INTERNATIONAL JOURNAL OF

PROJECT
MANAGEMENT
International Journal of Project Management 25 (2007) 485–493
www.elsevier.com/locate/ijproman

Modelling risk allocation decision in construction contracts

K.C. Lam *, D. Wang, Patricia T.K. Lee, Y.T. Tsang
Construction Management Research Group, Department of Building and Construction, City University of Hong Kong, 83 Tat Chee Avenue,
Kowloon Tong, Hong Kong, PR China

Abstract

The allocation of risk among the contracting parties in a construction contract is an important decision leading to the project success.
The decision-making process, based on the established risk allocation principles expressed in linguistic terms, requires qualitative judge-
ment and experiential knowledge of the construction experts. However, it is subjective and implicit. This paper presents a decision model
which transforms the linguistic principles and experiential expert knowledge into a more usable and systematic quantitative-based anal-
ysis by using the fuzzy logic. Seven risk allocation criteria and a set of knowledge-based fuzzy inference rules are established according to
the expert knowledge. Risk events are assessed on each criterion and the relevant rules. The corresponding risk allocation decisions
between the owner and contractor are then suggested by the model. A practical case to illustrate how the model works is presented.
 2006 Elsevier Ltd and IPMA. All rights reserved.

Keywords: Construction contract; Risk allocation; Fuzzy logic; Modelling; Expert system

1. Introduction Risk allocation, the definition and division of responsi-

No construction project is risk-free. Risk is manageable,
diminishable, transferable or acceptable but not ignorable
[1]. Rahman and Kumaraswamy [2] identified 41 risks in
construction projects. Risk management is thus an impor-
tant tool to cope with such substantial risks in construction
industry by: (a) assessing and ascertaining project viability;
(b) analyzing and controlling the risks in order to minimize
loss; (c) alleviating risks by proper planning; and (d) avoid-
ing dissatisfactory projects and thus enhancing profit mar-
gins [3,4]. A risk management process typically comprises
establishment of context, risk identification, risk analysis,
risk evaluation and risk response [3,5,6]. A risk can be
characterized by the risk event, its probability of occur-
rence and the amount of potential loss or gain. All factors
comprising a risk are to be identified, analysed and evalu-
ated so that response can then be given. Risk response is
a process of formulation of a management strategy leading
to identifying action owners and the risk management plan.
*
Corresponding author. Tel.: +852 2788 7238; fax: +852 2788 7612.
E-mail address: bckclam@cityu.edu.hk (K.C. Lam).
bility associated with a possible future loss or gain, seeks
to assign responsibility for a variety of hypothetical cir-
cumstances should a project not proceed as planned [5].
As a part of a risk management strategy, it is commonly
defined through the contractual documents. Usually, a ten-
der document of a construction project is prepared by the
contracting party, i.e. the owner, who initiates the project.
It is common that the owner tends to contractually pass the
responsibility for most of the risks to the contractor under
traditional procurement processes [7]. A contract can thus
be considered as a trade-off between the contractor’s price
for undertaking the work and his willingness to accept both
the controllable and uncontrollable risks [8]. Whether the
party is willing and aware to bear the risk will affect its
response to risk [9]. The cost of improper risk allocation
could be seen from the response from contractors such as
adding a high contingency (premium) to the bid price or
delivering low quality work. During the project, the owner
might spend more management resources for the increased
work disputes. Upon completion of the works, litigation of
contractual claims might come after. In the worst case, the
owner pays for the risks twice including one in bidding con-

0263-7863/$30.00  2006 Elsevier Ltd and IPMA. All rights reserved.

doi:10.1016/j.ijproman.2006.11.005
486 K.C. Lam et al. / International Journal of Project Management 25 (2007) 485–493
tingencies and the other one in court [10]. The allocation of
risk is thus one of the important decision-making processes
leading to project success.
Optimally, the goal of risk management should be to
minimize the total cost of risk to a project, not necessarily
the costs to each contracting party separately [11]. The
most challenging of the task is to decide what the equitable
risk allocation is such that the goal is effectively accom-
plished. While model or standard sets of general conditions
of contract are available, it is argued that the principles
behind the allocations in these documents have not been
clearly stated [12]. Problems can arise using any of them
if additional clauses affecting risk are applied to them.
Moreover, the nature and extent of risks tend to be pro-
ject-specific in today’s high-risk scenarios and multiparty
complex projects that adoption of tailor-made contract
strategies is more desirable [2].
Various risk allocation principles had been suggested by
a number of researchers such as Casey [13], Kuesel [14],
Barnes [15], Abrahamson [16] and Thompson and Perry
[12]. Adopting these principles as the basis for allocating
risks is useful in reaching an equitable decision. It would
be ultimately beneficial to both owners and contractors
[17]. Like most of the management doctrines, all these risk
allocation principles commonly use natural language in the
expression, which are nevertheless ambiguous in actual
application. For example, one of the principles mentioned
by Abrahamson [16] states that ‘a party should bear a con-
struction risk where it is in his control’. The term ‘‘in his
control’’ is difficult to be precisely interpreted as the ‘‘con-
trol’’ by a contracting party on a real situation could be
‘partial’. The application of those principles to final deci-
sion making thus heavily relies on the qualitative judgment
and experiential knowledge of construction experts. The
problem of this kind of decision making process is its
implicitness. Too often it is difficult to be analyzed and
retrieved by others. Human factors such as the attitude
of the parties [15] and bias in personal judgments may
impose significant variation on the decision outcome. Rah-
man and Kumaraswamy [2] had shown that there was a
divergence of perception on risk allocation in construction
contracts among different groups. It is not surprising that
improper risk allocation in construction contracts remains
a concern in the construction industry in Hong Kong [18].
The purpose of this paper is to provide a prototypic
quantitative model for risk allocation process. It aims to
support the decision making in risk management such that
the concerns of improper risk allocation can be addressed.
As linguistic principles and qualitative expert knowledge
are the essential ingredients of the risk allocation process,
the fuzzy set theory which relates to quantification and rea-
soning of natural language is utilized in the model building.
2. Fuzzy set theory
Fuzzy set theory is a branch of modern mathematics for-
mulated by L.A. Zadeh in the 1960s to model vagueness
intrinsic to human cognitive processes. Since then, it has
been adopted to tackle ill-defined and complex real-world
problems due to partial and imprecise information. It is
particularly suitable for uncertain or approximate reason-
ing that involve human intuitive thinking. Fuzzy set theory
uses linguistic variables and membership functions with
varying grades to model uncertainty inherent in natural
language [19].
In classical (crisp) set theory, each single element x in a
universal set X can either belong to or not belong to a sub-
set A, A ˝ X. The member elements can be defined by using
a characteristic function. It assigns a binary value (either 1
or 0) to each element x 2 X , in which 1 indicates member-
ship and 0 non-membership. In fuzzy set theory, the char-
acteristic function allows various degrees of membership
for the elements. Larger values denote higher degrees or
grades of set membership. Such a generalized function is
called a membership function, l(x), in fuzzy set theory with
a range of values within an unit interval [0, 1]. Then a fuzzy
subset A e in a universal set X is a set of ordered pairs
denoted by
e ¼ fðx; lA~ ðxÞÞ j x 2 X g
A
Fuzzy set theory allows the use of linguistic variables
whose values are not numbers but words or sentences in
a natural or artificial language which are less specific than
numerical ones [20].
For example, let X be a linguistic variable with the label
‘‘Age’’ with U = [0, 100]. The linguistic values called terms
of this variable could be called ‘‘old’’, ‘‘young’’, ‘‘very old’’
and so on. T(X) will define the term set,
T ðageÞ ¼ fvery old; old; not so old; more or less young;
quite young; very youngg
If a base variable u is equal to the age in years of life, then a
fuzzy subset Ne ðX Þ and its membership function of the term
‘‘old’’ can be [21]:
e ðoldÞ ¼ fðu; lold ðuÞÞ j u 2 ½1; 100g
N
(
0 u 2 ½0; 50
where lold ðuÞ ¼ 2 1
ð1 þ ðu50
5
Þ Þ u 2 ½50; 100
Based on fuzzy set theory, the fuzzy reasoning technique is
close to the approach to inference from commonsense
knowledge [22,23]. The fuzzy inference rules can be built
to represent the knowledge and heuristic rules of experi-
enced personnel (i.e. experts in the area). They are usually
in the form: ‘‘IF a set of conditions/premises is satisfied,
THEN a set of consequences can be produced’’. It could
be applied to establish a fuzzy inference engine of a deci-
sion supporting system with imprecise and uncertain
information.
3. Proposed risk allocation decision model
This study currently focuses on the main contract and its
risk allocation between owners and contractors. A mathe-
 K.C. Lam et al. / International Journal of Project Management 25 (2007) 485–493 487

matical model based on the fuzzy set theory is developed to
support the decision making of risk allocation.
The model mainly consists of three parts, namely: fuzz-
ification, inference engine and defuzzification. Fuzzifica-
tion is a procedure that converts raw data from the
practical world into membership values of corresponding
fuzzy subsets. The transformed data are then fed into
the inference engine containing a rule base. The fuzzy
mathematic operations are implemented, producing mem-
bership values belonging to the output variables. Defuzz-
ification is followed to convert the fuzzy inferences from
the engine to a single output action giving a clear indica-
tion to the human user. The model is illustrated by Fig. 1.
3.1. Input variables for risk allocation decision
An expert team is organized in the process of formu-
lation of the model. The team involves a deputy director,
a cost and quantity manger, a contracts manager, a
senior manager, two quantity surveyors and a cost con-
trol administrator from various organizations including
owners, consultants and contractors. They all have over
15-year experience in the construction industry and are
at the managerial level. Based on the previous works
of risk allocation [12–16] and through discussions with
the team members, risk allocation criteria were identified
as follows:
Identify input variables Establish risk
and linguistic terms allocation criteria
1. Whether the party is able to foresee the risk. A risk
receiving party can take appropriate measures to man-
age a risk only if the party can reasonably foresee the
risk he may encounter [14].
2. Whether the party is able to assess the possible magnitude
of consequences of the risk. If the risk receiving party is
not able to assess the possible magnitude of consequence
of the risk, he may underestimate the outcome and make
inadequate allowance at the time of tendering to cover
his obligation [13].
3. Whether the party is able to control the risk chance of
occurring. A party should bear the risk that he can exer-
cise best control on minimizing or avoiding the materi-
alization of the risk [12,14–16].
4. Whether the party is able to manage the risk in case of
occurring. Risk should be allocated to the party who is
best able to manage it in order to minimize the severity,
extra cost and delay once the risk occurs [12].
5. Whether the party is able to sustain the consequences if
the risk occurs. [12]. Chapman and Ward [24] pointed
out that the party able to manage a risk may not be
the party best able to bear the consequences of that risk.
6. Whether the party will benefit from bearing the risk. Risk-
bearing can be beneficial to a receiving party such as the
enhancement of credibility, reputation and efficiency in
planning and innovation [16].
7. Whether the premium charged by the risk receiving party
is considered reasonable and acceptable for the owner.
When an owner intends to transfer a risk to a contrac-
tor, he needs to consider whether the contractor will
charge a high price for taking the risk or he should
assume the risk himself so that a more realistic and rea-
sonable tender price will be finally obtained [12].

Conduct fuzzification –
convert linguistic terms
into membership functions
Examine the relevant base
rules in the Inference
Engine
Apply fuzzy mathematic
operations to obtain the
membership values of
inferences
Apply center-of-sum
method to calculate the
overall value of the output
variable
Build up base rules
by pooling expert’s
knowledge
As the model considers the risk allocation between the
owner and contractor of a traditional contract arrange-
ment, seven linguistic input variables, denoted by IV1–
IV7, are subsequently derived based on the risk allocation
criteria. To evaluate a risk event, the percentage is used
to indicate the likelihood of the risk being foreseeable,
assessable, controllable, manageable, sustainable by or
beneficial to the contractor, which forms the base variable
of the IV1–IV6. The percentage of premium in relation to
estimated cost of risk transfer is used for IV7. All base vari-
ables range from 0% to 100%.
Based on the fuzzy set theory, the linguistic values, i.e. terms
(TIV), are defined to describe the input variables to facilitate
the building of fuzzy inference rules in the next stage. For
IV1–IV6 the finalized term set is {very unlikely, unlikely, mod-
erate, likely, very likely} and for IV7 the term set is {acceptable,
moderate, unfavourable} through discussion with the expert
team. All details are listed in Table 1.

3.2. Fuzzification of the input variables

Determine final risk
allocation decision
The procedure for transforming input data such that
Fig. 1. Risk allocation decision model. they belong to a particular fuzzy subset is known as fuzzifi-
488 K.C. Lam et al. / International Journal of Project Management 25 (2007) 485–493

Table 1
Input variables and terms
Foreseeable by the contractor, IV1 Range of % of likelihood
TIV11 Very unlikely 0–25
TIV12 Unlikely 0–50
TIV13 Moderate 25–75
TIV14 Likely 50–100
TIV15 Very likely 75–100
Assessable by the contractor, IV2 Range of % of likelihood
TIV21 Very unlikely 0–25
TIV22 Unlikely 0–50
TIV23 Moderate 25–75
TIV24 Likely 50–100
TIV25 Very likely 75–100
Controllable by the contractor, IV3 Range of % of likelihood
Fig. 2. Fuzzification functions for ‘‘foreseeable by the contractor’’.
TIV31 Very unlikely 0–25
TIV32 Unlikely 0–50
TIV33 Moderate 25–75
They are adopted to represent the membership functions of
TIV34 Likely 50–100
TIV35 Very likely 75–100 the terms of input variables in the model. The ranges of
each TIV covering are defined by the expert team as listed
Manageable by the contractor, IV4 Range of % of likelihood
in Table 1. Fig. 2 illustrates the fuzzification functions for
TIV41 Very unlikely 0–25
TIV42 Unlikely 0–50 IV1,‘‘foreseeable by the contractor’’, in which S function
TIV43 Moderate 25–75 or p functions are used alternatively for the term set. The
TIV44 Likely 50–100 fuzzification functions for IV2 to IV6 are same as the IV1.
TIV45 Very likely 75–100 For the term set of IV7, the fuzzification functions are
Sustainable by the contractor, IV5 Range of % of likelihood shown in Fig. 3. As such each value of IVs is associated
TIV51 Very unlikely 0–25 with the membership values belonging to different TIVs
TIV52 Unlikely 0–50 within a unit interval.
TIV53 Moderate 25–75
TIV54 Likely 50–100
TIV55 Very likely 75–100 3.3. Output variable – risk allocation decision
Beneficial to the contractor, IV6 Range of % of likelihood
TIV61 Very unlikely 0–25 The output variable of the model is the risk allocation
TIV62 Unlikely 0–50 decision which is defined on the basis of the risk allocation
TIV63 Moderate 25–75 scale adopted by Hartman et al. [25] with modification by
TIV64 Likely 50–100 the expert team. The allocation scale proposes a range of
TIV65 Very likely 75–100
risk apportionment from fully bearing by the owner,
Premium of transfer, IV7 Range of % of premium in through a continuum to fully bearing by the contractor.
related to estimated cost Three values of output variable (OV) are identified: own-
TIV71 Premium charged by the 0–40
contractor is unfavourable
er’s risk (=1); shared risk (=2); and contractor’s risk
TIV72 Premium charged by the 0–80 (=3). The S and p functions are adopted again to trans-
contractor is moderate form the scale to the corresponding fuzzy subsets as shown
TIV73 Premium charged by the 40–100 in Fig. 4.
contractor is acceptable
3.4. The rule base
cation. The most popular fuzzification functions are either
the S function or p function. They are defined as follows By utilizing the fuzzy reasoning technique, a rule base is
[20]: built in the inference engine of the model based on the expe-
8
>
> 0 for m 6 a
>
>  2
>
> ma
< 2 ca for a 6 m 6 b
Sðm; a; b; cÞ ¼  2
>
> mc
>
> 1  2 ca for b 6 m 6 c
>
>
:
1 for m P c
( b
Sðm; c  b; c  2 ; cÞ for m 6 c
pðm; b; cÞ ¼ b
1  Sðm; c; c þ 2 ; c þ bÞ for m P c
ð a þ cÞ
with b ¼ and m; a; b; c are real numbers:
2 Fig. 3. Fuzzification functions for ‘‘premium of transfer’’.
 K.C. Lam et al. / International Journal of Project Management 25 (2007) 485–493
Fig. 4. Fuzzification functions for ‘‘risk allocation’’.
rience and knowledge of the expert team. The production
rules in IF–THEN format are used to represent the expert
knowledge, i.e. ‘‘if a set of conditions/premises is satisfied,
then a set of consequences can be produced.’’ The inference
engine is a mechanism for manipulating the encoded
knowledge and to form inferences and draw conclusion
[21]. A typical mathematical format for N number of input
variable (IV), S number of terms (TIV) associated with
each IV, and M number of output variable (OV) is [26]:
IF fIVi is TIVik g and . . . and fIVu is TIVuv g THEN OVj
where i 2 f1; . . . ; N g; k 2 f1; . . . ; S i g; u 2 f1; . . . ; N g;
v 2 f1; . . . ; Su g; j 2 f1; . . . ; Mg and u 6¼ i
Theoretically, the total number of possible fuzzy inference
rules would be SN rules, which can be enormous. The
advantage of fuzzy reasoning is that even when all the rules
are not present we can still get reasonable and usable re-
sults. The larger the rule base is, the better the result can
be [26,27].
Since the membership function of TIVik is given by lik
and so on, the membership function of the output variable
ðOVpj Þ of the pth rule is given by using the ‘‘minimum’’
operation of the fuzzy set theory:
lðOVpj Þ ¼ lpik ^    ^ lpuv ¼ min½lpik ; . . . ; lpuv 
3.5. Defuzzification and decision for risk allocation
When a case is evaluated, it will be examined by each
rule in turn. Defuzzification is followed to convert each
conclusion expressed in terms of a fuzzy set obtained by
the inference engine to a single output action. The center-
of-sum method is used to determine the overall level of risk
allocation of all examined rules. It is a popular defuzzifica-
tion method in most fuzzy control algorithms and tools
[21]. The idea of this method is to aggregate the informa-
tion about possible control actions that are represented
by the membership functions so as to reach a compromise
solution. If total T rules are examined, the numerical
defuzzified value, d(OV), is computed as
XT   X T  
dðOV Þ ¼ ðOVpj Þ  l OVpj = l OVpj
p¼1 p¼1
If d(OV) lies between two values of the set of OV, the final
decision is made by choosing the value of closet to it [28],
489
i.e. the OV having the highest membership grade of
d(OV) will be chosen (max[l(OVj)] where j 2 {1, . . . , M}).
It gives a clear linguistic indication to the human user of
overall decision of risk allocation.
4. Evaluation of the model
A railway company supported by the Hong Kong Gov-
ernment has been proceeding with the design, construction,
financing and operation of the railway network in Hong
Kong. To manage a latest large-scale railway construction
composed of 15 civil engineering contracts of awarded sum
ranging from HK$118 million to HK$2238 million, the
company specifically developed a set of conditions of con-
tract entitled as ‘‘Railway General Conditions of Contract
for Civil Engineering and Building Works’’ (RGCC).
16 risk events are identified from the RGCC as listed in
Table 2. Based on the Casey’s categorization [13], they are
categorized into five groups for clarity, namely, capability
related risks, contractual and legal risks, economic risks,
physical risks and political and societal risks.
The participants of the expert team are again asked to
assess each risk events against each risk allocation criterion
in terms of the percentage of likelihood for IV1–IV6 or pre-
mium in relation to estimated cost for IV7. Each partici-
pant works independently based on his/her experience
and professional judgment to determine the value of each
input variable. As this study focuses on the building of a
prototypic quantitative model based on fuzzy reasoning
techniques, the arithmetic mean of all participants’ value
is taken for simplicity. It is then used to derive the member-
ship values of the seven input variables for each risk event
using Figs. 2 and 3. The outcome is listed in Table 2. In
case participants have clear differences in estimating the
value in actual applications, these differences should be
explicitly discussed, e.g., through the approach of the Kei-
zer, Halman and Song’s work [29], before feeding into the
model.
The fuzzified data are then fed into the inference engine
in which the relevant base rules are examined for each risk
event. Total 128 (=27) base rules has been examined for
each case. To demonstrate how the model works, one of
the risk events, ‘‘Third Party Delay’’, is taken as an exam-
ple. For IV1, the mean percentage of likelihood of Third
Party Delay foreseeable by the contractor assessed by the
expert team is 38%. It is fuzzified into the fuzzy subsets
of the Terms TIV12 ‘‘Unlikely’’ and TIV13 ‘‘Moderate’’
with membership grade of 0.43 and 0.57, respectively, using
Fig. 2. All other values of the input variables for this risk
event are similarly transformed.
In the inference engine, for instances, the first relevant
base rule established is:
IF Third Party Delay is unlikely foreseeable by the con-
tractor: IV1 = TIV12
and it is unlikely assessable by the contractor:
IV2 = TIV22
 490
Table 2
16 Risk events and the corresponding membership values of input variables
IV1, Foreseeable by IV2, Assessable by IV3, Controllable IV4, Manageable by IV5, Sustainable by IV6, Beneficial by IV7, Premium of
the contractor the contractor by the contractor the contractor the contractor the contractor transfer
Capability
Contractor’s design TIV14 TIV15 TIV24 TIV25 TIV34 TIV35 TIV44 TIV45 TIV54 TIV55 TIV64 TIV65 TIV72 TIV73
0.07 0.93 0.09 0.91 0.08 0.92 0.21 0.79 0.17 0.83 0.08 0.92 0.86 0.14
Subcontractor failure TIV13 TIV14 TIV23 TIV24 TIV33 TIV34 TIV44 TIV45 TIV54 TIV55 TIV63 TIV64 TIV72 TIV73

K.C. Lam et al. / International Journal of Project Management 25 (2007) 485–493

0.23 0.77 0.11 0.89 0.28 0.72 0.12 0.88 0.21 0.79 0.33 0.67 0.31 0.69
Quality of work TIV13 TIV14 TIV24 TIV25 TIV34 TIV35 TIV44 TIV45 TIV53 TIV54 TIV63 TIV64 TIV72 TIV73
0.13 0.87 0.29 0.71 0.21 0.79 0.21 0.79 0.30 0.70 0.11 0.89 0.22 0.78
Safety at site TIV13 TIV14 TIV23 TIV24 TIV34 TIV35 TIV44 TIV45 TIV53 TIV54 TIV63 TIV64 TIV72 TIV73
0.15 0.85 0.29 0.71 0.79 0.21 0.90 0.10 0.35 0.65 0.45 0.55 0.34 0.66
Obtaining approval or consent TIV13 TIV14 TIV23 TIV24 TIV33 TIV34 TIV44 TIV45 TIV52 TIV53 TIV63 TIV64 TIV72 TIV73
from relevant authority 0.33 0.67 0.35 0.65 0.41 0.59 0.45 0.55 0.11 0.89 0.34 0.66 0.67 0.33
Contractual and legal
Conflicts in documents TIV11 TIV12 TIV21 TIV22 TIV31 TIV32 TIV42 TIV43 TIV51 TIV52 TIV61 TIV62 TIV71 TIV72
0.21 0.79 0.21 0.79 0.95 0.05 0.63 0.37 0.17 0.83 0.83 0.17 0.93 0.07
Third party delay TIV12 TIV13 TIV22 TIV23 TIV32 TIV33 TIV42 TIV43 TIV51 TIV52 TIV61 TIV62 TIV71 TIV72
0.43 0.57 0.46 0.54 0.82 0.18 0.45 0.55 0.31 0.69 0.21 0.79 0.81 0.19
Economic
Inflation TIV12 TIV13 TIV23 TIV24 TIV31 TIV32 TIV43 TIV44 TIV53 TIV54 TIV62 TIV63 TIV72 TIV73
0.79 0.21 0.83 0.17 0.99 0.01 0.88 0.12 0.67 0.33 0.78 0.22 0.21 0.79
Availability of labour and TIV13 TIV14 TIV23 TIV24 TIV34 TIV35 TIV44 TIV45 TIV53 TIV54 TIV63 TIV64 TIV72 TIV73
equipment 0.17 0.83 0.19 0.81 0.82 0.18 0.79 0.21 0.21 0.79 0.23 0..77 0.21 0.79
Physical
Ground conditions TIV12 TIV13 TIV22 TIV23 TIV31 TIV32 TIV43 TIV44 TIV52 TIV53 TIV62 TIV63 TIV71 TIV72
0.67 0.33 0.63 0.37 0.32 0.68 0.39 0.61 0.67 0.33 0.57 0.43 0.78 0.22
Rights of access to site TIV11 TIV12 TIV22 TIV23 TIV32 TIV33 TIV42 TIV43 TIV52 TIV53 TIV62 TIV63 TIV71 TIV72
0.21 0.79 0.59 0.41 0.89 0.11 0.77 0.23 0.88 0.12 0.87 0.13 0.93 0.07
Quantity variations TIV13 TIV14 TIV23 TIV24 TIV33 TIV34 TIV43 TIV44 TIV53 TIV54 TIV63 TIV64 TIV72 TIV73
0.21 0.79 0.21 0.79 0.34 0.66 0.32 0.68 0.18 0.82 0.82 0.18 0.68 0.32
Inclement weather TIV13 TIV14 TIV23 TIV24 TIV31 TIV32 TIV43 TIV44 TIV53 TIV54 TIV62 TIV63 TIV72 TIV73
0.19 0.81 0.21 0.79 0.45 0.55 0.17 0.83 0.24 0.79 0.32 0.68 0.31 0.69
Political and societal
Changes in laws and regulations TIV12 TIV13 TIV22 TIV23 TIV31 TIV32 TIV43 TIV44 TIV53 TIV54 TIV61 TIV62 TIV72 TIV73
0.29 0.71 0.27 0.73 0.15 0.85 0.16 0.84 0.37 0.63 0.16 0.84 0.58 0.42
Public disorder TIV11 TIV12 TIV21 TIV22 TIV31 TIV32 TIV41 TIV42 TIV51 TIV52 TIV61 TIV62 TIV71 TIV72
0.19 0.81 0.24 0.76 0.12 0.88 0.22 0.78 0.23 0.77 0.15 0.85 0.26 0.74
Labour disputes and strikes TIV13 TIV14 TIV23 TIV24 TIV33 TIV34 TIV44 TIV45 TIV53 TIV54 TIV63 TIV64 TIV72 TIV73
0.21 0.79 0.19 0.81 0.16 0.84 0.12 0.88 0.27 0.73 0.23 0.77 0.64 0.36
 K.C. Lam et al. / International Journal of Project Management 25 (2007) 485–493 491

and it is unlikely controllable by the contractor: dðOVÞ ¼ ½1  0:21ð1stÞ þ 1  0:19ð2ndÞ þ    þ 2

IV3 = TIV32
 0:19ð10thÞ þ 1  0:31ð11thÞ þ    þ 1
and it is unlikely manageable by the contractor:
IV4 = TIV42  0:18ð127thÞ þ 2  0:18ð128thÞ=½0:21ð1stÞ
and it is very unlikely sustainable by the contractor: þ 0:19ð2ndÞ þ    þ 0:19ð10thÞ þ 0:31ð11thÞ
IV5 = TIV51
and it is very unlikely beneficial by the contractor: þ    þ 0:18ð127thÞ þ 0:18ð128thÞ
IV6 = TIV61 ¼ 35:86=17:06 ¼ 1:33
and the premium of transfer is unfavourable:
IV7 = TIV71 With reference to Fig. 4, the d(OV) for ‘‘Third Party De-
THEN the risk should be allocated to the owner: lay’’ lies between the owner’s risk and shared risk with
OV1 = 1 membership grade of 0.77 and 0.23, respectively. To con-
solidate a final decision, the output variable with the high-
From Table 2, the value of membership function of each est membership value is chosen. Accordingly, the final risk
input variable devised for this risk event regarding this rule is: allocation decision for ‘‘Third Party Delay’’ is the owner’s
risk. The result calculated and recommendations from the
TIV12 (unlikely) = 0.43 model for all risk events are presented in Table 3.
TIV22 (unlikely) = 0.46
TIV32 (unlikely) = 0.82 4.1. Analysis of the results
TIV42 (unlikely) = 0.45
TIV51 (very unlikely) = 0.31 The actual risk allocation stipulated in the RGCC for the
TIV61 (very unlikely) = 0.21 16 risk events is also listed in Table 3. Comparing the two
TIV71 (unfavourable) = 0.81 patterns, the final decision obtained from the model basi-
cally ties in with those in RGCC. The major deviation is
The membership value of output variable is thus found in the risk events ‘‘Inflation’’ and ‘‘Inclement
obtained by the ‘‘min’’ operation: weather’’. The model suggests that these two risks shall be
shared by both parties while the RGCC allocates them to
lðOV1 Þ ¼ min½0:43; 0:46; 0:82; 0:45; 0:31; 0:21; 0:81 ¼ 0:21 the contractor. Actually the recommendation from the
model is more consistent with the usual contracting practice
The overall level of risk allocation after examining by all as ‘‘Inflation’’ and ‘‘Inclement weather’’ are commonly trea-
base rules is determined by applying the center-of-sum ted as shared risk in most contract arrangements. A possible
method: reason for this alternative arrangement in RGCC is the

Table 3
Comparison of risk allocation decision between the model and RGCC
Events Numerical results Final risk allocation Risk allocation in
obtained from the model decision from the model RGCC
Capability
Contractor’s design 3.0 Contractor Contractor
Subcontractor failure 2.99 Contractor Contractor
Quality of work 2.99 Contractor Contractor
Safety at site 2.99 Contractor Contractor
Obtaining approval or consent from 2.75 Contractor Contractor
relevant authority
Contractual and legal
Conflicts in documents 1.02 Owner Owner
Third party delay 1.33 Owner Owner
Economic
Inflation 2.33 Share Contractor
Availability of labour and equipment 3 Contractor Contractor
Physical
Ground conditions 1.39 Owner Owner
Rights of access to site 1.27 Owner Owner
Quantity variations 2.69 Contractor Contractor
Inclement weather 2.43 Share Contractor
Political and societal
Changes in laws and regulations 2.27 Share Share
Public disorder 1.0 Owner Owner
Labour disputes and strikes 2.83 Contractor Contractor
492 K.C. Lam et al. / International Journal of Project Management 25 (2007) 485–493
political consideration that maximizing the certainty of the
price and progress of the proposed project is more impor-
tant than seeking the minimum cost of risks of the whole
project. It is, however, out of the scope of the risk allocation
criteria adopted in the model. The model also gives signals
for some risk events to which special attention should be
paid. The numerical results of the risk events ‘‘Obtaining
approval or consent from relevant authority’’, ‘‘Third party
delay’’, ‘‘Ground condition’’, ‘‘Right of access to site’’,
‘‘Quantity variations’’ and ‘‘Change in laws and regula-
tions’’ lie somewhere between 2 risk allocation choices.
While the algorithm finalizes the decision by choosing the
one with the highest membership value, the users can focus
the resources to proceed further examination on these risks
where necessary. After reconciliation, the expert team
agreed to the result obtained from the model and recognised
that the model had enabled them to consider all those risk
allocation criteria in reaching the final decision.
5. Conclusions
New and project-specific risks emerged in today’s multi-
party complex construction projects that a tailor-made con-
tract strategy allocating risks among the contracting parties
based on the accepted allocation principles is more desirable.
The prototypic decision model proposed in this paper dem-
onstrates the extraction and transformation of these impre-
cise linguistic principles and experiential expert knowledge
into a more usable quantitative-based analysis by using the
fuzzy set theory. It has several advantages. It provides an
explicit and systematic framework in risk allocation practice
rather than a subjective and untraceable approach based on
individual’s professional judgement. It examines the alloca-
tion of risks more fundamentally based on accepted risk allo-
cation principles rather than simply using standard clauses
from typical conditions of contract without appreciating
the rationale behind. The outcome of the model can be in
numerical or linguistic indication which provides appropri-
ate signals to different users. As a decision support tool, it
provides a ground for the users in reaching a final decision.
Through the model building involving different experts from
the owners, consultants and contractors in the early stage of
the risk allocation decision, the final outcome will be more
agreeable by different contracting parties. Currently, the
scope of the proposed model is limited to risk allocation
between the owner and the contractor in a traditional con-
tract arrangement. There is room for upgrading when more
information and expert knowledge can be pooled. It can be
modified to suit a specific context or project strategy by
adjusting or expanding the risk allocation criteria, the mem-
bership functions of the variables and the inference rules.
Acknowledgement
The work described in this paper was fully supported by
a grant from the Research Grants Council of the Hong
Kong Special Administrative Region, China [Project No.
CityU 1293/04E].
References
[1] Latham M. Constructing the team: joint review of procurement and
contractual arrangements in the United Kingdom construction
industry. London: HMSO; 1994.
[2] Rahman MM, Kumaraswamy MM. Risk management trends in the
construction industry: moving towards joint risk management. Eng
Construct Archit Manage 2002;9(2):131–51.
[3] Edwards PJ, Bowen PA. Risk and risk management in construction: a
review and future directions for research. Eng Construct Archit
Manage 1998;5(4):339–49.
[4] Waring AI, Glendon AI. Managing risk. London: International
Thomson Business Press; 1998.
[5] Uff J. Risk, management and procurement in construction. London:
Centre of Construction Law and Management; 1995.
[6] Lyons T. Project risk management in the construction industry: a
review. Australia: The Australian Institute of Quantity Surveyor; 2003.
[7] Rutgers JA, Haley HD. Project risks and risk allocation. Cost Eng
AACE Trans 1996;38(9):27–30.
[8] Flanagan R, Norman G. Risk management and construction.
Oxford: Blackwell Scientific Publications; 1993.
[9] Ward SC, Chapman CB, Curtis B. On the allocation of risk in
construction projects. Int J Project Manage 1991;9(3):140–7.
[10] Fisk ER. Construction project administration. New Jersey: Prentice-
Hall; 2000.
[11] CII (Construction Industry Institute). Allocation of insurance related
risks and costs on construction projects. University of Texas at
Austin, Austin; 1993.
[12] Thompson P, Perry J. Engineering construction risks: a guide to
project risk analysis and risk management. London: Thomas Telford
Ltd; 1992.
[13] Casey JJ. Identification and nature of risks in construction project: a
contractor’s perspective. In: Proceedings: construction risk and
liability sharing conference. Scottsdale (USA): American Society of
Civil Engineers; 1979. p. 17–24.
[14] Kuesel TR. Allocation of risks. In: Proceedings: construction risk and
liability sharing conference. Scottsdale (USA): American Society of
Civil Engineers; 1979. p. 51–60.
[15] Barnes M. How to allocate risks in construction contracts. Int J
Project Manage 1983;1(1):24–8.
[16] Abrahamson MW. Risk management. Int Construct Law Rev
1984;1(3):241–64.
[17] Grove JB. Consultant’s report on review of general conditions of
contract for construction works for the government of the Hong Kong
special administrative region. New York: Thelen Reid & Priest Lip; 1998.
[18] CIRC (Construction Industry Review Committee). Construct for
excellence: report of the Construction Industry Review Committee.
HK SAR Government, HK; 2001.
[19] Baloi D, Price ADF. Modelling global risk factors affecting
construction cost performance. Int J Project Manage 2003;21:261–9.
[20] Zadeh LA. The concept of a linguistic variable and its application to
approximate reasoning. Inform Sci 1975;8:199–249.
[21] Zimmermann HJ. Fuzzy set theory and its applications. third ed.
USA: Kluwer Academic Publishers; 1996.
[22] Zadeh LA. Commonsense reasoning based on fuzzy logic. In: Proceed-
ings of the 1986 winter simulation conference. Washington: IEEE; 1986.
[23] Tanaka K. An Introduction to fuzzy logic for practical applications.
New York: Springer-Verlag Inc; 1997.
[24] Chapman C, Ward S. Project risk management: processes, techniques
and insights. Chichester: John Wiley and Sons; 1997.
[25] Hartman F, Snelgrove P, Ashraft R. Effective wording to improve
risk allocation in lump sum contracts. J Construct Eng Manage
1997;123(4):379–87.
 K.C. Lam et al. / International Journal of Project Management 25 (2007) 485–493 493

[26] Wong KC, So ATP. A fuzzy expert system for contract decision
making. Construct Manage Econ 1995;13:95–103.
[27] Lam KC, So ATP, Hu T, Ng T, Yuen RKK, Lo SM, Cheung SO,
Yang H. An integration of the fuzzy reasoning technique and the
fuzzy optimization method in construction project management
decision-making. Construct Manage Econ 2001;19:63–76.
[28] Klir GJ, Yuan B. Fuzzy sets and fuzzy logic – theory and
applications. New Jersey: Prentice Hall; 1995.
[29] Keizer JA, Halman JIM, Song M. From experience: applying the
risk diagnosing methodology. J Prod Innov Manage 2002;19(3):
213–32.