
while most common multivariate significance tests have been based on homogeneity and
normality assumptions.
Multivariate regression technique was seen as the most effective tool to manage multiple
project variables in the development of a model to determine relationships between projects,
project risks and project cost overrun. Multivariate regression has been the most common
method of modelling construction costs in the past (Koppula, 1981; Blair et al., 1993; Elhag
and Boussebaine, 1998). Multivariate regression is the estimation of the conditional
expectation of a random variable given other random variables and involves providing model
solutions which could be used as the basis for decision making (Mason et al., 1989; Neter et
al., 1990).
The multivariate statistical analysis method is designed for an assortment of descriptive and
inferential techniques that can either predict or measure sets of variables as found in
construction cost forecasting. Regression analysis can summarise data and qualify the nature
93
Risk factors leading to cost overrun in the delivery of highway construction projects.
and strength of relationships among the variables. New values of dependent variables based
on observed values can also be predicted by using regression analysis. The method was used
to assesses the strength of the relationship between each of a set of explanatory variables
(sometimes known as independent variables) and a single response (or dependent) variable.
Although there were several alternative techniques available in regression, the research
method adopted was the method of least squares.
Before using regression analysis, it was necessary to understand assumptions and limitations
of the processes. The following key factors were considered when the multivariate
regression model was developed (Carver and Nash, 2005):
problem formulation
adequate, high quality project data
selection of appropriate project variables.
Problem formulation: Studying the relationship of multiple project and cost overrun
variables is especially troublesome because the desired knowledge is complex in nature and
many different combinations of factors are needed to be examined. As well, it is impractical
and difficult to make use of a large number of parameters in the model. However, with the
complexities in construction, these parameters were amplified and it was necessary to
identify the key factors before they were used in the model. The focus of this part of the
research project was a decision-oriented model in the highway project construction
environment which had a statistical approach and which used the relationship between
qualitative project variables in order to define the correlation between project cost overrun
and the number of variables associated with the projects being analysed.
Adequate, high quality data: The project data consisted of two types quantitative and
qualitative. Quantitative data represented the quantity or amount of an element. As an
example, the dependent variable of the proposed model (cost estimate percentage overrun) is
a quantitative variable (Marco, 2000). In contrast, qualitative (or categorical) data, for
example the project delivery method or the reasons for the cost overrun, require no
quantitative interpretation and so are only classified into categories (Kantardzic, 2003).
Selection of appropriate variables: The aim of this stage of the research was to identify if
any project variables had a relationship to the accuracy of project cost contingency, for
example project size or location. Any project variables that were found to have a relationship
could then be used to predict more accurate project cost contingencies; or simply highlight to
estimators that when these variables were present, then there was a need to pay particular
attention to them in estimating.
Variable relationships can take on many different forms, most of which can be extremely
difficult to handle in a regression analysis. In many instances it may not be reasonable to
assume a linear relationship between the dependent variable and the independent variables.
This fact has often become obvious when constructing useful scatter diagrams of sample
observations. For this thesis, it has been assumed that the variables are linearly related and
the relationship between Y and each of the independent variables was linear.
Multivariate regression analysis was used and applied to the set of historical data to
investigate the development of a cost overrun model in terms of a regression coefficient for
each explanatory variable. The multiple regression took the form of the following model:
94
CHAPTER 3
Research methodology
where Y was a dependent variable (i.e. % or project cost overrun), a0 was a constant
indicating the intersect with Y axis, bn were partial regression coefficients, Xn were
independent variables, and e was the error term. The standardised versions of the bn
coefficients were the beta weights, and the ratio of the beta coefficients was the ratio of the
relative predictive power of the independent variables under consideration.
The purpose of the model was to predict the value of an observable quantity Y = (X) in
terms of a set of quantities X = (X, X2, X3,.Xn.), and the functional relationship ()
between Y and X. The function was typically established on the basis of a mixture of
accepted intuitive assumptions developed from the researchers experience in highway
construction regarding the data being analysed (Nilsen and Aven, 2003).
The usefulness of the multiple regression models was judged by the value of the multiple
correlation coefficients and by the examination of residuals. The performance of the model
was tested using Pearson's correlation coefficients.
Applying multivariate regression analysis to a set of data results in what are known as
regression coefficients, one for each explanatory variable. These coefficients give the
estimated change in the response variable associated with a unit change in the corresponding
explanatory variable, conditional on the other explanatory variables remaining constant. The
fit of a multivariate regression model can be judged in various ways, for example,
calculation of the multiple correlation coefficient or by the examination of residuals.
The reason is that the regression coefficients and their associated standard errors are
estimated conditional on the other explanatory variables in the current model. Consequently,
if a variable is removed from the model, the regression coefficients of the remaining
variables (and their standard errors) will change when estimated from the data excluding this
variable. As a result of this complication, other procedures have been developed for selecting
a subset of explanatory variables, most associated with the response.
The basic assumptions/requirements that needed to be satisfied in the model analysis using
least squares estimation to yield reliable estimates of a and b in the regression models, the
following are to be satisfied:
Normality: at each possible value of X , the random disturbances are normally
distributed
Zero mean: at each possible value of X, the mean of the random disturbances is zero
Homogeneity of variance: at each possible value of X, the variance of the random
disturbance is constant
Independence: at each possible value of X, the value of the variance is independent of
all other deviations.
If these conditions are not satisfied then there is the risk that any inferences about tests for
significance and confidence intervals will become misleading. Because random disturbances
cannot be directly observed in a model as the location of a true regression line cannot be
95
Risk factors leading to cost overrun in the delivery of highway construction projects.
known, then it was planned to examine the hypothesised relationship that is known as
residuals or the differences between an estimated regression line and observed Y values.
In order to identify any project variables that may be correlated with project cost overrun, a
correlation analysis is undertaken. This technique examines the performance of the models
and the relationship between variables. The correlation coefficient (R) is a measure of how
the actual and predicted correlate to each other in terms of direction. Values for R range
from 0 to 1. The closer the correlation value is to 1, the more correlated the actual, and
predicted values are. Pearson's correlation (R) was used to examine the relationship between
data that was collected on an interval or ratio scale (Norusis, 2005).
Whilst there are several different criteria that can be used for developing the rank order of
regression models in terms of goodness of fit, the most often used criteria was the R2 and
adjusted R2 statistics. The R2 was adopted in the research as it allowed direct comparison of
the best model identified (Neter et al., 1990). R2 (coeffiecient of multiple determination) is a
statistical indicator usually applied to multiple regression analysis. It compares the accuracy
of a regression model to the accuracy of a trivial benchmark model wherein the prediction is
the average of all the example output values.
Statistical regression is a way of computing the method of least squares in stages. In stage
one, the independent best correlated with the dependent is included in the equation. In the
second stage, the remaining independent with the highest partial correlation with the
dependent, controlling for the first independent, is entered. This process was repeated, at
each stage for previously-entered independents, until the addition of a remaining
independent does not increase R2 by a significant amount (or until all variables are entered).
The methodology involved testing within the established multivariate regression analysis for
sets of independent variables that explained a proportion of the variance in a dependent
variable at a determined significant level (R2). This established the relative predictive
importance of the independent variables by comparing beta weights. Hierarchical regression
analysis was adopted that determined how much variance in a dependent highway project
variable that was explained by one or a set of new independent highway project variables.
Testing of the significance of the difference of R2 was adopted and carried out in order to
determine if the addition or subtraction of targeted independent variables helped the model
significantly. The use of residual analysis was important because most assumptions of
multiple regressions cannot be tested explicitly but gross violations, such as outliers or
extreme cases, can be detected. When excluded, this can yield a better set of results.
The most commonly used multivariate analysis method is the stepwise selection method
(Norusis, 2005). However all three of the following methods were adopted in selecting the
included and excluded independent variables in the statistical regression analysis:
forward selection
backward selection
stepwise selection.
96
CHAPTER 3
Research methodology
Forward selection This method starts with a model containing none of the explanatory
variables. In the first step, the procedure considers variables one by one for inclusion and
selects the variable that results in the largest increase in R2. In the second step, the procedure
considered variables for inclusion in a model that only contained the variable selected in the
first step. In each subsequent step, the variable with the largest increase in R2 is selected,
until, according to an F-test, further additions are judged not to improve the model.
Backward selection This method starts with a model containing all the variables and
eliminated variables one by one. At each step the variable chosen for exclusion was that
leading to the smallest decrease in R2. The procedure was repeated until, according to an
F-test, further exclusions did not represent a deterioration of the model.
Stepwise selection This method involved the combination of the previous two approaches.
It starts with no variables in the model. Then variables are added, as with the forward
selection method. In addition, after each inclusion step, a backward elimination process is
carried out to remove variables that are no longer judged to improve the model. Caution in
this process is needed as the stepwise procedure is claimed to be controversial because it
includes independent variables based on statistical criteria and not theoretical ones (Bryman
and Cramer, 1999).
Automatic variable selection procedures are exploratory tools and the results from a
multivariate regression model selected by a stepwise procedure should be interpreted with
caution. Different automatic variable selection procedures can lead to different variable
subsets since the importance of variables is evaluated relative to the variables included in the
model in the previous step of the procedure. Care was required in the automatic procedures
for selecting subsets of variables (Agresti, 1996).
The regression analysis required the adoption of an adequate sample size of project data for
analysis (Tabachnick and Fidell, 2001)
For testing b coefficients, the sample size adopted was N >= 104 + m where m =
number of independent variables
For stepwise regression, N >= 40m was adopted since the stepwise method would
easily train to noise and not be generalised in a smaller dataset
There needed to be at least 10 to 20 times as many samples of data as there were
variables otherwise the estimate of the regression was very unstable and unlikely to
be replicable
For the significance test of R2, the sample size adopted was N >= 50 + 8m where m =
number of independent variables.
3.10.5 p-value
In the multivariate regression analysis, the statistical model required the reporting of
appropriate F-tests and t-tests. For each test, the p-value was reported. The p-value derived
depended on a given sample and provided a measure of the strength of the results of tests, in
contrast to a simple reject or do not reject criteria.
The p-value is a measure of how much evidence existed against a null hypothesis, with the
smaller the p-value, the more evidence of the null. The p-value is combined with the
97
Risk factors leading to cost overrun in the delivery of highway construction projects.
significance level to make decisions on a given test of hypothesis. In such a case, if the p-
value is less than some threshold (usually 0.05, sometimes a bit larger e.g. 0.1, or a bit
smaller e.g. 0.01) then the null hypothesis is rejected. If the null hypothesis is true and the
chance of random variation is the only reason for sample differences, then the p-value
provides a quantitative measure of the strength of the model.
Traditionally, p-values less than 0.01 are considered highly significant and less than 0.05
significant. A larger p-value means that a deviation can be due to chance. If the p-value is
measured less than some threshold (usually 0.05, sometimes a bit larger like 0.1 or a bit
smaller like .01) then the null hypothesis is rejected. Typical interpretation values for p are
shown in the following Table 3.5.
When a p-value is associated with a set of data, it is a measure of the probability that the data
could have arisen as a random sample from some population described by the statistical
model. A p-value is a measure of how much evidence you have against the null hypothesis
(H0). The smaller the p-value, the more evidence you have. It is possible to combine the
p-value with the significance level to make a decision on a given test of hypothesis.
The distribution of p-values under null hypothesis H0 is uniform, and thus does not depend
on a particular form of the statistical test. In a statistical hypothesis test, the p-value is the
probability of observing a test statistic at least as extreme as the value actually observed,
assuming that the null hypothesis is true. The value of p is defined with respect to a
distribution. Therefore, we could call it model-distributional hypothesis rather than the null
hypothesis. This means that if the null had been true, the p value is the probability against the
null in that case. The p-value is determined by the observed value.
3.10.6 t-test
The t-test is used to test that the regression coefficient is zero and to assess the significance
of individual b coefficients. All variables not significant at the 0.05 level or better are
removed from the equation. The t-test assumes randomly sampled data for the significance
tests and the t-test is not used for dummy variables.
98
CHAPTER 3
Research methodology
A major disadvantage of using the '% cost over' as the dependent variable is that it can
provide a wider range of data outliers in the regression analysis. One approach in reducing
this potential problem may be to divide the projects into different delivery types and then
recalculating the '% over-cost' variable according to each project delivery type. This may
require more data that are presently not available to the research project.
As well, one of the assumptions in using regression analysis is that the independent variables
have a constant variance. If this assumption is violated then the '% over-cost' model will be
required to be transformed in order to satisfy the assumption.
3.12 Conclusion
This methodology describes the process of investigating and assessing the correlations
between project risks, project types and cost overruns on highway projects procured within a
public highway agency. Aspects of the methodology adopted included: reviewing literature
on project risk and project cost overrun, determining and establishing a source of historic
project data, recognising project risk factors, determining highway project types and
undertaking statistical modelling to establish correlations between project cost overrun
elements and project attributes. These research methods were applied to research data
consisting of public records of highway construction projects, their budgeted estimates and
final costs at completion on projects which had exceeded budget by more than 10 % between
1995 and 2003. These public records were published by the QDMR.
The description of the highway project data collected for the research, the analysis
techniques using the methodology and the statistics are described in the following Chapter 4.
99
Risk factors leading to cost overrun in the delivery of highway construction projects.
100
CHAPTER 4
Data collection, analysis techniques and statistics
CHAPTER 4
4.1 Preamble
The research methodology has specified five stages for approaching the research questions.
Previous Chapter 2 has focused on Stage 1 of the methodology, namely the literature review
This chapter focuses on stages 2 through 5 of the research questions.
To address the first research question of: What client risks are present during the delivery of
highway construction projects in Queensland, Australia that lead to significant project cost
overruns? Stages 2, 3 and 4 of the research process have been developed as follows:
Stage 2: Establish data sources of highway construction projects
Stage 3 (a): Determine cost overrun factors from historic project data
Stage 3 (b): Use factor analysis (principal component analysis) and factor rotation on
cost overrun factors to consolidate data
Stage 4: Use nominal group technique (NGT) to elicit, review and prioritise principal
cost overrun risk groupings and highway project types.
To address the research question of: How does the amount of highway cost overrun in such
highway projects correlate with their project types, size, delivery processes and client
project risks when historical project data are analysed? Stage 5 was developed to undertake
data analysis and statistical modelling using multivariate linear regression analysis and to
establish correlations between client risks causing cost overrun, project attributes and project
programmed costs.
101
Risk factors leading to cost overrun in the delivery of highway construction projects.
It was ascertained that the data presented was true and factual representations of the highway
organisations historic data. since that data was contained within documents which
represented public statements of the organisations planning, construction and maintenance
activities of the state government and each document was authorised by both the state
government Minister and the Director General of the state highway organisation.
A sample of that data is given in Table 4.2 below. This sample data referred to highway
construction projects completed in the 20022003 financial year.
As shown in Table 4.2 above, the various columns contained important project data. These
data are described in the following sections.
The first column of Table 4.2 contains the Project number. This provides a systematised
identifier that includes the geographic Local Authority in which the project was constructed,
the name of the road on which it was built, and a sequential number that this project is the
x th project constructed on that particular road. It provides the following information about
project 13/25A/48:
The third column of Table 4.2 contains the Description of works and describes the type of
highway project constructed for the particular project. This description also indicates
whether the project was a highway or bridge-related project. A refined listing of works
descriptions was produced by combining all duplicate work type descriptions and by
consolidating like construction processes into representative processes. Table 4.3 below
shows the final 17 work types that were developed from the full listing of work descriptions
for use in further analysis.
102
CHAPTER 4
Data collection, analysis techniques and statistics
Column four of Table 4.2 shows the Method of delivery. All projects analysed used the
traditional procurement method of having separate design and construction steps. At the time
of this research, QDMR supported three viable and performing project delivery sectors
across Queensland the private contracting sector, local governments and RoadTek
which was a constituted government-owned-enterprise operating under the National
Competition Policy and Trade Practices Act 1974 which required evidence of procurement
best-value-for-money through competitive bids or comparisons for sole invitee work. Five
types of delivery methods were recorded in the project data that have been variously
employed on projects to satisfy the clients project delivery policy.
103
Risk factors leading to cost overrun in the delivery of highway construction projects.
4. The project is located in the Local Authoritys geographic area and uses the Local
Authoritys own workforce and its sub-contractors
5. Sole invitee Local Authority: This is a formal agreement entered into between the
client highway organisation and a particular Local Authority. The agreement is on the
basis that the construction of the project is carried out by the nominated Local
Authority in order to provide continuity to that clients workforce. (This form of
delivery was superseded by method 3 described above in 2001.)
6. Sole invitee RoadTech: This is a project delivery method where a formal agreement is
entered into between the client highway organisation and its internal workforce. The
contract price is a negotiated price based on agreed schedule of unit rates for the
project construction items. It is carried out by the clients own workforce in order to
provide continuity to their own workforce.
In the highway data analysed, there were 140 projects categorised as open tender and this
constituted 60.6% of the total projects analysed. Forty six (19.6%) APIC contracts were
identified and 23 (10%) were APPC contracts. The number of sole invitee Local Authority
projects and sole invitee RoadTech projects were 8 (3.5%) and 14 (6.0%) respectively. Table
4.4 shows the composition of the methods of delivery for the projects analysed.
The APIC project delivery method and the Sole invitee RoadTech delivery methods became
essentially the same over the duration of the research data period as the client organisation
used their workforce to deliver projects on a negotiated price basis. These two recorded
delivery methods were combined for the purpose of future analysis. As well, the APPC
project delivery method and the Sole invitee local authority delivery method were also
considered essentially similar for the purpose of the research. The client organisation
contracted particular local authority organisations to deliver the highway projects using their
workforce on a negotiated price basis. Therefore these two recorded delivery methods were
combined for the purpose of future analysis. A code was attached to each of the three revised
project delivery methods to allow for future analysis. These are shown in Table 4.5.
104
CHAPTER 4
Data collection, analysis techniques and statistics
Column three of Table 4.5 shows the combined revised percentages of projects representing
open tender and the negotiated price for internal and external workforce project delivery
methods.
The programmed cost shown in Column five of Table 4.2 was the client-allocated budget to
deliver the project in millions of A$. This was the project budget at the decision-to-build
stage of the project. This budget was derived from the traditional engineering analysis of the
design plans and drawings and the estimate was typically decomposed into line items
representing the major activities and acquisition costs of the project. These included:
conducting investigations and developing the design
detailing the design
acquiring land
altering public utility plant
construction
project administration and handover.
Project cost variability and project uncertainty was included in each programmed cost by the
incorporation of a contingency allowance in the budget. This was calculated as a percentage
on cost to the estimate totals of the six cost elements above. The policy of the client
organisation was that the same level of project contingency of 10% was attributed across all
project budgets over the analysis period. Table 4.6 shows the minimum, maximum, average
and standard deviations of the project programmed costs for the analysis period.
The actual cost shown in Column six of Table 4.2 was the projects reported information
about the actual outcome of the project as incurred by the client to deliver the project in
millions of A$. The reported actual cost included the total value of change orders etc. in
addition to the original programmed cost. It included the actual final costs of the component
costs included in the projects programmed cost. Table 4.7 shows the range of actual projects
costs over the analysis timeframe.
105
Risk factors leading to cost overrun in the delivery of highway construction projects.
The % cost overrun shown in Column seven of Table 4.2 was determined as the difference
between the clients actual project cost and their programmed cost, expressed as a percentage
of the programmed cost for each project. Table 4.8 shows the range of percentages of cost
overrun in the projects analysed.
In order to evaluate the size of highway projects by means of the individual project
expenditure over the full analysis period, all the reported project expenditures were indexed
up to 2003 project prices. This process involved the application of price indices to the client
project costs for the years 1997 through to 2002.
Price indexes for highway projects in Queensland have been available since 1984 and in
intervening years up to 30 June 2004 by the application of the Road Input Cost Index (RICI)
as detailed in Chapter 3. Price increases since 199900 included the impact of the
introduction of the Australian governments 10% Goods and Services Tax (GST). It should
be noted that the RICI was an input price adjuster index, rather than a cost of construction
index. (A price index is different from a cost index in that it does not provide an indication
106
CHAPTER 4
Data collection, analysis techniques and statistics
of the actual cost of constructing roads rather, its main purpose is to provide a way of
deriving constant prices for inputs to road expenditure.).
Appendix A contains the RICI indices that have been compiled form the Australian Bureau
of Statistics for the period 1984 to 2004.
Table 4.9 details the RICI indices adopted over the analysis period. Column 3 lists the
percentage factor used to factor up the historical project cost information. These factors
ranged from 3.4% to 16.3% and were applied to projects programmed and actual costs for
the corresponding financial years in which the projects were constructed.
All the historic project cost data of the clients programmed costs and actual costs were
adjusted to 200203 financial year figures by using the % factors in Table 4.9 above.
Appendix B contains a full listing of adjusted programmed and adjusted actual project costs
that were used in subsequent data analyses.
107
Risk factors leading to cost overrun in the delivery of highway construction projects.
The available highway data contained individual descriptions of all the reasons for individual
projects stated by the client as having caused the clients programmed budget for the project
to be exceeded by 10% or more. A sample of this data was shown under Columns two to
seven of Table 4.2.
It was considered that the reasons documented were true and factual representations of the
historic project data. Since that data was contained within documents which represented
public statements of the organisation's planning, construction and maintenance activities of
the state government, then it was considered that the individual reasons for projects
overrunning in costs contained within these documents were true and factual as each
document was authorised for publication by both the state government Minister and the
Director General of the state highway organisation.
These reasons were transferred from the public documents and recorded in an Excel
spreadsheet by the researcher for further analysis. Where common cost overrun factors
occurred across projects, single cost factors were recorded to cover incidences. All unique
reasons were recorded individually. This step in the research process drew on the experience
in highway construction and the professional judgement of the researcher to extract a
consistent group of reasons for the project overruns.
Where there were common cost overrun factors identified across a number of projects, then a
common cost factor was recorded to cover all incidences of those common factors. The
research identified 37 factors from the highway data analysed. The final list of cost overrun
variables, their symbols and the number of times they occurred across projects are shown in
Table 4.10.
Incidences across
Reason for cost overrun Code
projects
Project acceleration requirement A 5
Constructability difficulty costs C 10
Constructability under traffic CT 17
Design/project scope change D 95
Design scope change drainage DD 33
Design scope change environmental issues DE 19
Design scope change design error DF 2
Design scope change pavement materials/depth DM 23
Design preload requirement DPL 1
Design change to subgrade DSG 1
Design scope change safety audit requirement DSA 4
Extras unspecified EU 6
Government initiative employment continuity G 1
Government initiative contribution by developer GCD 3
Government initiative contribution by local government GCLG 1
Government initiative contribution by rail GCR 1
Cultural heritage issues H 4
Latent condition requires design change LD 4
Latent condition rock encountered LR 7
Latent condition additional stabilising LSG 3
Latent condition removal and replacement of unsuitable material LUS 21
Material cost increase asphalt MA 1
108
CHAPTER 4
Data collection, analysis techniques and statistics
Table 4.10: Project cost overrun factors derived from historic highway data
Column two of Table 4.10 shows the Code. This was an alphabetic code assigned to each
of the Reason for cost overrun entries. This was to allow further data analysis at a later
date. There was no logic to the allocation of the alphabetic code letters, other than a very
coarse alignment of the code letter(s) with the major descriptor of the reason. As an example,
A = Project Acceleration Requirements; WW = Wet Weather effects/rework.
For the purpose of the research, cost overrun variable EU Extras unspecified occurred
across six projects. These projects also had other identified variables also attributed to cost
overrun in those specific projects. Review of the project data revealed that those six
occurrences of EU Extras unspecified were all attributed to one type of construction
project, namely the upgrading from four lanes to eight lanes of highway. These projects
made up the various stages of the Pacific Motorway project that linked Brisbane to the Gold
Coast in the south-east corner of Queensland. An interview with a senior motorway project
staff member advised that more detailed cost overrun details for the specific projects were
not available as research materials as this information now formed parts of various contract
finalisation arrangements under specific deeds of agreements between the client and the
specific motorway constructors (Creedy, 2004). Therefore, those six motorway projects were
excluded from further data analysis of cost overrun reasons.
Factors which had produced an incidence of one (1) in Table 4.9 were excluded from the
factor analysis because they each related to only one incidence. Nine such factors had an
109
Risk factors leading to cost overrun in the delivery of highway construction projects.
incidence of one. The 37 identified cost overrun factors were reduced to 28 for the factor
analysis process. Those cost overrun factors were then formed into an index matrix using
MS Excel, with the 231 recorded projects displayed down the y-axis.
A scaling of zero (0) and one (1) was then established and the matrix completed for all the
projects. A one (1) was recorded against the cost overrun factor for a project when a
particular cost overrun factor was identified. A zero (0) was inserted in the data file when no
other factor occurred in the project. Where a project had multiple cost overrun factors
recorded against it, then multiple ones (1) were inserted in the matrix row. Table 4.11 shows
a sample of the data matrix developed in Excel for uploading into SPSS Version 12 for the
factor analysis.
Table 4.11: Sample display of project and cost overrun matrix data used in analysis
The Kaiser Meyer Olkin (KMO) test that measured sampling adequacy gave a sampling
adequacy of 0.460. This was much less than 0.6 that was sought in the Chapter 3
methodology (Kaiser, 1974; Hutcheson and Sofroniou, 1999). This low figure threw initial
doubt on the adequacy of the planned factor analysis technique in overcoming
multicollinearity in the data.
A further test using the Bartlett test of sphericity gave an approximate chi-square of 482.183
and 378 degrees of freedom The associated significance level was small at p = 0.000 for all
the factors as shown in Table 4.12.
Both the KMO and Bartlett tests indicated that any factor analysis process carried out on the
data would produce marginal results (the Bartlett Test was less than the adopted 700 even
though the Bartlett sphericity test is particularly relevant to small samples of data <100 and
not so relevant to the larger sample of around 230 in this application). After considering the
marginal sample characteristic test results, the principal component/factor analysis method
was still used as the preliminary method for reducing the large number of cost overrun
factors down to a more significant number.
110
CHAPTER 4
Data collection, analysis techniques and statistics
An initial analysis of the 28 factors ascertained the proportion of each variables variance
that could be explained by principal components. Column two of Table 4.12 shows that
initial communalities of the 28 cost overrun variables were 1.000 because, by definition, the
full orthogonal principal component analysis showed all of the variances in the variables to
be explained by all the factors. The extraction communality shown in Column three of Table
4.13 gave the percent of variance in a given variable Variables with high values were
represented in the common factor space, while variables with low values were moderately
represented.
Table 4.13: Initial and extraction communalities for cost overrun variables
The principal component analysis was undertaken with the view of reducing the number of
initial cost overrun factors by the redistribution of the variance to the first components
extracted in a correlation matrix. The method of eigenvalue decomposition was used. Table
4.14 shows the total variance derived from the principal component analysis before rotation.
111
Risk factors leading to cost overrun in the delivery of highway construction projects.
Column one of Table 4.14 shows the 28 components analysed. There were as many
components extracted during the principal components analysis as there were variables
initially used in the analysis, namely 28. The total % variance explained by each cost
overrun factor was listed in the 'Initial eigenvalue Column' of Table 4.14. The initial
eigenvalues derived were from standardised variables. Column two shows the total
eigenvalues in descending order of magnitude with component one accounting for the most
variance (96.76%) in the analysis and hence the highest eigenvalue. Component two
accounted for as much of the leftover variance as it could, and so on. Column three
contained the % of variance that was accounted for by each principal component. Column
four of Table 4.13 contained the cumulative percentage of variance accounted for by the
current and all preceding principal components. The factors were listed in descending order
112
CHAPTER 4
Data collection, analysis techniques and statistics
of variance explained. The third row showed a value of 18.61 which meant that the first three
components together account for 18.61 % of the total variance.
The three columns of the right half of Table 4.14 represented the extraction sums of squared
loadings and reproduced the values given on the same row on the left side of the table. The
number of rows reproduced on the right side of the table was determined by the number of
principal components whose eigenvalues were 1.00 or greater. There were 13 principal
factors that had eigenvalues greater than one (1.000). Column seven displayed the computed
variance of each variable explained by the 13 factor model. Since the factors were
uncorrelated, the total proportion of the variance explained was the sum of the variance
proportions of each factor. The figures in Column seven show that variation was relatively
even across almost all components and this indicated a low significance of the principal
components in reducing the dimensionality of the original set of cost overrun variables. It
showed that the first 13 components collectively accounted for only 61.7 % of the total
variability, while the other 15 components together accounted for the remaining 38.3 %.
A further test using a scree plot was carried out on the project data. As explained in the
methodology in Chapter 3, the scree plot graphed the eigenvalue against the number of
initial factors (component numbers) so as to support a particular eigenvalue cut-off
level. The plot in Figure 4.1 shows the plotted values that have been derived from the first
two columns of Table 4.14. From the 13th component point on, the slope of the line flattened
slightly indicating that each successive component was accounting for smaller and smaller
amounts of the total variance out to component 28. Components with an eigenvalue of less
than 1 accounted for less variance than did the original variable which had a variance of 1.
The scree plot indicated that a weak model, comprising of 13 component factors, was
feasible and warranted further investigation.
2.0
1.5
Eigenvalue
1.0
0.5
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28
Component number
113
Risk factors leading to cost overrun in the delivery of highway construction projects.
Factor rotation was performed based on the varimax method with 16 iterations. It was found
that a rotated and ordered factor loading matrix established 13 component groups for the 28
cost overrun variables as supported by the scree plot. The shaded portions of Table 4.15
show clearly the 13 groupings of the 28 cost-overrun factors.
114
CHAPTER 4
Data collection, analysis techniques and statistics
Component
1 2 3 4 5 6 7 8 9 10 11 12 13
LUS .717 -.067 .009 .048 -.130 -.010 -.063 .043 -.001 -.002 .079 .239 .028
EU -
.638 .044 -.020 -.020 .150 .006 .060 -.034 -.110 -.061 .021 -.125
.024
DM .375 -.213 -.099 .019 -.004 -.129 -.085 -.199 .186 .176 -.188 -.148 .275
H .016 .758 .026 .000 .014 .012 .014 -.088 -.180 -.216 -.063 -.036 .123
ME -
-.005 .689 -.029 .035 -.014 -.006 -.011 .022 .155 .131 -.010 .024
.007
O -
-.098 .509 -.042 -.038 -.083 -.122 -.020 .012 .250 .324 -.026 -.021
.065
R -
-.056 -.070 .706 -.015 .113 -.039 -.008 .160 -.011 .022 -.053 .054
.042
GCD -.065 -.061 .622 .071 -.208 .057 .025 -.085 .137 -.118 -.128 .024 .029
S .101 .129 .622 .022 .145 -.080 -.062 -.137 -.194 .021 .132 -.115 .098
TH -.220 -.058 -.165 -.744 .003 -.122 -.022 .070 -.084 -.080 -.141 .160 .049
D -.363 -.114 -.086 .606 -.073 -.180 -.058 -.102 -.391 -.128 -.032 -.068 .027
DE -
-.038 .237 -.141 .358 .332 .303 -.070 .076 .098 -.127 -.134 .249
.310
SC .049 -.035 .016 .007 .815 -.045 -.033 -.074 -.017 .024 .017 -.048 .049
MPS -
-.036 -.014 .237 -.256 .450 -.044 -.007 .345 -.105 .077 .055 -.094
.099
LR -
.051 -.022 -.016 .046 -.046 .774 -.117 .005 -.012 -.015 -.090 .027
.015
C -.162 -.066 -.042 -.022 -.028 .667 .295 -.054 -.070 .192 .152 -.112 .097
MQ -.056 .004 -.008 -.062 -.003 .072 .860 -.069 .008 -.048 .062 -.055 .032
WW -
.504 -.031 -.059 .123 -.067 -.124 .565 .152 -.045 .045 -.139 .061
.033
P -
.058 -.024 .019 -.170 .025 .020 .025 .787 .031 -.053 -.045 .023
.030
LD -.083 -.025 -.114 .270 -.133 -.082 -.086 .555 -.058 -.061 -.027 -.171 .297
MP -
-.065 .157 -.009 -.062 -.094 -.053 -.036 .003 .772 .028 .088 -.093
.039
DSA -.087 -.127 -.078 .197 .421 -.026 .064 -.103 .511 -.182 -.158 .090 .207
CT -.056 .061 -.059 -.052 .057 .168 -.095 -.108 -.030 .774 -.073 -.110 .086
Q -
.132 .030 .022 .155 -.056 -.122 .264 .057 .006 .539 .073 .482
.042
LSG -.148 -.053 -.056 .140 -.006 .067 .032 .069 -.006 .056 .734 .118 .006
DD .282 -.041 -.019 -.085 .018 -.114 -.014 -.180 .067 -.144 .695 -.135 .070
DF .019 -.021 -.012 -.168 -.045 -.010 -.081 -.085 -.062 -.070 .005 .756 .106
A -
-.017 -.086 -.058 .036 -.052 -.065 -.031 -.079 -.008 -.041 -.061 -.122
.823
115
Risk factors leading to cost overrun in the delivery of highway construction projects.
These 13 component factor groupings were expanded into their full descriptions in Table
4.16 below
A review of the 13 groupings lead to the conclusion that no common factor component
names were developed in the principal component analysis that could be allocated to the
individual component groups so as to allow consistency in purpose of the group of factors.
There appeared to be no evidence of strong correlations within any of the 13 groupings, even
after rotation. This was particularly evidenced by the diversity in the construction processes
associated with each of the principal component groups. These findings, supported by the
weak scree plot demonstrated the invalidity of the factor analysis process in this research.
This finding was further supported by the weak sampling adequacy of the initial highway
cost overrun data available as determined previously by both the KMO test and the Bartlett
test of sphericity.
The 13 principal component factor model derived in the factor analysis process could not be
used in any subsequent model development. Instead a process of using expert elicitation
116
CHAPTER 4
Data collection, analysis techniques and statistics
produced an acceptable number of reduced higher level component factors that could be
used in analysis.
4.5 Stage 4: Use nominal group technique (NGT) to elicit, review and
prioritise principal cost overrun risk groupings and highway
project types
This step of the research methodology was planned to confirm the principal component
factors from the factor analysis. Since the factor analysis was unsuccessful, this stage was
redefined so as to be used for the elicitation of a final grouping and ranking of cost overrun
factors. Nominal group technique (NGT) was used to obtain expert agreement on:
groupings of highway cost overrun factors and apply a ranking of importance of each
in terms of exposure to project cost overrun potential
a group of generic highway project types and apply a ranking of importance of each
to their exposure to cost overrun potential.
Potential group members were initially contacted via telephone to determine their
willingness to participate in the research. A questionnaire was then emailed to the potential
participants to ascertain their suitability within the group. Members were identified and
selected from within the highway organisation from which the cost overrun data was
obtained. A copy of the pro-forma questionnaire is contained as Appendix C and the
responses to the questionnaire have been summarised in Appendix D.
Clemen and Winkler (1985) have recommended five specialists and an additional two or
three generalists and two or three analysts comprise a group. For the purpose of the research,
a generalist was assumed to be one who had a broad range of experience across all of the
stated areas. A specialist was assumed to be a member having at least 10 years experience,
predominantly in one of the identified disciplines of estimating, design, construction and
management. The expert panel finally adopted included an analyst from a consulting
organisation, two generalists and five specialists those seven from the QDMR. Table 4.17
shows the final group composition of the nine participants.
Group type Member Member Member Member Member Member Member Member Member
1 2 3 4 5 6 7 8 9
Analyst *
Generalist * *
Specialist * * * * *
Researcher *
The primary qualifications of the nine members all were in civil engineering. Secondary
qualifications were also reported by seven panel members in their completed questionnaire
and included:
1 x PhD in Engineering
2 x Master of Engineering Science
117
Risk factors leading to cost overrun in the delivery of highway construction projects.
1 x Bachelor of Economics
3 x Business Management.
The reported spread of minimum, maximum and mean years of project-related experience of
the group from the questionnaires are shown in Table 4.18. The mean years shown in
Column four indicated a good balance of project management expertise across the group.
As recommended by Keeney and von Winterfeldt (1991), the group members were presented
with an overview of the complete study and the issues in preliminary form and given an
overview of the formal expert judgement process that they would be involved in. This was to
ensure that the members understood what was requested of them and how it fitted into the
project. As a way of providing preliminary training of the group in the elicitation process,
each member was emailed an overview of the NGT process as well as background
information on the project and the project data. A copy of Table 4.10 (Project cost overrun
factors derived from highway data) was also provided as well as the detailed explanation of
each highway cost overrun factor. Seven examples of high level risk groupings from
previous research identified in the literature review were provided to each member in order
to give them an idea of what was anticipated of the group.
This involved a presentation to the eight members by the researcher on the background to the
research project.
Group members were provided with details of the workshop aims and desired outcomes. As
well, a more detailed explanation was provided to the members on the NGT technique and
the importance of the process to the development of a cost overrun model.
The aim of agenda item three was to obtain expert agreement on groupings of highway cost
overrun factors.
The members were again provided with a copy of Table 4.10 (Project cost overrun factors
derived from highway data). The factor 'EUextras unspecified' had been excluded from the
list as previously discussed in Stage 2 of the research process.
118
CHAPTER 4
Data collection, analysis techniques and statistics
The workshop considered each of these 36 factors and came up with consolidated groupings.
The process adopted was as follows:
1. silent generation of grouping ideas in writing
2. round-robin feedback from group members to record each idea of groupings
3. discussion of each recorded member groupings for clarification and evaluation
4. individual voting on grouping options
5. agreement on groupings.
The following list of 10 groupings was agreed by the participants during the NGT process:
design and scope change (change on project definition)
services relocation
deficient documentation (specifications and designs)
right-of-way costs
insufficient investigations and latent conditions
constructability
environment
contractor risks
price escalation
client project management costs.
The next step in Item 3 was to group the 36 cost overrun factors into the 10 groupings. A
matrix grid of the risk factors and the groupings as shown in Table 4.19 was produced and
handed out in the workshop
Services relocation
Right-of-way costs
management costs
investigations and
Contractors risks
Design and scope
latent conditions
Constructability
Price escalation
documentation
Client project
(Spec/design)
Environment
Insufficient
Deficient
Analysed
change
risk
factors
A
C
CT
D
DD
DE
DF
DM
DPL
DSG
DSA
G
GCD
GCLG
GCR
H
LD
LR
119
Risk factors leading to cost overrun in the delivery of highway construction projects.
LSG
LUS
MA
MB
ME
MP
MPS
MQ
N
O
P
Q
R
S
SC
TH
TCI
WW
Members then went through the NGT process again and agreed on the matrix composition so
that each of the 36 cost overrun factors mapped to one of the 10 groupings. A Likert scale of
1 for poor fit, 2 for an acceptable fit and 3 for an excellent fit of each factor to its grouping
was agreed and then applied by the group. The final mapping and the corresponding agreed
Likert scale number are shown in Table 4.20.
Services relocation
Right-of-way costs
management costs
(Spec and design)
investigations and
Design and scope
latent conditions
Contractor risks
Constructability
Price escalation
documentation
Client project
Environment
Insufficient
Deficient
Analysed
change
risk
factors
A 3
C 3
CT 3
D 3
DD 3
DE 3
DF 3
DM 2
DPL 3
DSG 2
DSA 3
G 3
GCD 3
GCLG 3
GCR 3
H 3
LD 3
LR 3
120
CHAPTER 4
Data collection, analysis techniques and statistics
LSG 3
LUS 3
MA 3
MB 3
ME 3
MP 3
MPS 3
MQ 3
N 3
O 1
P 3
Q 3
R 3
S 3
SC 3
TH 3
TCI 3
WW 3
This ranking exercise required the workshop participants to provide numerical scores that
expressed their individual opinions on the level of importance of each of the 10 principal
cost overrun factors in terms of their cost impact to highway projects. A scale of 1 to 10 was
adopted, where a score of one indicated the least impact, and 10 indicated the most impact.
All scores of the eight participants were recorded and an overall ranking of each of the
groupings was derived.
A collation of the mean and standard deviation of all the participant scores that was provided
for each grouping was considered to not adequately determine an overall relative ranking as
such a method did not reflect any relationship between the groupings (Chan and
Kumaraswamy, 1997). The cost overrun group rankings were determined by calculating the
Importance index for each grouping, using the following formula:
Where w was the weighting, ranging from 1 to 10, given to each factor; W was the highest
weight, i.e. 10; f was the frequency of the response; and F was the total number of
respondents.
Each rank in Table 4.21 presents the degree of importance assigned to the groupings.
121
Risk factors leading to cost overrun in the delivery of highway construction projects.
Importance
Expert 1
Expert 2
Expert 3
Expert 4
Expert 5
Expert 6
Expert 7
Expert 8
index %
RANK
Cost overrun
grouping
Design and
scope change 8 10 10 10 9 10 10 9 95 I
Insufficient
investigations
and latent 9 9 7 8 8 7 9 10 84 II
conditions
Deficient
documentation
(Spec and 7 5 9 5 4 9 8 8 69 III
design)
Client project
management 10 8 8 7 10 8 1 1 66 IV
costs
Services
relocation 4 6 3 8 4 1 6 6 48 V
Constructability 4 6 3 4 3 4 6 5 44 VI
Right-of-way
6 3 6 6 6 1 2 2 40 VIII
costs
Contractor risks 1 1 2 2 1 5 5 6 29 IX
Environment 3 2 1 1 2 3 4 4 25 X
The following section describes the composition of the final groupings in their rank I
(highest) to X (lowest) order of their cost impact on highway projects.
122
CHAPTER 4
Data collection, analysis techniques and statistics
Design and project scope change was identified as the most important factor by participants
(a score of 95% in the importance index) and was ranked first. The group found that the
potential for cost increases in the design or scope change projects was very real and their
choice in ranking supported historical data previously analysed and provided in previous
Table 4.10. Below are the percentages of analysed projects which fell into this first ranked
grouping.
Design/project scope change (23.6% of projects represented)
Design scope change resulting from drainage, environmental issues, pavement
materials/depth (8.2% of projects represented)
Design scope change as a result of carrying out a safety audit on project
(1.0% of projects represented)
Quantity increase (7.7% of projects represented)
Specification change (1.7% of projects represented).
123
Risk factors leading to cost overrun in the delivery of highway construction projects.
124
CHAPTER 4
Data collection, analysis techniques and statistics
lead to unbudgeted costs for the client where project provisions were silent in such instances.
The relevant Table 4.10 percentages of project data have also been shown below.
Remote location costs (1.7% of projects represented)
Wet weather effects/rework (2.0% of projects represented).
The aim of this activity was to obtain expert agreement on a fully inclusive grouping of
highway project types that could be used as a means of collating historic projects for further
analysis.
Members were provided with a copy of Table 4.3 that contained a broad listing of highway
project works which had previously been compiled by the researcher from the initial
historical project data. In considering the projects types in Table 4.3, the workshop members
discussed the possibility of additional project types or the option of combining existing
project types. The workshop members agreed that the list of projects could then be
categorised into three broad highway project groupings of highway projects, bridge projects
and a miscellaneous category.
The NGT allowed for further round robin, workshop discussion and then a voting step that
developed the final groupings. The final list of project types consisted of nine road types,
two bridge types and one miscellaneous type. These are shown in Table 4.22.
125
Risk factors leading to cost overrun in the delivery of highway construction projects.
The group also agreed that it would be desirable to differentiate between certain project
types where the complexity or location of some projects might require identification at a
more specific level. The workshop members agreed that the following project aspects needed
to be considered if appropriate historical data became available in the future. Those aspects
were whether:
a rehabilitation project included pavement widening or not
a realignment project incorporated some of the existing road and was built under
traffic (The group referred to this type of project as being constructed at a brownfield
site.)
the project was on a completely new alignment away from the existing highway that
it had replaced, thus having none or only a small exposure to highway traffic (The
group referred to this type of project as being constructed at a greenfield site.)
an interchange/intersection or bridge and approaches were to be in an urban or rural
environment.
Table 4.23 shows the desired extensions to the agreed project types suggested by the
participants if appropriate project data were made available.
Table 4.23: Desired workshop project construction types for future grouping
126
CHAPTER 4
Data collection, analysis techniques and statistics
As part of the NGT process, an evaluation of the overall workshop process was carried out.
Workshop participants completed four questions about the workshop. A Likert scale was
also provided for each question as shown in Table 4.24. As well, a fifth question was
provided which gave the participants an opportunity to comment on the workshop in general.
The following details the responses of the four questions in the order presented. Seven out of
the eight participants completed the Workshop Evaluation document.
Question 2: The information provide in the week prior to the workshop was useful in
preparing for and also contributing to the workshop outcome.
Two participants strongly agreed that information provided prior to the workshop was useful
in preparing for and contributing to the workshop outcome and five participants said they
agreed (Table 4.25).
Question 3: The use of the NGT gave me adequate opportunity to provide input into the
workshop.
Two participants also strongly agreed that the use of the NGT provided adequate opportunity
to provide input into the workshop and five participants said they agreed (Table 4.26).
127
Risk factors leading to cost overrun in the delivery of highway construction projects.
Question 4: The range and number of attendees was adequate for the workshop outcome
Three participants strongly agreed that the range and number of attendees were adequate for
the workshop outcome and four participants said they agreed (Table 4.27).
The following general comments were also provided by five of the participants:
1. Difficult issue to have a one-on-one relationship. Many causes are multi-
dimensional
2. Not certain it consistently sorted out the conceptual levels of causes
3. Good
4. Good
5. Worked well.
Overall, the NGT workshop worked well as supported by the response rankings of the
workshop evaluation questions.
The expert elicitation in the research adopted the NGT process. The process was
straightforward and used the available time of the experts efficiently. Participants in the
overall expert elicitation process were very supportive of the application of NGT in the
difficult area of identifying client risk in the delivery of projects.
There were two important outcomes that were derived during the NGT process. Firstly, 10
project risk groupings were developed that were ranked in importance by the group with the
application of the importance index. Secondly, a generic set of 12 highway project types was
developed. Both sets of outputs were acceptable for use in the proposed modelling of project
cost overrun using multivariate regression analysis.
128
CHAPTER 4
Data collection, analysis techniques and statistics
The high-level (HL) risk groupings that were developed by the experts and mapped to the
initial 36 cost overrun factors, as contained in previous Table 4.20, were then mapped back
to the initial projects. This was accomplished with the aid of an Excel spreadsheet. Table
4.28 shows a small sample of the Excel spreadsheet that was produced to map the
associations of (say) the low level cost overrun risks of D Design/project scope change,
DD Design scope change for drainage and DE Design scope change for environmental
issues.
Project Number Risk D HL/1 Risk DD HL/1 Risk DE HL/1
47/16B/302 1 1 0 0
36/15A/301 &302 0 0 0
140/900/2 0 0 0
116/26B/31 1 1 0 0
148/17B/302 1 1 0 0
80/10A/772 1 1 1 1 0
150/10M/301 0 1 1 0
8/26A/302 0 0 0
94/36b/19 1 1 0 0
140/902/2 1 1 0 1 1
160/12B/5 0 0 0
30/6204/13 1 1 0 1 1
5/10K/19 1 1 0 1 1
55/99C/15 0 1 1 0
160/103/3 0 1 1 0
99/323/13 1 1 0 0
160/103/6 1 1 0 1 1
140/U18B/61 0 0 0
140/U18B/61 1 1 0 0
80/133/734 1 1 0 0
36/93E/706 1 1 0 0
Table 4.28: Sample of low level to high level risk group mapping to projects
For the purpose of analysis, 1 was inserted in the master data file for each low level risk and
its associated HL risk grouping for each project. As well, a 0 was inserted in all low level
risk columns when there was no association with a particular HL risk. This provided a
visually complete 'picture' of the project mappings.
129
Risk factors leading to cost overrun in the delivery of highway construction projects.
An analysis was carried out to ascertain which HL groupings of cost overrun factors had the
most occurrences across the projects analysed. There were 42 HL risk combinations
recorded across the project data. Table 4.29 shows the incidences of the HL groupings across
the projects.
Table 4.29: Incidences across projects for the HL/ risk groups
130
CHAPTER 4
Data collection, analysis techniques and statistics
Column one of Table 4.29 shows the HL/1. A decimal format has been used to record when
more than one HL risk grouping occurred in a particular project. Rows 1 to 10 of Column 2
of Table 4.29 display the incidences of single HL/1s. Whereas, Row 11 shows that there
were five projects where both HL/1 and HL/2 occurred.
HL/1 recorded the highest incidences with 92 projects. The next highest recorded 22
incidences recorded for HL/2. There were 19 recorded incidences where a combination of
HL/1 and HL/5 occurred in projects. As a further explanation of Table 4.29 above, the last
row shows that there was one project which had HL/1, HL/6, HL/ 8 and HL/10 risk
groupings that collectively contributed to cost overrun in the one recorded project.
It was identified that a significantly high number (92) of incidences were recorded for HL/1,
when compared with the rest of the incidences recorded. The composition of HL/1 consisted
of the following low-level risk factors that were determined in the previous expert elicitation
process:
Design/project scope change (23.6% of projects represented)
Design scope change resulting from drainage, environmental issues, pavement
materials/depth (8.2% of projects represented)
Design scope change as a result of carrying out a safety audit on project (1.0% of
projects represented)
Quantity increase (7.7% of projects represented)
Specification change (1.7% of projects represented).
There had been 31 incidences of quantity increase previously recorded against projects as
shown in previous Table 4.10. As this amount was comprised of 34% of the HL/1 grouping,
it was decided to extract the quantity increase incidences from the HL/1 risk group and
create an eleventh grouping of Quantity increase (HL/Q) for incorporation in the model
analysis.
The aim of this step was to analyse of historical project data based on statistical theories and
concepts that identified direct correlations between particular highway construction project
types and project cost overrun.
The dependent variable adopted in the model was the continuous variable % cost overrun
and was the difference between the clients actual project cost and their programmed cost,
expressed as a percentage of the programmed cost for each project.
The initial analysis of the highway cost data has indicated that substantial cost overrun
factors needed to be considered when looking at the reasons for project cost overrun. The
approach identified project variables that had relationships with projects reporting high cost
overrun. The following prediction variables were selected for model investigation and were
based on the availability of historical project data:
131
Risk factors leading to cost overrun in the delivery of highway construction projects.
The reason for including the geographic project type in the proposed model was because
there appeared to exist a strong relationship between the remoteness of a project from
established workforces and from proven materials and component manufactures that could
lead to increases in project costs above those estimated. Drew and Skitmore (1992) identified
the density of population and the extent of geographic area as important factors for
competitive bidding in building projects.
It was therefore postulated that the rural geographic type of highway projects had a higher
potential to overrun budgeted costs.
An analysis was carried out on the project data to split projects down into the geographic
area in which the project was constructed. This split was on the basis of a project being built
in an urban or rural environment. The criterion adopted was whether a project was
constructed in a city or a shire area. The adopted split was based on the Local Government
Regulation s. 5. This provided the following declaration of classes of local government (LA)
areas (Queensland Government, 1993):
Local government areas are classified as shires unless they meet the criteria to be
declared a City or Town and are so declared. They are generally rural areas with
significant rural uses or large tracts of undeveloped land in its natural state.
The criteria for City status require the area to be a regional centre with a total
population of at least 25,000, and a population in the urban centre of at least 15,000
at a population density of at least 150 persons for each square kilometre in that urban
centre.
There were 60 'rural' local authority areas identified in the data and 13 'urban' areas. The
splits into rural/urban are shown in Table 4.30 and were based on the previous shire/city
definitions.
132
CHAPTER 4
Data collection, analysis techniques and statistics
133
Risk factors leading to cost overrun in the delivery of highway construction projects.
For the purpose of analysis, 1 was inserted as a categorical variable in the master data file for
all projects constructed in a rural environment and a dummy value of' 0 for when the project
was not a rural project. Conversely, a 1 was inserted when a project was classified as urban,
and a dummy 0 for the null case. A sample of this data coding for the geographic area is
shown in Table 4.31.
134
CHAPTER 4
Data collection, analysis techniques and statistics
The type of project construction in the proposed model was included because it was thought
that a strong relationship existed between the degree of difficulty in scoping, designing and
constructing highway projects which had multiple lanes and high strength pavements. It was
therefore postulated that highway projects involving duplication to multiple lanes had a
higher potential to overrun budgeted costs. Conversely, it could have been postulated that
bridge projects would have less cost overrun potential because it was easier to design and
construct typically generic highway bridge types.
Twelve individual highway project construction types were developed as part of the expert
elicitation process. They were detailed in Table 4.22. For the purpose of further data analysis
individual reference numbers were coded to each of the 12 project construction types in
Table 4.32.
For the purpose of analysis, 1 was inserted as a categorical variable in the master data file for
all projects constructed in a project type 1 and a dummy value of' 0 for when the project was
not a project type 1. As well, 1s were inserted when a projects were successively project
types 2 to 12 inclusive, with the dummy value of' 0 inserted for when the projects were null
for the particular type.
135
Risk factors leading to cost overrun in the delivery of highway construction projects.
37/91A/20 0 0 0 0 0 0
160/203/3 0 0 0 0 0 1
76/153/3 0 0 0 0 0 0
45/655/18 0 0 0 0 0 0
5/10K/809 0 0 0 0 0 0
148/3042/2 1 0 0 0 0 0
140/U15/44 0 0 0 0 0 0
140/U16/819 0 0 1 0 0 0
37/90C/36 1 0 0 0 0 0
25/126/23 0 0 0 0 0 1
70/108/300 0 1 0 0 0 0
120/856/3 0 0 0 0 0 0
48/21A802 0 0 1 0 0 0
66/814/5 0 1 0 0 0 0
48/6404/3 0 0 0 1 0 0
34/1102/17 0 0 0 0 0 0
92/140/11 0 0 0 0 0 0
The reason for including the type of project delivery of the highway construction in the
proposed model was because there may be a strong relationship between the competitiveness
generated in an open tender environment that should produce lower than client estimate bids
as opposed to a negotiated price contract. As well, sometimes in high risk projects, clients
use of their own internal workforce can be the best option for assuming risks that may
eventuate and hence be managed such that all risk flows back to the client. Conversely, it
could be postulated that open tender contracts have the potential to generate excessive
change order variations brought about by the initial competitive contract price environment.
Three project delivery types were initially developed in Table 4.5 as open tender, negotiated
price internal workforce and negotiated price external workforce. The delivery codes are
shown in Table 4.34.
136
CHAPTER 4
Data collection, analysis techniques and statistics
For the purpose of analysis, 1 was inserted as a categorical variable in the master data file for
delivery code 1 and a dummy value of' 0 for when that delivery code did not apply to a
project. A similar treatment of 1s and 0s was applied for delivery code 2 and delivery code 3.
A sample of this data coding for the project delivery codes is shown in Table 4.35.
Generally there is a correlation between the cost of a project and the size of the project. In
this research it was adopted that, if projects costs are indexed to a common year, then the
project cost can be used as a surrogate for project size. The reason for including the indexed
highway programmed cost in the proposed model was because it was thought that there was
a strong relationship between the size of a project and the percentage a project might
overrun. For highway projects, the greatest risk lays below ground level due to the relatively
greater physical footprint of the project, and the larger the footprint, then the larger the risk
cost should be. Williams (2003) reported results that suggested that large highway projects
tended to have much higher relative cost increases than small projects.
Hence it was postulated that larger projects overrun budget more so than smaller projects
when indexed to a common year. The minimum project size was limited to A$1m.
For the purpose of the research analysis, the programmed cost data in A$m from financial
years 199596 through to 200102 were adjusted by way of a cost index (refer Table 4.9) up
to 200203 out-turn year was the continuous variable used.
137
Risk factors leading to cost overrun in the delivery of highway construction projects.
Indexed program
Project number
cost $m
37/91A/20 4.73
160/203/3 1.36
76/153/3 2.55
45/655/18 1.90
5/10K/809 1.30
148/3042/2 1.81
140/U15/44 2.30
140/U16/819 & 820 1.80
37/90C/36 1.77
25/126/23 1.28
70/108/300 1.30
120/856/3 3.81
48/21A802 1.13
66/814/5 2.06
48/6404/3 8.69
34/1102/17 2.00
92/140/11 1.70
The reason for including client risks associated with highway project construction in the
model was that there was very strong research and historical project data evidence that
suggested strong linkage to between the realisation of unplanned or under-budgeted project
risks and cost overrun of constructed projects. Smith (1995a) has confirmed that
inappropriate risk management can also lead to higher costs for the client. As well, El-
Choum (1994) identified project cost overrun parameters for public infrastructure projects.
Risks associated with varying types of highway construction projects were considered to
have possible effects on the differences observed in cost overrun. Possible correlations may
have existed between the prevalence of certain risks with certain levels of cost overrun. It
was postulated that particular groupings of historical risks which clients have been exposed
to in highway project delivery bear a relationship to the size of project cost overruns.It was
further postulated that risks that were ranked most important in terms of impact on project
cost overrun by a team of experts would produce higher cost overruns in the model.
Ten high level risk groupings were initially developed using the NGT process. A further
analysis demonstrated the need to have extracted quantity increase from HL/1 into its own
HL/Q. Table 4.37 shows the 11 groupings used in the model
138
CHAPTER 4
Data collection, analysis techniques and statistics
For the purpose of analysis, a 1 was inserted as a categorical variable in the master data file
against projects which had a HL/1 associated with them and a dummy value of' 0 for when
the high level risk design and scope change did not apply to the projects. A similar
treatment of 1s and 0s was applied for HL/2 through to HL/10 and HL/Q. A sample of this
data coding for the project delivery codes is shown in Table 4.38.
Project
HL/1 HL/2 HL/3 HL/4 HL/5 HL/6 HL/7 HL/8 HL/9 HL/10 HL/Q
number
37/91A/20 1 0 0 0 1 0 0 0 0 0 0
160/203/3 1 0 0 0 0 0 0 0 0 0 0
76/153/3 1 0 0 0 0 0 0 0 0 0 0
45/655/18 1 0 0 0 0 0 0 0 0 0 0
5/10K/809 1 0 0 0 1 0 0 0 0 0 0
148/3042/2 1 0 0 0 0 0 0 0 0 0 0
140/U15/44 1 0 0 0 0 0 0 0 0 0 0
140/U16/819 0 0 0 0 0 0 0 0 0 1 0
37/90C/36 1 0 0 0 0 0 0 0 0 0 0
25/126/23 1 1 0 0 0 0 0 0 0 0 0
70/108/300 1 0 0 0 0 0 0 0 0 0 0
120/856/3 0 0 0 0 0 0 1 0 0 0 0
48/21A802 1 0 0 0 0 0 0 0 0 0 0
66/814/5 1 0 0 0 0 0 0 0 0 1 0
48/6404/3 1 0 0 0 0 0 0 0 0 0 0
34/1102/17 0 0 0 0 0 0 0 0 0 1 0
92/140/11 1 0 0 0 0 0 0 0 0 0 0
Table 4.38 Sample of high level risk grouping codes against projects
As a preliminary step in the analysis process, the 231 project cases identified in Section 4.3
were analysed for random disturbance. For the purpose of specifically identifying any
139
Risk factors leading to cost overrun in the delivery of highway construction projects.
project outliers, a linear regression analysis was carried out using the dependent variable as
'project cost overrun $m' and the predictor variable as 'indexed programmed cost $m'. A
value of two standard deviations from the initial estimated line was set to measure the
variability of the random disturbance. It was postulated that about 95% of observed points
should have fallen within the adopted two standard errors, assuming normality. This analysis
generated a table of project cases with residuals more than the adopted value. This was
helpful in identifying potential outlying data values.
The following Table 4.39 details the casewise diagnostics that were produced from the
analysis of the two nominated variables.
As can be seen in Table 4.39, there are five cases identified with standardised residuals
exceeding 2 in absolute value. An analysis of the data behind the case numbers identified in
the diagnostics revealed the following projects in Table 4.40 as being potential data outliers
The projects * and # were reported on in successive years. It was noted that all five cases
above related to the Pacific Motorway project that had been constructed during the period
19972001 and all had been classified as R7 (Widen four lanes into six/eight lanes). Since
the standardised residuals of these five cases exceeded 2 in absolute terms, then it was
considered appropriate to exclude these five cases from any future model analysis. The
sample size was subsequently reduced from 231 highway projects down to 226.
A further analysis of the remaining project data for all R7 projects showed that only three
projects remained in this category. It was then decided to remove those remaining three R7
projects. This step removed all project type 7 from all further model analysis and provided
223 projects for further analysis.
140
CHAPTER 4
Data collection, analysis techniques and statistics
The statistical tool of multivariate regression was used to analyse the correlation between the
following project variables that were simultaneously associated with the dependent variable
of cost overrun percentage:
1. Highway geographic project type (urban/rural projects)
2. Highway project construction type (project types 1 to 12)
3. Highway project delivery type (delivery code 1, 2 or 3)
4. Indexed highway project programmed cost (indexed prog. cost $m)
5. Highway project high level risk grouping (HL/1 to 10 and HL/Q).
Due to wide spectrum of highway projects, delivery methods and client risks that caused cost
overrun, many regression models were possible. According to 5 variable listings above, there
were 2 geographic location types, 12 project types, 3 project delivery methods, and 11 high
level risk groupings that were hypothesised as influencing the amount of project cost overrun
for the projects analysed.
The null hypothesis adopted for Pearson's correlation test was that there was no correlation
between the size of cost overrun of projects (and their geographic location, construction
type, delivery process, project size, and client project risks).
Model testing was applied to confirm or otherwise that the given independent (prediction)
variables in the model did not correlate against the dependent variable of project cost
overrun with any significant level of accuracy.
For the purpose of research, the statistical models were considered linear normal models (i.e.
regression analysis with the appropriate f -tests and t-tests). For each test, the p-value was
reported as a measure for rareness if identity of groups was assumed. A p-value less than
0.01 was considered highly significant and less than 0.05 significant. Whereas a larger p-
value meant that the deviation could be due to chance.
Three stages of statistical regression were computed using the method of least squares
The following three methods were used to include and exclude independent variables
in the analysis:
forward selection
backward selection
stepwise selection.
In stage one, the independent best correlated with the dependent was included in the
equation. In the second stage, the remaining independent with the highest partial correlation
with the dependent, controlling for the first independent, was entered. This process was
repeated at each stage for previously-entered independents, until the addition of a remaining
independent did not increase R2 by a significant amount or until all variables were entered.
The next step was to inspect the residuals as surrogates for random disturbances and to make
further judgements on the validity of the data and the linear regression model. The following
sections detail the three selection modes applied to the project data for this purpose.
141
Risk factors leading to cost overrun in the delivery of highway construction projects.
Forward selection this method started with a model that contained none of the
explanatory variables. In the first step, the procedure considered variables one by one for
inclusion and selected the variable that results in the largest increase in R2. In the second
step, the procedure considered variables for inclusion in the model that only contained the
variable selected in the first step. In each subsequent step, the variable with the largest
increase in R2 was selected, until, according to an f -test, further additions were judged to not
improve the model.
The forward selection was next carried out on the data with each of the independent
variables entered separately. Table 4.41 shows the independent variables entered and
removed during the forward selection method. The dependent variable entered was % over
cost.
Variables
Model Variables entered Method
removed
Forward
1 Indexed prog cost $m Nil (Criterion: Probability-of-f-to-enter <=
0.050)
Forward
2 Project type 12 Nil (Criterion: Probability-of-f-to-enter <=
0.050)
Forward
3 HL/1 Nil (Criterion: Probability-of-f-to-enter <=
0.050)
As the forward selection mode had been used, there were no variables removed. Only those
variables which had a probability of f <= 0.050 were entered. Three models were produced
in the forward selection analysis.
The regression model summary for the forward selection mode is shown in Table 4.42. R
represents the correlation between the linear regression prediction and the actual % over
cost. The R values for models 1 and 2 were <0.2 and thus had negligible correlation. Model
3 had an R value of 0.245 which indicated a low correlation that was not very significant.
The R2 values 0.019, 0.037 and 0.060 for models 1, 2 and 3 were statistically very poor.
Column 4 shows the coefficient of multiple determination as adjusted R2 and was the
adjustment applied to R2 to account for the number of variables in the model and the sample
size used). The adjusted R2 for models 1, 2 and 3 were 0.015 (1.5%), 0.028 (2.8%) and 0.047
(4.7%) respectively. This meant that by moving to model three, there was an increase of only
3.2% in the total variation in the % over cost dependent variable.
142
CHAPTER 4
Data collection, analysis techniques and statistics
Table 4.42: Forward selection mode summary using dependent variable of % over cost
Unstandardised Standardised
coefficients coefficients
Model Std. t Sig.
B Beta
error
The variables excluded from the forward selection mode, along with beta, t and significance
(f) values are shown in Table 4.44. The significance values of excluded variables that are
shown in Column five of Table 4.44 were all >0.05. The observed slope of the model was
given by the t value in Column four and was found to be significantly greater than 0. The
results of the forward mode analysis indicated that the overall f results showed moderate
evidence that the three models were statistically significant.
143
Risk factors leading to cost overrun in the delivery of highway construction projects.
Excluded
Model Beta t Sig.
variable
1 Rural projects -0.101(a) -1.449 0.149
Urban projects 0.101(a) 1.449 0.149
Project type 1 -0.041(a) -0.612 0.541
Project type 2 -0.073(a) -1.098 0.273
Project type 3 -0.096(a) -1.415 0.158
Project type 4 -0.008(a) -0.116 0.907
Project type 5 -0.038(a) -0.573 0.567
Project type 6 0.000(a) -0.006 0.995
Project type 8 0.028(a) 0.424 0.672
Project type 9 0.104(a) 1.560 0.120
Project type 10 0.006(a) 0.097 0.923
Project type 11 0.038(a) 0.575 0.566
Project type 12 0.132(a) 2.000 0.047
Delivery code 1 0.058(a) 0.828 0.408
Delivery code 2 0.024(a) 0.351 0.726
Delivery code 3 -0.106(a) -1.579 0.116
HL/1 0.129(a) 1.950 0.052
HL/2 -0.016(a) -0.235 0.814
HL/3 -0.051(a) -0.762 0.447
HL/4 -0.107(a) -1.615 0.108
HL/5 -0.006(a) -0.083 0.934
HL/6 0.009(a) 0.128 0.899
HL/7 0.053(a) 0.794 0.428
HL/8 -0.024(a) -0.354 0.723
HL/9 -0.055(a) -0.823 0.411
HL/10 -0.062(a) -0.934 0.351
HL/ Q -0.090(a) -1.358 0.176
2 Rural Projects -0.068(b) -0.939 0.349
Urban Projects 0.068(b) 0.939 0.349
Project type 1 -0.026(b) -0.390 0.697
Project type 2 -0.062(b) -0.926 0.355
Project type 3 -0.075(b) -1.104 0.271
Project type 4 0.002(b) 0.024 0.981
Project type 5 -0.034(b) -0.512 0.609
Project type 6 0.015(b) 0.215 0.830
Project type 8 0.034(b) 0.509 0.612
Project type 9 0.115(b) 1.740 0.083
Project type 10 0.021(b) 0.313 0.755
Project type 11 0.044(b) 0.661 0.509
Delivery code 1 0.073(b) 1.043 0.298
Delivery code 2 -0.004(b) -0.059 0.953
Delivery code 3 -0.091(b) -1.357 0.176
HL/1 0.155(b) 2.333 0.021
HL/2 -0.006(b) -0.095 0.924
HL/3 -0.062(b) -0.934 0.352
144
CHAPTER 4
Data collection, analysis techniques and statistics
Backward selection This method started with a model containing all the variables and
eliminated variables one by one. At each step the variable chosen for exclusion was that
leading to the smallest decrease in R2. The procedure was repeated until, according to the f
test, further exclusions did not represent a deterioration of the model. Only those variables
which had a probability of f >= 0.100 were removed in the backwards selection process. A
145
Risk factors leading to cost overrun in the delivery of highway construction projects.
total of 24 models was produced and produced R2 values ranging from 0.110 (11%) down to
0.073 (7.3%), which indicated very poor coefficient of variations for the models developed.
The analysis output for the backwards selection mode was very large, and so it was not
reproduced in this section.
Stepwise selection This third method involved the combination of the previous two
approaches and started with no variables in the model. Variables were then added, as with
the forward selection method. In addition, after each inclusion step, a backward elimination
process was carried out to remove variables that were no longer judged to improve the
model. Caution in this process was required as the stepwise procedure was claimed to be
controversial because it included independent variables based on statistical criteria and not
theoretical ones (Bryman and Cramer, 1999). f statistics with probabilities of 5% and 10%
were employed for entry and removal criteria.
The stepwise analysis produced three models. Table 4.45 shows the variables entered, with
no variables being able to be removed under the above f criteria.
Variables Variables
Model Method
entered removed
Indexed prog. Stepwise (Criteria: Probability-of- f -to-enter <= .050,
1 Nil
cost $m Probability-of- f -to-remove >= .100).
Stepwise (Criteria: Probability-of- f -to-enter <= .050,
2 Project type 12 Nil
Probability-of- f -to-remove >= .100).
146
CHAPTER 4
Data collection, analysis techniques and statistics
The forward and stepwise analysis both produced three identical models, while the
backwards mode produced 13 models. The R2 values ranged overall from 0.019 (1.9%) up to
0.111 (11%) and these values indicated a very poor coefficient of variation. As well,
adjusted R2 values ranged overall from 0.002 to 0.061 which again indicated very weak
models. The standard error of the regression estimate shown in Column 6. The values were
consistently large and only showed a slight reduction in each of the models for the forward,
backwards and stepwise modes.
This fact, combined with the low estimates for R2 and adjusted R2, indicated that the models
had not fitted the data very well.
The Project type 12 in model 2 showed prima facie, that these 17 miscellaneous projects
were catch-alls for the undefined projects in the initial data categorisation. On review of the
project data, the majority of the Project type 12 (Miscellaneous) was associated with the
construction of sound attenuation structures to the boundary lines of urban highways or
traffic infrastructure such as signalisation and signage. In particular, Project 140/U16/301
was categorised as Project type 12 (Miscellaneous roadworks/traffic furniture and devices)
147
Risk factors leading to cost overrun in the delivery of highway construction projects.
project in the initial data compilation. A review of the project's specific data indicated that
this project was described as a traffic management devices project on the Ipswich Highway
(Cunningham Arterial). The open tender project included the installation of permanent
separator concrete barriers, guardrails and fencing as well as temporary traffic control
devices. The works were reported to have been carried out totally at night and under heavy
traffic flows. The programmed cost in 2001 was $7.43m and the project experienced a cost
overrun of 54.3%. Because of the unique nature of the project it was considered an outlier
and removed from further regression analysis.
The data were re-analysed with the omission of the two outlier projects using stepwise mode
and the analysis produced the following models.
Unstandardised Standardised
Model Variable coefficients coefficients t Sig.
B Std. error Beta
(Constant) 27.257 1.325 20.572 0.000
1
Project type 9 10.143 4.924 0.138 2.060 0.041
(Constant) 28.952 1.563 18.522 0.000
2 Project type 9 10.031 4.891 0.136 2.051 0.041
Indexed prog. cost $m -0.340 0.169 -0.134 -2.009 0.046
(Constant) 26.123 2.096 12.463 0.000
3 Project type 9 9.769 4.859 0.133 2.011 0.046
Indexed prog. cost $m -0.366 0.169 -0.144 -2.170 0.031
HL/1 5.137 2.558 0.133 2.009 0.046
As each of the revised models contained Project type 9, all these projects were investigated
for data outliers. It was identified that Project 76/153/3 was an open tender at-grade
intersection improvement project with an initial indexed programmed cost of $2.55m that
was completed in 1999 and had a cost overrun of 110.9%. This case gave a predicted value
of $40.09m and a residual of 70.1 which was outside two standard errors and it was
considered appropriate to discard the project as an outlier.
The data was re-analysed with the further omission of Project 76/153/3 using stepwise mode.
The following single model shown in Table 4.48 was derived.
148
CHAPTER 4
Data collection, analysis techniques and statistics
Unstandardised Standardised
Model coefficients coefficient t Sig.
B Std. Error Beta
(Constant) 29.249 1.472 19.870 0.000
1
IndexedProgCost$m -0.328 0.164 -0.135 -2.007 0.046
The variables excluded from the stepwise selection mode on the revised data set, along with
beta, t and significance (f) values are shown in Table 4.50. The significance values of
excluded variables that are shown in Column five were all >0.05. The observed slope of the
model was given by the t value in Column four and was found to be greater than 0. The
results of this stepwise mode analysis indicated that the overall f results showed moderate
evidence that the model was statistically significant.
149
Risk factors leading to cost overrun in the delivery of highway construction projects.
Table 4.50: Excluded variables in stepwise selection mode using revised data
The residual statistics for the stepwise model are detailed in Table 4.51 below and shows the
standard deviation for the residuals at a relatively wide 18.1581 for the revised 220 projects.
The stepwise analysis for the revised project data produced the one model. f statistics with
probability of 5% and 10% were employed for entry and removal criteria. Table 4.52 shows
the variables entered, with no variables being removed under the above f criteria.
Variables Variables
Model Method
entered removed
Indexed prog. Stepwise (Criteria: Probability-of-f-to-enter <= .0050,
1 Nil
cost $m Probability-of-f-to-remove >= 0.100).
150
CHAPTER 4
Data collection, analysis techniques and statistics
A normality of error test on the regression model was carried out to test the robustness
around the regression line. This was tested by plotting the normal probability of the sorted
values of the residuals versus the associated theoretical values from the standard normal
distribution. This was to check whether or not it was reasonable to assume that the random
errors inherent in the process had been drawn from a normal distribution. The assumption
adopted was that the variability of data around the regression line should have been constant
for all values of X in the regression model. The normal probability plot was ascertained to
see if the residuals were normally distributed by lying along a 45 degree upward sloping
diagonal line. The null hypothesis was that the residuals were normal, and to the extent that
the graph deviated moderately from the 45 degree pattern then the normality assumption
needed to be questioned. Figure 4.2 shows the Normal probability plot of the revised project
cases.
0.8
0.6
0.4
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0
The plot was inspected to see if the random errors were normally distributed with the plotted
points lying close to a straight line. The distinct curvature deviations from the straight line
indicated that the random errors were probably not normally distributed. As can be seen, the
151
Risk factors leading to cost overrun in the delivery of highway construction projects.
graph demonstrated poor conformance to a 45 degree diagonal line with some values
plotting away from the straight line.
The next section describes the test carried out for homegeneity of variance.
The next graph produced was the scatterplot of the standardised residuals versus the
estimated standardised (or fitted) values. This was to provide an insight into the assumption
of equal variances, as well as the assumption that X and Y had a linear relationship in the
regression equation. Figure 4.3 shows the scatterplot of the standardised residuals versus
predicted values of the original 220 cases.
-1
-2
-6 -5 -4 -3 -2 -1 0 1
Figure 4.3: Scatterplot of the standardised residuals versus predicted values of 220 cases
The plot is decreasingly dispersed across left to right. This suggested that the variances were
not constant along the regression line and indicated a violation of the homogeneity-of-
variance assumption (homoskedasticity). This rendered the inference unreliable by not
having a more randomly scattered plot in an even, horizontal band around the residual value
of zero. One reason for the shape of the plot was that the project data used in the research
was limited to the inclusion of projects => 10% cost overrun. As well, the lower project
programmed cost was limited to => $1m. In further examining the plot, a number of points
exceeded +2. Under the model assumption, 95% of the standardised residuals were expected
to be within the range of -2 to +2.
The bulking of the data to the right of the plot indicated that some form of data
transformation may have been required so as to satisfy random disturbances. This is
discussed in the following section.
152
CHAPTER 4
Data collection, analysis techniques and statistics
A histogram showing the frequency of the regression standardised residuals was next
produced (Figure 4.4) to ascertain whether the shape was consistent with the assumption of
normality.
60
50
Frequency
40
30
20
10
Mean =1.63E-16
0 Std. Dev. =0.998
N =220
-2 0 2 4
The histogram showed a relatively heavy positive skew of the data. This indicated that data
transformation was required to improve the model. Hutcheson and Sofroniou (1999)
suggested taking the reciprocal of X where X is an observed value for heavy positive
skewness.
The indexed programmed cost transformed using the reciprocal of the observed values.
Table 4.53 shows a sample from the 220 projects with transformed indexed programmed
costs.
Project number Indexed prog. cost $m Reciprocal indexed prog. cost $m % over cost
140/U88/3 47.79 0.02092 18.0
160/12A/8.566 46.02 0.02173 25.3
160/12B/5 42.55 0.02350 12.6
70/25A/13 40.42 0.02474 14.7
153
Risk factors leading to cost overrun in the delivery of highway construction projects.
A re-inspection of the residuals as surrogates for random disturbances was carried out so as
to make further judgements on the validity of the transformed data and an associated
regression model. Graphical plots were again used for testing the normality and homogeneity
for the revised model's validation. Figure 4.4 shows the revised Normal Probability Plot
when using the transformed project data.
0.8
0.6
0.4
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0
The scatter plot of the standardised residuals versus the estimated standardised transformed
values is contained in Figure 4.5.
154
CHAPTER 4
Data collection, analysis techniques and statistics
-1
-2
-2 -1 0 1 2
It was observed that the plot was dispersed uniformly from left to right. By having a
randomly scattered plot in an even, horizontal band around the residual value of zero, this
rendered the inference of homogeneity reliable. There were still a number of data points
which exceeded +2 for the standardised residuals.
Table 4.54: Model summary using reciprocal indexed prog. cost $m transformed data
155
Risk factors leading to cost overrun in the delivery of highway construction projects.
Unstandardised Standardised
T Sig.
Model Variable coefficients coefficients
B Std. error Beta
1 (Constant) 21.751 2.467 8.816 0.000
Reciprocal of prog.cost$m
14.519 4.872 0.198 2.980 0.003
The residual statistics for the model are detailed in Table 4.56 below and show the standard
deviation for the residuals at a relatively wide 18.746 for the 220 projects analysed.
4.13 Sensitivity testing using discrete rural and urban data sets
Further univariate regression testing was carried out on separate rural and urban project data
sets to ascertain if statistically stronger correlations could be developed by separating out the
project data into two discrete datasets. There were 134 rural project cases and 86 urban
project cases available for analysis and these agreed with the 220 cases previously derived
after the identified outliers were removed.
The predictor variable used in the analysis was reciprocal of prog. cost $m and the
dependent variable used was % over cost .
The rural project data were analysed and the model summary was derived as shown in
Table 4.57.
The R value for the model was <0.2 and thus had negligible correlation. The R2 value of
0.016 was statistically very poor. Column four shows the adjusted R2 at only 1.6%.
156
CHAPTER 4
Data collection, analysis techniques and statistics
Unstandardised Standardised
t Sig.
Model coefficients coefficient
B Std. error Beta
1 (Constant) 22.158 3.427 6.467 0.000
Reciprocal of prog.cost$m
11.534 6.476 0.153 1.781 0.077
The residual statistics for the model are detailed in Table 4.59 below and show the standard
deviation for the residuals at a relatively wide 17.9879 for the 134 projects analysed
An inspection of the residuals as surrogates for random disturbances was carried out on the
rural data so as to make further judgements on the validity of the transformed data and an
associated regression model. Graphical plots were again used for testing the normality and
homogeneity for the revised models validation as shown Figures 4.6 and 4.7.
157
Risk factors leading to cost overrun in the delivery of highway construction projects.
1.0
Expected cumulative probability
0.8
0.6
0.4
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0
The scatterplot of the standardised residuals versus the estimated standardised transformed
values is contained in Figure 4.7.
Scatterplot
-1
-2
-2 -1 0 1 2
Figure 4.7: Scatterplot of standardised residuals versus predicted values for rural projects
158
CHAPTER 4
Data collection, analysis techniques and statistics
The plot was dispersed uniformly from left to right and rendered the inference of
homogeneity reliable by having a randomly scattered plot in an even, horizontal band around
the residual value of zero. A number of data points still exceeded +2 for the standardised
residuals.
The 86 urban projects were analysed and the model summary was derived as shown in Table
4.60.
The R value for the model was slightly >0.2 and thus had minimal correlation. The R2 value
of 0.060 was statistically poor. Column four shows the adjusted R2 at only 1.6%.
Unstandardised Standardised
Model coefficients coefficients t Sig.
B Std. Error Beta
(Constant) 21.920 3.617 6.059 0.000
1 Reciprocal of prog.
17.465 7.574 0.244 2.306 0.024
cost $m
The residual statistics for the model are detailed in Table 4.62 below and show the standard
deviation for the residuals at a relatively wide 19.923 for the 86 projects analysed
159
Risk factors leading to cost overrun in the delivery of highway construction projects.
Figure 4.8 shows the revised Normal probability plot when using the urban project data.
1.0
Expected cumulative probability
0.8
0.6
0.4
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0
The scatterplot of the standardised residuals and the estimated standardised transformed
values is contained in Figure 4.9.
160
CHAPTER 4
Data collection, analysis techniques and statistics
Scatterplot
5.0
2.5
0.0
-1 0 1 2
Figure 4.9: Scatterplot of standardised residuals versus predicted values of urban projects
Again it was observed that the predicted value plot was dispersed relatively uniformly from
left to right and rendered the inference of homogeneity reliable by having a randomly
scattered plot in an even, horizontal band around the residual value of zero but with a slight
overflow of data points towards +2. A number of data points still exceeded +2 for the
standardised residuals.
4.14 Sensitivity testing using open tender and negotiated price data sets
Further univariate regression testing on separate open tender and negotiated price project
data sets was then carried out to ascertain if a statistically stronger correlation would be
developed within the separate project clusters. There were 131 open tender project cases and
89 negotiated price projects analysed.
All relevant predictor variables were used with the dependent variable % over cost in the
stepwise multivariate regression anlaysis on both individual data sets. The predictor variable
used was reciprocal of prog. cost $m with the dependent variable again being % over cost.
Both data set analyses were unremarkable in their statistical significance and again showed
no increased correlation properties over the full data set of 220 projects.
161
Risk factors leading to cost overrun in the delivery of highway construction projects.
The presence of the many considered variables here may create results contrary to what has
been posed and what would normally have been expected. It was anticipated that the results
of the research would contribute strongly in providing rational correlations between highway
projects and their cost overrun. The validation process was an important step to see if any
developed model best fitted the available data. All models were tested statistically to
determine whether they confirmed a correlation between real project attributes and real
project cost overruns. However, it was found that the regression models so developed
perform poorly in representing correlations within the current project data.
The regression analysis did demonstrate a weak correlation between the size of highway
projects, as measured in indexed programmed cost and the size of cost overruns. The
correlation evolved after data transformation was carried out to improve the model.
For the regression formulae y = bx + c, the correlated model so devised had the following
form, as extracted from Table 4.55.
% project cost overrun = 14.519 x reciprocal of indexed programmed cost $m.+ 21.751
In effect, this model has shown that as the size of a project increases in budgeted
programmed cost, then the percentage cost overrun reduces in value. It further suggests that
to counter the potential cost overrun, the over-and-above additional contingency percent that
should be factored into all types of highway projects by the client to compensate for
undefined (or under-defined) project risks can be varied depending on the project size. Such
indicative over-and-above contingency percentages that have been derived from the model
are shown in Table 4.63.
162
CHAPTER 4
Data collection, analysis techniques and statistics
The statistical tool of multivariate regression was used to analyse the correlation between the
following project variables that were simultaneously associated with the dependent variable
of cost overrun percentage:
Highway geographic project type (urban/rural projects)
Highway project construction type (project types 1 to 12)
Highway project delivery type (delivery code 1, 2 or 3)
Indexed highway project programmed cost (indexed prog. cost $m)
Highway project high level risk grouping (HL/1 to HL/10 and HL/Q).
A wide spectrum of highway projects, delivery methods and client risks were analysed using
three separate modes of variable selection in the regression analysis. There were 2
geographic location types, 12 project types, 3 project delivery methods, and 11 high level
risk groupings that were hypothesised to have a correlation to the amount project overrun in
cost for the various projects analysed. The null hypothesis adopted for Pearson's correlation
tests was that there was no correlation between the size of cost overrun of projects (and their
geographic location, construction type, delivery process, project size, and client project risks.
Model testing was applied to confirm or otherwise that the given independent (prediction)
variables in the model did not correlate against the dependent variable of project cost
overrun to any significant level of accuracy.
All models were tested statistically and practically to determine whether they are efficient in
predicting real project cost overrun results and based on the adopted statistical checks, all the
regression models perform poorly in representing the project data.
It was hypothesised that there existed a defined relationship between the measure of the
percentage of cost overrun in highway construction projects and:
construction risks
types of highway construction projects
project delivery methods
project location
163
Risk factors leading to cost overrun in the delivery of highway construction projects.
project size.
The correlation relationship of the multiple project variables was troublesome because the
sought-after knowledge about cost overrun in highway projects was complex and had many
different combinations of factors that had to be examined. The statistical modelling and
analysis of the research data has found no strong correlations between project types, work
types and project risk factors in producing cost overrun in highway construction projects.
There was, however a weak model that showed that as the size of a project increased in
budgeted programmed cost, then the percentage cost overrun reduced in value.
Chapter 5 follows and discusses the research findings in fulfillment of the research
objectives and documents its contribution to original research. It also suggests areas of
further research.
164
CHAPTER 5
Discussion, recommendations and conclusions
CHAPTER 5
This chapter discusses the research findings in fulfillment of the research objectives and
documents its contribution to original research. It also suggests areas of further research in
highway project management.
5.1 Discussion
The overall objective of this research has been to identify the factors that influence
significant project cost overruns for the client and to provide an analytical model that
correlates project attributes to the level of their cost overruns and client project risks relating
to decision-to-build budgets.
The research methodology has been structured to seek out answers to the following research
questions:
1. What client risks are present during the delivery of highway construction projects in
Queensland, Australia that lead to significant project cost overruns?
2. How does the amount of highway cost overrun in such highway projects correlate
with their project types, size, delivery processes and client project risks when
historical project data are analysed?
In order to provide adequate answers to these questions, the following research stages were
adopted:
1. the establishment of a data source for highway construction projects
2. the determination of project work types and cost overrun factors from historic case
study data
3. the utilisation of principal component analysis and factor rotation on cost overrun
factors in order to consolidate data
4. the use of the nominal group technique with highway construction experts to elicit
groupings of cost overrun factors and highway project types
5. the use of statistical analysis using multivariate linear regression analysis to
investigate correlations between cost overrun risk factors and project attributes by
using historic project data.
The client-allocated budgets to deliver the projects included the following costs:
165
Risk factors leading to cost overrun in the delivery of highway construction projects.
In order to carry out a valid analysis on the project data over the research period it was
necessary to bring all relevant project cost data to a common out-turn financial year of 2002
03. The Road Input Cost Index (RICI) was adopted and applied to the research cost data.
Table 5.1 details the RICI indices for the analysis period. The % adjustment in highway
input costs has increased from 1.3% in 199596 up to 3.4% in 200102. The average
increase over the seven-year analysis period was 2.4% per year and the index data shows a
16.3% increase in highway costs over the period.
Wilmot and Cheng (2003) carried out research to predict the cost escalation for future
highway construction projects in the US. They used the Bureau of Economic Analysis of the
Department of Commerce for predictions of future construction worker wages. The predicted
worker wages were transformed into an index of construction labor cost with a value of 100
in 1987. Data on the cost of highway construction equipment and material for the period
19842015 were purchased from the commercial supplier of industrial data, Data Resources
Incorporated and were transformed to an index with a value of 100 in 1987. Their model
predicted that future overall highway construction costs would double in the period 1998
2015, representing a growth rate of approximately 4.2% per year, considerably higher than
the current rate of general inflation of 2.5% in the US.
166
CHAPTER 5
Discussion, recommendations and conclusions
%
Cost overrun variable Occurrence
occurrence
1 Design/project scope change 95 31%
2 Contract tender price higher than original estimate 35 11%
3 Design scope change drainage 33 10%
4 Quantity increased measure 31 10%
5 Design scope change pavement materials/depth 23 7%
6 Latent condition remove and replace
unsuitable material
21 7%
7 Design scope change environmental issues 19 6%
8 Constructability under traffic 17 5%
9 Services relocation costs 12 4%
10 Material cost increase pavement materials 11 3%
11 Resumption/accommodation works 10 3%
12 Constructability difficulty costs 10 3%
Table 5.2: Top twelve highway risk factors causing cost overrun
Detailed descriptions of significant cost overrun variables have been provided in Appendix
E.
The number of component variables reproduced in the analysis was determined by the
number of principal components with eigenvalues of 1.00 or greater. Factor variation was
relatively even across all components and this indicated low success in reducing the
dimensionality of the original set of cost overrun variables using factor analysis.
In order to facilitate a better interpretation of factors, the factor analysis required the rotation
of axes that did not affect the goodness-of-fit of the factor solution. The varimax method of
rotation, as developed by Kaiser, was adopted and provided the best parsimonious analytical
solution. Computationally, the varimax rotation minimised the number of variables that had
high loadings on the factors and caused the factor loadings of each variable to be marginally
more clearly differentiated.
A review of the 13 groupings led to the conclusion that no common factor components
demonstrated consistency in purpose from the principal component analysis. There was no
167
Risk factors leading to cost overrun in the delivery of highway construction projects.
evidence of strong correlations within any of the 13 groupings, even after rotation. A weak
scree plot also indicated the derived model of 13 component factors was not a reliable
representation of the total factors for use in the model analysis.
The advantages of the NGT process were numerous. NGT provided a constructive, problem-
solving approach that permitted equal participation by all the group members, and avoided
the disproportionate influence by vocal individuals who are often present in group processes.
It also reduced the pressure for individual members to conform to group opinion. In the
feedback obtained during the process, participants appreciated the opportunity to give
meaningful input.
The NGT group composed of two generalists, five specialists, an analyst who took on the
role of moderator and a researcher. This membership is supported by Clemen and Winkler
(1985) who have suggested that five specialists are usually sufficient to cover most of the
expertise and breadth of opinion but with an additional two or three generalists and two or
three analysts for specific elicitation tasks.
The spread of project-related experience of the members as shown in Table 5.3 demonstrated
the good balance of project management expertise that existed across the group.
One problem encountered during the elicitation process was the poor definition of some cost
overrun factors provided to the group and the lack of specific project histories. To overcome
this issue, the definitions of the factors causing the cost overrun were redefined as necessary.
168
CHAPTER 5
Discussion, recommendations and conclusions
Further research has also identified the project phases in which the risks can highly impact
project cost. A Likert scale of L for low phase impact, M for a medium phase impact and H
for a high phase impact has been used to identify the most susceptible project development
areas. The final mapping is shown in Table 5.4. This supports the previous findings in Table
5.2 that the major propensity for significant cost overrun is predominantly located in the
design phase of project development.
Construction
Preliminary
Concept
Design
Risk group Cost overrun factors
169
Risk factors leading to cost overrun in the delivery of highway construction projects.
Further research findings have been developed in Table 5.5 to show risk sources capable of
causing potential project cost overrun. Highway clients need to take these into account when
they are developing new projects in order to reduce the possibility of their projects
significantly exceeding budget.
Project acceleration
Government initiative
employment continuity
Developer, local
Potential for poor quality and technical
government,
non-performance
rail contributions
Client project Uncertain design experience level
4 Material/process quality
management Poor team cohesiveness or composition
Contract failure
Estimates uncertain as they can be based on
new contract
past projected costs
establishment
Project administration
Tender price higher than
client estimate
Relocation of utility Approvals and environmental concerns for
Services
5 services alignments can cause time delays and/or
relocation
in road reserve cost escalation
6 Constructability Constructability Level of difficulty increases the potential
170
CHAPTER 5
Discussion, recommendations and conclusions
Twelve highway project types that collectively described all of the 231 highway projects in
the data have been developed using the NGT process as shown in Table 5.6. Clients need to
consider the standardized project types shown in order to be in a position to effectively
analyse highway project data across organizations. The non-adoption of a standardized group
of highway project types will only further perpetuate inefficiency in analyzing historical
highway project data.
Project type
Reference
Highway
R1 Construct to sealed standard (2 lanes)
R2 Widen shoulders and sealing
R3 Rehabilitate/Strengthen
R4 Realignment of two lanes
R5 Realignment of four lanes
R6 Duplicate two into four lanes
R7 Widen four into six/eight lanes
R8 Asphalt resurfacing
R9 Construct intersection/interchange
Bridge
B1 Bridges and approaches
B2 Widen bridges
171
Risk factors leading to cost overrun in the delivery of highway construction projects.
Miscellaneous
M1 Miscellaneous roadworks/traffic furniture and devices
The research also found it desirable to differentiate between certain project types where
complexity or location need to be identified at a more specific level. The research found that
the following project aspects needed to be considered in interrogating future details were
readily available:
rehabilitation projects include pavement widening or not
a realignment project incorporated some of the existing road and was built under
traffic this type of project is referred to as being constructed at a brownfield site
the project was on a completely new alignment away from the existing highway that
it had replaced, thus having none or only a small exposure to highway traffic this
type of project is referred to as being constructed at a greenfield site
an interchange/intersection or bridge and approaches were located in an urban or
rural environment.
The determination of the highway project types aligned well with the literature research
finding of Persad et al. (1995) who reported on a highway project classification developed
by the Texas Department of Transport (TDoT) in 1986. It reported 14 types of highway
projects as detailed in Table 5.7.
172
CHAPTER 5
Discussion, recommendations and conclusions
The dependent variable adopted in the model was the continuous variable percentage (%)
cost overrun and has been defined as the difference between the clients actual project cost
and their programmed cost, expressed as a percentage of the programmed cost.
The research has analysed the correlation between the following project variables:
1. Highway geographic project type (urban/rural projects)
2. Highway project construction type (project types 1 to 12)
3. Highway project delivery type (delivery code 1, 2 or 3)
4. Client project high level risk grouping (HL/1 to 10 and HL/Q)
5. Indexed highway project programmed cost (Indexed prog. cost $m).
Many regression models were possible due to the wide spectrum of highway projects,
delivery methods and client project risks that were associated with cost overrun. Two
geographic location types, 12 project types, three project delivery methods and 10 high level
client risks were hypothesised as influencing the amount of cost overrun in highway project.
The null hypothesis adopted was that there was no correlation between the size of cost
overrun of location of projects, construction type, delivery process, project size or client risk
factors.
Three stages of statistical regression were computed using the method of least squares. For
the purpose of research, the statistical models were considered linear normal models (i.e.
regression analysis with the appropriate f -tests and t-tests). For each test, the p-value was
reported as a measure for rareness if identity of groups was assumed. The forward selection,
backward selection and stepwise selection regression methods were used in the analyses.
The stepwise method delivered the most appropriate model after excluding outlier data and
data transposition.
Correlation analysis was undertaken to identify project variables that correlated with project
cost overrun. The technique examined the performance of various models and the
relationships between variables. Pearson's correlation coefficient (R) was used to examine
the relationship between the data and for developing the rank order of regression models in
terms of goodness of fit. The research used the coeffiecient of multiple determination R2
and adjusted R2 statistics as they allowed direct comparison of the best model identified.
A number of important analysis findings were revealed from the regression analyses as
follows:
1. The research found no strong correlation between the geographic location of the
highway projects and project cost overrun. This finding did not support Drew and
Skitmore (1992) who identified that the density of population and the extent of
geographic area were important factors for competitive tenders in building projects.
The difference may have been because the research focused on highways, whereas
Drew and Skitmore focused on building projects. As well, highway projects
researched were all located in the one area, Queensland and not a broader geographic
survey as was the case for Drew and Skitmore (1992).
2. The research found no strong correlation between highway project types and project
cost overrun. However, the research found that projects involved in constructing six
or eight lanes from four lanes demonstrated excessive cost overruns when analysed
for discordance. These were subsequently excluded as outliers so no data for six/eight
173
Risk factors leading to cost overrun in the delivery of highway construction projects.
3. The research found no strong correlation between the type of project delivery method
of highway projects and cost overrun. Projects that were delivered by open contract,
as apposed to negotiated price, were no less susceptible to significant cost overrun.
4. It was postulated in the analyses that strong correlations existed between the groups of
client project risks and degree of cost overrun. The research found that no correlation
between the client project risks and project cost overrun existed.
The research found correlation between indexed highway programmed cost and cost overrun
after data translation. For the regression formulae y = bx + c, the correlated model had the
following form:
% project cost overrun = 14.519 x reciprocal of indexed programmed cost $m.+ 21.751
This model indicates that, as the size of a project increases in budgeted programmed cost, so
then does the size of the percentage cost overrun reduce. In contrast to this finding, research
carried out in 1992 by the Transportation Research Board evaluated construction cost
overruns on 468 transport projects completed for the Washington State Department of
Transportation. Results of their analysis indicated that cost overruns, expressed as a
percentage of the original contract amount, increased with the size of the project.
This research finding has also been inconsistent with the research of Williams (2003) who
analysed transportation projects in both the UK and the US. He found a (log.) linear
relationship between contract size and cost overrun. The reasons for the inconsistency
appeared to be the fact that Williams analysed only contract elements of projects and also the
size of projects. Williams (2003) analysed projects which included contract values below the
equivalent of A$1m in predominately highway rehabilitation projects. As well, he initially
included dredging contracts in his data analysis in order to demonstrate linkages across
project types in the US. In contrast, only highway construction and rehabilitation projects in
Queensland that were >A$1m formed the basis for the data analysis. Both sets of projects
were based on the design-bid-build delivery model. Presumably the differing broad findings
are due to the unique geography of highway construction projects in Queensland, as opposed
to that of the US.
The research indicated that the percentage of project cost overrun is linked to the economy
of scale of projects, such that smaller dollar projects can attract larger percentages of cost
overruns, and larger dollar projects have the potential for smaller percentages of cost
overruns.
Smaller projects have a tendency towards a higher percentage increase in cost. The actual
dollar magnitude of the cost overrun for the large project may still be greater, even though its
174
CHAPTER 5
Discussion, recommendations and conclusions
percentage change is less. For example, a 5% cost increase on a $1 million dollar project is
$50, 000, whereas a 1% cost increase on a $100 million dollar project is $1,000,000.
The research also lends itself to the assertion that additional contingency percentages can be
derived from the model that can be applied to projects sizes when decisions have to be made
on the size of individual projects. These proportional contingency percentages to projects'
decision-to-build budgets of particular sizes are shown in Table 5.8.
A check on the regression model was carried out in order to ascertain if a wider range of data
outliers existed in the regression analysis. The approach undertook sensitivity testing on the
data by dividing it into discrete data types. Regression re-testing was carried out on four
separate data sets:
rural projects
urban projects
open tender projects
negotiated price projects.
There was no discernable improvement in the correlation model produced from any of the
four discrete datasets.
175
Risk factors leading to cost overrun in the delivery of highway construction projects.
time, so that corrective measures can be better taken into account in project design
and estimating
provided an in-depth post mortem analysis of client risks that have led to significant
project cost overrun and thus has strengthened the understanding of why highway
projects can overrun substantially above project decision-to-build budgets.
identified client risk factors that are aimed at simply providing a better understanding
of the risk parameters and management contingency requirements in project, thus
validating better risk contingency weightings in budget estimates
provided a standardised set of highway project types against which future highway
project data can be collated and analysed
provided importance ranked groupings of client project risks and descriptions of
potential risk sources which clients need to take into account when developing new
project budgets
established NGT as a suitable method for risk elicitation and used it in the research
project in a way that required minimal impact on the time of construction experts, but
at the same time providing an acceptable outcome to a difficult elicitation
requirement
accounted for uncertainty when developing client project cost estimates by proposing
a range of over and above % contingency to account for economy of scale in
highway projects for varying sizes of highway projects.
The construction industry is unique in that it rarely builds something the same way, at the
same place, or with the same set of people. Because construction professionals tend to
describe risk in linguistic terms based on their experience and judgement, the research has
aimed at providing the client with the background knowledge for representing project budget
cost estimates in an ideal situation where all epistemic (lack of knowledge) uncertainty has
been identified. As well, the construction industry is in need of project management tools
that are straightforward and take advantage of the industrys tendency to use engineering
judgement to solve problems. The use of a sound elicitation risk assessment technique, such
as in this research, provides a structured format to harness the experience and judgement of
experts in assessing the financial impact of risks on projects in an efficient and time
conscious manner.
The research has detected no real trend in improvement in cost estimating over the seven
years covered by the highway projects analysed. However, in the historic data analysis
period adopted for this research, there was evidence that a good proportion of highway
projects achieved their budget at project completion. This suggests that the project estimators
associated with those projects have done a good job in terms of estimate accuracy. In this
research, the project factors that have been analysed have not proven strong correlation to
project overrun and so it can be implied that highway project estimators are, on the whole,
making substantial allowances to cover project parameters that cause budget overrun.
176
CHAPTER 5
Discussion, recommendations and conclusions
While the research has been unable to supply any single and definitive answer to the
predominant reasons for project cost overruns, the persistent underestimation of costs and
the over-optimistic assumptions about performance of some project budget estimates raises
questions about why client highway organisations have not been able to adjust their
expectations over the years. Several possible remedial actions can be derived from the
research.
Firstly, the literature research has shown that contingency percentages of around 10% are
commonly allocated to project estimates at the decision-to-build time to allow for unforeseen
and unquantified client risks. The client organisation is reported to include a 10%
management contingence to its decision-to-build budget estimates in the research period to
take into account any underestimated client risks. The research now indicates that a higher
order percentage in management contingency may be required for all highway projects
(excluding six and eight-lane projects). This higher order contingent percentage is required
to minimise the protential for projects to significantly exceed the client budgeted cost
because of undefined, or under-defined, client risks. Therefore, an estimating procedure
which includes revised levels of management reserve contingencies, ranging from 35% for
projects up to $3m, down to 25% for larger projects, should be considered. This revised
contingency would reduce the potential for projects to be reported as having incurred
significant budget over-cost during delivery.
Secondly, client estimating procedures for highway projects should take into account the
potential cost impacts from the risk factors identified in this research. Failure to
systematically evaluate the project risk potential of all of the risk factors could severely
impact on the favourable outcomes of projects and on a client's overall highway construction
program.
Thirdly, the research found that design and scope changes are the highest contributing risk
factor to project cost overrun in highway projects analysed. This finding is supported by
Tilley et al. (2000) who found that the quality of project design and design documentation
has fallen considerably over the past 15 to 20 years. McLennan and Jorss (2006) reported
that poor design documentation includes the following root causes:
inadequate project briefs based on unrealistic expectations of time and cost
lack of integration along the supply chain linking parties, and between project phases
poor understanding and low skills in risk assessment and management
inadequate use of CAD/computer technology in the design process and in the
compilation of specifications.
To further manage cost overruns from design and scope changes the following qualitative
measures need to be considered:
177
Risk factors leading to cost overrun in the delivery of highway construction projects.
The tight control of any scope creep by the client can limit variation works during
construction and this is absolutely necessary for successful financial outcomes of projects.
The statistical model assumed error free measurement, at least of independent (predictor)
variables. However, measurements are seldom perfect. Therefore, close attention is needed
to be paid to the effects of measurement errors. This is especially important when dealing
with noisy data or processes which are difficult to measure precisely (Helberg, 1996).
This research was confined to the study of highway projects in the state of Queensland,
Australia. Since construction costs can be specific to geographical/economic areas and time
periods, caution is required when comparing the data and results to another situation.
The research findings lend significance to a broad range of highway client organisations.
However, it needs to be said that the historic data used in this research were founded on the
following:
178
CHAPTER 5
Discussion, recommendations and conclusions
projects were all of the design-bid-build project delivery method which may have
lead to higher client exposure to design risks
project contractors were all pre-qualified under a maximum capacity rating
assessment and this pre-qualification process had the potential to limit client risk
exposure to contract default
contract payments were on the basis of schedule-of-rates and projects used bill of
quantities in their contract elements
specifically tailored contract provisions can have potential to limit the client highway
organisation to the exposure of adverse physical and latent conditions events which
sometimes occur in highway projects under certain circumstances.
Further research could be undertaken by testing the model by implementing two contingency
allocation percentages on half the number of highway programme projects each, and the
differences in budget cost reporting observed and reported on. Figure 5.1 shows this future
research proposal that is designed to see if project managers try harder to keep payments
down.
179
Risk factors leading to cost overrun in the delivery of highway construction projects.
Highway
construction projects
=>$1m budget
The proposal could also be modified further by the inclusion of a blind placebo test to a
small number of projects by the inclusion of unmodified contingency % within the modified
contingency program. This mt ascertain if there is any a placebo effect that leads to further
project cost savings.
The linkage between the risk assessment and determining the appropriate probability
distributions for various highway project types should be researched for a more analytical
connection. Perhaps an Analytical Hierarchy Process (AHP) or another decision technique
could be used. The level of effort required to produce this tighter linkage needs to be
balanced against the potential increase in accuracy of the results. Project risk assessment is
readily adaptable to fuzzy set theory where it is preformed in linguistic terms and then
reduced to mathematical expressions. This may provide a convenient linkage between the
evaluation of various highway risks and a numerical cost control technique for differing
highway project types.
This research effort has identified and quantified the risk drivers that affected the degree that
projects significantly overrun in the highway infrastructure sector. Future research is needed
to adequately identify and quantify the project cost overrun drivers as they relate to the
construction of buildings and other infrastructure project types.
Finally, there is a need for further studies into the storage and retrieval of historical project
data. Historical data of finished projects have to be made more accessible. In view of the
nature of construction projects with data being stored in different formats such as site diaries,
calculation sheets, payment systems and even annotated drawings, the use of sophisticated
information technology is very desirable. This further research into insufficient access to
historical data will ultimately allow estimators to generate and deliver better project cost
estimates.
180
CHAPTER 5
Discussion, recommendations and conclusions
The research into high cost overrun required a robust sample of highway infrastructure
projects that was appropriate for this area of research. The data was required to be large
enough to allow statistical analyses of cost overrun factors and project costs and research
investigations found that data on actual costs in transportation infrastructure projects were
relatively difficult to come by. One reason is that such data is time consuming to generate.
The publication of confidential of commercial-in-confidence project data has been scarce as
it can be of concern to organisations that have provided the information.
The research utilised price movement indicators of construction outputs as a valuable tool in
the analysis output price indexes so as to enable the measurement of changes over time in the
price of new roads and bridges.
RICI indices were used over the analysis period to factor up the historical project cost
information. These factors ranged from 3.4% to 16.3% and were applied to the project
programmed and actual costs for the corresponding financial years in which the project was
delivered.
The focus of the research analyses were based on client exposure to project cost overrun, not
that of contractors delivering the projects. The literature research identified a number of
considerations to be taken into account when adopting the cost overrun factors. These
included:
the use of design-bid-build contracts that could lead to higher client exposure to
design risks
pre-qualification of contractors that has the potential to limit client risk exposure to
contract default
contract payment types that focused on schedule of rates and bill of quantities
contract clauses that were designed to reduce the clients exposure to certain
construction risks
tender evaluation techniques
contract provisions that limit the client's exposure to adverse physical and latent
conditions and wet weather events.
The risk identification process in this research has been beneficial as it focused the attention
of the client on the detection and control strategies of risks. The research has provided an
analytic procedure that allows the recognition of realistic cost overrun contingencies to
projects and investigates the influences of risk factors on clients' risk handling decisions
181
Risk factors leading to cost overrun in the delivery of highway construction projects.
The research has not looked at the whole project development life cycle but rather that part
where major cost overrun occurs, namely from the time the decision-to-build has been made
by the client and sound lessons have been drawn from the researchers post-mortems of the
many highway projects considered. The analysis has produced important findings about the
reasons that highway projects have overrun and has provided conclusive evidence of the
most important risks that highway agencies need to focus their efforts on. Of particular
concern is that of changes in project designs and scope changes during project development.
The research process used the experience in highway construction and the professional
judgement of the researcher to determine the listings of work types and reasons for cost
overrun using the NGT process. From initial interviews by the researcher, it was important to
take note of interviewees concerns about time management. In the rapidly paced world of
construction management, it was reinforced that an expert elicitation method needed to be
efficient because experts in construction management were continually confronted with
insufficient time. The selection of the NGT elicitation technique was therefore determined to
a large extent by the time that experts were prepared to allocate to the research project.
The final stage of the research process has involved the investigation into statistical models
which can explain the correlation between the cause, effect and other relationships relating to
cost overrun in highway construction projects. While the term model can be used in many
ways, however for this research it has referred to the dynamic framework or schema that
helps portray key concepts and propositions of the research phenomenon.
The regression analysis demonstrated a weak correlation between the size of highway
projects, as measured in indexed programmed cost and the size of cost overruns. The
correlation evolved after data transformation was carried out to improve the model. It can
also be concluded from the research that the arbitrary application of a base contingency
percentage figure, such as 10%, to accommodate project risk can lead to those projects
reporting substantial budget overrun. The provision of more realistic contingency
percentages across such projects will go a long way in providing for better reporting of
highway project and programme results and associated key project performance
indicators.Sensitivity testing was investigated for the stepwise regression models developed
in order to ascertain whether additional outlier data still unduly influenced the models.
The model developed has demonstrated that as the size of a project increases in budgeted
programmed cost, then the percentage cost overrun reduces in value. It further suggests that
to counter the potential cost overrun, the over-and-above additional indicative contingency
percent that should be factored into all types of highway projects by the client to compensate
for undefined (or under-defined) project risks can be varied depending on the project size
To aid in the identification of future highway risks, this research has identified and compiled
listings of highway project risks that have lead to cost overrun, as well as the incidence of
each risk across the sample highway projects. The development of such listings should help
to prevent projects risks from being overlooked by clients. The research has also illustrated
how highway projects have overrun and have done so beyond any reasonable
expectations. The research has considered historical projects where the cost overrun was
significantly beyond what might have ever been anticipated. The aim of the research was to
contribute to the understanding of how highway projects go wrong, when they do, and in
particular to draw some lessons from this exploration.
182
BIBLIOGRAPHY
BIBLIOGRAPHY
Adler, M. and E. Ziglio. 1996. Grazing into the oracle, The Delphi method and its
application to social policy and public health. London: Cromwell press.
Ahmed, S. M., R. Ahmed and D. E. Saram. 1999. Risk management trends in the Hong
Kong construction industry: a comparison of contractor and owners perceptions. Journal of
Engineering, Construction and Architectural Management 6/3:225-234., 6.
Akinci, B. and M. Fischer. 1998. Factors affecting contractors' risk of cost overburden.
Journal of Management in Engineering, 14 (1): 67-76.
Akpan, E. O. P. and O. Igwe. 2001. Methodology for determining price variation in project
execution. Journal of Construction Engineering and Management, 127 (5): 367-73.
Al-Sobiei, O. S., D. Arditi and G. Polat. 2005. Managing Owner's Risk of Contractor
Default. Journal of Construction and Engineering Management, 131 (9): 973-78 (accessed
September 2005).
183
Risk factors leading to cost overrun in the delivery of highway construction projects.
Anderson, S. D., D. Fischer and S. Rahman. 1999. Constructability issues for highway
projects. Journal of Management Engineering, 15 (3): 60-68.
Ang, A. H.-S. and W. H. Tang. 1984. Probability Concepts in Engineering Planning and
Design, Volume II Decision, Risk and Reliability. Vol. II - Decision, Risk and Reliability.
New York, NY: John Wiley and Sons.
Arndt, R. 1999. Efficient risk allocation in privately funded infrastructure projects. PhD,
Civil Engineering, University of Melbourne, Melbourne.
ASCE. ed. 1989. Avoiding and Resolving Disputes in Underground Construction: Successful
Contracting Practices and Guidelines. New York: American Society of Civil Engineers.
Ashley, D. B. 1989. Project risk identification using inference subjective expert assessment
and historical data. In Transactions of Internet International Expert Seminar, 9-28. Atlanta.:
Atlanta Institute of Technology.
Ashworth, A. 1994. Cost Studies of Building. London: Longman Scientific and Technical.
Ayyub, B. M. and A. Haldar. 1984. Project Scheduling Using Fuzzy Set Concepts. Journal
of Construction Engineering and Management, 110(2).
Ayyub, B. M. and R. H. McCue. 1997. Probability, Statistics and Reliability for Engineers.
Boca Raton, FL.: CRC Press.
Baker, S., D. Ponniah and S. Smith. 1999. Survey of risk management in major U.K.
companies. Journal of Professional Issues in Engineering Education and Practice, 125 (3):
94-102.
184
BIBLIOGRAPHY
Banks, J., J. S. Carson and B. L. Nelson. 1996. Discrete - Event System Simulation 2nd
edition. 2nd ed. Upper Saddle River, NJ.: Simon and Schuster.
Barnes, M. 1981. Project management by motivation. In Proc., PMI/ INTERNET 1981 Joint
Symposium, 101-123. Boston.
Barnett, V. and T. Lewis. 1984. Outliers in Statistical Data. 2nd Ed. ed. New York: John
Wiley & Sons, Inc.
Bathurst, E. P. and D. A. Butler. 1977. Building Cost Control Techniques and Economics.
London: Heinemann.
Bennett, R. 1991. How is management research carried out? In The Management Research
Handbook, eds. N. C. Smith and P. Danity, 85-103. London: Routledge.
Berny, J. 1989. A new distribution function for risk analysis. Journal of the Operational
Research Society, 40: 1121-7.
Blair, A. N. 1999. Risk Analysis of Cost and Schedule of Complex Engineering Systems.
Doctoral Thesis, University of Maryland, College Park, MD.
Blair, A. N., B. M. Ayyub and W. J. Bender. 2000. Fuzzy Stochastic Risk-Based Decision
Analysis: MOB Case Study. Marine Structures Journal, Society of Naval and Marine
Engineers, Sept. 2000.
Blair, A. N., L. M. Lye and W. J. Campbell. 1993. Forecasting construction cost escalation.
Canadian Journal of Civil Engineering, 20: 602-12.
Blake, E. H. 1900. Extras and omissions in building contracts. The Surveyors' Institutuion
transactions, XXXII: 525-72.
185
Risk factors leading to cost overrun in the delivery of highway construction projects.
Bryman, A. and D. Cramer. 1999. Qualitative Data Analysis with SPSS Release 8 for
Windows: A Guide for Social Scientists. London: Routledge.
BSI. 1999. Guide to project management: in British Standard BS 6079. Vol. 2. London.:
British Standards Institution.
Carelli, G. 1989. The Manus Opera House. Icaro, Varig Inflight Magazine, 17.
Carver, R. H. and J. G. Nash. 2005. Doing Data Analysis with SPSS Version 12. Ontario:
Brooks/Cole Thompson Learning.
Chang, A. S.-T. 2002. Reasons for cost and schedule increases for engineering design
projects. Journal of Management in Engineering, January 2002: 29-36.
Chapman, C. B. and S. Ward. 1997. Project Risk Management: Processes, Techniques, and
Insights. Chichester: Wiley N.Y.
Charette, R. 1989. Software Engineering Risk Analysis and Management. New York:
McGraw-Hill.
Chee, T. S. and K. T. Yeo. 1995. Risk analysis of a build-operate transfer (BOT) power plant
project. In IEEE Annual International Engineering Management Conference, 275-80.
CII. 1989. Management of project risks and uncertainty; Publication 6-8. Cost/Schedule
Controls Task Force. Austin, Texas: Construction Industry Institute.
186
BIBLIOGRAPHY
CII. 1996. The Partnering Process - Its Benefits, Implementation, and Measurement, .
Austin, TX.: Construction Industry Institute,.
Clark, F. D. and A. B. Lorenzoni. 1985. Applied Cost Engineering; 2nd Edition. New York:
Marcel Dekker, Inc.
Clemen, R. B. and R. L. Winkler. 1985. Limits in the precision and value of information
from dependent sources. Operations Research, 33: 427-42.
Clough, R. H. 1986. Construction Contracting. New York: John Wiley & Sons, Inc.
Cowie, G. 1987. Enough scope yields better estimates. Journal of Construction Engineering
and Management, 28 (8): 21-27.
Curtis, B., S. Ward and C. B. Chapman. 1991. Roles, responsibilities and risks in
management contracting: Construction Industry Research and Information Association
(CIRIA) Special Publication 81.
Cvethovich, G. and T. C. Earle. 1992. Environmental hazards and the public. Journal of
Social Issues, 48 (4): 1-20.
Dake, K. 1992. Myths of nature and the public. Journal of Social Issues, 48 (4): 21-38.
Dalcher, D. 1993. The new project management mindset for the 21st century. In Proceedings
of 1st British Project Management Colloquium. Henley-on-Thames, UK.
Dalkey, N. C. 1968. The Delphi Method: An experimental study of group opinion. Santa
Monica, California: The Rand Corporation.
187
Risk factors leading to cost overrun in the delivery of highway construction projects.
del Cano, A. 1992. Continuous project feasibility study and continuous project risk
assessment. International Journal of Project Management, 10 (3): 165-70.
Delbecq, A. L., A. H. Van de Ven and D. H. Gustafson. 1975. Group Techniques for
Program Planning: A Guide to Nominal Group and Delphi Processes. Glenview, IL.: Scott,
Foresman and Company.
Delbecq, A. L., A. H. Van de Ven and D. H. Gustafson. 1986. Group Techniques for
Program Planning: A Guide to Nominal Group and Delphi Processes. Wis.: Green Briar
Press.
Dey, P. K. 2001. Decision support system for risk management: a case study. Management
Decisions, 39 (8): 634-49.
Dias, A., and P. G. Ioannou. 1995. A desirability model for the development of privately-
promoted infrastructure projects.' UMCEE Report No. 95-09. Ann Arbor, Mich: Civil
Engineering Dept., University of Michigan.
Dozzi, S. P., S. M. AbouRizk and S. L. Schroeder. 1996. Utility theory model for bid
markup decisions. Journal of Construction and Engineering Management, 122 (2): 119-24.
Dyer, D. and J. H. Kagel. 1996. Bidding in common value auctions: how the commercial
construction industry corrects for the winner's curse. Management Science Journal, 42 (10.):
1463-75.
Eden, C., F. Ackerman and T. Williams. 2005. The amoebic growth of project costs. Project
Management Journal, 36 (2): 15-27.
188
BIBLIOGRAPHY
Egan, S. J. 1998. Rethinking Construction. The report of the Construction Task Force:
Department of the Environment, Transport and the Regions.
Elkington, P. and C. Smallman. 2002. Managing project risks: a case study from the utilities
sector. International Journal of Project Management,, 20 (1): 49-57.
Elsam, R. 1826. The Practical Builder's Perpetual Price Book. London: Kelly.
Emhjellen, M., K. Emhjellen and P. Osmundsen. 2003. Cost estimation overruns in the
North Sea. Project Management Journal, March: 23-31.
Emory, C. W. 1980. Business Research Methods, 2nd Edition. Illinois: Irwin & Co..
Farnum, N. R. and L. W. Stanton. 1987. Some results concerning the estimation of beta
distribution parameters in PERT. Journal of the Operational Research Society, 38 (3): 287-
90.
Fayek, A., D. M. Young and C. F. Duffield. 1998. Survey of tendering practices in the
Australian construction industry. EMJ - Engineering Management Journal, 10 (4): 29-34.
FDOT. 2003. Chapter 3: General management principles for transit capital projects. In
Project and Construction Management Guidelines, Vol. 2003 Update. New York: US
Department of Transportation.
Fisk, E. R. 1997. Construction Project Administration. 5th Edition ed. Englewood Cliffs, NJ:
Prentice-Hall.
Flanagan, R. and G. Norman. 1993. Risk Management and Construction. Oxford: Blackwell
Scientific Press.
Flyvbjerg, B., N. Bruzelius and W. Rothengatter. 2003. Megaprojects and Risk: An anatomy
of ambition. Cambridge, UK: Cambridge University Press.
189
Risk factors leading to cost overrun in the delivery of highway construction projects.
Flyvbjerg, B., M. S. Holm and S. Buhl. 2002. Underestimating Costs in Public Works
Projects: Error or Lie? Journal of the American Planning Association, 68 (3): 279-95
(accessed 2002).
Freudenburg, W. R. 1988. Perceived risk, real risk: social science and the art of probabilistic
risk assessment. Science, 242: 44-49.
Freudenburg, W. R. and S. K. Pastor. 1992. NIMBY's and LULU's: stalking the syndromes.
Journal of Social Issues, 48 (4): 39-62.
Gransberg, d. and M. Ellicot. 1997. Best-value contracting criteria. Cost Engineering, 39 (6):
31-4.
Gregory, R. and R. Mendelson. 1993. Perceived risk, dread, and benefits. Risk Analysis, 13
(3): 259-64.
Hair, J. F., J. Anderson, R. L. Tatham and Black. 1998. Multivariate Data Analysis. 5th
edition ed. Upper Saddle River, N.J: Prentice Hall, c1998.: Prentice Hall.
Hair, J. F. and R. L. Tatham. 1995. Multivariate Data Analysis. 4th edition ed. New York:
Macmillan.
Hairston, D. E., P. DeLise and W. J. Skerpan. 1997. Delivered into providence. Civil
Engineering (New York), 67 (2): 54-56.
190
BIBLIOGRAPHY
Halpin, D. W. 1985. Financial and Cost Concepts for Construction Management. New
York, N.Y.: John Wiley and Sons.
Harman, H. H. 1967. Modern Factor Analysis. 2nd ed. Chicago: University of Chicago
Press.
Hartgen, D. T., V. B. Bowman and M. W. Horner. 1997. Baseline market forecasts for state
highway construction engineering, maintenance and administration programs 1997-2001:
Transportation Publication Number 151. Charlotte, N.C: Centre for Interdisciplinary
Transportation Studies, University of North Carolina.
Hartman, F., P. Snelgrove and R. Ashra. 1997. Effective wording to improve risk allocation
in lump sum contracts. Journal of Construction Engineering and Management, 123 (4): 379-
87.
Hartman, F. T. 2000. Don't Park Your Brain Outside-a practical guide to improving
shareholder value with SMART management: NC:PMI Publications.
Hatush, Z. and M. Skitmore. 1998. Contractor selection using multicriteria utility theory: an
additive model. Building and Environment Journal, 33 (105-115).
Hegazy, T. and A. Ayed. 1998. Neural network model for parametric cost estimation of
highway projects. Journal of Construction Engineering and Management, May/June 1998:
210-18.
Hegazy, T. and O. Moselhi. 1995. Elements of cost estimation: a survey in Canada and the
United States. Cost Engineering, 37 (5): 27-33.
Helberg, C. 1996. Pitfalls of data analysis. Practical Assessment, Research and Evaluation, 5
(5). http://PAREonline.net/getvn.asp?v=5&n=5.
Hertz, D. B. 1964. Risk analysis in capital investment. Harvard Business review, 42 (1): 95-
106.
191
Risk factors leading to cost overrun in the delivery of highway construction projects.
Hester, W. T., J. A. Kuprenas and T. C. Chang. 1991. Construction Changes and Change
Orders: Their Magnitude and Impacts. Berkeley: University of California.
Hetland, P. W. 1994. Towards ultimate control of megaprojects in the North Sea. In Global
Project Management Handbook, eds. D. I. Cleland and R. Gareis. New York: McGraw-Hill
Book Co. Ltd.
Hibbert, P., D. Merrifield and A. Taylor. 1990. Key Factors On Contractual Relationships.
London: RICS.
Hillson, D. 1999. Developing effective risk responses. In PMI Annual Seminar and
Symposium. Philadelphia: Project Management Institute.
Hillson, D. 2002. Extending the risk process to manage opportunities. International Journal
of Project Management, 20 (3): 225-40.
Hinze, J. and K. Walsh. 1997. Shoring as a bid item. Managing Engineered Construction in
Expanding Global Markets; ASCE Construction Congress Proceedings: 226-32.
HM Treasury. 1993. CUP Guidance: No. 41 - Managing risk and contingency for works
projects. London: HM Treasury Central Unit on Procurement.
192
BIBLIOGRAPHY
Ibbs, W. C. and D. B. Ashley. 1987. Impact of various construction contract clauses. Journal
of Construction Engineering and Management, 113 (3): 501-21.
ICE. 1998. Institution of Civil Engineers: Risk analysis and management for projects
(RAMP). London: Thomas Telford.
In't veld, J. and J. Peeters. 1989. Keeping large projects under control: the importance of
contract type selection. Journal of Project Management, 7 (3): 155-62.
Iseley, T. and S. Gokhale. 2003. Equipment Productivity. South Carolina: CRC Press LLC.
Jaafari, A. and A. Schub. 1990. Surviving failure: Lessons from field study. Journal of
Construction and Engineering Management, 116 (1): 68-86.
Jones, D. S. 1996. Philosophies of risk allocation - the case for foreseability. International
Construction Law Review, 13 (4): 570-85.
Kaiser, H. F. 1958. The varimax criterion for analysis rotation in factor analysis.
Psychometrika, 23: 187-200.
193
Risk factors leading to cost overrun in the delivery of highway construction projects.
Kaplan, S. 1991. The General Theory of Quantitative Risk Assessment. In Proceeding of the
Fifth Conference - Risk-Based Decision Making in Water Resources. Reston, VA: ASCE.
Kaplan, S. and B. J. Garrick. 1984. On the quantitative definition of risk. Risk Analysis., 1:
11-28.
Kayode, M. O. 1979. The Art of Project Evaluation. Ibadan, Nigeria: University Press.
Keeney, R., L and D. von Winterfeldt. 1991. Eliciting probabilities from experts in complex
technical problems. IEEE Transactions in Engineering Management, 38 (3): 191-201.
Khan, Z. 1998. Risk premiums associated with exculpatory clauses. Masters, University of
Calgary, Canada, Alberta.
Khosrowshahi, F. 1999. Neural network model for contractors' pre-qualification for local
authority projects. Journal of Engineering, Construction and Architectural Management, 6
(3): 315-28.
Kirszner, L. G. and S. R. Mandell. 1992. The Holt Handbook. 3rd Ed. ed. New York.:
Harcourt Brace Jovanovich,.
Koskinen, K. U., P. Pihlanto. and H. Vanharanta. 2002. Tacit knowledge acquisition and
sharing in a project work context. International Journal of Project Management, 21: 281-
290.
Kozek, J. and C. Hebberd. 1998. Contracts: share the risk. Journal of Construction
Engineering and Management, 111 (2): 356-61.
Kumamoto, H. and E. J. Henley. 1996. Probabilistic Risk Assessment and Management for
Engineers and Scientists. 2nd Edition ed. New York, NY: IEEE Press.
194
BIBLIOGRAPHY
Lam, K. C., T. Ng, M. Skitmore and O. Cheung. 2002. Decision support system for
contractor pre-qualification - artificial neural network model. Journal of Engineering,
Construction and Architectural Management, 7 (3): 251-66.
Lansdown, J. 1982. Expert Systems: Their Impact on the Construction Industry. London:
RIBA.
Latham, S. M. 1994. Constructing the Team - Final Report of the Government / Industry
Review of Procurement and Contractual Arrangements in the UK Construction Industry.
London.
Law, A. and W. D. Kelton. 1991. Simulation Modeling and Analysis. 2nd ed. New York:
McGraw-Hill, Inc.
Leavitt, D., S. Ennis and M. P. 1993. The cost escalation of rail projects: Using previous
experience to re-evaluate the Calspeed estimates; Working Paper No. 567. Berkeley:
Institute of Urban and Regional Development, University of California.
Leu, S. S., A. T. Chen and C. H. Yang. 2001. A GA-based optimal model for construction
time-cost trade-off. International Journal of Project Management, 19: 47-58.
Levine, H. 1995. Risk management for dummies: managing schedule cost and technical risks
and contingencies. PM Network, October: 30-32.
Linstone, H. A. and M. Turoff. 1975. The Delphi Method Techniques and Applications.
Reading, MA.: Addison-Wesley.
Loftus, J. 1999. Project Management of Multiple Projects and Contracts. London: Thomas
Telford Ltd.
195
Risk factors leading to cost overrun in the delivery of highway construction projects.
Macneil, I. R. 1980. The New Social Contract: An Inquiry into Modern Contractual
Relations. New Haven, NJ.: Yale University Press.
Mak, S. and D. Picken. 2000. Using risk analysis to determine construction project
contingencies. Journal of Construction Engineering and Management, 126 (2): 130-36.
Mak, S., J. Wong and D. Picken. 1998. The effect of contingency allowances of using risk
analysis in capital cost estimating: a Hong Kong case study. Journal of Construction
Management and Economics, 16: 615-19.
Marco, D. 2000. Building and Managing the Meta Data Repository: The Full Lifecycle
Guide. New York: John Wiley.
Mason, R. L., R. F. Gunst and J. l. Hess. 1989. Statistical Design and Analysis of
Experiments - With Application to Engineering and Science. New York: John Wiley and
Sons, Inc.
Masterman, J. W. E. 1994. A study of the basis upon which clients in the construction
industry choose their building procurement systems. PhD, Institute of Science and
Technology, University of Manchester, Manchester, UK.
McCuen, R. H. 1996. The Elements of Academic Research. New York: ASCE Press.
McDermott, P. ed. 1999. Strategic and emergent issues in construction procurement. Edited
by S. Rowlinson, E. McDermott and F. N. Spoon, CIB W92 "Procurement Systems: A Guide
to Best Practice in Construction". London: E and F.N. Spon,.
196
BIBLIOGRAPHY
McKim, R. A. 1992. Risk Management Back to Basics. Cost Engineering, Association for
the Advancement of Cost Engineering, 34(12).
McLennan, A. and P. Jorss. 2006. Getting it right the first time: A plan to reverse the
declining standards in project design documentation within the building and construction
industry. In Clients Driving Innovation: Moving Ideas into Practice: Second International
Conference 2006. Gold Coast, Queensland, Australia: Cooperative Research Centre for
Construction Innovation.
Minato, T. and D. B. Ashley. 1998. Data driven analysis of "corporate risk" using historical
cost data. Journal of Construction Engineering and Management, 124 (1): 42-47.
Molenaar, K. R. 2005. Programmatic cost risk analysis for highway megaprojects. Journal of
Construction Engineering and Management, 131 (3): 343-53 (accessed March 1, 2005).
Morgan, D. L. 1997. Focus groups as qualitative research. Second ed. London: Sage
Publications.
Morris, P. W. G. 1988. Managing project interfaces - Key points for project success. In
Project Management Handbook, eds. D. I. Cleland and W. R. King, 06-55. New York: Van
Nostrand Reinhold.
Morris, P. W. G. and G. Hough. 1987. The Anatomy of Major Projects. New York: John
Wiley.
Moselhi, O. 1997. Risk assessment and contingency estimating. In AACE Transactions 13-
16 July Conference. Dallas: AACE.
Musgrave, R. A. and P. B. Musgrave. 1984. Public Finance in theory and practice. New
York: McGraw-Hill.
NAS. 2003. Completing the Big Dig: Managing the final stages of Boston's Central
Artery/Tunnel Project. In Report to the National Academy of Sciences. Washington D.C: The
National Academies Press.
197
Risk factors leading to cost overrun in the delivery of highway construction projects.
Neter, J., W. Wasserman and M. Kutner. 1990. Applied Linear Statistical Models. Third Ed.
ed. Homewood, Boston: Richard Irwin Inc.
Newton, S. and B. S. Logan. 1988. Causation and its effect: the blackguard in CAD's
clothing. Design Studies, 9 (4): 196-201.
Ng, S. T. 1996. Case-based reasoning decision support for contractor prequalification. PhD
Thesis, University of Manchester of Institute of Science and Technology, Manchester,
England.
Nijkamp, P. and B. Ubbels. 1999. How reliable are estimates of infrastructure costs: a
comparative analysis? International Journal of Transport Economics, 26 (1): 23-53.
Nilsen, T. and T. Aven. 2003. Models and model uncertainty in the context of risk analysis.
Reliability Engineering & System Safety, 79 (3): 309-17.
Niwa, K. and M. Okuma. 1982. Know-how transfer method and its application to risk
management for large construction projects. IEEE Transactions on Engineering
Management, 29/4 (4): 146-53.
Norusis, M. J. 2005. SPSS 13.0 Statistical Procedures Companion. Chicago: SPSS, Inc.
Oakes, M. 1986. Statistical Inference: A Commentary for the Social and Behavioural
Sciences. Chichester, UK: Wiley.
Oberlender, G., D and S. Trost, Michael. 2001. Predicting accuracy of early cost estimates
based on estimate quality. Journal of Construction Engineering and Management
(May/June): 173-82.
OGC. 2002. Managing Successful Projects With PRINCE 2. 3rd ed. London: The Stationary
Office.
O'Reilly, M. 1995. Risk, construction contracts and construction disputes. Construction Law
Journal, 11 (5): 343-54.
198
BIBLIOGRAPHY
Paek, J. H., Y. W. Lee and J. H. Ock. 1993. Pricing Construction Risk: Fuzzy Set
Application. Journal of Construction Engineering and Management, ASCE, 117(3).
Panagiotokopoulos, D. 1977. Cost-time model for large CPM project networks. Journal of
Construction Engineering and Management, ASCE, 103 (C02): 201-11.
Paul, A. and G. Gutierrez. 2005. Simple probability models for project contracting.
European Journal of Operational Research, 165 (2): 329-38.
Perry, J. G. and R. W. Hayes. 1985. Risk and its management in construction projects.
Proceedings of the Institute of Civil Engineering, 78 ( 1): 499-521.
Peurifoy, R. and G. Oberlender. 1989. Estimating Construction Costs. 4th ed. New York:
McGraw-Hill Book Co. Inc.
Potter, S. 1987. On the right lines? The limits of technological innovation. London.: Frances
Pinter.
199
Risk factors leading to cost overrun in the delivery of highway construction projects.
Queensland Department of Main Roads. ed. 2000. Main Roads' Project Delivery System -
Major Works Pre-qualification System. Vol. Three. Brisbane: Queensland Department of
Main Roads.
Raftery, J. and S. T. Ng. 1993. Knowledge Based Approaches to Construction Risk Analysis.
In International Symposium on Electronic Evaluation and the Built Environment, CIB W55
and W95, 152-65. Lisbon, Portugal.
Reutlinger, S. 1970. Techniques for project appraisal under uncertainty. In World Bank Staff
Occasional Papers. Baltimore, Md: Johns Hopkins University.
RIBA. 1984. Handbook of Architectural Practice and Management. 4th ed. London: The
Royal Institute of British Architects.
Robinson, R. B., C. D. Cox and K. Odom. 2005. Identifying outliers in correlated water
quality data. Journal of Environmental Engineering, 131 (4): 651-57 (accessed April 2005).
Rosenberg, L. H., T. Hammer and A. Gallo. 1999. Continuous risk management at NASA.
http://satc.gsfc.nasa.gov/support/ASM-FEB99/crm-at-nasa.html.
200
BIBLIOGRAPHY
Rutgers, A. and D. H. Haley. 1997. Project risk and risk allocation. Cost Engineering, 39 (9):
27-9.
Sackman, H. 1972. Delphi Critique: Expert Opinion, Forecasting and Group Process.
Lexington, MA: Lexington Books.
Salapatas, J. N. and W. S. Sawte. 1986. Measuring success of utility projects past, present
and future. In Proceedings of the 18th Annual Seminar/symposium of the Project
Management Institute. Montreal,.
Schexnayder, C. and L. G. Ohrn. 1997. Highway specifications quality versus pay. Journal
of Construction Engineering and Management, 123 (4,): 437-43.
Shen, L. Y., W. lu, Q. Shen and H. Li. 2003. A computer-aided decision support system for
assessing a contractor's competitiveness. Automation in Construction, 12: 577-87.
Shepherd, B. 1997. Managing Risk. In The 29th Annual Project Management Institute
Seminars and Symposium, 79-87. Chicago: Project Management Institute.
Simon, P., D. Hillson and K. Newland. 1997. PRAM-Project risk analysis and management
guide. High Wycombe, UK: Association for Project Management.
Slovic, P., B. Fischhoff and S. Lichenstein. 1980. Facts and fears: understanding perceived
risk. In Societal Risk Assessment: How safe is enough?, eds. R. C. Schwing and W. A.
Albers (eds.), 181-214. New York: Plenum Press.
201
Risk factors leading to cost overrun in the delivery of highway construction projects.
Smith, M. A. ed. 1996. Construction Insurance, Bonding, and Risk Management. Edited by
W.J Palmer, J. M. Maloney and J. l. Heffron. New York: McGraw-Hill.
Smith, R. J. 1995b. Risk identification and allocation: saving money by improving contracts
and contracting practices. International Construction Law Review, 12 (1): 40-71.
Soanes, C. and A. Stevenson. eds. 2003. Oxford Dictionary of English. Oxford: Oxford
University Press.
Standards Australia. 1999. AS/NZS 4360: Risk Management. Strathfield NSW: Standards
Association of Australia.
Staugus, J. 1995. Variations. Journal of Building and Construction Law, 11 (3): 156-58.
Stevens, J. D. 1995. Cost estimating and forecasting for highway work in Kentucky:
Research Report KTC 95-12. Lexington, Ky.: Kentucky Transportation Centre, University of
Kentucky.
Summers, R. 1967. Cost estimates as predictors of actual costs: A statistical study of military
developments.. In Strategy for R&D: Studies in the microeconomics of development, eds. T.
Marschak, T. K. Glennan and R. Summers, 14089. Berlin: Springer-Verlag.
Sykes, A. 1982. Reducing neglected risks on giant projects. In New Dimensions of Project
Management, ed. A. J. Kelley(ed). Lexington, MA: Lexington Books.
202
BIBLIOGRAPHY
Szyliowicz , J. S. and A. R. Goetz. 1995. Getting realistic about megaproject planning. The
case of the Denver International Airport. Policy Sciences, 28: 347-67.
Tah, J. H. M. and V. Carr. 2001. Towards a framework for project risk knowledge
management in the construction supply chain. Advances in Engineering Software, 32: 835-
46.
Taha, M. A., S. C. Park and J. S. Russell. 1995. Knowledge based DSS for construction
contractor pre-screening. European Journal of Operational Research, 84: 35-46.
Tilley, P. A., S. L. McFallen and S. N. Tucker. 2000. Design and documentation quality and
its impact on the contract process. Brisbane: CSIRO.
Trost, S., Michael and G. Oberlender, D. 2003. Predicting accuracy of early cost estimates
using factor analysis and multivariate regression. Journal of Construction and Engineering
Management, March/April: 198-204.
203
Risk factors leading to cost overrun in the delivery of highway construction projects.
Tuohey, P. 2002. The use of risk management processes in construction contracts. In The
Contract in Successful Project Management Conference. Christchurch, NZ: Worley Group.
Tversky, A. and D. Kahneman. 1981. The framing of decisions and the psychology of
choice. Science, 211: 453-58.
Uff, J. and A. Odams. 1995. Contract Documents and the Division of Risk. In Risk
Management and Procurement Practices, eds. J. Uff and A. Odams. Kings College, London:
Centre for Construction Law and Management.
Uher, T. 1994. A general risk classification model for construction. Australian Institute of
Building Papers, 5: 25-33.
Under water, o. b. 1989. Under water, over budget. The Economist, October: 37-8.
Velicer, W. F. and D. N. Jackson. 1990. Component analysis versus common factor analysis:
Some issues in selecting an appropriate procedure. Multivariate Behavioral Research, 25
(1): 1-28.
Vojir, C.P., K.R. Jones, R. Fink. And E. Hutt. 2006. Scientific Enquiry: What a difference a
point makes. Journal of Specialists in Pediatratic Nursing, 11 (4) 257-59.
Walker, D. H. T., K. Hampson and R. Peters. 2000. Project alliancing and project partnering
- what's the difference? - Partner selection on the Australian National Museum project: a
case study. In Proceedings of the CIB W92 Conference, 641-55. Chile.
Wallace, I. N. D. 1995. Hudson's Building and Engineering Contracts. 11th edition ed.
London: Sweet & Maxwell.
Wang, J. X. and M. L. Roush. 2000. What Every Engineer Should Know About Risk
Engineering and Risk Management. New York, NY.: Marcel Dekker.
Wang, M.-T. and H.-Y. Chou. 2003. Risk allocation and risk handling of highway projects in
Taiwan. Journal of Management in Engineering, 19 (2): 60-8.
204
BIBLIOGRAPHY
Wang, W.-C. 2003. SIM-UTILITY: Model for project ceiling price determination. Journal
of Construction and Engineering Management, 1218 (1): 76-84.
Ward, S. 1999. Requirements for an Effective Project Risk Management Process. Project
management Journal, 30(3).
Ward, S. and C. Chapman. 2003. Transforming project risk management into project
uncertainty management. International Journal of Project Management, 21 (2): 97-105.
White, D. and J. Fortune. 2002. Current practice in project management: An imperial study.
International Journal of Project Management, 25 (4).
Williams(b), T. M. 1993. Using the risk register to integrate risk management in project
definition. International Journal of Project Management, 12: 17-22.
Williams, T. M. 1999. The need for new paradigms for complex projects. International
Journal of Project Management, 17 (5): 269-73.
Williams, T. M., F. R. Akermann and C. L. Eden. eds. 1997. Project risk: systemicity, cause
mapping and a scenario approach. Edited by K. Kahkonen and K. A. Arto, Managing Risks
in Projects. London: E & F.N. Spoon.
Williams, T. P. 2003. Predicting final cost for competitively bid construction projects using
regression models. International Journal of Project Management, 21: 593-99.
Wilmot, C. G. and C. Cheng. 2003. Estimating future highway construction costs. Journal of
Construction Engineering and Management, May/June: 272-79 (accessed May/June 2003).
Winch, G. 1989. The construction firm and the construction project: a transaction cost
approach. Construction Management and Economics,. 7: 331-45.
205
Risk factors leading to cost overrun in the delivery of highway construction projects.
World Bank. 1994. World Development Report 1994: Infrastructure for development.
Oxford, UK: Oxford University Press.
Yin, R. K. 2003. Case Study Research: Design and Methods, 3rd Ed. 3rd ed. ed. Thousand
Oaks, Calif.: Sage Publications.
Zeitoun, A. A. and D. O. Oberlender. eds. 1993. Early Warning Signs of Project Changes:
Document 91. Edited by C. I. Institute. Austin, Texas: Construction Industry Institute.
206
APPENDIX A
APPENDIX A
Road Input Cost Indices for the period 1995-96 to 2002-03
207
Risk factors leading to cost overrun in the delivery of highway construction projects.
208
APPENDIX A
209
Risk factors leading to cost overrun in the delivery of highway construction projects.
400.00
Financial year average 100.00 1.66
September 1998 101.50 0.79
December 1998 101.98 0.47
March 1999 102.41 0.42
June 1999 102.69 0.27
408.58
Financial year average 102.15 2.15
September 1999 103.50 0.79
December 1999 104.05 0.53
March 2000 104.56 0.49
June 2000 105.05 0.47
417.16
Financial year average 104.29 2.10
September 2000 105.36 0.30
December 2000 106.41 1.00
March 2001 107.04 0.59
June 2001 107.60 0.52
426.41
Financial year average 106.60 2.22
September 2001 108.54 0.87
December 2001 109.10 0.52
March 2002 109.78 0.62
June 2002 110.54 0.69
437.96
Financial year average 109.49 2.71
September 2002 111.85 1.19
December 2002 112.87 0.91
March 2003 113.80 0.82
June 2003 114.38 0.51
452.90
Financial year average 113.23 3.41
September 2003 115.21 0.73
December 2003 116.10 0.77
March 2004 116.79 0.59
June 2004 118.18 1.19
466.28
Financial year average 116.57 2.95
210
APPENDIX B
APPENDIX B
Programmed and actual project costs indexed to year 2003
211
Risk factors leading to cost overrun in the delivery of highway construction projects.
Project number Year project Indexed prog cost Indexed actual cost
completed $m $m
12/46C/803 1996 2.152 2.849
107/852/804 1996 1.396 1.997
80/140/13 1996 2.908 4.380
160/11A/2 1996 7.443 10.395
25/126/22 1996 1.251 1.605
100/851/26 1996 5.338 6.891
24/162/13 1996 2.239 2.744
122/31A/804 1996 2.326 2.873
30/8202/803 1996 2.291 2.777
146/166/15 1996 1.215 1.700
113/35A/807 1996 1.977 2.710
94/36b/19 1996 2.093 2.353
128/48/15 1996 3.256 3.838
66/814/5 1997 2.059 3.337
140/U16/819 & 820 1997 1.803 3.131
158/10P/811 1997 2.300 2.574
22/18E/807 1997 1.854 2.293
25/10A/802 1997 3.565 4.879
130/45B/14 1997 1.550 1.880
5/10K/19 1997 1.495 1.691
150/10L/38 & 39 1997 9.051 10.567
160/103/2 1997 5.290 7.418
160/12B/5 1997 42.550 47.909
24/10C/22 & 64/10C/3 1997 1.351 1.692
25/126/23 1997 1.283 2.214
158/20A/2 1997 2.990 4.156
70/25A/11 1997 22.749 29.937
92/133/11 1997 2.415 2.887
45/655/18 1997 1.898 3.908
158/647/15 1997 3.283 4.016
37/90B/43 1997 2.482 2.828
30/8202/303 1997 2.645 3.148
148/17B/20 1997 2.530 3.146
25/126/24 1997 1.196 1.680
S140/U12B/28 1997 1.021 2.933
140/U18B/61 1997 1.567 1.796
134/14C/301 1997 4.165 6.138
23/78A/17 1998 16.980 18.678
70/108/300 1998 1.301 2.177
36/15A/301 &302 1998 6.905 7.647
25/1204/301 1998 2.151 2.390
89/41A/17 1998 2.830 3.113
24/1632/12 1998 1.366 1.519
40/22A/28 1998 3.034 3.588
150/10L/38 & 1998 8.857 11.030
39;150/10M/26
160/102/3 1998 4.528 6.012
212
APPENDIX B
213
Risk factors leading to cost overrun in the delivery of highway construction projects.
214
APPENDIX B
215
Risk factors leading to cost overrun in the delivery of highway construction projects.
216
APPENDIX C
APPENDIX C
Questionnaire format to establish NGT membership
217
Risk factors leading to cost overrun in the delivery of highway construction projects.
Questionnaire:
Date of Completion: ..
Dear interviewee,
The aim of this questionnaire is to elicit expert information on the process of preparing total project
cost estimates at the time of design, ideally in road construction.
For the purpose of this questionnaire and research, the total project cost estimate includes the
estimated costs of all component activities from the initiation of the project proposal to finalisation.
This total project cost estimate is usually prepared at the design stage of a project and can often
include the costs of the following:
conducting investigations and developing altering public utility plant,
the design, construction,
detailing the design, project administration, and handover
acquiring land,
The information supplied in this completed questionnaire will be used for broad research purposes
only. All specific company and interviewee information will be kept confidential at all times.
Only generalised analysis of the information contained within this completed questionnaire will be
utilised in the research process.
Regards,
Garry Creedy
7 December 2004
218
APPENDIX C
(This section refers to the whole of the project estimate not just the construction estimate.)
Section 3: Total Project Estimating
3.1 In your opinion, what are the four most important factors/issues in project estimating?
3.2 In your experience, what role does value engineering and/or value management play in total
project cost estimating and also in design?
3.3 What are the estimating methods most often used in your organisation for total project
estimating at the design stage?
e.g. parametric
factored
basic cost
unit rate
Explain.
3.5 Describe the typical project scope definition information available to you for project design
estimating.
3.6 What approach do you use for contingency analysis in project design estimates?
Probability based
Historical values
Decision analysis
Experience based
Other (explain)
3.7 List items at the project level that you normally include as contingencies?
e.g. - Design allowances
- Force majeure
-
-
3.8 List the main factors that you take into account for contingency analysis and
evaluation ranked according to their importance.
e.g. Nature of project
Level of design completeness
219
Risk factors leading to cost overrun in the delivery of highway construction projects.
3.10 Please list four of the most frequent caused errors, mistakes, omissions in project design
estimating.
3.11 What are four main problems or constraints encountered in project design estimates?
3.12 What is your percent confidence in the expected accuracy level of project design estimates
for the following?
Roads
Bridges
Miscellaneous (road related)
3.13 Do you have any comments on how to improve total project estimates and how to achieve
this?
Other (specify)
220
APPENDIX C
Section 5: Other:
5.1 Would you like to participate in a personal interview to expand your answers or provide
more
nformation about total project design estimating?
Yes
No
Garry Creedy
Queensland University of Technology
School of Construction Management and Property,
Level 4, L Block, Gardens Point Campus
GPO Box 2434,
Brisbane 4001.
or
Fax 38641170
221
Risk factors leading to cost overrun in the delivery of highway construction projects.
222
APPENDIX D
APPENDIX D
Completed Questionnaire details of NGT membership
223
Risk factors leading to cost overrun in the delivery of highway construction projects.
Questionnaire:
Dear interviewer,
The aim of this questionnaire is to elicit expert information on the process of preparing total
project cost estimates at the time of design, ideally in road construction.
For the purpose of this questionnaire and research, the total project cost estimate includes the
estimated costs of all component activities from the initiation of the project proposal to
finalisation. This total project cost estimate is usually prepared at the design stage of a
project and can often include the costs of the following:
The information supplied in this completed questionnaire will be used for broad research
purposes only.
All specific company and interviewee information will be kept confidential at all times.
Only generalised analysis of the information contained within this completed questionnaire
will be utilised in the research process.
Regards,
Garry Creedy
7 December 2004
224
APPENDIX D
Estimating (years)
2. 10
3. 10
4. * embodied in construction and maintenance
5. 14 (Continuous, ongoing role)
7. 5
Design (years)
2. 17
3. 10
4. * embodied in construction and maintenance
5. 1
6. 12
7. 5
225
Risk factors leading to cost overrun in the delivery of highway construction projects.
Construction (years)
1. 10
2. 3
3. 10
4. 10*
6. 2
7. 15
Management (years)
2. 16
3. 15
4. 29*
5. 6
6. 3
7. 15
(This section refers to the whole of the project estimate not just the construction estimate.)
Section 3: Total Project Estimating
3.1 In your opinion, what are the four most important factors/issues in project estimating?
2. Detailed geotechnical investigation. Competent design. Detailed quantity
calculations. database for tender bids on similar projects over >5 years.
3. Knowing and understanding project scope. Historical costs.
Identification and management of risks.
4. Knowledge of construction processes/techniques/materials etc.
Understanding the real needs of the project. Clients expectations. Risk
assessment.
5. Knowing the scope of the project. Suitable allowances for risks and
contingencies. Breaking the project down into significant detail to
estimate cost of works; i.e. not using $/km. Using correct items and
quantities in breakdown.
6. Understanding the design and specifications, and knowing what work
the designers intend to be included in each of the items in the schedule.
Having sufficient contingency where the nature of the work is not well
understood. Experience not only in estimating but in all phases of the
project eg. planning, design, and construction. Local knowledge of
materials, conditions, prices.
7. Completeness of scope. Ensuring estimate is in out-turn $s. Adequate
time allowed for estimating costs. Experienced estimators.
3.2 In your experience, what role does value engineering and/or value management play in total
project cost estimating and also in design?
2. Clarify scope. Set design performance standards.
3. Very important.
4. Identifying likely issues/problems and bringing together a broadly based
view on likely solutions; i.e. facilitating an understanding across the
supply chain.
5. Useful in ensuring scope is adequately understood and defined.
7. Allows completeness of adequate design and completeness of scope.
3.3 What are the estimating methods most often used in your organisation for total project
estimating at the design stage?
e.g. parametric 7.
factored
226
APPENDIX D
Explain.
4. Contracts are basically schedule of rates type. Separation of the
organisation into purchaser/provider has resulted in less ability of the
client to do basic cost estimating.
5. Mainly unit rate. Some major items are calculated using basic cost.
Many staff have insufficient experience to effectively use basic cost
estimates.
7. Bridges initial estimate using parametric method. design estimate for
road and bridge projects use unit rate and bill of quantities.
3.5 Describe the typical project scope definition information available to you for project design
estimating.
2. Bridge fixings width, length, height of bridge.
3. Extent of physical project scope. Survey data. Planning layouts.
4. DMR is introducing a formalised project management approach with a
tight scope definition profile.
5. Extremely variable. Often fairly limited at concept phase, usually fairly
well defined by detailed design.
7. Services. resumption requirements. Planning studies. Native title
clearances. Consultation outcomes.
3.6 What approach do you use for contingency analysis in project design estimates?
Probability based 4.
Historical values 2., 7.
Decision analysis 3.
Experience based 2., 3.
Other (explain)
5. Risk identification. Risk register. Assessment. Management plan.
Re-assessment. Likelihood. Potential cost. Contingency pool.
3.7 List items at the project level that you normally include as contingencies?
e.g. - Design allowances
- Force majeure
1. Resumptions. Services Relocation. Administration Costs.
227
Risk factors leading to cost overrun in the delivery of highway construction projects.
3.8 List the main factors that you take into account for contingency analysis and evaluation
ranked according to their importance.
e.g. Nature of project
Level of design completeness
2. Nature of the project
3. Impact of risk to project scope, duration, cost (in $ terms, not as a
percentage). Probability of being able to mitigate the risks.
5. Identified risks are rated according to: a) the likelihood of the risk
occurring, b) the effect to the project id the event (risk) occurs. This
gives us a ranking. Risk management is considered for all extreme, high
and medium risks.
7. Type of delivery contract. Constructability.
3.10 Please list four of the most frequent caused errors, mistakes, omissions in project design
estimating.
1. Amount of unsuitable material. Pipe length specified. Concrete quantities. No
allocation for rideability bonus or rise in the cost of bitumen.
2. Errors in quantities. Misinterpretation of borelogs/foundation reports. Lack of
data on similar projects eg remote location projects. Lack of market suppliers
one bidder, so no competition.
3. Overlooking whole items of cost eg. resumptions, public utilities, removal of
existing bridge of pavement after new installation. poor and conflicting scheme
documentation (plans-V-specifications). Scope increase. Using inappropriate
historical data.
4. Quantity errors. Constructability issues. inconsistent specifications.
5. Final scope of project is not what was originally estimated. Very imprecise
global estimating used at concept phase and this determines project budget.
Inadequate or no contingency allowance. Higher than expected cost escalation.
228
APPENDIX D
6. Use of historical rates not appropriate to the scale of the works. Inexperience.
Time pressure to fast track projects can lead to design errors.
7. Quantity errors. Inexperienced estimators. Poor project scope definition. Plan
and document errors in contracts.
3.11 What are four main problems or constraints encountered in project design estimates?
2. Estimates made several years ago and not updated. Rapidly changing market
prices when construction program increases rapidly.
3. Lack of input from construction experts. Pressure of time. Misuse of expert
systems.
4. Quantity errors. Constructability issues. inconsistent specifications.
5. Political pressure to produce estimates in unrealistic short timeframes. Political
influences and lobby groups resulting in change in project scope.
6. Insufficient geotechnical investigation/lack of understanding of ground
conditions. Lump sum items for traffic management and environmental
management different understanding of what is required. Need to break these
down into sub-items. In a hot market, prices can rise rapidly due to demand
exceeding supply. Design estimate may be out of date by the time the job is
tendered.
7. Time allowed. Contracting climate at time of tender v- time of estimate. Time
delay for approvals to project elements. Inexperienced estimators in road and
bridge projects.
3.12 What is your percent confidence in the expected accuracy level of project design estimates
for the following?
Roads
1. 75%
3. Within +20% and -5%.
4. Within +- 10% for design estimates
5. +- 10% at detailed design.
7. 85%
Bridges
1. 90%
2. Coefficient of variation 10% on standard projects.
3. Within +15% and -5%.
4. Within +- 10% for design estimates
5. +- 10% at detailed design
7. 95%
Miscellaneous (road related)
3. within +30% and -5%.
7. 85%
3.13 Do you have any comments on how to improve total project estimates and how to achieve
this?
1. Use a quantity surveyor who specialises in Road/Bridge estimates.
2. Adequate geotechnical info. Competent design. Accurate schedule of quantities.
3. Knowing and understanding project scope. Historical costs. Identification and
management of risks.
4. Better risk assessment. Better documentation i.e. plans, specifications, etc. Use
of basic cost estimating by client. Adequate time allowed for design. Clear scope
definitions.
5. Ensure project scope is adequately defined. Allow sufficient time and resources
to estimate the cost of the project.
229
Risk factors leading to cost overrun in the delivery of highway construction projects.
230
APPENDIX D
Alliance
2. 5%
3. 2%
4. 1%
5. 5%
7. 1%
Other (specify)
Section 5: Other:
5.1 Would you like to participate in a personal interview to expand your answers or provide
more
information about total project design estimating?
Yes
No
2. The issue in MRD appears to revolve around our inability to develop a program
over several years, and accept that initial project concept estimates can vary
widely from final project cost, and that the estimate should be refined through
the following stages: Concept Estimate; Preliminary Design Estimate; Final
Design & Documentation Estimate; Contractors Tender Price and
Contingencies Estimate. Being realistic about the project development and cost
development phases, and adjusting the RIP to fit, would solve many problems.
4. I think you need to send the questionnaire to hands on people. I am too far and
too long removed from the detail to effectively answer the questions accurately.
231
Risk factors leading to cost overrun in the delivery of highway construction projects.
232
APPENDIX E
APPENDIX E
Client risks in highway project delivery
233
Risk factors leading to cost overrun in the delivery of highway construction projects.
The following client project risks which have prevailed in historical highway projects and
caused significant cost overruns.
Detailed descriptions of those risks are included here so that project estimators and project
managers can be made more aware of their potential to impact costs in highway project
budgets.
Poorly coordinated designs, plans and specifications can lead to contractors claims during
construction because of conflicting project information that, when clarified, causes the
contractor to perform differently than was planned at the time of preparing the bid.
When multiple contractors are planned to be working simultaneously on a project site, the
designer should develop a site utilisation plan that will define the area(s) each contractor has
jurisdiction over and how to accommodate the other contractor(s). If this has not been done
properly, there will be contractor claims usually based on delays that will result in costs to
the owner as well as potential time delays in achieving project completion.
At the project estimating stage, studies are often required for significant items of work to
determine the feasibility and efficiency of alternative production methods. In estimating the
construction component of a project for example, the project manager may need to examine
the mass haul diagram for earthworks to evaluate haul distances, borrow and spoil
requirements and the most effective construction fleet for the particular site conditions.
Similarly, for major road projects in high traffic areas, it may be necessary to develop traffic
management and construction-staging plans in order to evaluate traffic management
requirements.
234
APPENDIX E
process continuity and efficiency, situations can arise in contracts where the contractors
obligations are compromised by client instructions and directions that require the contractor
to incur costs other than provided for in the contract provisions. Limiting a contractors work
schedule to out-of-peak traffic times as a result of unforseen peak hour traffic congestion is
one example where lower daily productivity and additional project costs might need to be
borne by the client. In urban areas, even say a traffic problem occurring on a nearby (arterial)
road can force traffic through construction projects which normally would not have attracted
higher flows, resulting in subsequent traffic congestion and safety issues through the project
site.
Side tracking of traffic is some urban and rural projects is also a safe and efficient means of
managing traffic during construction. Again, there are times when contingent emergencies or
processes can occur that will disrupt this orderly traffic management strategy that results in
construction processes having to be carried out under flowing traffic conditions. This may be
outside the control of the contractor and then could be deemed the cost responsibility of the
client.
This variation category includes such design scope changes as vertical and horizontal
alignment changes that can result in changes in quantities, as well as rework. It can include
scope changes in traffic guardrail, traffic signage and kerb and channel and traffic island
changes. Changes in this category are generally limited to one or a number of locations
within the project and would not generally include a major design change such as the
addition of further traffic lanes to the full length of the project.
235
Risk factors leading to cost overrun in the delivery of highway construction projects.
Drainage design scope changes can occur when an increase is required in the flow capacity
due to changes in hydraulic gradients due to conflicts with underground service installations
or other expensive obstructions. In some urban and semi-urban areas, changes in zoning in
upstream catchments can result in older drainage designs requiring redesign due to revised
runoff coefficients as a result of planned land use.
A small factor that sometimes requires design consideration is when new metric drainage
installations are required to join to very old imperial drainage pipes and cells. The design
modifications are small, but real and often overlooked by designers.
236
APPENDIX E
Pavement Depth: Changes in pavement depth occur when field measured strength of
pavement subgrades are weaker than the assumed strength that the pavement designs were
based on. Revised pavement designs usually result in increased pavement depths. In most
cases, this also requires the raising of the highway vertical grade line by the same amount as
that of the pavement. Minor regrading back down to original design grade levels at bridges
and new drainage culverts may be required if the pavement redesign extends over a
substantial distance.
EU Extras unspecified
This is a category that is intended to be a catch all for all cost increase factors that cannot be
categorised elsewhere due to inadequate project documentation or because of there being
large number of small project variations.
237
Risk factors leading to cost overrun in the delivery of highway construction projects.
Highway construction contracts generally contain unforseen, latent site condition clauses
that spell out the circumstances under which they can be deemed to be additional to the
contract, and thus incur additional client costs. Four discrete latent condition occurrences are
identified and described below:
238
APPENDIX E
1. LD Latent condition requiring design change: This includes the discovery of any
unidentified underground obstructions that require modifications to the design. This
can include modifications to new drainage designs and partial to full foundation
redesigns to structures due to unsuitable foundation conditions. This can include
increases or reduction in foundation levels that can lead to adjustments to the
physical dimensions of structures, etc.
2. LR Latent condition rock encountered: This risk factor is self explanatory and
can occur in earthworks operations in cuttings with the presence of rock bars which
may not have been identified by a site inspection or bore log. Rock is normally
defined as sound and solid mass and of such hardness that it cannot be loosened or
broken down by a certain defined size and capacity of excavation equipment under
the assessment of the client. Unrecorded rock outcrops can occur in excavations for
drainage structures, underground ducting etc. and such installations might not be
able to be raised or deviated to miss the obstruction.
3. LSG Latent condition additional stabilising: During the process of new
pavement construction or the strengthening an existing pavement with a structural
overlay of paving material, an assessment should be made of the strength of the
insitu subgrade or sub-layer material so that the design strength of the paving layer is
not exceeded during the pavement service life. In most cases, in-situ testing against
the design axle loading, on which the pavement depth is designed, is carried out.
During the construction process there is a risk that the exposed subgrade material
can test weaker than its design assumption. In such a case, the in-situ strength of
such material can be improved by the addition of cement or lime by in-situ
stabilisation of the subgrade material. This process has the advantage that the design
pavement depth can still be installed without the need to regrade the pavement
levels.
4. LUS Latent condition removal and replacement of unsuitable material: This
process is used to overcome situations where foundation materials for pavement,
drainage structures and structures are found to be inferior and the material is ordered
to be removed to a specified depth and replaced with material of a certain quality. In
some situations, the knowledge of the presence of unsuitable material can be
ascertained by bore probes or other testing methods and adequate provision made in
the schedule of quantities to cover such treatments. In cases where none or limited
testing is carried out in the design stage, then it is appropriate to include only a
nominal provision in the quantities to cover the risk of discovering and having to
replace unsuitable material. Usually, contract provisions allow for such variations as
a cost to the client.
239
Risk factors leading to cost overrun in the delivery of highway construction projects.
usually made in projects for the risk of price increases but instances can occur where
higher than predicted price increases can impact on the final cost of bitumen.
2. ME Material cost increases earthworks: Road designers generally endeavour to
obtain a balance between the volumes of excavated material as opposed to the
compacted volume of material required for fill for roadway embankments. Instances
can occur where the risk of not achieving an economic balance can lead to large
unplanned costs for either the disposal of excess material or the cost of importing
additional material onto the project. The major reason for such a net imbalance is that
certain excavated material may not be able to be incorporated into the project as fill
because of its inferior quality or because of its contamination from some site source
that was not known at the time of design. Where measured quantities of both
excavation and fill materials substantially differ from the scheduled quantities, then
there can be circumstances demonstrated when additional costs may need to be borne
by the project.
3. MP Material cost increase paving materials: Pavement design changes can
substantially increase paving material volumes which can lead to the risk of increased
pavement materials and hence increased costs. In addition, for certain project
locations, there may be no economically suitable source of paving materials and
hence the material then has to be transported large distances from established
quarries to the project. In some rural projects, there may be raw material suitable for
paving gravel production close to a rural project but no suitable crushing equipment.
It may then be a matter of establishing crushing facilities at the material source, if
suitable equipment can be enticed there. In this case, the volume of paving material
to be produced can be the important economic factor in determining the supply price
of the paving material. Either supply option can be significantly more expensive than
conventional quarries that are close to projects. This can have significant influence
on the final cost of a project.
240
APPENDIX E
completed using a dedicated work force on a cost-plus basis. The method of completing the
contract works will generally be determined by the client by taking into account factors such
as the amount of works outstanding, the potential for deterioration of the already provided
works and public and works safety provisions. A number of factors can lead to a contract
termination, however there are also procedures usually adopted to minimise the risk of such
an occurrence. These include:
pre-qualification of contractors that includes financial capability checks
tender evaluation procedures to check that submitted contract rates are commercially
realistic.
Provisions to protect the client in the event of a contract failure can include the provision of
an insurance bond on the value of the work and/or the retention of a certain percentage of
progress payments during the conduct of the contract. The additional costs to a project as a
result of a contract failure usually includes the need for a complete and certified measure up
of completed works and works in progress, the making safe of the works, ongoing traffic
control provisions where the public traverse the project, legal costs associated with, say,
bankruptcy of the contractor, re-documentation and calling of a new project contract. All
these activities will consume the time of a project manager and administration support and
can add up to a considerable unplanned project cost.
241
Risk factors leading to cost overrun in the delivery of highway construction projects.
All net increases in measured quantities can have a substantial effect on the final cost of a
project. Initial inaccurate quantity take-offs for schedule of rates contracts can mislead
contractors resulting in costly errors in bid pricing in other areas such as quality assurance
requirements.
R Resumption/accommodation works
Where private land is required on which all or part of the project is to be built, the land can
be purchased outright from the owner or can be compulsorily acquired under the powers of
the government. In the latter case, the land owner also has recourse to an appeal mechanism
that can take a considerable period of time. The outcome of this appeal mechanism can in
some circumstances increase the cost of acquiring the land for the project. In addition to
compensation for the taking part land parcels, the government is obliged to carry out
accommodation works adjacent to the resumed land so as not to affect the amenity of that
land. This can be costly and can increase project costs over that estimated at the time of
commencement of any land procurement.
Property acquisition costs can be impacted by a design that does not consider the effects that
the project has on adjoining properties. Poor or restricted access to residual and adjacent
properties can have a direct bearing on the property acquisition cost and on the potential
consequential damages to adjacent properties. Construction sequencing required to build a
project that causes multiple disruptions to the activities on abutting property can often result
in claims and/or litigation by the adjacent owners to recover damages to his operations.
Early and careful planning and coordination with the owners of the affected utilities is the
key to minimising the cost overrun of relocating services and maintaining services to
adjacent properties. Depending on the regulations and agreements that allow the utility
owners to be in their pre-project locations, the cost of relocation could be borne by the utility
owner or the project. Poor coordination of utility relocations (i.e. requiring multiple moves
of the same utility) will result in additional costs that usually have to be borne by the project.
It should be noted that there are circumstances where multiple moves cannot be avoided, and
when recognised by both parties, the cost is usually included in the project budget. It is the
unexpected multiple relocations, which could have been avoided, that have the impact on the
project budget. The project design team usually has a major responsibility to plan for the
242
APPENDIX E
coordination of the utility relocations with the utility owner. This can have a serious effect
on the project budget if it has not been coordinated adequately well in this area.
A utility activity that involves substantial management at the design stage is the locating of
existing utilities and is often shared with the utility owner that has been designated for an
area. Regardless of any shared responsibility, the project designers should ensure that they
are getting the best information available as there can be substantial cost saving when
designing a project with the intention of not requiring relocation of certain types of utilities.
When it is discovered during construction that a utility is in fact in conflict with the
construction, then the cost impact can be large since it often requires the contractor to shut
down activities in that area of conflict or develop some expensive work-around scheme to
avoid the shutdown. In this case, these costs come back to the client who may or may not
seek to have the designer share in paying for the cost. Another utility cost overrun area is the
lack of appreciation of long lead times necessary to procure special materials, equipment, as
well as system testing of the relocated utilities before they can be placed in service.
Telecommunication and power facilities are often candidates that require longer lead times.
When this long lead time requirement is not recognised or not appreciated then there is often
a schedule as well as a cost impact to the client..
SC Specification change
The quality of specified works and/or materials can be varied by the client during the
construction of a project, however this is not done as a matter of course. This can be done to
rectify an error in the specification or drawings. It can also be done to improve the specified
quality of certain material properties that may not perform adequately under anticipated
service conditions. For example, specifications can be changed to vary the specified finishes
on concrete surfaces from wood to steel float finish for durability. As well, finishing
construction tolerances for pavements can be changed to provide improved the completed
running surface finish than that documented. In most cases, where a specification change has
been requested by the client, then it is treated as a variation under the terms of the contract.
243
Risk factors leading to cost overrun in the delivery of highway construction projects.
parts of a project are saturated from rain. Generally contract conditions shift the risk of the
occurrence of wet weather on to the contractor and the cost of rework are deemed to be
included in the cost of particular construction processes that are contracted. The probability
of the occurrence of rain can generally be derived from the use of historic rainfall charts and
probability. A contractor can also insure against losses from the effects of rain. Where
intense rainfall periods can be predicted, contracts can be let such that the project works can
be reasonably completed outside the windows of anticipated rains. A prudent contractor
would plan to construct those portions of a road project vulnerable to the effects of
prolonged wet weather during periods of lower wet weather probability. Contract provisions
normally provide only for an extension of time to the construction period for wet weather or
its effects. Generally, the contract does not allow for any contractor cost reimbursement for
wet weather. Where there is demonstrated a case of extremely un-seasonal or exceptional
wet weather, then there may be a case where the client may incur such certain extra costs.
244