Automation in Construction 65 (2016) 920

Contents lists available at ScienceDirect

Automation in Construction
journal homepage: www.elsevier.com/locate/autcon

Impact of considering need for work and risk on performance of

construction contractors: An agent-based approach
Sadegh Asgari a,, Rita Awwad b, Amr Kandil c, Ibrahim Odeh a
a
b
c

Department of Civil Engineering and Engineering Mechanics, Columbia University, New York, NY 10027, United States
Department of Civil Engineering, Lebanese American University, P.O. Box 36, Byblos, Lebanon
School of Civil Engineering, Purdue University, 550 Stadium Mall Drive, West Lafayette, IN 47907, United States

a r t i c l e

i n f o

Article history:
Received 3 February 2015
Received in revised form 25 December 2015
Accepted 22 January 2016
Available online 16 February 2016
Keywords:
Agent-based modeling
Competitive bidding
Computer-aided simulation
Multi-attribute decision making
Markup
Need for work
Risk attitude

a b s t r a c t
Competitive bidding is the main mechanism of allocating projects in the construction market. In the traditional
single criterion bidding method, the markup decision has a signicant impact on a contractor's business success.
Contractors usually take into consideration several factors in the process of determining their markup. This study
has reviewed the literature and identied a range of contractors' behaviors when making their markup decision
within a competitive bidding environment. An additive markup function consisting of three components, namely
competition, risk, and need for work, was developed in order to replicate markup behaviors of contractors. Then,
agent-based modeling has been employed for simulating the bidding process within a market formed of a set of
heterogeneous contractors with different risk attitudes and dened markup behaviors. This model was used to
study the impact of considering need for work and risk allowance in markup determination on nancial performance of contractors in various market scenarios. Results suggest that the optimal policy is moderation in both
dimensions of risk attitude and need for work.
2016 Elsevier B.V. All rights reserved.

1. Introduction
Once a contracting rm has made the decision to bid on a project, it
is then faced with a more challenging decision which is to select a bid
price that is low enough to win the project and at the same time high
enough to make a good prot. A reasonable objective for each construction company is maximizing its prot in order to ensure survival in the
market. However, while prot maximization is usually the most frequently used bidding objective [7], many researchers argue that it
should not be the sole criterion upon which the markup is based. A
contractor's need for work might sometimes play a conicting role
with short-term prot maximization and thus is identied as a major
factor when determining the optimum markup for a project [1,12,39].
For instance, if a contractor is in a bad nancial situation and is in urgent
need of cash ow to cover its overhead and its general and administrative expenses, winning the project becomes a priority even if it comes at
the expense of a much lower prot. Thus, the need to continuously
cover some xed costs such as ofces' rentals, utilities, personnel salaries, insurance payments on property and equipment, and others
might sometimes drive the contractor to submit a very low bid in
order to at least breakeven and stay in the business. Moreover, the attitude of a contractor towards risk is another primary factor in
Corresponding author.
E-mail addresses: sadegh.asgari@columbia.edu (S. Asgari), rita.awwad@lau.edu.lb
(R. Awwad), akandil@purdue.edu (A. Kandil), odeh@columbia.edu (I. Odeh).

http://dx.doi.org/10.1016/j.autcon.2016.01.004
0926-5805/ 2016 Elsevier B.V. All rights reserved.

determining the optimum markup for a project. Contractors have varying perceptions about market conditions and projects' uncertainties according to their inner state, bidding preferences and risk tolerance [33].
In fact, a construction project can be considered as a lottery with different prot outcomes and with a level of uncertainty resulting from the
expected variance in the nal cost of the project. The value of this lottery and its desirability differs from one contractor to another depending on their risk attitudes. Thus, it is important to consider the behavior
of a contractor towards risk when making decisions under uncertainty
such as the optimum markup decision, and in complex and risky environments such as the construction eld. A third major component
that denes the markup choice by contractors and that is outside their
control is the intensity of competition in the market determined by
the number of potential opponents bidding on the same project and
their observed competitiveness degree. Contractors need to adjust
their prot margins from one project to the other depending on the
change in the number or identity of competitors [9]. Some other less important factors affecting the optimal markup decision were also identied in the literature such as type of project and inherent complexity,
client character and record of payment, reliability of subcontractors,
and the degree of uncertainty in cost estimates.
Several researchers listed and ranked the different factors a contractor takes into consideration when deciding whether to bid on a certain
project or not and determining which markup to use. Moreover, many
analytical and simulation models were presented in the literature that
provided insight into the complexities and dynamics of the bidding

10

S. Asgari et al. / Automation in Construction 65 (2016) 920

process in the construction industry. These models reected the multiattribute nature of the bid/markup decision and devised approaches to
help contractors in making bidding decisions carefully. However, ndings from such models remain mere theoretical recommendations
since top management in contracting rms keep relying on their subjective assessment, their experience and their intuition in determining
their optimum markup [1,11,16].
The main contribution of multi-attribute bidding models is that they
reected the invalidity of the homogeneous bidding behavior assumption adopted in previously developed bidding models and paved the
way for a new set of studies emphasizing heterogeneity in bidding preferences among contractors operating in the same environment and between groups of bidders operating in different markets. For instance, Oo
et al. [32,33] designed bidding experiments to imitate the decision to
bid and optimal markup selection in two construction markets, Hong
Kong and Singapore, and showed that the two nations exhibited different behaviors in response to four bidding factors, namely number of
bidders, market conditions, project type and size. Kim and Reinschmidt
[25,26] addressed the heterogeneity in the risk behavior of contractors
and used an evolutionary simulation approach to investigate its effect
on the markup decision, the contractor's success, and the market structure within a competitive environment. Ho and Hsu [23] also considered heterogeneity among contractors in their competitive advantage
over each other and followed a game theoretic approach to explore
the effect of several bid compensation strategies on encouraging bidders
to put more effort into project planning and tender preparation.
This paper focuses on three of the most inuential bidding factors
identied in the literature namely, the market competition, the
contractor's risk behavior and his need for work. None of the prior studies discussed the interaction among the aforementioned markup dening components and the relative dominance of one on the other within a
market of heterogeneous contractors exhibiting different bidding behavior and strategies. How does considering a contractor's need for
work in his markup decision affect his longterm performance in the
market? To what extent should a markup be adjusted to account for
need for work without compromising protability? How is this process
affected by the risk behavior of a contractor and how do these two components interact? To answer the latter questions, this study presents an
agent-based simulation model of the construction bidding process
where contractors are modeled as autonomous agents competing on
projects over time which allows observing the long-term effect of a
contractor's risk attitude, his need for work and the market competition
intensity on his optimal markup, his nancial growth and survival, and
his bid success rate. Hence, the three main objectives of this paper can
be summarized as follows: (1) observing the impact of a contractor's
risk attitude on his markup decision taking into account his need for
work and the market competition, (2) assessing if and to which extent
considering need for work in the markup decision affects the nancial
growth of a contractor and his market share, and (3) identifying emergent market patterns in a diverse competitive environment exhibiting
different degrees of risk tolerance and different levels of need for work
considerations.
2. Previous works
Following the very rst statistical construction bidding models that
focused on the sole criterion of prot maximization [8,19,21] in determining the optimal markup for a contractor, several multi-attribute
models were presented in the literature that emphasized the importance of considering multiple factors when making bidding decisions.
Indeed, the rst addition to the aforementioned prot maximization
models was the introduction of the risk notion based on utility theory
and concepts. A group of studies emerged in this regard and focused
on the impact of the risk behavior of a contractor on his bidding decisions [13,24,43]. Then, the literature on this topic went in two directions: a set of empirical studies that focused on dening the most

inuential factors in a contractor's optimal markup decision, and another set of articles that attempted at modeling the multi-attribute nature
of the optimal markup decision using different approaches.
2.1. Multi-attribute nature of markup decision
Ahmad and Minkarah [1] presented the results of a survey administered to US contractors to determine the main factors considered in
their bid/no bid and markup decisions. This study identied the
contractor's need for work, type of job, degree of hazard, economic conditions, competition, degree of uncertainty in cost estimate, and reliability of subcontractors among others as important factors in determining
the project markup value. Also, Shash [39] presented the ndings of a
questionnaire that collected feedback from 85 top UK contracting
rms about the most prominent factors that affect their bidding decisions and their weights. Degree of difculty, risk involved, current
workload and need for work were identied as the most inuential
ones among several other listed factors. Chua and Li [12] conducted interviews with competitive bidding experts and top contractors in the
Singapore construction industry through which they identied the potential level of competition, the inherent project risk, the contract
type, the company's bidding position and its need for the job as essential
considerations in bidding decisions. In turn, Dulaimi and Shan [16] identied forty factors that medium and large size contractors in Singapore
perceive to be important when considering their bid markup decision.
These factors were classied into project characteristics, company's attributes, bidding situation, economic environment, and project documentation. Furthermore, Ye et al. [45] presented a study that dened
key factors contractors in China consider when preparing tender prices
for public projects. These factors were ranked and classied under different categories including construction cost, contractor heterogeneity,
payment terms, potential competitors, client requirements, market conditions, and third-party stakeholders.
All of the aforementioned empirical studies identifying factors that
contractors take into consideration when faced with bidding decisions
highlighted that the rm's internal conditions (including need for
work, risk behavior, current workload, nancial capacity, rm size and
others) are equally and sometimes more important than project attributes and market characteristics. In a study conducted by Ahmad and
Minkarah [1], need for work was identied as the most inuential factor
in making a contractor desperate to get the work. Also, in his survey
about UK contractors, Shash [39] reiterated the observation that 90%
of the respondents identied need for work as the main incentive that
would make them take any measures to win the job. In fact, more
than 70% of contractors pass through situations where they are in urgent need for the job and have to bid accordingly [12]. Hence, the
focus of this paper will be on studying to which extent the need for
work component affects the bid markup decision and the survival of a
contractor in a risky and competitive market.
2.2. Multi-attribute bidding models
Inspired by the former studies which highlighted key attributes that
form the basis of the bid markup decision, some researchers went further to develop multi-attribute bidding models that can assist contractors in determining their optimal markup. For example, Seydel and
Olson [38] presented a quantitative method based on the analytical hierarchy process that can be used by contractors to determine their optimal markup for a particular project given their bidding preferences at
that time. This model took into consideration three factors in the markup decision which are prot maximization, risk exposure minimization
and workforce continuity. On the other hand, DeNeufville and King [14]
focused on two inuential factors in a contractor's bidding behavior
which are risk and need for work, and developed prot markup utility
functions for different possible combinations of these two factors
based on an empirical investigation of the bidding decisions of 30

S. Asgari et al. / Automation in Construction 65 (2016) 920

selected contractors in Boston. Moreover, Hegazy and Moselhi [22] developed an optimum markup estimation model based on articial neural networks technique which determines the best optimum markup for
a certain project based on analogy with previous similar projects. The
bidding factors considered in their model were based on the ones identied in the survey conducted by Ahmad and Minkarah [1]. As for Dozzi
et al. [15], they also used some of the bidding criteria identied in
Ahmad and Minkarah [1] including market conditions, competition,
current workload and other factors based on which they developed a
utility theory model that can assist contractors in their markup selection. Fayek [18] used fuzzy set theory to model the margin-size decision
based on 93 bidding factors compiled from previous research [1,39].
Christodoulou [11] also combined neurofuzzy systems and multidimensional risk analysis algorithms to develop a bidding model that assists
contractors in their optimal markup decision. Then, Liu and Ling [30]
used fuzzy neural network technique to model the markup estimation
process by contractors. The model was based on data collected through
a survey of 29 large size contractors about the importance of several attributes when determining the bid markup component. The study revealed that payment record of client, competitiveness of other
bidders, availability of work, project complexity and the presence of
owner's special requirement were respectively the top ranked attributes in the bid markup decision. Furthermore, Wang et al. [42] presented a simulation-based cost model that takes into account cost
uncertainties and a multi-criteria evaluation model that uses fuzzy integrals to reect the bidding preferences of the contractor in question. The
model recommends an optimal bid price to be tendered by the contractor based on several bidding factors classied under environmental,
project and company conditions as determined by previous studies.
These models do have insightful implications for the management
side of a construction rm and do reveal important information about
bidding practices and patterns in different construction markets; however, they are rarely used by construction professionals in actual
decision-making processes. Traditionally, contractors have not been relying on any of the empirical or analytical bidding models presented in
the literature in estimating their bid markup values on tendered projects but they rather depend on their gut feeling, judgment and previous
experience [1,12,16,18,29]. Hence, the purpose behind this study is not
to offer one additional bidding model that serves the same objective as
the previous literature but rather to create a virtual experimental setting that mimics the construction bidding dynamics to a large extent
in order to observe and understand how the different identied bidding
attributes interact together and inuence each other. Specically, this
paper uses the benets of agent-based modeling and simulation to observe how do three of the most inuential bidding attributes
(contractor's risk attitude, need for work, and market competition)
make up the bid markup decision, affect the contractor's survival and nancial growth on the long-term, and shape the market patterns.
3. Methodology & description of experiments
3.1. Simulation model
Agent-based modeling (ABM) is relatively a new development that
has found extensive use in areas such as economics, transportation, sociology, biology, marketing, sales, and others [37]. ABM is a powerful
tool for modeling and analyzing complex environments where complexity emerges from a large number of unknown and known variables
and autonomous interacting decision makers. In such problems, game
theory and mathematical models, considered as prescriptive approaches in decision science, cannot help alone and need to be integrated with agent-based modeling. For further reading on integration of
game theory and mathematical modeling with agent-based modeling,
Unsal and Taylor [41] and Awwad et al. [3] are suggested. Also, a comprehensive review of game theory applications in construction management can be found in Asgari et al. [2].

11

Recently, ABM has been used in some construction management applications as well such as infrastructure management [35], dispute resolution [17], subcontractor selection [41], sustainable design and
management of the built environment [4,5,10,34], and competitive bidding [3,25,26]. ABM proves to be very helpful when it comes to modeling
a collection of independent decision-making entities as agents that sense
and stochastically respond to emerging conditions in their local environments. Each agent assesses its individual situation, analyzes it in relation
to the other agents, and then makes decisions based on a set of developed rules [20]. Hence, ABM is a promising tool to mimic the complex
bidding process where contractors are considered as agents having different bidding behavior and strategies, and competing over projects
with varying characteristics for the purpose of achieving growing prots
and survival. In this paper, the construction bidding environment is simulated using an agent-based approach where two classes of objects are
dened, namely Projects and Contractors, for the purpose of studying
the interaction between different markup components and their longterm inuence on contractors' performance. Projects are generated sequentially over a simulation period of ten years and are assigned a set
of characteristics such as the project budget chosen uniformly over the
range [80, 120] M $, the project duration also selected uniformly between 20 and 30 weeks, the project complexity (low or high), and its actual cost based on the inherent uncertainty. On the other hand, a set of
contractors is created in the market along with their attributes including
their attitude towards risk, expertise, cost estimation skills, bidding competitiveness, nancial status, work backlog and need for work, in addition to their decision making rules with respect to bidding on projects
and markup choice. Fig. 1 shows a schematic representation of the bidding process simulation and reects the interaction among the different
agents over time within one competitive environment. Several contractor agents are created in the market with different attributes that inuence their decision to bid or not on a certain project and their
subsequent selection of the optimal markup. For instance, a variable
called working capital is assigned to every contractor and may increase
or decrease depending on the accumulated prots or losses from won
projects. This variable is used to assess the nancial situation of each contractor in the market and thus affects directly his need for work, as indicated through the unidirectional link between the two variables in Fig. 1.
Besides, a contractor's resources and expertise dene the scope and
amount of work that can be performed at one point in time. Hence, if a
contracting rm has all its resources assigned to projects' tasks, then it
will not be in need for work. A third major attribute dened under the
contractor agent is its risk attitude which is typically the byproduct of
the rm's culture and leadership. The rm's risk attitude determines
the way it perceives the inherent risk and complexity in a project and
hence affects the corresponding risk contingency embedded in the rm's
bid price. The aforementioned internal attributes of each contractor inuence his markup choice in addition to other external factors such as
the market competition and the project complexity, as shown in the Contractor X box in Fig. 1. Looking at the big picture, the developed simulation imitates the construction market environment which consists of a
group of heterogeneous contractors competing over a set of projects
with different characteristics generated throughout the simulation
time. Each project has a certain complexity level (low or high) which affects the contingency component of the markup used by bidders on this
project as well as the project actual cost. The Contractor agent also includes among its attributes a markup decision function that serves to determine the optimal markup when bidding on a new project as will be
illustrated further in Section 3.2. This markup percentage is a summation
of three components: (1) market competition inuenced by the number
and competitiveness of opponents, (2) the contractor's need for work affected by his nancial situation and the amount of work at hand, and
(3) the risk allowance which is a result of internal and external characteristics, i.e., the contractor's risk tolerance and the project inherent uncertainty. It is worth noting that the agentenvironment boundary
represents the limit of the agent's absolute control, not of its knowledge.

12

S. Asgari et al. / Automation in Construction 65 (2016) 920

Fig. 1. An abstraction of a contractor in the bidding environment and the network structure of the market.

Readers can refer to the virtual bidding laboratory developed in [3] for
further details about the abstraction and denition phases of the simulation model used in this study.

3.2. Markup function

Using the agent-based approach to simulate the construction bidding mechanism focuses on the heterogeneity feature among competitors in their internal attributes such as their behavior towards risk,
nancial capacity, need for work, management expertise and others,
as well as their different perceptions of and reactions to external conditions like market competition, availability of projects, and project risk
level among others. A set of parameters, variables and functions for
the two dened agents, Contractors and Projects, are dened as illustrated in the previous section. Besides, behavioral rules of agents are
also dened in a simple and reasonable manner [31].
Every time a project is generated in the market, the present contractors will evaluate their interest and capability to bid on this project
through a function called Shallbid(). This function checks if the
contractor's specialty and size can accommodate the project type and
complexity, and if the contractor's bonding capacity and current
amount of work allows for taking one more project with the respective
budget. The second critical decision that a contracting rm has to make
after deciding to bid on a project is the optimal markup to use in its bid
price. Thus, another main function developed for the agent Contractor is
the markup decision function which describes how a contractor chooses
his markup value based on present external and internal circumstances.
A contractor's markup for a project is dened as the sum of contingencies and prot and is expressed as a percentage applied to the project's
estimated cost including all direct and overhead costs [30]. A slight increase in the markup value could lower the contractor's chance in winning the project and consequently affect his protability and survival in
the market. Since the main purpose of this study is to understand the interaction between different markup components and their inuence on
the contractors' performance, it was assumed that contractors can estimate a project cost with the same level of accuracy. This is to remove
any variability in bid prices and winning chances among bidders
resulting from cost estimation inaccuracies. It is also assumed that all
contractors have the same level of management capability and therefore uncertainties in the actual cost at which a project is completed result mainly from different levels of project complexities. Note that the

actual cost is set to follow a triangular distribution with a mean equal

to the project's market budget.
This paper assumes that a typical contractor considers three criteria
when trying to determine the markup for a certain project which are the
market competition, the inherent risk, and the need for work. The latter
components were identied in the literature among the most inuential
factors in a contractor's markup decision. DeNeufville and King [14] conducted an empirical study which showed that each of the two components, risk and need for work, causes independently and additively a
rough increase of 3% in the bid price of a contracting rm. Based on
the former study and to maintain simplicity in behavioral rules, an additive markup function was developed in this study consisting of three
components described in the following subsections:
3.2.1. Competition component
Contractors consider the market competition as their prime concern
when pricing their work [40]. Bidders should adjust their markups to reect the quality and quantity of competition displayed by their opponents [9]. Competitiveness of opponents is measured in this model
using the bid competitiveness ratio (BCR) expression introduced in
the study by Oo et al. [33] which determines the gap between an
opponent's bid and the lowest bid submitted for a project divided by
the lowest bid.
BCRip

Bip  Blp
Blp

where BCRip is the bid competitiveness ratio for opponent i on project p,

Bip is the bid price submitted by opponent i for project p, and Blp is the
lowest submitted bid for project p.
This ratio reects how close an opponent's bid was to the lowest bid
for a previous project and thus the smaller this ratio is, the more competitive the opponent is. During each bidding cycle, a contractor determines the average bid competitiveness ratios for his potential
opponents over the last ten projects tendered in the market using the
following code excerpt from the developed model:
//This piece of code shows how the BCRip of each contractor i is computed
and archived after determining the winning bid for every past project p//
for (Contractor c: get_Main().allContractors).
{

S. Asgari et al. / Automation in Construction 65 (2016) 920

c.bidCompetitivenessRatio = (c.bidPrice/winningPrice) 1;
c.bidCompetitivenessDataset.add(this.projectID,c.bid
CompetitivenessRatio);
}
//This piece of code shows how the average BCRip over the last ten projects, named bidCompetitivenessTen, is computed for each potential opponent i on the current project p//
for (Contractor c: Opponents).
{
sum = 0;
for(int n = 0; n b c.bidCompetitivenessDataset.size(); n++)
{sum + = c.bidCompetitivenessDataset.getY(n);}
c.bidCompetitivenessTen = sum/c.bid
CompetitivenessDataset.size();
}
Then, once the average bid competitiveness ratios of potential opponents are determined, the contractor will be able to compare and rank
himself among his competitors and hence evaluate the percentage of
opponents who have been more competitive than him in the recent
past projects. The lower this percentage is, the higher markup a contractor can afford to use in his bid price for the current project and thus the
higher the competition component will be in the developed additive
markup function.
3.2.2. Risk component
The risk attitude of a contracting rm is part of its organizational culture and it may vary depending on the internal conditions of the rm
and the external conditions of the market within which it operates.
The construction industry encompasses a lot of hazards and uncertainties, and thus, contractors have to consider all project complexities
when deciding to bid on a project and determining the bid price to tender. However, contractors try to limit inating their bid prices to keep
their competitive edge and rely on other ways to mitigate risk including
subcontracting large portions of their work, enhancing their cost estimation and management skills, tailoring contract conditions such that
risks are shared with clients, and relying on claims after winning the
project [27]. Also, some identied risks at the bid pricing stage such as
possible inaccuracies in quantities or price estimates and anticipated
operational difculties during project execution are typically accounted
for in the unit price estimates of the project pay items. But even after applying all former risk mitigation strategies, there remains some residual
risk that cannot be predicted or quantied and thus, has to be evaluated
and accounted for as a risk allowance in the bid price by the company's
decision makers. This type of risk is usually apportioned in a bid price as
part of the markup applied to the project cost and is dependent on a
combination of external factors such as market conditions and internal
factors such as the rm's circumstances. The range identied in the literature for residual risk allowance percentage in bids is in the order of
[03] % which was adopted in the presented model [14,27,40]. The
risk tolerance of a contracting rm affects the way it perceives the inherent risk in a project and the impact it can have on the rm's bid price.
This study considers three levels of contractors' risk aversion (mild,
moderate and extreme) and two degrees of project risk (low and
high) which will be the two main factors dening the risk contingency
component in the markup function as such:
if (riskAversion == mild)
{
if (project.riskLevel == low)
riskcontingency = 0.00;
if (project.riskLevel == high)
riskcontingency = 0.01;
}

13

else if (riskAversion == moderate)

{
if (project.riskLevel == low)
riskcontingency = 0.01;
if (project.riskLevel == high)
riskcontingency = 0.02;
}
else (riskAversion == extreme)
{
if (project.riskLevel == low)
riskcontingency = 0.02;
if (project.riskLevel == high)
riskcontingency = 0.03;
}
3.2.3. Need for work component
A contracting rm needs to maintain a certain work backlog in order
to cover its general and administrative costs and to be able to retain
skilled personnel. In the presented model, we dened the parameter,
Work-in-Progress Limit, which indicates the average annual workload
a contractor can have during a year given its size and capabilities. Consequently, the need for work ratio is determined through comparing the
current work volume of a contractor against its Work-In-Progress Limit.
The need for work ratio (NWR) for a contractor i is determined through
the following expression:

NWRi 1 


Current Work Volume
:
Work  in  Progress Limit

The closer this ratio is to 1, the higher the need for work and the
lower the markup is expected to be; and the closer this ratio is to 0,
the lower is the contractor's need for work and the higher markup it
can afford. Through surveying contractors in the market, DeNeufville
and King [14] showed that a low need for work could cause a rough decrease of 3% in the bid price of a contracting rm. The need for work
component in this study is selected within the range [04] % and its specic value is chosen based on two variables namely the nancial status
of the contractor and the need for work ratio dened in Eq. (2). It is important to note here that this component, in contrast with the two previously discussed components, is a negative value indicating a decrease
in the markup in order to enhance the winning chance of a contractor
and thus accommodate his need for work. The nancial situation of a
contractor is assessed through the positive or negative change to its initial working capital. Accordingly, there are six possible scenarios considered when determining the need for work component of the markup
value: a poor nancial situation with a low, moderate or high need for
work ratio, and a good nancial situation with low, moderate or high
need for work. The ideal situation is to have a contractor who is in a
good nancial status and with enough work at hand and thus can afford
a zero decrease in its markup. And the worst combination would be a
contractor who is in a bad nancial shape and who has minimal or
zero work at hand. In this case, a contracting rm would be willing to
apply a decrease up to 4% to its markup in order to enhance its chance
in winning the project and covering its overhead costs. The main focus
of this paper is to investigate the long-term impact of considering
need for work within a risky and competitive environment, and to observe to what extent the need for work should be considered.
To sum up, the main assumptions of the paper are the following:
All contractors have the same size, initial working capital (which is
zero), and general and administrative (G&A) costs.
All contractors can estimate a project cost with the same level of
accuracy.
All contractors have the same level of management capability and
expertise.

14

S. Asgari et al. / Automation in Construction 65 (2016) 920

The actual cost is set to follow a triangular distribution with a mean

equal to the project's market budget.
All contractors consider three criteria when trying to determine the
markup for a project: the market competition, the inherent risk, and
the need for work.
All contractors have access to their competitors' bidding history.
3.2.4. Example application
In this section, a numerical application is conducted to illustrate the
markup determination by a Contractor agent in the simulated bidding
model. We consider the case of Contractor C1 who is informed about a
newly adverstised project in the market for which he decides to submit
a bid after evaluating his bidding capacity using the Shallbid() function.
C1 expects to be competing on this project with three other contractors
in the market C2, C3, and C4 for which he computes the average bid
competitiveness ratio (BCR) based on the last ten tendered projects in
the market as shown in Table 1. Based on the former computations,
C1 ranked second among his three opponents in recent prior projects
and thus he has been more competitive than 2/3 of his opponents.
Assuming that C1 typically chooses his markup within a range of [5
9] %, the algorithm takes competition into effect through choosing a specic value from this range based on the competitiveness of C1 relative to
his opponents. In other words, if C1 has a lower BCR (i.e., is more competitive) than 80% of his opponents, he can afford to use the highest value in
his range, 9%, as his markup competition component. On the other hand,
if C1 has a BCR lower than only 20% of his opponents, he has to go down
to the lowest end of his range, 5%. Following the same logic, the markup
values between the two range boundaries are chosen in accordance with
20% increments of BCR values. As such, having a current BCR that is lower
than 2/3 of his opponents, C1 chooses 8% as the competition component
of his markup value. This value is now adjusted by C1 to account for his
need for work and the project risk. Assuming that C1 is an extremely risk
averse contractor and the current project he is bidding on incorporates
lot of hazards and uncertainties, the risk contingency that will be applied
by the developed algorithm is 3% which is the upper bound of the risk
component. Finally, the need for work component depends on C1's
need for work ratio (NWR) and his nancial capacity. C1 is currently
working on 4 projects and his Work-in-Progress Limit is 5, hence, his
NWR is equal to 0.2. This means that C1 has a minimal need for work
and thus can afford a high markup. Assuming a good nancial capacity
for C1 and given the need for work range of [4, 0] %, a 0% decrease is
chosen by the model as the need for work component. In summary, contractor C1 will use a markup of 11% (=8 + 3 0) on this project after
taking into account all internal and external conditions.

market and are set to have the same value for initial working capital,
Work-in-Progress Limit, and cost estimation and project management
skills. Each contractor can simultaneously work on four projects at most.
Given the number of contractors in the market and their work limit, the
market is considered very competitive in all sets of experiments since
most contractors participate in all biddings. Each experiment set has
been run for 100 times in order to ensure the consistency of the results.
Table 2 presents a summary of the experiments sets and their purposes.
The purpose of the rst set of experiments is to examine what level
of risk attitude results in the contractor's best nancial performance on
the long run. Results will be compared with ndings in the literature in
order to check and verify capability of the simulation model. There are
nine contractors competing with each other in the market including
three slightly risk averse contractors, three moderately risk averse contractors, and three extremely risk averse contractors.
Table 3 shows the risk attitude assigned to each of the nine contractors. In the rst set of experiments, projects are generated in the market
one at a time unit under three different scenarios. Under scenario 1, all
projects have a low level of complexity and uncertainty. Under scenario
2, the market comprises a mix of projects with low and high level of
complexity and uncertainty. In this case, every time a project is generated, its complexity is chosen randomly with a probability of 50% being
low and 50% being high. As for scenario 3, all projects have a high
level of complexity and uncertainty. These three scenarios are considered only in the rst set of experiments in order to analyze the impact
of project risk on the results.
The second set of simulation experiments investigates the importance of the component Need for Work. It aims at assessing whether
considering Need for Work impacts the business success of contractors on the longterm or not. In this experiment set, two types of contractors were considered with ten contractors in total. As shown in Table 3,
there are ve contractors who do not consider Need for Work and ve
other contractors who consider it and accordingly discount their markup up to 2%. Financial performance of the two sets of contractors is observed and compared in order to evaluate the impact of considering
Need for Work in markup selection on the contractor's performance.
As for the third set of experiments, they aimed at determining to what
degree a markup discount should be considered in order to account
for need for work without compromising the business success of the
contractor. Five types of contractors were considered in this experiment
with ten contractors in total. These ve sets differ by the discount level
they apply to their markup in order to take into consideration the Need
for Work component.

Table 2
Summary of the experiment sets 1, 2, and 3.

3.3. Description of the experiments

This subsection explains the details of three sets of simulation experiments conducted in this study. A set of contractors are dened in the

Experiment Purpose
set

Experiment conditions

Finding the optimal level of risk

attitude for contractors

Finding whether considering

Need for Work impacts business
success of contractors

Finding to what degree a markup

discount should be considered to
account for need for work

Scenario 1: the market

comprises only projects with
low level of complexity
Scenario 2: the market
comprises a mix of projects
with low and high level of
complexity
Scenario 3: the market
comprises only projects with
high level of complexity
The market comprises a mix of
projects with low and high
level of complexity.
All contractors are moderately
risk averse
The market comprises a mix of
projects with low and high
level of complexity.
All contractors are moderately
risk averse.

Table 1
Bid prices of opponents on last ten tendered projects.
Project

C1 bid
price

C2 bid
price

C3 bid
price

(M $)

(M $)

(M $)

P1
85
92
78
P2
95
105
102
P3
112
107

P4
75
80
82
P5
83
80
90
P6
120
110
125
P7

88
83
P8
105
110
123
P9
90
78
83
P10
115
95
110
Average bid competitiveness ratio:

C4 bid
price

BCR
for C1

BCR
for C2

BCR
for C3

BCR
for C4

0.089
0
0.047
0
0.037
0.091

0
0.154
0.211
0.069

0.179
0.105
0
0.066
0
0
0.060
0.048
0
0
0.046

0
0.074

0.093
0.125
0.136
0
0.171
0.064
0.158
0.091

0.158
0.121
0.2
0.075
0.027
0.084
0.143
0.089
0.105
0.111

(M $)

110
120
90
86
113
90
120
85
105

S. Asgari et al. / Automation in Construction 65 (2016) 920

15

Table 3
Contractors' characteristics in the experiment sets 1, 2, and 3.
Contractor

Risk attitude in the experiment set 1

Need for work in the experiment set 2

Need for work in the experiment set 3

#1
#2
#3
#4
#5
#6
#7
#8
#9
#10

Slightly risk averse

Slightly risk averse
Slightly risk averse
Moderately risk averse
Moderately risk averse
Moderately risk averse
Extremely risk averse
Extremely risk averse
Extremely risk averse
N.A.

Ignored
Ignored
Ignored
Ignored
Ignored
Discounted up to 2%
Discounted up to 2%
Discounted up to 2%
Discounted up to 2%
Discounted up to 2%

Ignored
Ignored
Discounted up to 1%
Discounted up to 1%
Discounted up to 2%
Discounted up to 2%
Discounted up to 3%
Discounted up to 3%
Discounted up to 4%
Discounted up to 4%

Table 3 shows the varying levels of Need for Work consideration in

the markup function of the ten dened contractors. It is worth mentioning that the second and third sets of experiments were conducted under
three levels of risk attitude separately in order to ensure that results are
consistent regardless of the risk behavior of contractors. Also, the market consisted of a mix of low and high risk projects (similar to the scenario 2 under the rst set of experiments) in both sets.
4. Results & discussion
This section presents and discusses the results that were obtained in
the three sets of experiments described in Section 3.3.
4.1. Experiment set 1
Figs. 2 to 5 as well as Table 4 and Table 5 present the results obtained
for the rst experiment set. In particular, Figs. 2 to 4 show the progress
of the average working capital for the three levels of risk aversion versus
the project ID under the three different described scenarios which exhibit varying degrees of project risk. As shown in each of these gures,
moderately risk averse contractors nancially outperform others in
highly competitive markets in the long run. This result is consistent
across all scenarios and is aligned with results from the literature [3,
25,26]. Also, slightly risk averse contractors outperform extremely risk
averse contractors in all scenarios. One of the methods for validating
simulation models is comparing obtained results with real-world observations or ndings from the literature and in this case, we can consider
this experiment set as a validation tool for the simulation model.
Comparing Figs. 2, 3, and 4, it is observed that slightly risk averse
contractors can do better in riskier markets. Slightly risk averse, on average, have the highest difference in working capital with moderately
risk averse contractors when the market comprises low-risk projects
(Scenario 1). This gap decreases when moving from low risk to high
risk market; namely, from scenario 1 to scenario 2 and then scenario
3. This can be due to the fact that when facing high uncertainty in projects, the more risk averse a contractor is, the higher the allocated

contingency (risk allowance) in his bid which reduces his competitiveness. This gives a winning edge to slightly risk averse contractors over
moderate ones, and thus the gap between the growing working capitals
for both decreases from scenario 1 to scenario 3. Another observation
about Figs. 2, 3, and 4 is that, given a certain risk aversion level, the generated working capital representing an accumulation of actual project
prots over time has increased from low to high risk market since all
contractors are expected to allocate a higher risk allowance in their
markup to hedge against high level of risk in projects.
Table 4 presents the working capital of contractors in the three scenarios. It also compares each contractor's performance with the best
performance in the market in terms of working capital. It can be observed that average working capitals of contractors have increased
from scenarios 1 to 3. In other words, in riskier markets, contractors increased their markups to cover higher risks and thus were able to earn
higher prots. It is also observed that moderate risk averse contractors
had the smallest average gap (%) with the best performer in the market in all scenarios which again indicates that moderation in risk attitude is the optimal strategy regardless of the project risk level. Note
that contractors #5 and #6 exhibited the best nancial performance in
all three scenarios ( = 0%). Another important observation that was
emphasized earlier is that slightly risk averse contractors performed signicantly better under scenario 3 where the average % dropped from
41% in scenario 2 to 24% in scenario 3.
Table presents the market share of the nine contractors under the
three different scenarios. The market share of each contractor i is dened as such:
Market share i

#of won projects by i

 100:
#of tendered projects in the market

This table supports again the conclusion that moderately risk averse
contractors, on average, have a better performance under all market
scenarios through having the highest share of projects among the
three types of contractors.

Fig. 2. Financial performance of contractors in the experiment set 1 under scenario 1.

16

S. Asgari et al. / Automation in Construction 65 (2016) 920

Fig. 3. Financial performance of contractors in the experiment set 1 under scenario 2.

Fig. 5 shows the success rate of contractors throughout the simulation period. The success rate of a contractor i is dened as per the following equation:
Success ratei

#of won projects by i

:
total#of projects i has bid on

This gure conrms that moderately risk averse contractors, on average, outperform other contractors. Moreover, there exists a convergence to a somehow constant success rate for each contractor after 2
3 years (in the simulation time scale). It is worth noting that the result
in Fig. 5 is obtained from several simulation runs of experiment 1
under scenario 2.
4.2. Experiment set 2
The results of the second set of experiments, as reected in Fig. 6,
show that contractors #6, 7, 8, 9 and 10 who considered Need For
Work as one of the criteria for determining their markup have a better
nancial performance in comparison with the contractors who did not
(#1, 2, 3, 4, and 5). This could be explained by the fact that these contractors lower their markup strategically depending on their need for
work and nancial situation which allow them to acquire new projects
when they have their resources idle, despite that it might be at the expense of lower expected prots. Specically, this exibility is of more
help to contractors when risk level of projects is low and market competition is intense. It is worth noting that this experiment set is conducted
with the same risk aversion degree for all contractors in order to isolate
the Need for Work effect on contractors' growing capital. Fig. 6 shows
the results obtained for moderate risk aversion level. However, the simulation was repeated another two times, once with mild risk aversion
condition for all contractors and the other with extreme risk attitude.
Both scenarios showed similar results.

4.3. Experiment set 3

While considering Need for Work is a reasonable strategy for contractors, the extent to which this consideration should be accounted for
in the markup percentage has not been studied in the literature. The experiment set presented in this section addresses this gap through having ve different groups of contractors in the market who perceive the
importance of Need for Work consideration differently and thus allocate different boundaries to this criterion in their markup functions.
Fig. 7 shows the experiment results where the growing capitals for the
ve groups of contractors are presented versus the project Id. It is observed that contractors 1 and 2 who do not take their need for work
into consideration in their markup decisions performed the worst,
whereas, contractors 7 and 8 who adopted a need for work upper margin of 3% had the best nancial performance. Contractors 5 and 6 came
in the second rank with using an upper limit of 2%. As for contractors 9
and 10 who discounted their markup up to 4%, they increased their
chance of winning the project while decreasing the prot margin radically. On the other hand, contractors 3 and 4 who discounted up to 1%
were most probably not able to immediately secure a contract when
they need it. Based on the aforementioned, it can be concluded that considering Need for Work strategically and discounting the markup up
to 23% is the optimal policy for contractors in a competitive market
on the long run. All contractors whose working capitals are shown in
Fig. 7 were assigned a moderate risk aversion degree. It should be
noted again that this experiment was conducted under different contractors' risk behavior (slightly and extremely risk averse) and the results were consistent regardless of the risk attitude.
4.4. Summary of research ndings
The aforementioned conducted experiments and scenarios revealed
several ndings which can be of signicant use to contractors, their

Fig. 4. Financial performance of contractors in the experiment set 1 under scenario 3.

S. Asgari et al. / Automation in Construction 65 (2016) 920

17

Fig. 5. Contractors' success rate under scenario 2.

understanding of the construction market and their shaping of their

bidding strategies. Disregarding the need for work component, it was
shown that moderately risk averse contractors consistently outperform
other contractors in all market conditions including all levels of project
risk whereas extreme risk aversion is the least effective strategy under
all circumstances. Another interesting and not so intuitive observation
is that the higher the projects' risk and uncertainty within the market,
the more competitive slightly risk averse contractors are where they
tend to exhibit a close nancial performance to moderately risk averse
contractors. This implies that in a riskier market, a contractor can afford
to be more risk tolerant and still maintain a protable business.
Furthermore, the study showed that a contractor who takes into
consideration his need for work in his markup choice has a signicantly
better nancial performance on the long term than one who does not.

Based on the simulations studying different consideration levels for

the need for work component, it was observed that 2 to 3% markup decrease accounting for the contractor's need for work led to higher
contractor's chance of nancial growth. A higher discount may allow
the contractor to win more projects but will have a detrimental effect
on his nances. Therefore, the optimal policy can be concluded to be
moderation in both dimensions of risk attitude and need for work. At
the limit, not considering need for work and being extremely risk averse
appeared to be the least effective strategy a contractor can adopt. Thus, a
contractor needs to be cautious when discounting his markup to account for his need for work in such a way to create a balance between
earning enough prot and securing the needed amount of projects. Finally, the results obtained from all sets of experiments converged towards the same observation of a market equilibrium reached under

Table 4
Contractors' working capital in the experiment set 1 under scenarios 1, 2, and 3.
Working capital ($M)

Contractor 1
Contractor 2
Contractor 3
Average of Contractors 1, 2, 3
Contractor 4
Contractor 5
Contractor 6
Average of Contractors 4, 5, 6
Contractor 7
Contractor 8
Contractor 9
Average of Contractors 7, 8, 9

Scenario 1

% with the best performance

Scenario 2

% with the best performance

Scenario 3

% with the best performance

237
90
155
161
260
318
269
283
163
167
112
147

25%
72%
51%
49%
18%
0%
15%
11%
49%
47%
65%
54%

170
292
198
220
346
271
372
330
207
158
194
186

54%
22%
47%
41%
7%
27%
0%
11%
44%
58%
48%
50%

294
299
345
313
321
382
408
371
202
234
224
220

28%
27%
16%
24%
21%
6%
0%
9%
50%
43%
45%
46%

The bold values are basically the average of the three above values.
The underlined value shows the highest value in a scenario.

18

S. Asgari et al. / Automation in Construction 65 (2016) 920

Table 5
Contractors' market share in the experiment set 1 under scenarios 1, 2, and 3.
Market share

Contractor 1
Contractor 2
Contractor 3
Average of Contractor 1, 2, 3
Contractor 4
Contractor 5
Contractor 6
Average of Contractor 4, 5, 6
Contractor 7
Contractor 8
Contractor 9
Average of Contractor 7, 8, 9

Scenario 1

Scenario 2

Scenario 3

12.3%
9.0%
10.9%
10.7%
12.9%
13.1%
12.9%
13.0%
9.6%
10.2%
9.0%
9.6%

10.6%
11.9%
10.9%
11.1%
12.3%
12.1%
13.4%
12.6%
9.8%
9.2%
9.6%
9.5%

10.9%
10.9%
12.3%
11.4%
11.9%
13.1%
12.3%
12.4%
9.6%
9.4%
9.4%
9.5%

The bold values are basically the average of their three above values.

competitive bidding where all contractors have gained enough information about their competitors and adjusted their strategies accordingly.
4.5. Verication & validation
Verication is the process of determining whether the programming
implementation of the abstract model is correct whereas validation is
the process of determining whether the conceptual model is a reasonably true representation of the real world for the purpose of answering
the research questions [28]. In other words, verication is concerned
with solving the problem right while validation is concerned with solving the right problem [44]. There are numerous methods with different
levels of rigor for verication and validation of simulation models, particularly agent-based models. Depending on the type of the model and

available data, a method may be applicable for verication and validation; a model should be veried and validated to the degree needed
for the model's intended purpose or application [36]. The agent-based
simulation platform used in this study, AnyLogic, allows the user to
breakdown the model into several computation steps and verify the
programming component of each step due to its capability of collecting
information on any parameter or process at any time through the simulation. For some specic bidding cycles in several simulation runs, the
corresponding calculations of all the process steps were computed manually, compared and veried with the model calculation.
One of the most prevalent validation methods is model-to-model
comparison where the results of the simulation are compared with previous studies on the subject. As mentioned previously, results of the rst
set of experiments were aligned with the literature [3,25,26]; moderately
risk averse contractors outperform other contractors in competitive markets given any level of project risk. Kim and Reinschmidt [25,26], using an
evolutionary model in which contractors choose their markup according
to their risk attitude and project uncertainty, showed that moderately
risk averse contractors are more successful than the others. Also,
Awwad et al. [3] reached the same conclusion. They developed an
agent-based environment where contractors use a mathematical model
that integrates utility theory and Friedman's model (1956) to determine
an optimal markup according to contractors' risk attitude and the market
competition. Another validation method used in this study was parameter variability (sensitivity analysis) where we closely examined how uncertainty in the values of the main input parameters can impact the
model output. In addition, different distributions were used for project
budget, estimated duration, actual cost, and actual duration in addition
to the fact that all experiments were conducted under different scenarios
in order to make sure results are consistent. For example, experiment sets
2 and 3 were conducted with three different contractors' risk attitudes.

Fig. 6. Working capital of moderately risk averse contractors in the experiment set 2.

S. Asgari et al. / Automation in Construction 65 (2016) 920

19

Fig. 7. Working capital of moderately risk averse contractors in the experiment set 3.

The simulation model developed in this study, like any other simulation
model, is based on some assumptions. One important assumption was
that contractors have access to other contractors' bidding history. With
respect to validation of model assumptions, it is worth mentioning that
many contractors have a specic unit in their business development, typically a Research & Development department, that is in charge of
collecting market information, tracking and analyzing their competitors.
On the other hand, it was assumed that all contractors have the same
size, initial working capital, cost estimation accuracy, construction management skills, expertise, and G&A costs in order to avoid any possible impact of these factors on the results. Concerning validation of the markup
determination, the selection of these three additive and independent
criteria is supported by a thorough literature review presented in the related work and methodology sections of the paper. As for the structural
validation of the simulation, the bidding process used in the simulation
model imitates the actual bidding process happening in reality.
5. Conclusion and future works
Markup decision can massively impact a contractor's nancial performance because the primary channel of securing work in the construction industry is still through a variety of competitive bidding
mechanisms. The presented study used agent-based modeling as an appropriate framework for analyzing and investigating the impact of risk
behavior and need for work on construction contractors' performance
in a competitive bidding environment. Prior to applying agent-based
modeling in this study, the authors had rst to conduct an extensive literature review in order to build a rule-based and descriptive markup
decision model that replicates the behavior of contractors. The developed model built on previous analytical and empirical studies about
contractors' bidding parameters and markup considerations and combined all these attributes and functions into a comprehensive agentbased model that reects many of the complexities of the bidding process. Different combinations of risk behavior, need for work and market
competition were simulated and their implications on the contractor's
business and the market as a whole were observed and analyzed. This
model can serve as a virtual simulation lab that can be used by any potential user (contractor, consultant or owner) to evaluate and compare
different bidding strategies, markup behaviors, and bid tendering approaches and to learn about resulting emergent market patterns. In particular, this study provided contractors with invaluable insights into the
impact of considering need for work on their nancial performance and
to what extent this consideration should be taken into account. It was
shown that moderation in both dimensions of risk allowance and

need for work is the optimal bidding strategy for contractors in competitive environments. At the limit, this model can be used by a contractor
to compare different markup functions that he may wish to adopt in
order to evaluate which would most benet his business on the long
run.
The authors suggest that future works can focus on adaptive risk attitude in two main directions. The rst step is to study contractors' organizational culture and risk behavior in order to nd out whether and when
contractors change their risk attitude. This would help identifying what
information and signals contractors look for in the market or their organization in order to act more or less risk averse. Second, new methodologies
such as reinforcement learning can be used to help contractors design the
optimal dynamic strategy according to their organization, competition,
and market conditions. This strategy can be later experimented and compared against other strategies in the virtual lab designed and developed
by the authors using agent based modeling. Furthermore, some of the assumptions used in this paper can be enhanced in the future version of the
model. For instance, the number of contractors in the market was kept
xed in this study and there was no newcomer or quitter from the market. Also, the size of a contracting rm was considered constant throughout the simulation while it is not always the case in reality. In future
studies, new behaviors for contractors such as entry, exit, expansion, contraction, alliance, and merging can be dened and added to the model if
aligned with the purpose of the research. As we tried to apply a variety
of validation methods, validation of the model through expert judgment
is still crucial [6]. As part of future work, the simulation model can be validated by comparing its results with real case studies or by presenting it to
and getting feedback from practitioners and professionals in the eld of
construction bidding.

Acknowledgments
The second author wishes to thank the Council for International Exchange of Scholars (CIES) and the Fullbright commission for funding her
visiting scholar program at Purdue University which allowed this collaborative work.

References
[1] I. Ahmad, I. Minkarah, Questionnaire survey on bidding in construction. J. Manag.
Eng., ASCE 4 (3) (1988) 229243.
[2] S. Asgari, A. Afshar, K. Madani, Cooperative game theoretic framework for joint resource management in construction, J. Constr. Eng. Manag. 140 (3) (2014)
(04013066).

20

S. Asgari et al. / Automation in Construction 65 (2016) 920

[3] R. Awwad, S. Asgari, A. Kandil, Developing a virtual laboratory for construction bidding environment using agent-based modeling, J. Comput. Civ. Eng., ASCE (2014)
(04014105).
[4] E. Azar, C. Menassa, Agent-based modeling of occupants and their impact on energy
use in commercial buildings, J. Comput. Civ. Eng. 26 (4) (2012) 506518.
[5] E. Azar, C. Menassa, A framework to evaluate energy saving potential from occupancy interventions in typical US commercial buildings, J. Comput. Civ. Eng. 28 (1)
(2014) 6378.
[6] E. Bonabeau, Agent-based modeling: methods and techniques for simulating human
systems, Proc. Natl. Acad. Sci. U. S. A. 99 (Suppl. 3) (2002) 72807287.
[7] P.D. Boughton, The competitive bidding process: beyond probability models, Ind.
Mark. Manag. 16 (1987) 8794.
[8] R.I. Carr, General bidding model, J. Constr. Div., ASCE 108 (4) (1982) 639650.
[9] R.I. Carr, Impact of number of bidders on competition, J. Constr. Eng. Manag., ASCE
109 (1) (1983) 6173.
[10] J. Chen, R.K. Jain, J.E. Taylor, Block conguration modeling: a novel simulation model
to emulate building occupant peer networks and their impact on building energy
consumption, Appl. Energy 105 (2013) 358368.
[11] S.E. Christodoulou, Optimum bid markup calculation using neurofuzzy systems and
multidimensional risk analysis algorithm, J. Comput. Civ. Eng., ASCE 18 (4) (2004)
322330.
[12] D.K.H. Chua, D. Li, Key factors in bid reasoning model, J. Constr. Eng. Manag., ASCE
126 (5) (2000) 349357.
[13] R. DeNeufville, E.N. Hani, Y. Lesage, Bidding models: effects of bidders' risk aversion,
J. Constr. Div. 103 (CO1) (1977) 5770.
[14] R. DeNeufville, D. King, Risk and need for work premiums in contractor bidding, J.
Constr. Eng. Manag., ASCE 117 (4) (1991) 659673.
[15] S.P. Dozzi, S.M. AbouRizk, S.L. Schroeder, Utility-theory model for bid markup decisions, J. Constr. Eng. Manag., ASCE 122 (2) (1996) 119124.
[16] M.F. Dulaimi, H.G. Shan, The factors inuencing bid mark-up decisions of large and
medium size contractors in Singapore, Constr. Manag. Econ. 20 (7) (2002) 601610.
[17] I. El-Adaway, A. Kandil, Multiagent system for construction dispute resolution
(MAS-COR), J. Constr. Eng. Manag. 136 (3) (2010) 303315.
[18] A. Fayek, Competitive bidding strategy model and software system for bid preparation, J. Constr. Eng. Manag. 124 (1) (1998) 110.
[19] L. Friedman, A competitive bidding strategy, Oper. Res. 4 (1) (1956) 104112.
[20] R. Garcia, Uses of agent-based modeling in innovation/new product development
research, J. Prod. Innov. Manag. 22 (5) (2005) 380398.
[21] M. Gates, Bidding strategies and probabilities, J. Constr. Div., ASCE 93 (1) (1967)
75107.
[22] T. Hegazy, O. Moselhi, Analogy-based solution to markup estimation problem, J.
Comput. Civ. Eng. 8 (1) (1994) 7287.
[23] S.P. Ho, Y. Hsu, Bid compensation theory and strategies for projects with heterogeneous bidders: a game theoretic analysis, J. Manag. Eng. 30 (5) (2014) (04014022).
[24] C.W. Ibbs, K.C. Crandall, Construction risk: multi-attribute approach, J. Constr. Div.,
ASCE 108 (2) (1982) 187200.

[25] H.J. Kim, K.F. Reinschmidt, Effects of contractors' risk attitude on competition in construction, J. Constr. Eng. Manag., ASCE 137 (4) (2011) 275283.
[26] H.J. Kim, K.F. Reinschmidt, Association of risk attitude with market diversication in
the construction business, J. Manag. Eng., ASCE 27 (2) (2011) 6674.
[27] S. Laryea, W. Hughes, Risk and price in the bidding process of contractors, J. Constr.
Eng. Manag., ASCE 137 (4) (2011) 248258.
[28] A.M. Law, W.D. Kelton, W.D. Kelton, Simulation Modeling and Analysis, 2McGrawHill, New York, 1991.
[29] A. Lesniak, E. Plebankiewicz, Modeling the decision-making process concerning participation in construction bidding, J. Manag. Eng. (2014) (04014032).
[30] M. Liu, Y.Y. Ling, Modeling a contractor's markup estimation, J. Constr. Eng. Manag.,
ASCE 131 (4) (2005) 391399.
[31] S. Metcalfe, J. Foster, Evolution and economic complexity, Edward Elgar Publishing,
Northampton, MA, 2004.
[32] B.L. Oo, D.S. Drew, H.P. Lo, Heterogeneous approach to modeling contractors'
decision-to-bid strategies, J. Constr. Eng. Manag., ASCE 134 (10) (2008) 766775.
[33] B.L. Oo, D.S. Drew, H.P. Lo, Modeling the heterogeneity in contractors' mark-up behavior, J. Constr. Eng. Manag., ASCE 136 (7) (2010) 720729.
[34] H. Said, A. Kandil, S. Nookala, H. Cai, M. El-Gafy, A. Senouci, H. Al-Derham, Modeling
of the sustainability goal and objective setting process in the predesign phase of
green institutional building projects, J. Archit. Eng. 20 (2) (2014) (04013007).
[35] K.L. Sanford Bernhardt, S. McNeil, Agent-Based Modeling: Approach for Improving
Infrastructure Management, J. Infrastruct. Syst. 14 (3) (2008) 253261.
[36] R.G. Sargent, Verication and validation of simulation models, J. Simul. 7 (1) (2013)
1224.
[37] A. Sawhney, H. Bashford, K. Walsh, A.R. Mulky, Agent-based modeling and simulation
in construction, Proc., 2003 Winter Simulation Conference,IEEE 2003, pp. 15411547.
[38] J. Seydel, D.L. Olson, Bids considering multiple criteria, J. Constr. Eng. Manag., ASCE
116 (4) (1990) 609623.
[39] A.A. Shash, Factors considered in tendering decisions by top UK contractors, Constr.
Manag. Econ. 11 (2) (1993) 111118.
[40] G.R. Smith, M.C. Bohn, Small to medium contractor contingency and assumption of
risk, J. Constr. Eng. Manag. 125 (2) (1999) 101108.
[41] H. Unsal, J. Taylor, Modeling interrm dependency: game theoretic simulation to examine the holdup problem in project networks, J. Constr. Eng. Manag. 137 (4)
(2011) 284293.
[42] W.C. Wang, R.J. Dzeng, Lu Y.H., Integration of simulation-based cost model and
multi-criteria evaluation model for bid price decisions, Comput. Aided Civ.
Infrastruct. Eng. 22 (3) (2007) 223235.
[43] J. Willenbrock, Utility function determination for bidding models, J. Constr. Div.,
ASCE 99 (1) (1973) 133153.
[44] X. Xiang, R. Kennedy, G. Madey, S. Cabaniss, Verication and validation of agentbased scientic simulation models, Agent-Directed Simulation Conference 2005,
pp. 4755.
[45] K. Ye, B. Li, L. Shen, Key factors considered in compiling tender prices for China's
public works projects, J. Manag. Eng. 29 (3) (2013) 206215.