International Journal of Construction Management

ISSN: 1562-3599 (Print) 2331-2327 (Online) Journal homepage: https://www.tandfonline.com/loi/tjcm20

System dynamics modeling approach to quantify

change orders impact on labor productivity 1:
principles and model development comparative
study

Zain Ghazi Al-Kofahi, Amirsaman Mahdavian & Amr Oloufa

To cite this article: Zain Ghazi Al-Kofahi, Amirsaman Mahdavian & Amr Oloufa (2020): System
dynamics modeling approach to quantify change orders impact on labor productivity 1: principles
and model development comparative study, International Journal of Construction Management,
DOI: 10.1080/15623599.2020.1711494

To link to this article: https://doi.org/10.1080/15623599.2020.1711494

Published online: 11 Jan 2020.

Submit your article to this journal

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=tjcm20
INTERNATIONAL JOURNAL OF CONSTRUCTION MANAGEMENT
https://doi.org/10.1080/15623599.2020.1711494

System dynamics modeling approach to quantify change orders impact on labor

productivity 1: principles and model development comparative study
Zain Ghazi Al-Kofahi, Amirsaman Mahdavian and Amr Oloufa
Department of Civil and Environmental Engineering, University of Central Florida, Orlando, FL, USA

ABSTRACT KEYWORDS
In construction projects, change orders are commonly issued by the owner and may cause interruption System dynamic; change
to the contractor’s work, resulting in loss of labor productivity, delay, and cost overruns, which may lead orders; labor productivity;
to claims. Because of the complexity in linking the cause of the productivity loss to the change order, the model development;
construction projects
relationship is not well understood. The contractor must prove with a credible calculation that the prod-
uctivity loss was a result of the change order issued by the owner to be able to receive compensation.
Compared to all available productivity loss quantification methods, the measured mile approach is con-
sidered the most acceptable and popular in litigation. This study is a discourse on the understanding,
use, and potential application of system dynamics within quantifying change orders’ impact on labor
productivity. The model boundary, including Endogenous and Exogenous variables, are identified. After
determining the parameters used in modeling, the CLD is constructed to understand the system structure
better. A system dynamics model is developed using Vensim software, verified, validated, and utilized to
quantitatively measure the influence of the change in the project scope on labor productivity. To exam-
ine the applicability of the proposed model, it has been implemented to a real highway construction
case study, and the impacts of various variables on labor productivity loss have been studied using the
data for two completed highway construction projects from FDOT. The study can be applied to construc-
tion projects that faced change orders, and as a result of these changes, the contractor’s performance
was affected. The proposed model can be used to determine the primary sources that result in labor
productivity loss on a specific type of task.

Introduction credible estimate, proving that the loss of productivity is a result

of a change order issued by the owner. To have a reasonable
In several countries, the construction sector captivated critique
estimation of the cost variation and make sound project deci-
for a low level of productivity and inadequate quality of the sions, it is necessary to understand the types of costs that gener-
work (Eriksson and Westerberg 2011). Tran (2011) conducted a ally occur in construction projects. The main expenses that
comprehensive investigation of labor productivity in the New impact construction projects are fixed and variable costs. Fixed
Zealand construction sector. He asserted that productivity costs do not depend on the activity level. They are unavoidable
increase is considerably correlated to the economic extension. and must be paid regardless of the level of output and the
The productivity of labor is of remarkable concern to construc- resources used in various projects. The risks associated with fixed
tion services as they are profoundly labor-intensive. Yi and Chan costs are relatively low. On the other hand, variable costs are
(2014) affirmed that labor productivity as a vital productivity defined as the costs that change during the project’s lifecycle.
metric because of the labors required to perform a particular job. Labor rates are the main variable cost item in any construction
Change orders are common issues in construction projects project, determining most of the project cost. As a result, labor
which may have a positive or negative impact on the construc- productivity is considered a significant factor in determining
tion projects’ duration, cost, and quality. Although change orders profit on a project by contractors.
can increase the efficiency by using a better method of work, The literature review section investigates the causes of labor
substituting a higher strength/cost material with the planned productivity and their relationship to liability and change orders.
material, the expected norm in most construction projects ranges Moreover, the methods used by the industry to quantify the loss
from an increase of between 5 and 10 percent to the project’s in labor productivity are also covered. The literature review sec-
cost due to such changes (AACE 2004). Olsen et al. (2012) clas- tion helps to provide a better understanding of the variables
sified the causes of change orders in six categories, including affecting productivity loss due to the change orders. Since most
unforeseen/existing conditions, plans and/or specifications errors of the change orders lead to negative impacts, the primary goal
and omissions, owner directed scope change (additions/dele- of this research is to develop a system dynamics model that can
tions), value engineering, force majeure, and acceleration/delay be used to quantitatively measure the negative impact of the
in the project schedule. The causal relationship between a change owner changes in the project scope on labor productivity. Thus,
order and loss of productivity is not well understood. Therefore, the methodology section describes the basics of how the system
to be entitled to compensation, the contractor must provide a dynamics model is built and then presents steps to develop the

CONTACT Amirsaman Mahdavian amirsaman@knights.ucf.edu

ß 2020 Informa UK Limited, trading as Taylor & Francis Group
2 Z. G. AL-KOFAHI ET AL.

model. Ultimately, the model is verified and validated within sev- schedule acceleration methods while experiencing a delay and
eral steps by utilizing two completed highway project sections are forced to speed up the project. Schedule acceleration may
from FDOT. lead to long periods of mandatory overtime, the addition of
second shifts, stacking of trades, and overcrowding of laborers
beyond the site’s saturation point, making it challenging to man-
Literature review age or coordinate efficiently. Moreover, rework adds a substantial
There are several different definitions of labor productivity. In amount of material and several tasks to be re-planned and re-
the construction industry, it is merely the working hours sequenced. Moreover, it results in delays, modifications in the
required to accomplish a given unit of work. In general, con- learning curve, an increase in engineering errors and omissions,
struction contractors are paid after the completion of the work adjustments of specifications. Regarding project management fac-
by the terms of the contract implying that productivity is related tors, deficiencies in the management could adversely affect the
to project cash flow and project profitability. Therefore, a con- availability of critical resources, including materials, machines,
tractor must concentrate on obtaining the highest output from and personnel, at the necessary time, cause inefficient dynamics
the lowest input resources, striving for the highest possible prod- between team members and crews, and result in a failure to cor-
uctivity in a project (Dolage and Chan 2013). A loss in a con- rectly schedule and coordinate the work and an incorrect mix-
tractor’s labor productivity directly affects the associated profit. ture and use of labor crews. Productivity might also halt while
In any construction project, labor productivity loss may occur awaiting approval, redirection, or information when there is a
due to numerous circumstances, and certain events create a chal- delay in response time among owners, designers, and construc-
lenge in estimating the impact of changes on productivity. This tion managers.
complexity is possibly the reason for the difference in productiv- Regarding the learning curve, at the outset of any project,
ity for the same construction item on different projects, there is a common learning curve while the laborers become
Herbsman and Ellis (1990). Common causes of productivity loss familiar with the project, its location, the quality standards
include site or work area access restrictions, project location and imposed, laydown area locations, etc. However, if the method of
external conditions, contractors using schedule acceleration the work of the project changes, labors may have to go through
when a delay happens, rework, learning curve impact, and pro- another learning curve. This is probably an unanticipated impact
ject-mismanagement-related issues (AACE 2004). Soekiman et al. on labor productivity. If this occurs more than once, then each
(2011) investigated multiple variables influencing labor product- time a work stoppage happens, another learning curve productiv-
ivity in Indonesia; financial difficulties, labor strikes, delay in the ity loss may result.
arrival of material, lag of equipment and design change, higher
absenteeism of labor, supervisors absenteeism, no administration
system, unclear instruction to labor, lag of materials, are among Construction change orders and construction claims
the shortlist he prepared. Hickson and Ellis (2014) organized
Changes usually increase construction claims between the con-
productivity variables into four principal groups, including man-
tracted parties; managing change orders is difficult for both con-
agement, human/Labor, technological, and external.
tractors and owners (Habibi et al. 2018). While it is evident that
Approximately 24 variables were considered in management var-
change orders have a direct influence on project performance
iables, the essential variables based on importance index were a
(Safapour et al. 2018), it is often challenging to recognize the
shortage of labor supervision, unreliable scheduling, and expect-
ation of labor production, communication, leadership deficit, indirect impacts (Kermanshachi et al. 2016). Ibbs and Vaughan
and payment delay. Robles et al. (2014) classified a collection of (2015) assert that changes in a construction site could be over-
35 variables to distinguish variables influencing labor productiv- whelming and have an essential influence on project success and
ity in Spain considering their relative importance. Five various profitability. The resulting changes not only disturb the work
classes were grouped, including project, human, management, that is directly impacted by them, but they may also have indir-
material, and tools, environmental. ect impacts. The relationship between change orders and loss of
In some projects, having access to the working area is not labor productivity is not well understood because the link
easily provided. Various factors including the danger of working between the two instances is not evident. In order to be entitled
at height and existence of chemical substances, the difficulty of to compensation, the contractor needs to prove with a credible
reaching to the site location due to the distance or road condi- calculation that the productivity loss is a result of the change
tions, and limited space of the site for movement of labors, stor- order issued by the owner. Several methods were developed by
ing materials and installing equipment, are among the reasons the construction industry to estimate and measure the loss in
related to the site access restrictions. Site or work area access labor productivity. The following section illustrates these meth-
restrictions can lead to a shortage of employees, materials, and ods with their advantages and limitations.
equipment needed on-site to work efficiently, delays, and a time
crunch for the contractor to complete work. Direct repercussions
include reduction of direct supervision, and reduced integration Methods used to quantify loss in labor productivity
between trades and may ultimately lead to excessive travel time The current methods being used to quantify the loss in labor
between the assembling area and working area (AACE 2004). productivity are divided into three main groups, based on the
Project location and external conditions include physical condi- applied input data by Nelson (2011): project practice-based
tions (saturated soils), logistical conditions (low-hanging power method, cost-based method, and industry-based method.
lines), environmental and legal conditions (permit requirements
affecting when construction can take place), availability of skilled
labor, and the local economy (AACE 2004; Serag 2006). Project practice-based methods
Moreover, adverse or unusually severe weather conditions can Project practice-based approaches are the most preferred by the
negatively affect the productivity of the projects. Contractors use courts, boards of contract appeals, and other legal forums
 INTERNATIONAL JOURNAL OF CONSTRUCTION MANAGEMENT 3

(AACE 2004) since their calculations depend on the contempor- Industry-based methods
aneous documentation of the disputed project. This method includes specialty industry and general industry
studies. Industry-based analysis methods can be used when there
Measured mile analysis. Compared to all available productivity is a lack of project documentation. Also, they can be used with
loss quantification methods, the measured mile is considered the another method to augment supportive evidence of damages.
most acceptable (Ibbs 2005). The analysis in the measured mile What industry-based analysis methods gain in speed of calcula-
method is done by comparing the unaffected period of the pro- tion and cost savings, they lose due to lack of contextual rele-
ject with the impacted period (Shwartzkoph 1995). The basic vance and statistical measures about the specific project in
approach is to identify an unaffected or least impacted period of question; hence, other methods are preferred (Nelson 2011).
construction activity, linearly extrapolate the cumulative
unaffected hours to the end of an impacted period and use the Specialty industry studies. These are about specific topics such as
difference between the projected unaffected hours and the actual acceleration, learning curve, overtime, and weather. These studies
cumulated hours as the number of damage hours (Gulezian and examine matters directly related to the causation of change
Samelian 2003). orders and resultant claims.

Baseline productivity analysis. The baseline productivity analysis General industry studies. These studies are based on industrywide
was proposed to overcome some of the limitations associated manuals and/or reports.
with the measured mile analysis. Similar to the latter, the base-
line productivity method depends on the actual performance of
the contractor in the project under investigation. In this analysis System dynamics modeling
method, the main point is to define a baseline period when the System dynamics (SD) analytical modeling is derived from Jay
contractor performs the best productivity. It is not necessarily a Forrester’s work on industrial dynamics at the Massachusetts
continuous, unaffected timeframe, nor is it a purely unaffected Institute of Technology. SD tends to the qualitative and quantita-
period; owner- and contractor-caused inefficiencies may be pre- tive aspects of an identified problem (Forrester 1968). This aims
sent throughout the project. A research was conducted by to provide a deeper understanding of the problem and its varia-
Thomas and Zavrski (1999) to develop the theoretical basis for bles. Forrester used computer simulation to highlight the impact
the baseline productivity measurement. After analyzing a forty- of time on the problem. According to Forrester, SD demon-
two-project database, the main findings show that baseline prod- strates how most decision-making policies are the cause of the
uctivity depends on the complexity of the design. As the design problems that are usually blamed on others and how to identify
becomes more complicated, the baseline productivity worsens.
policies to follow to improve these situations. In the context of
Also, it shows that the baseline productivity depends on the
project management, SD approach a holistic view of the process
management, craft skills, and technology used.
of project management, and reveal how most project manage-
ment decision-making policies are the cause of the problems
Earned value analysis. The earned value method can be used to (Forrester 1968).
measure productivity when there is a problem in obtaining pre-
cise data for the physical units of work installed on the project,
which prevents the use of one of the most reliable methods, such Other methods
as a measured mile or baseline analysis. In this method, the value
of payments and the unit price are used to determine the earned Moselhi et al. (2005) developed a model employing a neural net-
hours (Ibbs et al. 2007). The difference between the calculated work for predicting the productivity loss that happened due to
earned hours and the actual hours expended for an impacted the effect of the change orders on construction operations. They
period can be used to compute the productivity loss experienced concluded that the neural network gives better results as com-
(Ibbs et al. 2007). Errors can arise when using this analysis pared with the other models that have been generated using vari-
method due to uneven dollar values for different types of work ous techniques. Cheng et al. (2015) developed an Evolutionary
(Schwartzkopf 1995). Fuzzy Support Vector Machines Inference Model (EFSIM) to
predict change-order-related productivity losses more precisely.
Simulation outcomes and their validation against prior studies
Cost-based methods prove that the EFSIM predicts the impact of change orders
Total cost method. In this method, the contractor claims to remarkably better than other AI-based tools, including the artifi-
recover its entire man hours overrun, which can be calculated by cial neural network (ANN), support vector machine (SVM), and
subtracting the actual man hours from the estimated man hours Evolutionary Support Vector Machine Inference Model (ESIM).
on a given scope of work. The total cost method is an imprecise El-Gohary et al. (2017) introduced a practical and standardized
method. Thus, it is not usually accepted in the courts, and it is engineering approach to control, predict, and then improve con-
not recommended for claims (Serag 2006). struction labor productivity. The technique of ANN was used to
quantify and map the relationship between construction labor
Modified total cost method. The modified total cost method is an productivity and the relevant influencing factors. Also,
excellent alternative to the overall cost method; this is especially Kermanshachi et al. (2018) asserted that if the project falls
true in claims in which changes have had a significant impact. behind schedule due to change and the project deadline contin-
This method examines individual cost codes rather than the costs ues fixed, schedule pressure drives a rise in productivity at the
of the entire project (Serag 2006). A strength of the modified beginning but ultimately may lead to disappointment. This dis-
total cost method is that it considers unreasonable estimations appointment has the potential to expand the project duration
and contractor inefficiencies in its calculations (Nelson 2011). farther and causes an unsatisfactory situation. Schedule pressure,
4 Z. G. AL-KOFAHI ET AL.

moreover, worsens the problem by typically extending the Agreement) was issued by the owner to increase the scope of
rework fraction or the errors on the built work packages. work by 3% due to the plan’s modification.
After data was collected from the projects, the next step was
to prepare the data to start building the model. Through the
Methodology data extracting and preparation stage, it was observed that the
The model developed in this study was created using an SD soft- contractor was working inefficiently even before the issuance of
ware package called Vensim. Vensim is an industrial-strength change orders. This study will focus on measuring the loss of
simulation software for enhancing the performance of real sys- productivity for road projects. The productivity measure will be
tems. Vensim’s rich feature set intends for modeling quantities expressed as working hours per unit installed. The level of aggre-
that vary over time, and highlights model quality, connections to gation of the data used in the model stock and flow is appropri-
data, adjustable distribution, and high-level algorithms. This ate in terms of what was being measured: working hours,
study employs Vensim to investigate the impact of change orders productivity, and work done over some time. In construction
on labor productivity, quantifying productivity loss, and linking projects, work is performed daily, while the performance reviews
and work inspection are performed monthly. Based on that, the
this loss to the causes. To develop the framework, this research
data was configured to show results weekly; the working hours
follows some steps, including building the model, specifying
are summed up weekly while the quantities installed were div-
inputs, running the model, examining the output, and finally,
ided into a weekly performance instead of a monthly one. The
interpretation and analysis. Firstly, the scope and the boundary
researcher used Florida Automated Weather Network, FAWN,
of the including the Endogenous and Exogenous Variables are
and website, and from the archived weather data, he reported
identified by studying the literature. After determining the
the temperature values during the project periods based on the
parameters used in modeling, the CLD is constructed to under-
project site and the particular days of work.
stand the system structure better. Based on highly established
Alvanchi et al. (2012) developed a revised dynamics of fatigue
and validated models, an SD model, the stock-flow diagram
models that could be utilized by the work manager to determine,
(SFD), was developed. Then, according to Sterman (2000), five
with a fair degree of accuracy, the predicted productivity in sev-
tests were applied to verify the SD model to ensure that the
eral different construction jobs under various combinations of a
model is correctly implemented according to the conceptual
working hour. This research studied four key variables impacting
model. Ultimately, regarding model behavior validation, different labor productivity loss, including mental and physical fatigue,
tests were suggested to evaluate SD behavior quantitatively, and overtime work, and time of day. Chang et al. (2007) examined
such tests focused on major time patterns rather than individual how the schedule compression could affect construction labor
data points, including reference mode and statistical behavioral productivity and provided a model quantifying the influence of
validation. Work done over the project period is selected as the schedule compression on labor productivity. Their model com-
reference mode to illustrate the multidimensional patterns of pared two distinct schedule compression conditions; one sample
behavior for this system arising through the nonlinear inter- study is when the construction period is compressed, and the
action of the subsystems with one another. It also used in the other sample study is when the contractor is required to conduct
validation of the SD model. Actual quantities installed are com- further work within the original period. Love et al. (2011)
monly observed and are an easily recorded reference mode of studied the dynamics of rework and the issues related to it in
behavior. The problem is dynamic and complex because the ref- complex hydrocarbon (oil and gas) projects. They built a general
erence mode varies over time, is affected by many factors in the CLD that could be used by the managers to provide them with a
system, and is affected by the feedback loops in the system. better understanding of the interrelationships between factors
Some of the factors that influence the reference mode are the that contributed to reworking; thus, rework prevention might be
amount and timing of change orders, weekly worked hours, the included in future projects. Considering that design errors are
number of laborers on site, laborers’ productivity, overtime widespread, most design and construction companies fail to
hours, rework, fatigue, labor crowding, and temperature. measure the errors they make, which leads to less familiarity
The first and most crucial step is to determine the projects’ with the procedures that undermine project performance. Han
criteria under study. The critical projects criteria are the Heavy et al. (2013) developed a system dynamics model to catch the
Construction Road Projects type, with public owners. The dynamics of design errors that drive to rework and design modi-
Projects that were 100% completed and encountered change fications. While design errors are deemed prevalent, most design
orders were studied. The data source to collect the data were the and construction companies do not consider the number of
daily work reports, contract documents, changes, and inspection errors they produce, thereby having insufficient understanding
reports from the Florida department of transportation (FDOT). concerning their mechanism to undermine project performance.
Ultimately, two case studies were selected for this study. In the Nasirzadeh and Nojedehi (2013) designed an SD model to
1st case study, FDOT engaged a primary contractor to widen and describe the impact of various interrelated influencing variables
add lanes for a road in the state project “XXXXXXX.” This pri- that affect labor productivity. Hence, the project manager can
mary contractor hired a sub-contractor to build the cement evaluate the impact of various answers to enhance labor product-
concrete pavement of this road. The scope of work of this sub- ivity. There are many factors positively influencing labor prod-
contract included placing 74781SY plain cement concrete pave- uctivity, including project management competence, skillfulness,
ment (12 1/2”). After 12 weeks of work, a change order and motivation. There are some other factors negatively affecting
(Supplemental Agreement) was issued by the owner to increase labor productivity, including lack of operating area, labor fatigue,
the scope of work by 11%, due to the plan’s modification. In the and unfamiliarity with a novel method. Oliva and Sterman
2nd case study, FDOT engaged a primary contractor to perform (2010) developed a dynamic model with a wide boundary to
an 8-mile road resurfacing in the state project number catch the structural properties of the service delivery means,
“XXXXXXX.” The total bid amount equals to $3,396,600.00. stakeholders’ decision-making manners, and the development of
After 28 weeks of work, a change order (Supplemental expectations for employees and clients. Extra scenarios are
 INTERNATIONAL JOURNAL OF CONSTRUCTION MANAGEMENT 5

Table 1. Model boundary (endogenous and exogenous variables).

Endogenous variables (Affected) Exogenous variables (Not affected)
Available time Average daily workers on site
Desired completion rate Baseline productivity
Desired workforce Fatigue onset time
Effect of learning on productivity Labors on site
Fatigue Learning rate
Man-hours Maximum workweek
Maximum completion rate Scope change
Productivity Standard workweek
Remaining work Target delivery time
Scheduled overtime Temperature
Work being done Time for inspection and discovered error
Work really done Work to do
Workweek

required to be tested to strengthen the model. They created the

model up in steps to accentuate the dynamics of each notable div-
ision – budgeting, workflow, human resources, hiring, standards
for quality, and client feedback to quality. Warhoe (2013) devised
an SD model with which construction projects are delivered using
the design-bid-build (DBB) methodology can be simulated to
show how the factors that can affect the construction projects can
Figure 1. Causal Loop Diagram (CLD) of labor productivity.
influence labor productivity. His model has both exogenous and
endogenous variables. Endogenous variables include workers hired,
overtime hours, error generation rate, total effort applied, work 2. Work To Do ! þ Schedule Delay ! þ Overtime ! þ
performed, and workers laid off. Exogenous variables include daily Fatigue ! - Productivity ! þ Work Done ! Work To Do
worker productivity, estimated project labor hours, scope change (Reinforcing 1)
identification, the time it takes to hire a worker, timing of 3. Work To Do ! þ Schedule Delay ! þ Hire Labors ! þ
changes, and estimated baseline work rate. This study also ignored Crowding ! - Productivity ! þ Work Done ! Work To
and removed some variables from the model including resource Do (Reinforcing 2)
availability, other project delivery methods, major force issues, 4. Work Done ! þ Learning ! þ Productivity ! Work Done
weather impacts, disputed changes, equipment hours and product- (Balancing 2)
ivity, and safety influences on fatigue and morale. 5. Work To Do ! þ Hire labors ! þ Work Done ! - Work
The devised model in this study, modified the previously To Do (Balancing 3)
mentioned models, by incorporating the broad set of endogenous
and exogenous variables, while considering scenarios for the Once a change order is issued, the number of units needed to
inclusion of the impact of learning curve rate. Different tests accomplish the project is changed, and some finished work might
were suggested to evaluate SD behavior quantitatively, and such need to be reworked because the change order also increased the
tests focused on major time patterns rather than individual data number of items required to perform. To avoid delay in the project,
points, including reference mode and statistical behavioral valid- and based on the targeted delivery time, the contractor might use a
ation. To examine the applicability of the proposed model, it has
work acceleration technique, such as implementing overtime hours.
been implemented to a real highway construction case study,
However, too much overtime for successive weeks will cause fatigue
and the impacts of various variables on labor productivity loss
in the workers, resulting in productivity loss and, therefore, decreas-
have been studied using the data for two completed highway
ing the number of accomplished units per week (reinforcing loop).
construction projects from FDOT. Table 1 summarizes the scope
The contractor might choose to hire more laborers to increase the
of the model by presenting the most critical endogenous varia-
number of items accomplished per day (balancing loop). However,
bles, which are the dynamic variables involved in the feedback
too many laborers onsite lead to overcrowding, implying that the
loops of the system, so they enable the exploration of the pat-
available space per laborer to work efficiently decreases and will
terns of behavior created by the rules among them and the dis-
generate productivity loss and a consequent decrease in the number
covery of how the behavior might change if those rules are
of accomplished units per day (reinforcing loop). As the number of
altered. The model also contains several exogenous variables
items finished increases, the labor learning effect increases. This will
whose values are not directly affected by the system.
lead to the reduction of time required to complete the activity,
After determining the parameters used in modeling, the CLD
is constructed to understand the system structure better. Based increasing labor productivity, which means increasing the number
on cause-effect relationships among parameters, the causal (feed- of accomplished units per day (balancing loop).
back) loops are determined to outline the relationships among
system variables, which are not necessarily linear but circular Model formulation
chains of cause and effect, as shown in Figure 1; this figure dis-
plays five causal loops, three of them balancing loops and the An SD model, the stock-flow diagram (SFD), was developed
others reinforcing loops. An explanation of the defined loops is based on highly established and validated models in the previous
presented below. A positive sign indicates a reinforcing impact, studies (Chang et al. 2007; Oliva and Sterman 2010; Alvanchi
and a negative sign indicates a balancing influence. et al. 2012; Han et al. 2013; Nasirzadeh and Nojedehi 2013;
1. Work To Do ! þ Schedule Delay ! þ Overtime ! þ Warhoe 2013). When the contractor wins the bid, he or she
Work Done ! - Work To Do (Balancing 1) must do a certain amount of work as specified in the bid
6 Z. G. AL-KOFAHI ET AL.

Figure 2. Stock and flow diagram.

documents’ “work to do.” During the simulation, this amount of that has been input into the model generates the output or if
work will flow through “work being done” toward the “work the model is producing the output. When the input parameters
done” stockpile at a prescribed work rate, depending on the con- are modified, can it be seen that the model is making the out-
tractor’s best productivity. When the project owner issues change put? The significant endogenous and exogenous variables used
orders that add work items to the project scope, it is hypothe- in the model are summarized in Table 1. All significant varia-
sized that the productivity can be impacted. Factors such as bles (productivity, work done, overtime, fatigue, crowding, and
scope change, temperature, labor crowding, overtime, fatigue, learning) are generated endogenously, whereas scope change,
and learning effects create feedback loops that affect the contrac- work to do, temperature and hiring of laborers are the exogen-
tor’s productivity, as well as the work rate flow, as shown in ous variables. All the variables used are based on concepts that
Figure 2. This model studies the impact of these different factors have existed and have been in use in the construction industry
on productivity, labor hours needed to finish and deliver the for years.
project on time, and the rate of work done. 2. Does the behavior of the model change significantly when
boundary assumptions are relaxed? For the model in this study,
Model verification the behavior changes as would be expected when the boundary
assumptions are relaxed. Increasing or decreasing the boundary
Model testing is a critical step in the modeling process to ensure assumptions will create a cause-and-effect chain that will impact
that the model is correctly implemented according to the con- other aspects of the model. For example, if the overtime bound-
ceptual model. During the verification process, errors are
ary is relaxed, the fatigue impact on labor will be affected; there-
revealed and fixed to create the best model available to study the
fore, labor productivity and the amount of work done will
case in question, in addition to understanding the model limita-
be affected.
tions. This section presents the process used to verify and test
the suitability of the model for its specific purpose, build confi-
dence in the simulation model, discover flaws, fix it, and
improve the model so that it can be useful.
According to Sterman (2000), five tests were applied to verify
Structure verification
the SD model:
A necessary and essential test in the verification process is struc-
ture verification. “Structure assessment tests ask whether the
Boundary adequacy
model is consistent with knowledge of the real system relevant to
According to Sterman (2000), the boundary adequacy test the purpose” (Sterman 2000). A CLD represents the cause-and-
answers the following questions: effect relationships among variables. Once a change order is
1. Are the essential concepts for addressing the problem issued, the number of units needed to accomplish the project is
endogenous to the model? This question asks whether the data changed, and some finished work might need to be reworked.
 INTERNATIONAL JOURNAL OF CONSTRUCTION MANAGEMENT 7

Dimensional consistency variations: (1) spot overtime and (2) scheduled or extended over-
time. The former is used to address unexpected changes or to
The Vensim software has a built-in function that notifies the complete time-sensitive work. The latter is offered as an incen-
user if there are any inconsistencies with dimensions. tive to attract labor or to complete a project earlier than would
Additionally, each mathematical equation used to formulate the be done with the standard workweek. Overtime is commonly
SFD is tested by the researcher to confirm that the measurement selected for acceleration when the impact of a change is not fully
units of the used variables and constants are dimension- known. Other techniques should be considered, however, because
ally consistent. overtime is costly. Part of this cost is because workers typically
The following are two of these equations with their dimen- receive remuneration at 1.5 times the standard wage. An add-
sional consistency analyses: itional reason for the cost is that hourly productivity does not
desired completion ratecontractor productivity increase and declines due to physical fatigue.
desired workforce ¼
standard workweek Overmanning. Overmanning means increasing the number of
workers within the same trade on a project. Using overmanning,
This equation describes that the required number of workers the contractor can achieve a higher production rate without the
(desired workforce) to deliver the project on time depends on fatigue issue that is inherent with overtime. However, labor con-
the required completion rate, actual contractor productivity, and gestion and decreasing supervision will lead to a decrease in
the standard two (working hours/week, standard workweek). labor productivity.
Therefore, the dimensional analysis for this equation is: Temperature. A study by Fox (1967) explored the effect of
unit laborhour temperature on labor productivity. This study reported that
week  unit
worker ¼ hour
¼ worker “there is a critical hand surface temperature (HST) below which
week performance is significantly affected and above which there are
Thus, the equation is dimensionally consistent. few effects.” Moreover, for tactile sensitivity, the critical HST is
near 8  C (46  F). For manual dexterity, the critical HST is some-
Man hours
Work Being Done ¼ what higher, between 12  C (54  F) and 16  C (61  F)
productivity (Schwartzkopf 1995).
This equation describes the number of units accomplished Learning Effect. Repetition promotes familiarity. Therefore,
per week (work being done) as dependent on the number of repeated actions require less effort and can be completed in less
hours spent by the laborers per week (man hours) and actual time as the number of these repetitions increase.
labor hours spent to produce one unit of work (productivity). Do all parameters have real-world counterparts? All parame-
The dimensional analysis for this equation is: ters and values used in model building, both CLD and SFD, are
hour
consistent with the knowledge of the construction industry and
unit labor week unit have counterparts in the real world, as discussed in the previ-
¼ laborhour ¼
week unit
week ous question.

Extreme condition test

Parameter verification
“Models should be robust in extreme conditions. Robustness
According to Sterman (2000), the parameter verification test
under extreme conditions means the model should behave realis-
answers the following questions:
tically no matter how extreme the inputs imposed on it may be”
Are the parameter values consistent with relevant descriptive
(Sterman 2000). In this test, extreme values were given to
and numerical knowledge of the system? All parameters values
selected model parameters, and then the simulation-generated
used in the model building of the CLD and SFD are consistent
behavior was compared to the behavior of the real system. The
with the knowledge of the construction industry. The values
most important variables on the results were evaluated at their
given to the parameters of this model are based on the existing
extreme values during sensitivity analysis.
project data and knowledge found in published papers and
research. The following is an illustration of some of the parame-
ters used and their values. Model validation
Acceleration techniques are used to increase the productive
labor hours on a job. These techniques include overtime, over- Model behavior validation, and SD model validation, in particu-
manning and trade stacking, and shift work, all of which are lar, is possibly the essential part of simulation validation in gen-
used by the contractor to complete work earlier than scheduled, eral. The processes by which model validation is accomplished
to overcome the delay in project schedule due to changes, or to include model conceptualization that follows the implementation
make up for a material delay. Although each one of these accel- of policy recommendations.SD models have specific characteris-
eration techniques will increase the daily production on a job, it tics that make standard statistical tests inappropriate. Different
is not true that doubling the number of hours or doubling the tests were suggested to evaluate SD behavior quantitatively, and
worker power will double the output. There is an inherent loss such tests focused on major time patterns rather than individual
of labor productivity with each of these acceleration techniques. data points, including reference mode and statistical behav-
Overtime. The simplest way to accelerate a project is to ioral validation.
increase man hours with the use of overtime. This versatile tech-
nique retains the original number of workers who are already
Reference mode (behavior reproduction)
familiar with the project. It also maintains the original sequence
so that there are not new needs for coordinating multiple trades Here, the historical fit is evaluated. As indicated in Figure 3, the
and workers within a specific area. Overtime has two main results of the simulation, including all the effects, relatively
8 Z. G. AL-KOFAHI ET AL.

Figure 3. Validation result, including all effects (1st case study). Figure 4. Validation results, without learning effect (2nd case study).

Figure 5. (a) Work really done statistical model curve. (b) Residual plots for work really done statistical model.
 INTERNATIONAL JOURNAL OF CONSTRUCTION MANAGEMENT 9

the original data was split into two parts. The first part was used
to build the statistical model and estimate its parameters, and
the second part was used to assess the fitted model’s predictive
ability and to run the model, tabulate the estimated values, and
compare them to their counterparts from the Vensim model.
Using statistical analysis, a regression equation for cumulative
man hours is developed as follows:
y ¼ Work Really Done, x ¼ Week Count
y ¼ 0:02036 þ 0:01367 x 0:000115 x2 þ 0:000001 x3
S ¼ 0:0118483 R ¼ 99:8%, R2 ðadjustedÞ ¼ 99:7%
2

The coefficient of determination R2 has a value of 99.8 per-

cent, implying that the model equation relating work really done
to the week can explain that 99.8 percent of the variation
Figure 6. Work really done results of system dynamics model vs. statistical presents in the data values of work really done. The plotted
model (1st Case). cumulative man-hours graph, along with its residuals plots, are
shown in Figures 5a and 5b, respectively.
accurately reproduce the project data regarding the work really The second part of the split data was used to validate the
done. Where this variable is endogenously generated in the model as presented in Table 2A in Appendix A.
model and sufficiently serves the purpose of the investigation As shown in Figure 6, the statistical model predicted data
(because its value is affected by the value of labor productivity), closely resembling the SD model data, and thus our model has
weekly man hours are used as well as the decision made by the been statistically validated.
project contractor regarding the number of laborers used. The
proposed SD model was simulated using exogenous scope change
values; these values were added into the “work to do” stock. The Conclusions
model considered the new work scope endogenously and In productivity loss claims, the analysis assumes that the differ-
addressed the effect on labor productivity and, therefore, on the ence between actual outcomes and the measured-mile prediction
needed work hours to complete the project. To compute the cor- is all claimable, whereas some or all of this difference may be
respondence between model output and actual project data, the under the contractor’s responsibility. Most of the published mod-
R2, as well as the adjusted R square, were calculated as follows: els that study the problem of loss of productivity suggest linear
R-Square ¼ ((COVAR (Data 1, Data 2))/(STDEV (Data models, while in this research, the nonlinear relationships among
1)(STDEV (Data 2)))) different factors and productivity are taken into account during
the model formulation. For the contractor to be entitled to com-
 
ð1 R2 Þðn 1Þ
R2adj: ¼ 1 pensation, the productivity claim must prove that the contractor
n k 1
suffered a loss, the owner caused the loss and the value of it.
Where: The available methods used to quantify the loss in labor product-
N: The number of points in data sample ivity assume that all the loss during the impacted period is due
K: The number of variables in the model to the owner’s part, which is not always a valid assumption.
R2, equal to 94.63 percent, and adjusted R2, equal to 94.33 Thus, the inefficiencies on the contractor’s part can also be con-
percent, were observed, allowing the conclusion that the refer- sidered during the calculations. Many different factors must be
ence mode provided by the proposed SD model is valid; the considered when trying to examine any change in a project.
same pattern of behavior is shown with the actual cumulative These factors affect productivity, and it is vital to recognize
man hours. which ones can have an impact on any particular job site. This
The model was rerun without including the effect of the research highlighted the problem of productivity loss resulting
learning curve; results are shown in Figure 4. R squared is equal from owner-liable change orders in construction projects. The
to 95.34 percent, and adjusted R squared is equal to 95.15 per- focus was on disagreements on quantification methods of
cent. It must be noted that both runs are compared to the actual changes and resultant productivity loss costs in road construction
project output. The higher R-squared and adjusted R-squared projects. Therefore, an SD model was developed and validated to
values imply that the model assumes that the laborers follow a analyze the changes in labor productivity due to the owner
particular learning curve as they become more experienced with change orders to address the disputes between the owner and
the project, but in reality, they did not acquire the experience at contractor. The model boundary, including Endogenous and
the same rate that the model assumed. Exogenous variables, are identified. After determining the param-
eters used in modeling, the CLD is constructed to understand
the system structure better. An SD model, the stock-flow dia-
Statistical behavioral validation
gram (SFD), was developed. Different tests were suggested to
In this section, Vensim model behavior is assessed using statis- evaluate SD behavior quantitatively, and such tests focused on
tical model validation, which implies how successful the model major time patterns rather than individual data points, including
will be when applied to new data. Different statistical validation reference mode and statistical behavioral validation. The results
techniques have been designed using a comparison of the values of the reference mode simulation, including all the effects, rela-
predicted from the Vensim model and an independent validated tively accurately reproduce the project data regarding the work
statistical model (Fahmy 2015). A statistical model was built for done. The proposed SD model was simulated using exogenous
the actual work done within the project period. Since the avail- scope change values; High R-squared value allowing the conclu-
able data was limited, and it was impossible to collect new data, sion that the reference mode provided by the proposed SD
10 Z. G. AL-KOFAHI ET AL.

model is valid. Vensim model behavior is assessed using statis- Gulezian R, Samelian F. 2003. The productivity baseline. AACE International
tical model validation. Using statistical analysis, a regression Transactions. EST.02. Morgantown, WV, USA: AACE International.
Habibi M, Kermanshachi S, Safapour E. 2018. Engineering, procurement and
equation for cumulative working hours is developed. The coeffi- construction cost and schedule performance leading indicators: state-of-
cient of determination R2 is implying that the model equation the-art review. Proceedings of Construction Research Congress (CRC);
relating to work done to the week can explain that 99.8 percent New Orleans (LA): ASCE. p. 378–388.
of the variation presents in the data values of work done. Han S, Love P, Pena-Mora F. 2013. A system dynamics model for assessing
the impacts of design errors in construction projects. J Math Comp
Additionally, the model accounts for the nonlinear behavior Model. 57(9–10):2044–2053.
that real projects experience. Ultimately, the model provides Herbsman Z, Ellis R. 1990. Research of factors influencing construction prod-
causal logic to explain why the impact of disruptions and delays uctivity. J Constr Manage Econ. 8(1):49–61.
would lead to productivity loss and result in an increase in work- Hickson B G, Ellis L. 2014. Factors affecting construction labour productivity
in Trinidad and Tobago. J Assoc Prof Eng Trinidad Tobago. 2(1):4–11.
ing hours needed to complete the project. The study can be Ibbs W. 2005. Impact of change’s timing on labor productivity. J Constr Eng
applied to construction projects that faced change orders, and as Manage. 131(11):1219–1223.
a result of these changes, the contractor’s performance was Ibbs W, Nguyen L, Lee S. 2007. Quantified impacts of project change. J Prof
affected. The proposed model can be used to determine the pri- Issues Eng Educ Pract. 133(1):45–52.
Ibbs W, Vaughan C. 2015. Change and the loss of productivity in construc-
mary sources that result in labor productivity loss on a specific tion: a field guide. Ibbs Consulting Group.
type of task. For future studies, more variables should be studied Kermanshachi S, Dao B, Shane J, Anderson S. 2016. An empirical study into
to check their impact on productivity loss. Moreover, the positive identifying project complexity management strategies. Proc Eng. 145:603–610.
impacts of the change orders can be quantified and be compared Kermanshachi S, Thakur R, Govan P. 2018. Discovering the impact of late
change orders and rework on labor productivity: a water treatment case
to the negative impacts.
study analysis using system dynamics modeling. ASCE, Construction
Research Congress 2018, New Orleans, US.
Love P, Edwards D, Irani Z, Goh Y. 2011. Dynamics of rework in complex
Disclosure statement offshore hydrocarbon projects. J Constr Eng Manage. 137(12):1060–1070.
Moselhi O, Assem I, El-Rayes K. 2005. Change orders impact on labor prod-
No potential conflict of interest was reported by the authors. uctivity. J Constr Eng Manage. 131(3):354–359.
Nasirzadeh F, Nojedehi P. 2013. Dynamic modeling of labor productivity in
construction projects. Int J Proj Manage. 31(6):903–911.
References Nelson D. 2011. The analysis and valuation of disruption. Philadelphia (PA):
Hill International, Inc.
Alvanchi A, Lee S, AbouRizk S. 2012. Dynamics of working hours in con- Oliva R, Sterman J. 2010. Death spirals and virtuous cycles: human resource
struction. J Constr Eng Manage. 138(1):66–77. dynamics in knowledge-based services. The handbook of service science.
Association for the Advancement of Cost Engineering (AACE) International, London: Springer. p. 321–358.
Inc. 2004. Estimating lost labor productivity in construction claims. AACE Olsen D, Killingsworth R, Page B. 2012. Change order causation; who is the
International Recommended Practice No. 25R-03. Morgantown, WV, guilty party? 48th ASC Annual International Conference Proceedings.
USA: AACE International. Birmingham, UK: Birmingham City University.
Chang C, Hanna A, Lackney J, Sullivan K. 2007. Quantifying the impact of Robels G, Stifi A, Ponz-Tienda JL, Gentes S. 2014. Labor productivity in the
schedule compression on labor productivity for mechanical and sheet construction industry-factors influencing the Spanish construction labor
metal contractor. J Constr Eng Manage. 133(4):287–296. productivity. Int J Civ Env Constr Arch Eng. 8(10):1021–1030.
Cheng M, Wibowo D, Prayogo D, Roy A. 2015. Predicting productivity loss Safapour E, Kermanshachi S, Ramaji I. 2018. Entity-based investigation of
caused by change orders using the evolutionary fuzzy support vector project complexity impact on size and frequency of construction phase
change orders. Proceedings of Construction Research Congress (CRC),
machine inference model. J Civ Eng Manag. 21(7):881–892.
New Orleans, LA.
Dolage D, Chan P. 2013. Productivity in construction - a critical review of
Schwartzkopf W. 1995. Calculating lost labor productivity in construction
research. J Inst Eng. Sri Lanka. 46(4):31–42.
claims. New York (NY): Wiley.
Eden C, Williams T, Ackermann F. 2005. Analyzing project cost overruns:
Serag E. 2006. Change orders and productivity loss quantification using verifiable
comparing the measured mile; analysis and system dynamics modelling. site data [Ph.D. dissertation]. Orlando (FL): University of Central Florida.
Int J Proj Manage. 23(2):135–139. Soekiman A, Pribadi K S, Soemardi BW, Wirahadikusumah R D. 2011.
El-Gohary K, Aziz R, Abdel-Khalek H. 2017. Engineering approach using Factors relating to labor productivity affecting the project schedule per-
ANN to improve and predict construction labor productivity under differ- formance in Indonesia. Proc Eng. 14:865–873.
ent influences. J Constr Eng Manag. 143(8):04017045. Sterman J. 2000. Business dynamics: systems thinking and modeling for a
Eriksson P E, Westerberg M. 2011. Effects of cooperative procurement proce- complex world. Boston (MA): Irwin/McGraw-Hill.
dures on construction project performance: A conceptual framework. Int J Thomas H R, Zavrski I. 1999. Construction baseline productivity: Theory
Proj Manage. 29(2):197–208. and practice. J Constr Eng Manage. 125(5):295–303.
Fahmy T. 2015. Sustainability assessment of a municipal utility complex: a Tran V. 2011. Labor productivity in the New Zealand construction industry:
system of systems approach [Ph.D. dissertation]. Orlando (FL): University A thorough investigation. Constr Econ Build. 11(1):41–60.
of Central Florida. http://purl.fcla.edu/fcla/etd/CFE0005944. Warhoe S. P. 2013. Applying earned value management to design-bid-build
Forrester J W. 1968. Industrial dynamics-after the first decade. J Manage Sci. projects to assess productivity disruption: a system dynamics approach.
14(7):398–415. Boca Raton, FL, USA.
Fox WF. 1967. Human performance in the cold, human resources research Yi W, Chan A P C. 2014. Critical review of labor productivity research in
office. The George Washington University. Hum Factors. 9(3):203–220. construction journals. J Manage Eng. 30(2):214–225.
 INTERNATIONAL JOURNAL OF CONSTRUCTION MANAGEMENT 11

Appendix A
Model boundary
Table 1A presents the scope of the model by showing the parameters and their related information including description of parameters, type
and units.

Table 1A. Model boundary - overview of variables.

Parameter Description Type Units
Available time The remaining time budget before reaching project deadline Endogenous Weeks
Average daily workers on site Number of laborers on site Exogenous Laborers
Baseline productivity Contractor best productivity Exogenous Labor(Hours/Unit)
Desired completion rate Desired work rate to deliver project on time Endogenous Units/Week
Desired workforce Desired number of laborers to deliver project on time Endogenous Laborers
Fatigue Workers’ physical fatigue Endogenous –
Labors on site Number of available laborers on site Exogenous Laborers
Learning rate User leverage on past experiences to gain the intuition Exogenous –
on how to do the work better
Man-hours man-hours spent each week Endogenous Labor(Hours /Week)
Maximum completion rate Work To Do/minimum delivery delay Endogenous Units/Week
Maximum workweek Maximum hours/week, depends on the organization’s policy Exogenous Hours/Week
Productivity Actual labor hours spent to produce one unit Endogenous Labor(Hours/Unit)
Remaining work Remaining work to be done Endogenous Units/Week
Rework due to changes Amount of units needed rework due to change order Endogenous Units/Week
Rework due to contractor’s errors Amount of units needed rework due to contractor’s error Endogenous Units/Week
Scheduled overtime Scheduled overtime hours per week to deliver project on time Endogenous Hours/Week
Scope change Amount of added/deleted units from the project scope Exogenous Units/Week
Target delivery time Scheduled time to deliver the project Exogenous Units/Week
Temperature The degree or intensity of heat present in the job site Exogenous Fahrenheit
Time for inspection and discovered error Time needed to inspect work and discover errors Exogenous Weeks
Work being done Rate of the work execution over time Endogenous Units/Week
Work really done Total number of accomplished units Endogenous Units
Work to do Total number of units not accomplished yet Exogenous Units
Work week Adjusted standard workweek by the effect of overtime Endogenous Hours/Week

Appendix B
Validation data set
A statistical model was built for the actual work done within the project period. Since the available data was limited, and it was impossible
to collect new data, the original data was split into two parts. The first part was used to build the statistical model and estimate its parame-
ters, and the second part was used to assess the fitted model’s predictive ability and to run the model, tabulate the estimated values, and
compare them to their counterparts from the Vensim model. The second part of the split data was used to validate the model as presented
in Table 1B.

Table 1B. Validation data set for work really done statistical model.
a b c d e
Actual cumulative Estimated cumulative % Error ¼ (b-c)/c
Time (man-hours) man-hours ABS % Error
2 0.0342 0.047248 0.2762 0.276159837
5 0.0855 0.08596 0.0054 0.005351326
6 0.1026 0.098456 0.04209 0.042089868
9 0.1482 0.134804 0.099 0.099002997
10 0.1595 0.14656 0.08829 0.088291485
13 0.1923 0.180832 0.06342 0.063417979
15 0.2125 0.20291 0.04726 0.047262333
17 0.2295 0.224428 0.02238 0.022376887
20 0.25 0.25576 0.0225 0.022521114
21 0.2579 0.265976 0.0304 0.030363642
23 0.2737 0.286102 0.0433 0.043348177
25 0.2895 0.30586 0.0535 0.053488524
28 0.3129 0.334912 0.0657 0.065724728
29 0.3223 0.344464 0.0643 0.064343444
31 0.3411 0.363406 0.0614 0.061380384
33 0.3778 0.382172 0.0114 0.011439875
34 0.3962 0.391504 0.01187 0.011867056
36 0.4226 0.410096 0.03049 0.030490422
38 0.4388 0.428632 0.02372 0.023721981
39 0.4439 0.437894 0.01372 0.013715648
(continued)
12 Z. G. AL-KOFAHI ET AL.

Table 1B. Continued.

a b c d e
Actual cumulative Estimated cumulative % Error ¼ (b-c)/c
Time (man-hours) man-hours ABS % Error
41 0.4541 0.456436 0.0051 0.005117914
43 0.4663 0.475042 0.0184 0.018402583
44 0.4734 0.484384 0.0227 0.022676224
46 0.4876 0.503176 0.031 0.030955371
48 0.5073 0.522152 0.0284 0.028443825
50 0.527 0.54136 0.0265 0.026525787
51 0.5306 0.551066 0.0371 0.037138927
53 0.5378 0.570712 0.0577 0.057668316
55 0.5505 0.59071 0.0681 0.068070627
57 0.5687 0.611108 0.0694 0.069395262
58 0.5778 0.621472 0.0703 0.070271871
60 0.6066 0.64256 0.056 0.055963645
64 0.6463 0.686344 0.0583 0.058343921
66 0.6805 0.709136 0.0404 0.040381535
67 0.6902 0.720778 0.0424 0.042423603
71 0.729 0.769126 0.0522 0.052170906
72 0.7359 0.781688 0.0586 0.0585758
74 0.7497 0.807424 0.0715 0.071491558
75 0.7566 0.82061 0.078 0.078002949
79 0.7916 0.875614 0.096 0.096005774
Average % Error 0.051609603