International Journal of Scientific Engineering and Research (IJSER)

ISSN (Online): 2347-3878

Impact Factor (2018): 5.426

Deterministic and Probabilistic Engineering Cost

Estimating Approaches for Complex Urban Drainage
Infrastructure Capital Improvement (CIP) Programs
Thewodros K. Geberemariam
P. O. Box 23195 Brooklyn, New York 11202, USA
tgm[at]nyu.edu

Abstract: Accurate and reliable project cost estimates are fundamental to achieve successful municipal capital improvement (CIP)
programs. Engineering cost estimates typically represent critical information for key decision makers to authorize and efficiently allocate
the necessary funds for construction, budgeting, to generate a request for proposals, contract negotiations, scheduling, etc. for these
reasons, cost estimators are using different estimating methods and approaches that allow for required levels of accuracy. As the project’s
scope becomes more detailed and the potential risks are identified and/or the project design stage progresses these cost estimates are
revised and updated. In this paper, the most common project cost estimation methods and approaches were collected and categorized into
two main groups of (1) probabilistic and (2) deterministic methods. Under these groups overall ten different methods were identified and
discussed addressing their requirements, advantages, and shortcomings, including the potential risk that can positively or negatively affect
the project’s cost outcome. This paper will be a good resource for professionals who are in budget development and/or are seeking to a
better understanding of different methods in determining an appropriate base cost margin and produce a meaningful and reliable project
cost estimate.

Keywords: Risk Management; Deterministic; Probabilistic; Engineering Cost Estimating; Uncertainty; Cost Estimating Methods; Urban
Drainage Infrastructure; Capital Improvement (CIP) Programs

1. Introduction available or can be estimated. However, the probabilistic

approach used best fit probability distributions to model the
The conventional deterministic cost estimation methods for uncertainties and risk in the cost estimate. The main
capital improvement projects in most municipal agencies and advantages of the probabilistic cost estimating approaches are
the local governments are based on preparing a its ability to provide insight in the accuracy of the estimate
single-point-estimates. A single-point-base-estimate is based and the impact of uncertainties and risks of cost overruns will
on typically on the level of a project’s scope definition and be known [34, 65-67].
the project design phase, available historical data, current
contractor rates and preliminary quotes from sub-contractors The aim of this paper is to provide a resource for
and other vendors [7, 34]. Moreover, to adjust for inflation professionals who are involved in budget development and/or
costs of labor, material, and equipment additional Consumer are seeking to a better understanding of different methods in
Price Index (CPI) is added to each cost item every year. This determining an appropriate base cost margin to produce a
poses a challenge on the accuracy of the project cost estimate meaningful and reliable project cost estimate. It highlights the
and/or budget(s) and may cause cost overruns [11, 50]. various methods of engineering cost estimating approaches
Accurately estimating the costs of complex infrastructure advantages, and shortcomings, including the potential risk
projects in the design, and construction phases have typically that can positively or negatively affect the project’s cost
become a unique challenge for engineers, architects, owners, outcome.
municipal agencies, and contractors [8]. Complex and
technologically advanced projects are usually contained However, it should be recognized that any technical books,
much uncertainty and related challenges than other projects. municipalities’ guidelines, etc. cannot provide entirely
Therefore, engineering cost estimates must adequately complete and practical engineering cost estimation guidance
address uncertainty at the preliminary stages of projects applicable to all capital improvement projects and cannot
where neither the exact quantities nor specific costs or replace experience and sound judgment. As such, deviations
ultimate prices are known. However, dealing with risks and from those books or guidelines may, sometimes, be
uncertainties are usually a problem [11, 54]. unavoidable or otherwise justifiable.

The sources of risks and uncertainties in a project are several. 2. Accuracy of Cost Estimates
At the early stage of the project, the uncertainty in a cost
estimate increases due to the available information quantity The overall purpose of an accurate cost estimate is its use in
and quality. As the design progress, more and better establishing the budget for a project and as a tool used for
information becomes available; the uncertainty in the cost scheduling and monitoring and controlling of the project cost.
estimate is gradually reduced [38, 42-43, 49].In the The level of accuracy of engineering cost estimates increases
deterministic approach, information about uncertainties and as the project phase progresses and the potential risks are
their characteristics such as higher or lower values, ranges of identified. The earlier the estimate in the life of the project the
quantities, and potential costs cannot easily be taken into lower its accuracy consequently, assessments of conceptual
consideration even though this information is generally estimate accuracy are quite low [2, 10, 19, 45]. Figure1
Volume 8 Issue 10, October 2020
www.ijser.in
Licensed Under Creative Commons Attribution CC BY
Paper ID: 18092001 1 of 13
 International Journal of Scientific Engineering and Research (IJSER)
ISSN (Online): 2347-3878
Impact Factor (2018): 5.426
below shows the Characteristic curve of accuracy vs. time to
make estimates.

Figure 1: Characteristic Curve of Accuracy vs. Time to

Figure 2: Classifications of Cost Estimation Methods
Make Estimates
Generally, in the deterministic approach, information about
The target cost estimate accuracy set calculated from uncertainties and their characteristics such as higher or lower
programmatic data, prior to design generally assumed to be values, ranges of quantities, and potential costs cannot easily
around +/- 30%. However, Experts assert that this variance be taken into consideration even though this information is
allows conceptual estimates to be useful for determining generally available or can be estimated [1, 6, 10, 41, 46, 48].
feasibility but not for establishing a control budget. There are However, the probabilistic approach used best-fit probability
a different factor that affects the accuracy of conceptual distributions to model the uncertainties and risk in the cost
engineering cost estimates, such as basic process design estimate [4, 12, 28, 38]. The main advantage of the
(23.2%), team experience and cost information (13.3%), time probabilistic cost estimating approach is its ability to provide
to estimate (12.1%), site requirements (11.5%), and bidding insight into the accuracy of the estimate and the impact of
and labor climate (10.2%). In general, in schematic and/or uncertainties and risks of cost overruns [18, 26, 42, 54, 56].
preliminary stage (order-of-magnitude) cost estimates
accuracy are in between ±20% of actual costs and in detailed The fundamental difference between these two-cost
estimates are in range of ±5% of actual costs [9, 14, 21, 24, estimation approaches (probabilistic and deterministic) is that
36]. by using the probabilistic cost estimation approach, we are
enabling explicitly model the uncertainties and risk
associated with it using appropriate statistical distributions
3. Classifications of Cost Estimation Methods [39, 59, 62, 66]. Moreover, the selection of these cost
estimation techniques will be adopted for different projects,
3.1. Deterministic and Probabilistic Cost Estimating depending upon the aims and objectives sought by the client
Methods [28].
There are several different deterministic methods of 3.1.1. Deterministic Cost Estimating
preparing a cost estimate depending on the purpose, the level
of planning, and/or design, as well as the project type, size, Under this category, Parametric, Detailed, Comparative,
complexity, circumstances, schedule, and location. In (Unit, cost, and Power law and sizing method), and Factored
general, regardless of whether the project technical scope is Estimates methods have been discussed below:
traditional (capital funded, construction, equipment
purchases, etc.) or nontraditional (expense funded, research I. Parametric Cost Estimating (Top-Down Estimating)
and development, operations, etc.). The levels of
requirements and techniques used are the common This method is generally used during the earliest stage of the
characteristics of most project cost estimates [45, 56, 60-61, project [48]. However, it also can be used to establish a
67]. These includes (1) Status of Project life cycle, (2) the baseline at any stage, where the comparison or validation of
detail information available, (3) cost estimation methods other estimating methods are needed or estimation of the use
(e.g., parametric vs. definitive), and/or (4) Constraints and of resources required to perform for a new project [13, 29,
other estimating variables such as time [50]. Preparing cost 52-53]. This model has a mathematical representation of the
estimate also depending on the purpose, level of planning, cost estimating relationships (CERs) that able to predict and
and/or design, as well as the project type, size, complexity, provides a logical correlation between the physical and
circumstances, schedule, and location. These methods can functional characteristics of a project [5, 24].
fall into categories such as parametric, historical bid-based,
unit cost/quantity based, range, and probabilistic risk-based A particular cost or price can be established and estimated
estimates [15, 34, 43, 52-53]. Figure2 below shows, the two using cost estimating relationships (CERs) with an
major Classifications of Cost Estimation approaches namely independent variable. The cost estimating relationship
deterministic and probabilistic method. (CER), mathematical ratio or equation can be developed
using an independent variable that demonstrates a measurable
relationship between contract cost and price. It usually
derived from regression analysis of historical systems or
subsystems. Equations (1) and (2) are the associated linear

Volume 8 Issue 10, October 2020

www.ijser.in
Licensed Under Creative Commons Attribution CC BY
Paper ID: 18092001 2 of 13
 International Journal of Scientific Engineering and Research (IJSER)
ISSN (Online): 2347-3878
Impact Factor (2018): 5.426
and nonlinear form of cost estimation relationships (CER). manageable tasks, or when works breakdown structure is
These equations are called cost estimating relationships available [17, 18]. A work breakdown structure is used to
(CER's) framework. The CER uses quantitative techniques to divides project deliverables into a series of work packages
quantify a relationship between an independent variable and and each work package comprised of a series of tasks [22, 53,
contract cost or price [6, 46, 49]. 58]. During detailed cost estimate, the project teams of cost
estimators work with engineers, Architects etc. to complete
n
each itemized task, work packages, and develop the total
Tc = � PCR PI (1) detailed cost estimate for the entire proposed project. The cost
i estimator’s quantity estimates have to be validated by the
professional engineers to make sure this cost estimation
Where: 𝐓𝐓𝐂𝐂 = Total Cost; 𝐏𝐏𝐂𝐂𝐂𝐂 = parameter cost ratio; 𝐏𝐏𝐈𝐈 = process is leads to a consistent and reproducible result. The
parameter of an interest. general mathematical formula for detailed cost estimating
method is given by Equation-3. However, this method is
Equation (2) for CER with associated nonlinear form cost different for each project depends on complete work
estimation relationships breakdown structure availability, [40, 57].
n

Tc = � PCR P ni (2)
i
TC = � qi (Mi + Wi + Li ) + � Ij �UCj � (3)
Where: 𝐏𝐏𝐈𝐈 = parameter of independent variable of interest; i j
𝐧𝐧𝐢𝐢 = exponent used to transform𝐏𝐏𝐈𝐈
n= is the number of units (Suppose that a construction Where: 𝐓𝐓𝐂𝐂 = Total Cost; 𝐪𝐪𝐢𝐢 = quantity of work; 𝐌𝐌𝐢𝐢 = Unit
project is divided into “n” elements material cost; 𝐖𝐖𝐢𝐢 = Unit Wage rate; 𝐋𝐋𝐢𝐢 = Unit Labor Rate;𝐈𝐈𝐣𝐣
for cost estimation). = measure of work in indirect cost elements; 𝐔𝐔𝐔𝐔𝐣𝐣 = Unit cost
of in indirect elements.
The exponents ni (E.q-2) used to transform and normalize the
temporal effects of cost including an inflation, rapid increases Certainly, the detailed cost estimating is the most accurate
of material cost, and for an independent variable and other and provides insight into the major cost contributors, all cost
metrics. In general, Parametric cost estimating can also be components and make sure nothing can be overlooked [33,
incorporated with probabilistic estimating to form range 20]. However, it can also be time-consuming, and requires a
estimating that predict uncertainties and potential risks [34, lot of effort to establish especially in large and complex
39, 46, 52, 66]. Table-1 below shows the advantages and projects with numerous work breakdown structure
disadvantages of Parametric Cost Estimating Method. components [16, 22, 57]. Table-2 below shows the
advantages and disadvantages of Detailed Cost-Estimating
Table 1: Parametric Cost Estimating Method Method.
Advantage Disadvantage Requirement
Relatively quick Documentation of Cost Sufficient Table 2: Detailed Cost-Estimating Method
and accurate way Estimating Relationships historical data for Advantage Disadvantage Requirement
to estimate costs (CERs) can be difficult statistical analyses A greater level of More time needed Collaboration of
Reduced Database with confidence Very high to develop the the engineers that
likelihood of Traceability of CERs can historical data that accuracy possible estimate conduct the work
serious cost be difficult can be updated more costly to
overruns regularly All cost components develop than Work Breakdown
Reduced cost of are taken into account relationship Structure
Historical data must be
preparing project Database estimating
available
proposals Additional ‘sanity’
Multiple decision Improper use of CERs Nothing can be Historical data must
check or
options for can lead to serious overlooked be available
benchmark
project managers estimating errors Project’s scope
Model can be Parts of the estimates must be determined
used as basis for Maintenance of database can be reused and understood
uncertainty with historical data considerably
andrisk analyses Actual cost data of
Easy to combine Periodically updated to Confidence level
ongoing project can
with other capture the most current difficult to
be used as predictor
estimating cost, technical, and determine
for future
methods programmatic data.

III. Comparative Cost Estimating/Analogous Estimating

II. Detailed Cost Estimating /Bottom-up/ Analytical Method
Estimating Method
The comparative estimating method can be used to make a
The detailed cost estimating requires is the most accurate quick comparison when a new project is similar to another
estimating technique when, the project is decomposed into project recently completed. During this process the major
cost components that were used on previous similar projects
Volume 8 Issue 10, October 2020
www.ijser.in
Licensed Under Creative Commons Attribution CC BY
Paper ID: 18092001 3 of 13
 International Journal of Scientific Engineering and Research (IJSER)
ISSN (Online): 2347-3878
Impact Factor (2018): 5.426
and direct and recent experience is needed [35]. Adjustment estimate for the proposed project, especially when the
shall be made on the proposed cost estimate factoring the proposed project has minor deviations from an appropriate
differences in project size and complexity, performance comparative similar past project that has been completed[3,
requirements, duration, location and available technology [9, 15, 37].The shortcomings of this method are its dependent on
44]. This relation factors are not usually linear. Cost capacity a single data point, its requirement of normalization in order
factors and economies of scale are the main factors that to create baseline and ensure a good accuracy of the estimate,
determine the nonlinear form of cost estimation relationships and the difficulties of finding an appropriate comparative data
(CER) [3, 15, 31]. Commonly used technique for preliminary for similar past project and experts to make judgment to
design stage cost estimates are Unit Method, cost indexes, adjustment factors[25, 62-63].Table-3 below shows the
Cost-Capacity Equation or power law and sizing model, and advantages and disadvantages of Comparative Cost
Factored Estimates [39, 64]. The general mathematical Cost Estimating Method.
estimate equations are presented below.
Table 3: Comparative Cost Estimating Method
I. Unit Method Advantage Disadvantage Requirement
𝑛𝑛 Easy to generate Uncertainty due to
Requires analogous
Tc = � 𝑈𝑈 ∗ 𝑁𝑁 (4) and estimate, subjective
product and
provided historical evaluations made by
𝑖𝑖 program data.
data is available. estimator.
Where: 𝑻𝑻𝑪𝑪 = Total Cost; U= per unit cost; N= quantity of Provides better
Requires a detailed
credibility than Difficult to apply for
work, n= is the number of units (Suppose that a construction program and
plain detailed differences in scope
project is divided into “n” elements for cost estimation). estimating. Can be of work, design,
technical definition
of the analogous
used early in project configuration and
II. Cost Indexes system as well, as
even if scope of the number of aircraft or
the system being
project is not aircraft programs.
estimated.
Cost Index (CI) is the ratio of cost to date versus cost in the complete
past. The CI change in cost over time to account the impact of Quick and
Once the technical
inflation and it is dimensionless [63]. The general assessment has
reasonable accuracy Experience or data
mathematical formula used to calculate the total Cost identified the
for similar systems, of a
estimate is: analogous system,
or end items. relevant
actual cost data on
Estimate is easy to comparative project
that system must be
𝐼𝐼𝑡𝑡 understand
𝑇𝑇𝐶𝐶 = � 𝐶𝐶0 � � obtained.
𝐼𝐼0 (5)
Accuracy is limited,
𝑡𝑡 Good accuracy for
Cost impacting
similar systems if
factors have to be
comparative and Comparison factors
determined, and
Where: 𝑻𝑻𝑪𝑪 = Estimated total cost of present time; 𝑪𝑪𝟎𝟎 = Cost recent data is
Normalization
at previous time; 𝑰𝑰𝒕𝒕 = Index value at timet; 𝑰𝑰𝟎𝟎 = Index value available
required
at base time 0.
IV. Ratio or Factored Estimates Method
III. Cost-Capacity Equation or Power Law and Sizing
Model In this method, scaling relationships used to forecast the cost
of new project when historical and component data are
The general mathematical formula used to calculate the total available from similar project [19]. However, this scaling
Cost estimate is: relationship does not include economical factor, location and
the timing of the work. Generally, this method is used in
𝑄𝑄2 𝑥𝑥 estimating total plant cost in the processing industries. Both
𝐶𝐶2 = 𝐶𝐶1 � � (6)
𝑄𝑄1 direct and indirect costs can be included [20, 24, 37, 41]. The
general mathematical formula used to calculate the total Cost
Where: 𝑪𝑪𝟏𝟏 = Cost at Capacity𝑸𝑸𝟏𝟏 ; 𝑪𝑪𝟐𝟐 = Cost at Capacity𝑸𝑸𝟐𝟐 ; estimate can be expressed as:
x= Correlating Exponent.
Where: x = 1, relationship is linear; x< 1, economies of scale
(larger capacity is less costly than linear); x> 1, diseconomies 𝑇𝑇𝐶𝐶 = 𝐶𝐶𝐸𝐸 �� 𝑓𝑓𝑖𝑖 ∗ 𝐶𝐶𝐸𝐸𝑖𝑖 � ∗ (𝑓𝑓𝐼𝐼 + 1) (8)
of scale. 𝑖𝑖
Cost-Capacity Combined with Cost Index: Multiply the
𝐼𝐼 Where: 𝑪𝑪𝑬𝑬 = The total cost of major equipment item; 𝒇𝒇𝒊𝒊 =
cost-capacity equation by a cost index � 𝑡𝑡 � to adjust for time
𝐼𝐼0 Overall cost factor which can be determined using two basis;
differences and obtain estimates of current cost (in Delivered equipment cost including purchase cost of major
constant-value dollars) equipment Installation cost; (𝒇𝒇𝑰𝑰 + 𝟏𝟏) = the cost factor
(commonly the sum of a direct cost component and an
𝑄𝑄2 𝑥𝑥 𝐼𝐼𝑡𝑡 indirect cost component) for i = 1, 2… n components,
𝐶𝐶2 = 𝐶𝐶1 � � � � (7)
𝑄𝑄1 𝐼𝐼0 including indirect costs.

Some of the advantages of this method are its ability to

generate quick, easily, very accurate and understandable cost
Volume 8 Issue 10, October 2020
www.ijser.in
Licensed Under Creative Commons Attribution CC BY
Paper ID: 18092001 4 of 13
 International Journal of Scientific Engineering and Research (IJSER)
ISSN (Online): 2347-3878
Impact Factor (2018): 5.426
3.1.2. Probabilistic Cost Estimating Method
The formula that can be used to calculate the covariance of
The probabilistic cost estimating techniques focus on the two random variables X and Y, denoted by 𝑪𝑪𝑪𝑪𝑪𝑪(𝑿𝑿, 𝒀𝒀)is
risks and uncertainties involved in the project and attempt to defined as:
quantify the project cost variability based on one or more
parameters. It addresses the concerns regarding the chance of 𝐶𝐶𝐶𝐶𝐶𝐶(𝑋𝑋, 𝑌𝑌) = 𝐸𝐸(𝑋𝑋𝑋𝑋) − 𝜇𝜇𝑋𝑋 𝜇𝜇𝑌𝑌 (13)
exceeding a particular cost in the range of possible costs, the
possible amount of the cost overrun, and the different types of Therefore, the Pearson’s correlation coefficient between data
uncertainties and how they drive cost [4, 38, 42, 60, 66-67]. sets X and Y can be calculated:
The probabilistic cost estimating techniques uses probability
distribution to consider range estimation rather than point ∑𝑛𝑛𝑖𝑖=1(𝑋𝑋𝑖𝑖 − 𝑋𝑋�)(𝑌𝑌𝑖𝑖 − 𝑌𝑌�)
estimates to reflect the different outcomes [17, 26, 30]. The 𝑟𝑟 =
(14)
Expected value, Variance, Covariance and the Central Limit �∑𝑛𝑛𝑖𝑖=1(𝑋𝑋𝑖𝑖 − 𝑋𝑋�)2 �∑𝑛𝑛𝑖𝑖=1(𝑌𝑌𝑖𝑖 − 𝑌𝑌� )2
are some of the key aspects of the mathematical application of
probabilistic cost estimating techniques. � = Mean of
Where: r = Pearson’s correlation coefficient 𝑿𝑿
� = Mean of data set Y.
data set X; 𝒀𝒀
I. Expected Value
IV. Central Limit Theorem
The expected value of a cost parameter can be defined as the
weighted average of all possible values. The term expected
The Central Limit is the second fundamental theorem of the
value in essence means the same as the often used term
probability function that allows us to develop a process to
average [46]. The expected value equals:
estimate and test the mean of a population using a sample.
∞ The central limit theorem of statistics states that the sample
𝐸𝐸(𝑋𝑋𝑖𝑖 ) = 𝜇𝜇𝑥𝑥 = � 𝑥𝑥 𝑓𝑓(𝑥𝑥) 𝑑𝑑𝑑𝑑 (9) mean X¯ follows approximately the normal distribution with
−∞ mean µ and standard deviation √σ n, where µ and σ are the
mean and standard deviation of the population from where
Where: 𝒇𝒇(𝒙𝒙) = The probability density functions of cost the sample was selected. In order to be able to give lower and
parameter 𝒊𝒊. If all cost parameters of 𝒊𝒊 are correlated such upper bounds on the total cost we use confidence limits.
that 𝒀𝒀 = 𝒙𝒙𝟏𝟏 + 𝒙𝒙𝟐𝟐 , then Confidence limits are the probability that the interval estimate
𝐸𝐸(𝑌𝑌) = 𝐸𝐸(𝑋𝑋1 ) + 𝐸𝐸(𝑋𝑋2 ) (10) will include the lower and upper bound of cost parameter
[22-23, 55]. Table-4 below shows the Confidence Level
The variance in this case is given by 𝛿𝛿𝑌𝑌 2 = 𝛿𝛿12 + 𝛿𝛿22 + 2𝛿𝛿1,2 Using the Standardized Normal Distribution.
in this formula 𝛿𝛿1,2 is the covariance of random variables
of𝑥𝑥1 𝑎𝑎𝑎𝑎𝑎𝑎𝑥𝑥2 . If the random variables are independent then 𝛿𝛿1,2 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿 = 𝐸𝐸𝐸𝐸𝐸𝐸 − 𝑍𝑍 ∗ 𝜎𝜎 (15)
is equal to zero. If the total cost is the product of independent, 𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈 = 𝐸𝐸𝐸𝐸𝐸𝐸 + 𝑍𝑍 ∗ 𝜎𝜎 (16)
continuous, random variables, such that= 𝑥𝑥1 ∗ 𝑥𝑥2 , then
Where: 𝑳𝑳𝑳𝑳𝑳𝑳𝑳𝑳 =Lower bound on Total Cost; 𝑼𝑼𝑼𝑼𝑼𝑼𝑼𝑼= Upper
𝐸𝐸(𝑌𝑌) = 𝐸𝐸(𝑋𝑋1 ) + 𝐸𝐸(𝑋𝑋2 ) (11) bound on Total Cost; 𝑬𝑬𝑬𝑬𝑬𝑬 = Expected Total Cost; 𝝈𝝈 =
𝛿𝛿𝑌𝑌 2 = 𝑋𝑋12 𝛿𝛿12 + 𝑋𝑋22 𝛿𝛿22 + 𝛿𝛿12 𝛿𝛿22 (12) Standard Deviation; Z= is determined by the confidence level
II. Variance using the standardized Normal distribution.

In probability theory, variance gives a measure of how much Table 4: Confidence Level Using the Standardized Normal
the values of a function of a random variable x vary as we Distribution
sample x from a probability distribution. When the variance is Confidence Level Value of Z
low, values of f(x) cluster around its expected value. The 90% 1.28
square root of the variance is known as the standard deviation 95% 1.65
and usually indicated with the symbol σ [38, 43]. 98% 2.05
99.9% 3.09
III. Covariance
3.1.3. Probability Distributions
Covariance measures how two values are linearly related, as
well as scale of variables. Calculating correlation is an Different cost parameters coupled with several simple
important to analyze the correlation between two or more cost probability distributions are useful in many engineering cost
components that can have a large impact on the degree of risk estimation modeling and risk analysis. Normal, Lognormal,
associated with using the variance. If two random variables Beta, Triangular and Weibull are typical probability
have no correlation with covariance equal to zero they are distributions that are commonly used in the construction
called independent [11]. The covariance can be high absolute, industry [4, 11, 17, 43, 56, 60, 65-66]. Below are summary of
positive, zero or negative. High absolute values of covariance discussion together with the probability density function
means values change very much and are both far from their (PDF), the cumulative density function (CDF), the expected
mean. Positive value means both variables take relatively value (E(X)) and the variance (Var(X)) of each distributions.
high values far from mean. Negative value means one Table-5 below shows the advantages and disadvantages of
variable takes on high values & another takes low values [59, Probabilistic Cost Estimating Method.
63].
Volume 8 Issue 10, October 2020
www.ijser.in
Licensed Under Creative Commons Attribution CC BY
Paper ID: 18092001 5 of 13
 International Journal of Scientific Engineering and Research (IJSER)
ISSN (Online): 2347-3878
Impact Factor (2018): 5.426
I. Uniform Distribution ⎧ 2(𝑥𝑥 − 𝑎𝑎)
⎪(𝑏𝑏 − 𝑎𝑎)(𝑚𝑚 − 𝑎𝑎) 𝑖𝑖𝑖𝑖 𝑎𝑎 ≤ 𝑥𝑥 ≤ 𝑚𝑚
The uniform distribution is a continuous probability 𝑓𝑓𝑥𝑥 (𝑥𝑥) = (21)
distribution the assumption: the random event is equally ⎨ 2(𝑏𝑏 − 𝑥𝑥)
⎪(𝑏𝑏 − 𝑎𝑎)(𝑚𝑚 − 𝑎𝑎) 𝑖𝑖𝑖𝑖 𝑚𝑚 ≤ 𝑥𝑥 ≤ 𝑏𝑏
likely in an interval. It is defined by two parameters, the ⎩
minimum possible value (a) and the maximum possible value
(b).A variable X is said to be uniformly distributed if the
density function is: The cumulative probability distribution of the triangular
distribution is given by
1
𝑓𝑓(𝑥𝑥) = �𝑏𝑏 − 𝑎𝑎 𝑓𝑓𝑓𝑓𝑓𝑓 𝑜𝑜𝑜𝑜 − ∞ < 𝑎𝑎 ≤ 𝑥𝑥 ≤ 𝑏𝑏 < ∞ (17) 𝑓𝑓𝑥𝑥 (𝑥𝑥)
0 0
⎧
⎪ (𝑥𝑥 − 𝑎𝑎)2
Figure-3, below shows the sample of the uniform distribution ⎪ 𝑖𝑖𝑖𝑖 𝑎𝑎 ≤ 𝑥𝑥 < 𝑚𝑚
(𝑏𝑏 − 𝑎𝑎)(𝑚𝑚 − 𝑎𝑎)
curve graph =
⎨ (𝑏𝑏 − 𝑥𝑥)2
⎪ 1− 𝑖𝑖𝑖𝑖 𝑚𝑚 ≤ 𝑥𝑥 < 𝑏𝑏
⎪ (𝑏𝑏 − 𝑎𝑎)(𝑏𝑏 − 𝑚𝑚)
⎩1
(22)
The expected value is given by:
𝑎𝑎 + 𝑚𝑚 + 𝑏𝑏
𝐸𝐸(𝑋𝑋) = (23)
3

The variance is given by:

𝑎𝑎2 + 𝑚𝑚2 + 𝑏𝑏 2 + 𝑎𝑎𝑎𝑎 + 𝑎𝑎𝑎𝑎 + 𝑚𝑚𝑚𝑚
V(𝑋𝑋) = (24)
Figure 3: Sample of a Uniform Distribution Graph 18
The standard deviation is given by:
The mean and variance of X following a uniform distribution 𝑎𝑎 2 + 𝑚𝑚2 + 𝑏𝑏 2 − 𝑎𝑎𝑎𝑎 − 𝑎𝑎𝑎𝑎 − 𝑚𝑚𝑚𝑚
is: 𝛿𝛿 = � (25)
18
(𝑎𝑎 + 𝑏𝑏)
𝐸𝐸(𝑋𝑋) = (18)
2 III. Beta Distribution
2
(𝑏𝑏 − 𝑎𝑎)
𝑉𝑉(𝑋𝑋) = (19)
12 One of its most common uses of this distribution is to model
The standard uniform density has parameters a = 0 and b = 1,
uncertainty and bounded continuous random variables based
so the PDF for standard uniform density is given by:
on expert’s judgment. The Beta (α, β) distribution is a
1, 0 ≤ 𝑥𝑥 ≤ 1
𝑓𝑓(𝑥𝑥) = � (20) continuous probability that is defined by two shape
0, 𝑜𝑜𝑜𝑜ℎ𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒
parameters α and β [5, 27, 32]. The general formula for the
II. Triangular Distribution probability density function of the beta distribution is:

𝑓𝑓 (𝑥𝑥)
In this method, it is assumed that a Triangular or Beta 1 𝛤𝛤(𝛼𝛼 − 𝛽𝛽) 𝑥𝑥 − 𝐿𝐿 𝛼𝛼−1 𝐻𝐻 − 𝑥𝑥 𝛽𝛽−1
distribution can be used to describe each item T (a, m, b). This = � � � �
𝐻𝐻 − 𝐿𝐿 𝛤𝛤(𝛼𝛼)𝛤𝛤(𝛽𝛽) 𝐻𝐻 − 𝐿𝐿
� �
𝐻𝐻 − 𝐿𝐿
� 𝐿𝐿 < 𝑥𝑥 < 𝐻𝐻
means that the user gives an optimistic estimate a, a most = 0 𝑜𝑜𝑜𝑜ℎ𝑒𝑒𝑒𝑒
likely estimate m and finally a pessimistic estimate b [5, 32,
(26)
56]. Figure-4, below shows sample of triangular distribution
The shape parameters: 𝜶𝜶 > 𝟎𝟎, 𝜷𝜷 > 𝟎𝟎Figure-5, below shows
graph.
sample of beta distribution graph.

Figure 4: Sample of Triangular Distribution (a) =Lowest (b)

= Highest, and (M) = Most Likely Values
The PDF of the triangular distribution is given by:

Figure 5: Sample of Beta Distribution

Volume 8 Issue 10, October 2020

www.ijser.in
Licensed Under Creative Commons Attribution CC BY
Paper ID: 18092001 6 of 13
 International Journal of Scientific Engineering and Research (IJSER)
ISSN (Online): 2347-3878
Impact Factor (2018): 5.426
Most schedule or cost estimates follow right skewed pattern. (L, H, M) Distribution using L, M, and H to calculate the
The value of 𝜶𝜶𝜶𝜶𝜶𝜶𝜶𝜶𝜶𝜶 can be determined using Beta-PERT expected value mean and standard deviation as [5, 27, 32]:

(𝐿𝐿 + 4𝑀𝑀 + 𝐻𝐻) (𝐻𝐻 − 𝐿𝐿)2 (𝐻𝐻 − 𝐿𝐿)

𝐸𝐸(𝑋𝑋) = 𝜇𝜇 = 𝑉𝑉𝑉𝑉𝑉𝑉 (𝑋𝑋) = 𝜎𝜎 = (27)
6 36 6
⎧ (𝜇𝜇 − 𝐿𝐿) (𝜇𝜇 − 𝐿𝐿)(𝐻𝐻 − 𝜇𝜇)
⎪ 𝛼𝛼 = ∗ −1
(𝐻𝐻 − 𝐿𝐿) 𝜎𝜎 2
𝛼𝛼 𝑎𝑎𝑎𝑎𝑎𝑎 𝛽𝛽 = (28)
⎨ (𝐻𝐻 − 𝜇𝜇)
⎪𝛽𝛽 = (𝜇𝜇 − 𝐿𝐿) ∗ 𝛼𝛼
⎩

Or, from the expected value (μ) and the distribution P (L, M, H) the parameters α and β can be derived by

⎧𝛼𝛼 = (𝜇𝜇 − 𝐿𝐿)(2𝑀𝑀 − 𝐿𝐿 − 𝐻𝐻)

⎪ (𝐻𝐻 − 𝐿𝐿)(𝑀𝑀 − 𝜇𝜇)
𝛼𝛼 𝑎𝑎𝑎𝑎𝑎𝑎 𝛽𝛽 = (29)
⎨ 𝛼𝛼(𝐻𝐻 − 𝜇𝜇)
⎪ 𝛽𝛽 =
⎩ (𝜇𝜇 − 𝐿𝐿)

Where: 𝜶𝜶 > 𝟎𝟎, 𝜷𝜷 > 𝟎𝟎; (L), lowest; (H) Highest and (M) Most likely values.

IV. Normal Distribution N (µ, σ). Here µ is its mean,𝜹𝜹𝟐𝟐 its variance, and σ is its
standard deviation [10, 12, 25]. The normal distribution is a
The normal distribution is a continuous probability continuous distribution with probability density function of:
distribution and it has two parameters, µ and σ and is denoted

1 1 𝑥𝑥−𝜋𝜋 2
𝑓𝑓𝑥𝑥 (𝑥𝑥) = 𝑒𝑒 −2� 𝜎𝜎
�
(30)
𝛿𝛿√2𝜋𝜋

The cumulative probability distribution of the normal distribution is given by:

𝑥𝑥
1 1 𝑥𝑥−𝜋𝜋 2
−� � � � 𝑑𝑑𝑑𝑑
𝑓𝑓𝑥𝑥 (𝑥𝑥) = 𝑃𝑃(𝑋𝑋 ≤ 𝑥𝑥) = � 𝑒𝑒 2 𝜎𝜎 (31)
−∞ 𝛿𝛿√2𝜋𝜋

The expected value is given by

𝐸𝐸(𝑋𝑋) = 𝜇𝜇 (32)
The expected value is given by
𝑉𝑉𝑉𝑉𝑉𝑉(𝑋𝑋) = 𝛿𝛿 2 (33)

Figure-6, below shows sample of standard normal V. Lognormal Distribution

distribution graph. The lognormal distribution is the probability distribution
where the natural log of the sample values has a normal
distribution. The probability density function (pdf) is given
by ln (N (μ, σ2)) [5, 32, 43].

1 1 𝑙𝑙𝑙𝑙 𝑥𝑥 − 𝛼𝛼 2
𝑓𝑓(𝑥𝑥) = 𝑒𝑒 �− � � � (34)
𝑥𝑥𝑥𝑥√2𝜋𝜋 2 𝛽𝛽

Where: α is the mean of ln(x) and β is the standard deviation

of ln(x). These are related to the mean and standard
deviation of random variable x (μ and σ respectively) as
follows:

1
𝜇𝜇 = 𝑒𝑒 �𝛼𝛼 + 𝛽𝛽 2 � (35)
2
Figure 6: Sample of Standard Normal Distribution Graph 𝛿𝛿 = �𝑒𝑒[2𝛼𝛼 + 𝛽𝛽 2 ](𝑒𝑒[𝛽𝛽 2 ] − 1) (36)
1 𝛼𝛼 2
𝛼𝛼 = 𝑙𝑙𝑙𝑙 𝜇𝜇 − 𝑖𝑖𝑖𝑖 �� � + 1� (37)
2 𝜋𝜋

Volume 8 Issue 10, October 2020

www.ijser.in
Licensed Under Creative Commons Attribution CC BY
Paper ID: 18092001 7 of 13
 International Journal of Scientific Engineering and Research (IJSER)
ISSN (Online): 2347-3878
Impact Factor (2018): 5.426
The Expected value is given by:
𝛼𝛼 2 1
𝛽𝛽 = �𝑙𝑙𝑙𝑙 �� � + 1� (38)
𝜋𝜋 μ = E(X) = δΓ �1 + � (44)
β
The Variance is given by:
Figure-7, below shows sample of standard lognormal δ2 = V(X)
distributions graph 2
= δ2 Γ �1 + � (45)
β
− μ2 , where X~ Weibull(δ, β)
The cumulative probability function (CDF) for the standard
Weibull (δ,β) probability distribution is given by:
x β
F(x) = 1 − e �− � � � , F(x) = 0, for x ≤ 0 (46)
δ

Figure-8, below shows sample of standard sample of

Weibull distributions graph

Figure 7: Sample of Lognormal Distributions Graph

P is the cumulative probability function (CDF) of the log

normal distribution, is given by
𝑥𝑥
1 1 𝑙𝑙𝑙𝑙 𝑥𝑥 − 𝛼𝛼 2 Figure 8: Sample of Weibull Distribution Graph
𝑃𝑃(𝑥𝑥 ≤ 𝑋𝑋) = � 𝑒𝑒 �− � � � 𝑑𝑑𝑑𝑑 (39)
0 𝑥𝑥 √2𝜋𝜋 2 𝛽𝛽
Table 5: Probabilistic Cost Estimating Method
𝑥𝑥
Note that F is the cumulative probability function (CDF) for
Advantage Disadvantage Requirement
the standard normal probability distribution Probability of cost Probability
𝑙𝑙𝑙𝑙 𝑋𝑋 − 𝛼𝛼 Additional
𝑃𝑃(𝑥𝑥 ≤ 𝑋𝑋) = 𝐹𝐹 � � (40) overrun is insightful distribution of cost
analysis that
𝛽𝛽 and defined as components based on
requires additional
The Expected value is given by: opportunities or
effort
historical data or
𝜎𝜎2 threats experience
𝐸𝐸(𝑋𝑋) = 𝑒𝑒 𝜇𝜇+ 2 (41)
Determining
The variance is given by: Software to run
Improved reliability probability
2 2 Monte Carlo
𝑉𝑉𝑉𝑉𝑉𝑉(𝑋𝑋) = �𝑒𝑒 𝜎𝜎 − 1�𝑒𝑒 2𝜇𝜇+𝜎𝜎 (42) of the estimate distributions may
Simulation
be difficult
Range of outcomes is
VI. Weibull distribution
available
Uncertainties and
Generally, the Weibull distribution is one of the most risks are mapped and
commonly used statistical model for project cost estimations quantified
and many other applications. The Weibull Distribution
(1939) was first published to represent the probability of 3.2. Sample Project to demonstrate and compare different
failure and has proven to be extremely useful for data Cost Estimating Methods for small urban drainage system
analysis in many engineering applications such as aerospace, improvement project.
automotive, electric power, nuclear power, medical, dental,
electronics and every industry [5, 18, 32, 36, 43]. VII. P50 and P90 cost estimation

A continuous function X is said to have a Weibull From the director of the public work's point of view, the
distribution with parameters δ> 0 and β> 0 if the PDF of X expected cost of a portfolio of projects is of more interest than
is: the costs of projects cost separately. Individual projects are
considered at the mean of simulated cost distribution,
𝛽𝛽 𝑥𝑥 𝛽𝛽−1 𝑥𝑥 𝛽𝛽 typically the P50 estimate. The P50 cost value is an estimate
𝑓𝑓(𝑥𝑥) = � � 𝑒𝑒 �− � � � ,
𝛿𝛿 𝛿𝛿 𝛿𝛿 of the project cost based on a 50% probability that the cost
(43)
𝑓𝑓𝑓𝑓𝑓𝑓 𝑥𝑥 > 0, 𝑎𝑎𝑎𝑎𝑎𝑎 𝑓𝑓(𝑥𝑥) = 0, 𝑓𝑓𝑓𝑓𝑓𝑓 𝑥𝑥 will not be exceeded. The P90 value is an estimate of the
≤0 project cost based on a 90% probability that the cost will not
be exceeded. Project proponents (and their management)
Volume 8 Issue 10, October 2020
www.ijser.in
Licensed Under Creative Commons Attribution CC BY
Paper ID: 18092001 8 of 13
 International Journal of Scientific Engineering and Research (IJSER)
ISSN (Online): 2347-3878
Impact Factor (2018): 5.426
often prefer to have less commercial (and political) exposure aggregating inputs. However, aggregating inputs that reflect
in respect of capital budgets and often look for a P90 figure the type of risk exposure or other logical model structures
(or equivalent if done deterministically), meaning the such as aggregation based on geographically discrete work
contingency allowance on top of the base estimate is packages may be more suitable. Costs representing the most
sufficient to ensure that there is a 90% chance that the amount likely, the best case and worst case in this example are
will not be exceeded [69]. hypothetical only.

4. Validation through Case Studies The selected typical capital improvement project
demonstrates the validity, usefulness, and benefit of the
An example of deterministic approach, range based approach different deterministic and risk-based probabilistic cost
and Risk Based – Probabilistic Approach have been prepared estimation approach. It compares the actual costs with
as an accompaniment to this journal with all formulae intact. traditional deterministic methods of cost estimation (such as
Practitioners are welcome to utilize this model for training or single-point base-case estimates inclusive of contingency)
as a template, modified as appropriate for their own and provide valuable insights that can aid management in
circumstances. evaluating alternatives and make informed decisions when
estimating and allocating budgets to a portfolio of Urban
The example uses typical capital improvement project Cost drainage system and associated drainage infrastructure
Breakdown (PCB) template structure as the basis for rehabilitation projects.

Table 6: Drainage Rehabilitation Capital Improvement Project (Base Cost Only) case study-1

Quantity Unit Price Amount

Item Description Distribution min ml max unit Distribution min ml max Q (ml) * UP (ml)
1 Remove existing storm drainage structures Triangular 6 6 6 ea. Triangular $438.75 $450.00 $465.75 $2,700.00
2 Remove surface materials - asphalt Triangular 87.75 90 92.97 lf. Triangular $8.53 $8.75 $9.06 $813.49
3 Construct single grate inlet catch basin box Triangular 4 4 4 ea. Triangular $2,656.88 $2,725.00 $2,820.38 $10,900.00
4 Construct 15" diameter rcp storm drain pipe Triangular 87.75 90 92.97 lf. Triangular $38.27 $39.25 $40.62 $3,649.07
5 Construct 4' diameter cleanout manhole Triangular 2 2 2 ea. Triangular $2,486.25 $2,550.00 $2,639.25 $5,100.00
6 Furnish trench backfill materials Triangular 48.75 50 51.65 tons Triangular $4.39 $4.50 $4.66 $232.43
7 Furnish bedding materials Triangular 29.25 30 31.5 tons Triangular $10.24 $10.50 $10.87 $330.75
8 Construct concrete curb and gutter Triangular 117 120 124 lf. Triangular $19.01 $19.50 $20.18 $2,417.22
9 Construct double hooded inlet catch basin box Triangular 2 2 2 ea. Triangular $3,607.50 $3,700.00 $3,829.50 $7,400.00
10 Roadway patching Triangular 98.48 101 104.3 sy. Triangular $19.01 $19.50 $20.18 $2,034.44
11 Traffic control and flagging Triangular 13.65 14 14.46 days Triangular $146.25 $150.00 $155.25 $2,169.00
12 Landscaping & surface restoration Triangular 95.55 98 101.2 sy. Triangular $195.00 $200.00 $207.00 $20,246.00
$57,992.39

A. Deterministic Approach The Square Root approach delivers one single figure which
is the sum of all base costs plus the square root of the sum of
The deterministic approach delivers a single figure that is the squares of the risk contingencies as shown below
the sum of the products of the most likely quantity
multiplied by the most likely price as shown below:
𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂 = �� 𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂 𝐈𝐈𝐈𝐈𝐈𝐈𝐈𝐈 (𝐦𝐦𝐦𝐦𝐦𝐦 − 𝐌𝐌𝐌𝐌𝐌𝐌)𝟐𝟐

= � 𝐌𝐌𝐌𝐌𝐌𝐌𝐌𝐌 𝐥𝐥𝐥𝐥𝐥𝐥𝐥𝐥𝐥𝐥𝐥𝐥 𝐪𝐪𝐪𝐪𝐪𝐪𝐪𝐪𝐪𝐪𝐪𝐪𝐪𝐪𝐪𝐪 𝐱𝐱 𝐦𝐦𝐦𝐦𝐦𝐦𝐦𝐦 𝐥𝐥𝐥𝐥𝐥𝐥𝐥𝐥𝐥𝐥𝐥𝐥 𝐩𝐩𝐩𝐩𝐩𝐩𝐩𝐩𝐩𝐩] + � 𝐌𝐌𝐌𝐌𝐌𝐌 = 𝟓𝟓𝟓𝟓, 𝟑𝟑𝟑𝟑𝟑𝟑. 𝟗𝟗𝟗𝟗 𝐔𝐔𝐔𝐔𝐔𝐔
= $ 𝟓𝟓𝟓𝟓, 𝟗𝟗𝟗𝟗𝟗𝟗. 𝟑𝟑𝟑𝟑 𝐔𝐔𝐔𝐔𝐔𝐔
B. Risk Based – Probabilistic Approach
B. Range approach with Minimum, Most Likely, and
Maximum Cost. Risk Based - Probabilistic Approach include base cost
uncertainty, the monetary impact of discrete risks as defined
Range approach with Minimum, Most Likely, and in the risk register plus escalation, as well as the monetary
Maximum Cost Approach delivers three results as the sums impact of schedule delays through extended overhead and
of the following elements: additional escalation caused by schedule delays. The
non-escalated base-cost estimate for this contract has been
= ∑ 𝐌𝐌𝐌𝐌𝐌𝐌. 𝐐𝐐 ∗ 𝐌𝐌𝐌𝐌𝐌𝐌. 𝐔𝐔𝐔𝐔 = $𝟓𝟓𝟓𝟓, 𝟕𝟕𝟕𝟕𝟕𝟕. 𝟗𝟗𝟗𝟗 𝐔𝐔𝐔𝐔𝐔𝐔 determined through the estimating process to be $57,992.36
= ∑ 𝐌𝐌𝐌𝐌𝐌𝐌𝐌𝐌 𝐋𝐋𝐋𝐋𝐋𝐋𝐋𝐋𝐋𝐋𝐋𝐋 (𝐦𝐦𝐦𝐦)𝐐𝐐 ∗ 𝐦𝐦𝐦𝐦𝐦𝐦𝐦𝐦 𝐥𝐥𝐥𝐥𝐥𝐥𝐥𝐥𝐥𝐥𝐥𝐥. 𝐔𝐔𝐔𝐔 = USD. Assuming no risk or uncertainty on this value and
$𝟓𝟓𝟓𝟓, 𝟗𝟗𝟗𝟗𝟗𝟗. 𝟑𝟑𝟑𝟑 𝐔𝐔𝐔𝐔𝐔𝐔 incorporating projected escalation provides an escalated
=∑ 𝐌𝐌𝐌𝐌𝐌𝐌. 𝐐𝐐 ∗ 𝐌𝐌𝐌𝐌𝐌𝐌. 𝐔𝐔𝐔𝐔 = $𝟔𝟔𝟔𝟔, 𝟎𝟎𝟎𝟎𝟎𝟎. 𝟏𝟏𝟏𝟏 𝐔𝐔𝐔𝐔𝐔𝐔 base-cost estimate of $1,077 USD. Further incorporating
risk associated with: (i) the base cost estimate, (ii) specific
A. Square root Approach event risks and (iii) potential schedule delay provides a
risk-based estimate of total Project costs. As shown in
Volume 8 Issue 10, October 2020
www.ijser.in
Licensed Under Creative Commons Attribution CC BY
Paper ID: 18092001 9 of 13
 International Journal of Scientific Engineering and Research (IJSER)
ISSN (Online): 2347-3878
Impact Factor (2018): 5.426
Figure 9, there is an 80% probability that this cost will lie (90th percentile). At the 70th percentile, the risk-based cost
between $1,075 million (10th percentile) and $1,235 million estimate is $1,188 million.

0.9
$58,000.0
0.8 $63,220.00
Probablity of Not Exceeding

0.7

0.6

0.5 $62,060.00
0.4

0.3 $58,580.00
0.2

0.1

Cost (US $)

Figure 9: Graphical Representation of Cost-Risk Analysis Results

To compare this to a traditional cost estimating approach (in allowance/contingency over the escalated base cost estimate
which allowance and/or contingency are set at fixed of $1,077 million (to ensure a level of confidence of 70
proportions of the base cost estimate), this outcome suggests percent).Figure -9 provides a graphical representation of
the Project should budget a 10.3 percent cost-risk results.

Table 7: Drainage Rehabilitation Capital Improvement Project (Base Cost Only) case study-2

Design Engineer’s Risk-based

Drainage Projects Initially Allocated Design Engineer’s Preliminary Estimate ($) Single-Point Base
Budget ($) Estimate Inclusive of 10-20% Contingency (Max. Value) Preliminary Estimate ($) Actual Final Cost ($)
P50 P90

PR-1 $ 295,800.00 $ 300,000.00 $ 297,900.00 $ 300,588.00 $ 305.075.00

PR-2 $ 355,000.00 $ 406,048.00 $ 380,524.00 $ 413,195.00 $ 421,400.00
PR-3 $ 520,000.00 $ 595,000.00 $ 557,500.00 $ 605,500.00 $ 596,418.00
PR-4 $ 450,000.00 $ 495,958.00 $ 472,979.00 $ 502,391.00 $ 509,927.00

Table 7 illustrates the difference between the P50 and P90

cost estimation and the importance of adopting a P90 5. Conclusions
estimate. The P50 value predicted at the preliminary stage
(when uncertainty is highest), considering only the rates of In this paper, the most common project cost estimation
the allocated contractor, predicted the closest value for methods and approaches were collected and classified into
budgeting purposes when compared with the actual final two main categories of: (1) deterministic and (2)
cost. The incorporation of risk should be a necessary step in probabilistic methods. Then, the two main categories were
project cost estimation at all stages of a project, by applying further divided into four and six sub categories respectively:
the P90, rather than the P50, value. The application of the (1) Parametric cost estimating method, (2) Detailed cost
P90 value will assist the municipalities/ cost estimator estimating methods, (3) Comparative cost estimating
professionals in their understanding and appraisal of project methods, (4) Ratio or Factored estimates method, and (1)
proposals, and improve the preparation of cost estimates that Uniform distribution methods, (2) Triangular distribution
form part of the documentation to make them more methods, (3) Beta distribution method, (4) Normal
transparent, reliable and consistent. Distribution methods, (5) Lognormal distribution methods,
and (6) Weibull distribution method. Overall, ten different
methods were identified and discussed. Table 8 demonstrate
the advantages and shortcomings of these methods including
the potential risk that can positively or negatively affect the
project’s cost outcome. This paper will be a good resource

Volume 8 Issue 10, October 2020

www.ijser.in
Licensed Under Creative Commons Attribution CC BY
Paper ID: 18092001 10 of 13
 International Journal of Scientific Engineering and Research (IJSER)
ISSN (Online): 2347-3878
Impact Factor (2018): 5.426
for professionals who are in budget development and/or project cost estimate for their projects. Summarized.
seeking to maximize and produce a meaningful and reliable

Table 8: Pros and Cons of Deterministic and Probabilistic Cost Estimations Methods
Deterministic Probabilistic
Pros Cons Pros Cons
Varity of techniques may be Does not determined residual risk.
more complex, still
used including engineering Unknown risk, can be inconsistent known risk, handles areal
wide-spread
judgement, factor of safety, between sites. For areal sources, selection sources in a consistent way
misunderstanding
etc . of deterministic event is uncertain
simple to use , Doesn’t rely Needs probabilistic thinking
away from measured or interpreted data, Uses impartial statistical rules,
on statistics, Maintains & understanding Needs
Not statistical, Exhaustive cases can be run,
dependencies software support
No probability information of single All potential risks are included,
One single figure Well-known
value No Value at Risk information More best estimation assumptions,
& accepted Quick Can be Time consuming
often than not on the unsafe side (high, and follow well established
performed “manually”
unknown probability of cost overruns) methodology.
Good accuracy for similar Accuracy is limited, Cost impacting Multiple and common cause of
systems if comparative and factors have to be determined, and failures can be easily assessed Comparison factors
recent data is available Normalization required and addressed at the early stage

It should be emphasized that a forecast is only good as the [9] Ballard, G., &Pennanen, A. (2013). Conceptual
probabilistic model and the quality of data that is fed to the estimating and target costing. In Proc. 21st Ann. Conf.
model. A probabilistic analysis does not necessarily imply of the Int’l. Group for Lean Construction (pp. 31-2).
precluding the use of a deterministic analysis. In fact, [10] Bates, J., Burton, C. D. J., Creese, R. C., Hollmann, J.
deterministic analysis is often required to provide input to K., Humphreys, K. K., McDonald Jr, D. F., & Miller, C.
probabilistic analysis. The key point is that deterministic A. (2005). Cost estimate classification system–as
analyses alone can have significant disadvantages and in applied in engineering, procurement, and construction
such cases, should be complemented by probabilistic for the process industries.
analyses as is seen through the fiscal results of the case [11] Beck, J. V., & Arnold, K. J. (1977). Parameter
studies. estimation in engineering and science. James Beck.
[12] Bier, V. M. (1997). An overview of probabilistic risk
References analysis for complex engineered systems. Fundamentals
of risk analysis and risk management, 67.
[1] AACE International 2003, “Cost Estimate Classification [13] Black, J. (1984). Application of parametric estimating
System” to cost engineering. AACE transactions, B-10.
http://www.aacei.org/technical/rps/18r-97.pdf [14] Blank, L., & Tarquin, A. (2005). Engineering economy.
[2] Abou Rizk, S., & Mohamed, Y. (2002, December). McGraw-Hill.
CEPM 1: optimal construction project planning. [15] Burak Evrenosoglu, F. (2010). Modeling Historical Cost
In Proceedings of the 34th conference on Winter Data for Probabilistic Range Estimating. Cost
simulation: exploring new frontiers (pp. 1704-1708). Engineering, 52(5), 11.
Winter Simulation Conference. [16] Burns, T. J., Page, E. C., Gregory, R. A., & Pryor, G.
[3] Akintoye, A. (2000). Analysis of Factors Influencing M. (1993). U.S. Patent No. 5,189,606. Washington, DC:
Project Cost Estimating Practice. Construction U.S. Patent and Trademark Office.
Management and Economics, 18(7) 77-89 [17] Chou, J. S. (2009). Generalized linear model-based
[4] Anderson, S. D., Molenaar, K. R., & Schexnayder, C. J. expert system for estimating the cost of transportation
(2007). Guidance for cost estimation and management projects. Expert Systems with Applications, 36(3),
for highway projects during planning, programming, 4253-4267.
and preconstruction (Vol. 574). Transportation Research [18] Chou, J.-S. 2011. ‘Cost Simulation in an Item-Based
Board. Project Involving Construction Engineering and
[5] Ayyub, B. M., & McCuen, R. H. (2016). Probability, Management.’ International Journal of Project
statistics, and reliability for engineers and scientists. Management 29 (6): 709–17.
CRC press. [19] Christensen, P., & Dysert, L. R. (2003). AACE
[6] Bajaj, A., Gransberg, D. D., & Grenz, M. D. (2002). international recommended practice no. 17R-97 cost
Parametric estimating for design costs. AACE estimate classification system. AACE International,
International Transactions, ES81. USA.
[7] Bakhshi, P., & Touran, A. (2012). A Method for [20] Clark, F., & Lorenzoni, A. B. (1996). Applied cost
Calculating Cost Correlation among Construction engineering. CRC Press.
Projects in a Portfolio. International Journal of [21] Council, P. S. R. (2009). Benefit-Cost Analysis:
Architecture, Engineering and Construction, 1 (3), General Methods and Approach. The Council.
134-141. [22] Dell’Isola, M. D. (2003). Detailed cost
[8] Bakhshi, P., & Touran, A. (2014). An overview of estimating. Excerpt from Archit. Handb. Prof. Pract,
budget contingency calculation methods in construction 1-13.
industry. Procedia Engineering, 85, 52-60.

Volume 8 Issue 10, October 2020

www.ijser.in
Licensed Under Creative Commons Attribution CC BY
Paper ID: 18092001 11 of 13
 International Journal of Scientific Engineering and Research (IJSER)
ISSN (Online): 2347-3878
Impact Factor (2018): 5.426
[23] Diekmann, J. E. (1983). Probabilistic estimating: review. International Journal of Advanced Research in
mathematics and applications. Journal of Construction Computer Science and Software Engineering, 3(7).
Engineering and Management, 109(3), 297-308. [41] Lemmens, S. (2016). Cost engineering techniques and
[24] Dysert, L. R. (2003). Sharpen your cost estimating their applicability for cost estimation of organic
skills. Cost Engineering, 45(6), 22. Rankine cycle systems. Energies, 9(7), 485.
[25] Edwards, D. J., Holt, G. D., & Harris, F. C. (2000). A [42] Modarres, M. (2016). Risk analysis in engineering:
comparative analysis between the multilayer perceptron techniques, tools, and trends. CRC press.
“neural network” and multiple regression analysis for [43] Moergli, A., Sander, P., & Reilly, J. (2015).
predicting construction plant maintenance costs. Journal Risk-Based, Probabilistic Cost Estimating Methods.
of Quality in Maintenance Engineering, 6(1), 45-61. In International Tunneling Association, World Tunnel
[26] Elkjaer, M. (2000). Stochastic budget Congress, Dubrovnik May.
simulation. International Journal of Project [44] Nijkamp, P., & Ubbels, B. (1999). How reliable are
Management, 18(2), 139-147. estimates of infrastructure costs? A comparative
[27] Erkoyuncu, J. A., Durugbo, C., Shehab, E., Roy, R., analysis. International Journal of Transport
Parker, R., Gath, A., & Howell, D. (2013). Uncertainty Economics/Rivistainternazionale di
driven service cost estimation for decision support at the economiadeitrasporti, 23-53.
bidding stage. International Journal of Production [45] Ogilvie, A., Brown Jr, R. A., Biery, F. P., & Barshop, P.
Research, 51(19), 5771-5788. (2012). Quantifying Estimate Accuracy and precision
[28] Evans and Peck Report for Department of for the process industries: a review of industry
Infrastructure, Transport, Regional Development and data. Cost Engineering-Morgantown, 54(6), 28.
Local Government (2008), Best Practice Cost [46] Ostwald, P. F. (1974). Cost estimating for engineering
Estimation for Publicly Funded Road and Rail and management. Prentice-Hall.
Construction, viewed 10 Jan 2013, http://www.nation [47] PMI 2018, “Project Management Body of Knowledge,”
[29] Fad, B., Summers, R., DeMarco, A., Geiser, T., Walter, Project Management Institute, Pennsylvania
J., Chackman, B., & King, E. (1998). U.S. Patent No. [48] Qian, L., & Ben-Arieh, D. (2008). Parametric cost
5,793,632. Washington, DC: U.S. Patent and Trademark estimation based on activity-based costing: A case study
Office. for design and development of rotational
[30] FHWA (January 2004). Guidelines on Preparing parts. International Journal of Production
Engineer’s Estimates, Bid Reviews and Evaluation. Economics, 113(2), 805-818.
http://www.fhwa.dot.gov/programadmin/contracts/ta508 [49] Reilly, J. J. (2001, March). Managing the costs of
046.pdf complex, underground and infrastructure projects.
[31] Flyvbjerg, B., Holm, M. S., & Buhl, S. (2002). In American Underground Construction Conference.
Underestimating costs in public works projects: Error or [50] Reilly, J., McBride, M., Sangrey, D., MacDonald, D., &
lie?. Journal of the American planning Brown, J. (2004). The development of
association, 68(3), 279-295. CEVP®-WSDOT’s Cost-Risk Estimating
[32] Garvey, P. R., Book, S. A., & Covert, R. P. Process. Proceedings, Boston Society of Civil
(2016). Probability methods for cost uncertainty Engineers, http://www. wsdot. wa.
analysis: A systems engineering perspective. Chapman gov/projects/projectmgmt/riskassessment.
and Hall/CRC. [51] Rosse, J. N. (1970). Estimating cost function parameters
[33] Grayson, J., Nickerson, J., & Moonin, E. (2015). without using cost data: Illustrated
Partnering through Risk Management: Lake Mead methodology. Econometrica: Journal of the
Intake No. 3. Risk Management Approach. RETC June. Econometric Society, 256-275.
[34] Gregory, N. (2012) Improving Cost Estimation with [52] Rush, C., & Roy, R. (2000). Analysis of cost estimating
Quantitative Risk Analysis. Vose Consulting, processes used within a concurrent engineering
www.voseconsulting.com. environment throughout a product life cycle. In 7th
[35] Griffith, A., Stephenson, P., & Watson, P. ISPE International Conference on Concurrent
(2014). Management systems for construction. Engineering: Research and Applications, Lyon, France,
Routledge. July 17th-20th, Technomic Inc., Pennsylvania USA (pp.
[36] Heldman, K. (2018). PMP: project management 58-67).
professional exam study guide. John Wiley & Sons. [53] Rush, C., & Roy, R. (2001). Expert judgement in cost
[37] Humphreys, K. K. (1995). Basic cost engineering. CRC estimating: Modelling the reasoning
Press. process. Concurrent Engineering, 9(4), 271-284.
[38] Jensen, U. (2002). Probabilistic risk analysis: [54] Sander, P. (2016). Risk Management–Correlation and
foundations and methods. Dependencies for Planning, Design and
[39] Kermanshachi, S., Anderson, S., Molenaar, K. R., & Construction. ITA WTC 2016 Proc. April.
Schexnayder, C. (2018). Effectiveness Assessment of [55] Shaheen, A. A., Fayek, A. R., & Abou Rizk, S. M.
Transportation Cost Estimation and Cost Management (2007). Fuzzy numbers in cost range estimating. Journal
Workforce Educational Training for Complex Projects. of Construction Engineering and Management, 133(4),
In International Conference on Transportation and 325-334.
Development (p. 82). [56] Shane, J. S., Strong, K. C., & Gad, G. M.
[40] Kumari, S., & Pushkar, S. (2013). Performance analysis (2015). Risk-Based Engineers Estimate (No. MN/RC
of the software cost estimation methods: a 2015-10). Minnesota Department of Transportation,
Research Services & Library.

Volume 8 Issue 10, October 2020

www.ijser.in
Licensed Under Creative Commons Attribution CC BY
Paper ID: 18092001 12 of 13
 International Journal of Scientific Engineering and Research (IJSER)
ISSN (Online): 2347-3878
Impact Factor (2018): 5.426
[57] Shen, Z., & Issa, R. R. (2010). Quantitative evaluation
of the BIM-assisted construction detailed cost estimates.
[58] Sonmez, R. (2004). Conceptual cost estimation of
building projects with regression analysis and neural
networks. Canadian Journal of Civil Engineering, 31(4),
677-683.
[59] Touran, A. (2003). Calculation of contingency in
construction projects. IEEE Transactions on
Engineering Management, 50(2), 135-140.
[60] Trost, S. M. and Oberlender, G. D. (2003). "Predicting
Accuracy of Early Cost Estimates Using Factor
Analysis and Multivariate Regression." Journal of
Construction Engineering and Management 129(2), 198
- 204.
[61] Tsagkari, M., Couturier, J. L., Kokossis, A., & Dubois,
J. L. (2016). Early‐Stage Capital Cost Estimation of
Biorefinery Processes: A Comparative Study of
Heuristic Techniques. ChemSusChem, 9(17),
2284-2297.
[62] Whitesides, R. W. (2005). Process equipment cost
estimating by ratio and proportion. Course notes, PDH
Course G, 127.
[63] Williams, T. P. (1994). Predicting changes in
construction cost indexes using neural networks. Journal
of Construction Engineering and Management, 120(2),
306-320.
[64] Wilmot, C. G., & Cheng, G. (2003). Estimating future
highway construction costs. Journal of Construction
Engineering and Management, 129(3), 272-279.
[65] WSDOT (July 2009). Cost Estimating Manual for
WSDOT Projects.
[66] WSDOT (July 2010). Project Risk Management
Guidance for WSDOT Projects.
[67] WSDOT 2009, “Cost Estimating Manual for WSDOT
Projects,” Guideline Document
[68] Ashworth, A., & Perera, S. (2015). Cost studies of
buildings. Routledge.
[69] Evans and Peck Report for Department of
Infrastructure, Transport, Regional Development and
Local Government (2008), Best Practice Cost
Estimation for Publicly Funded Road and Rail
Construction, viewed 10 Jan 2013,
http://www.nationbuildingprogram.gov.au/publications/
administration/pdf/Best_Practice_Cost_ Estimation.pdf

Volume 8 Issue 10, October 2020

www.ijser.in
Licensed Under Creative Commons Attribution CC BY
Paper ID: 18092001 13 of 13