A dynamic modelling of labor productivity impacts arising from change orders in road projects

Abstract

It is vital to investigate the system dynamics (SD) between change orders and labor productivity to identify the causes of productivity loss in construction projects. Most productivity loss studies are financed by the contractor and rely on the contractor’s data. There is a gap in the literature concerning the labor productivity impacts of change orders on the part of the owner. This research utilizes a previously developed SD model to highlight the problem of productivity loss resulting from issuance of a change order. We conducted a sensitivity analysis to evaluate the impact of overtime, overmanning, temperature, and learning on the behavior of the SD model quantifying change orders’ impact on labor productivity when their effects are simulated. Based on the results, SD is more reliable than the measured mile analysis (MMA) approach for the compensation request, considering the leading factors affecting the productivity loss other than the change order. The model developed in this study can accept or reject the responsibility of a change order for the occurrence of productivity loss.

Full text

1
A dynamic Modelling of Labor Productivity Impacts
Arising from Change Orders in Road Projects
Zain Ghazi Al-Kofahi, Amirsaman Mahdavian and Amr Oloufa
Zain Ghazi Al-Kofahi: Ph.D. Graduate, Department of Civil and Environmental Engineering,
University of Central Florida, Orlando, FL, USA, e-mail: zgkofahi@knights.ucf.edu
Amirsaman Mahdavian: Ph.D. Candidate, Department of Civil and Environmental Engineering,
University of Central Florida, Orlando, FL, USA, Phone: +1 (407)808-3580; ORCID:
https://orcid.org/0000-0003-2146-4405; e-mail: amirsaman@knights.ucf.edu
Amr Oloufa: Professor, Department of Civil and Environmental Engineering, University of
Central Florida, Orlando, FL 32816, e-mail: amr.oloufa@ucf.edu
Corresponding author: Amirsaman Mahdavian, Department of Civil and Environmental
Engineering, University of Central Florida, Orlando, FL 32816. ORCID: https://orcid.org/0000-
0003-2146-4405. Email: amirsaman@knights.ucf.edu
Word count: 9930 word equivalents.

2
ABSTRACT
It is vital to investigate the system dynamics between the change orders and labor productivity
to identify the causes of the productivity loss of the construction projects. Most productivity loss
studies were financed from the contractor’s part and rely on the contractor’s data. There is a
gap for a study investigating the labor productivity impacts of change orders from the owner’s
part. This research highlighted the problem of productivity loss resulting from issuance of a
change order by utilizing a previously developed SD model. It conducted a sensitivity analysis
to evaluate the impact of overtime, overmanning, temperature and learning on the behavior of
the SD model quantifying change orders' impact on labor productivity when their effects are
simulated. Based on the results, system dynamics (SD) provides more reliable results
comparing with the measured mile analysis (MMA) approach for the compensation request,
considering the leading factors affecting the productivity loss other than the change order. The
model developed in this study can accept or reject the responsibility of a change order for the
occurrence of productivity loss.
Keywords: Labor productivity, change orders, system dynamics, sensitivity analysis,
construction management

3
INTRODUCTION
Change orders have become a common issue in construction projects and are typically issued
to incorporate variations in the scope of work. Both owners and contractors generally believe
that change orders harm construction productivity, leading to a decline in labor productivity.
Change orders then continue to pose a severe challenge to both contractors and owners; the
factors that are hard to quantify and frequently result in disputes.
Most productivity loss studies were financed from the contractor’s part and rely on the
contractor’s data. There is a gap for a study investigating the labor productivity impacts of
change orders from the owner’s part. The goal of this study is to investigate the labor productivity
impacts arising from change orders in road projects by employing the data from the owner’s part
incorporating the effect of overtime, overmanning, temperature, and learning curves. To reach
the study's goal, this article employed a previously developed system dynamics model by Ghazi
Al-Kofahi et al. (2020) to conduct a sensitivity analysis to evaluate the impact of overtime,
overmanning, temperature, and learning curve on the model's behavior. To test the developed
sensitivity analysis model, two different road construction projects encountered due to owner-
directed change orders issued for modification were analyzed using both methods, namely, an
MMA and an SD model. The projects studied are fully completed and are potential cases for
claims due to change orders' issuance. Ultimately, the conclusions and contributions of this
research, as well as recommendations for future work, and are presented.
LITERATURE REVIEW
Change orders represent a vital mechanism to fulfill the construction requirements throughout
the project delivery process. On the other hand, change orders regularly pose severe difficulties
to both contractors and owners, resulting in cost and time overruns and multiple disputes (Habibi

4
et al. 2018; Safapour 2018; Kermanshachi 2016). Ultimately, a “change order” alters the number
of resources needed to complete the construction project.
The general causes of productivity loss resulting from change orders’ issuance includes
problematic project location, contractors using schedule acceleration when a delay happens,
changing the construction method and its impact on the learning curve, payment delay, and
rework (Hickson and Ellis 2014; Robles et al. 2014).
Regarding the learning style, there is a natural learning curve for the laborers to become familiar
with the defined methods, machines and materials, instructions and quality requirements, and
the working area. Change orders that include a scope change requiring new work methods and
quality standards or different expertise require more time so that laborers must go through a
new learning curve to get familiar with the updated scope. Accelerating schedule techniques are
also prevalent between contractors while experiencing a delay and/or are required to fast-track
the project; however, it can be hard for the project management team to determine the effects
of these techniques on the productivity of the labors and to make suitable decisions to optimize
project performance. Schedule acceleration may result in long periods of compulsory overtime
and overcrowding of the resources. The overcrowding beyond the capacity of the construction
work area, could increase the risk of the occurrence of the accidents, and decrease the
productivity level.
Three method groups, namely project practice-based method such as measured mile analysis
(MMA), a cost-based method such as total cost method (TCM), and industry-based method such
as system dynamics (SD) are being employed to quantify the loss in labor productivity in
construction projects (Nelson 2011).

5
Measured Mile Analysis (MMA)
The MMA is recognized as the most acceptable tool to quantify the productivity loss in
construction projects, in comparison with other available methods (Zhao and Dungan 2014; SCL
2017). To employ the MMA for the construction projects, mostly, the unimpacted period of the
construction activity is recognized and would be linearly extrapolated to the end of an impacted
period. Then, the number of damage hours would be the difference between the projected
unimpacted hours and the actual cumulated hours (Gulezian and Samelian 2003). Zhao and
Dungan (2018) also reviewed the methods to quantify lost labor productivity used in the US
construction industry and examine the practical concerns in electing the proper method for a
particular case. They also discussed quantifying lost productivity in the setting of international
claims examining the advantages and disadvantages of the MMA. Measured mile analysis
limitations regarding reviewing the literature are as follow:
 The need for the existence of a non-impacted period for the same type of work being
evaluated.
 The length of the non-impacted period should be significant compared to the impacted period.
 Enough contemporaneous project data is needed for the analysis. Formatting of the data is
also necessary to perform the analysis, so this method of analysis can be cumbersome and
costly.
 Choosing the timeframe for the MMA is very subjective, which means different timeframes
chosen from the MMA period might generate distinct numbers (Gulezian and Samelian
2003).
 All disruption during the impacted period is assumed to be due to one party.
 This method does not produce causal logic for explaining why a change occurrence results
in productivity loss.

6
Cost-based methods
Total cost method: In this approach, the contractor demands to redeem the whole overrun by
subtracting the actual work from the planned work on a contracted scope of work. Since the total
cost approach is imprecise, it is not usually recommended for claims (Serag 2006).
System Dynamics (SD)
Regarding the use of SD in construction claims and project management, Howick and Eden
(2001) investigated disruption and delay (D&D) that resulted from client demand for earlier
project delivery in an attempt to help managers and contractors in deciding whether to go for
incentives for early delivery or not. To accomplish this task, an SD model, based upon a large
model, was developed to describe the complexity of a claim for D&D about a specific mega-
project. The impact of compressed delivery date on interruption and delay was assessed about
two specific and typical options: pressure and overtime. The findings show that employing
overtime and managerial pressure to a compressed delivery date can have adverse side effects
to an otherwise well-planned project. Increased fatigue and decreased morale will likely harm
both productivity and quality of work. This will then loop and cause further disruption and delay.
This again would cycle because the delay would cause more significant strain on the system.
Therefore, managers must use caution when choosing acceleration methods of overtime or
pressure since the ramifications can be costlier than initially considered, and once the loop
begins, the trajectory is difficult to repair. An analysis is crucial to consider the advantages and
disadvantages of potential incentives and acceleration.
Eden et al. (2005) compared two approaches, the MMA and SD simulation modeling, which are
often used in the forensic analysis of failed projects to analyze the reasons behind project cost
overruns. The comparison reveals the following problems associated with the MMA approach:
• It supposes the presence of an unaffected beginning to the project.

7
• It assumes that the entire difference between observed values and predictions is claimable,
though the contractor might be liable for some or all of the difference.
• It lacks causal logic about the extra work that might be precipitated by D&D.
• It relies on a linear extrapolation of the measure mile to estimate future progress.
On the other hand, the SD modeling approach:
• Shows causality with causal loop diagrams (CLD) in building the simulation model.
• Permits statistical validation of causal logic between the modeling and actual data at any
desired point in the project.
• Directly addresses both constructive acceleration and management’s actions made in effort
of project compression.
• Accounts for D&D for which the contractor is liable.
• Includes nonlinearities that emerge in real projects.
Eden et al. (2005) concluded that despite the popularity of the MMA approach in litigation, its
results can be untrustworthy in cases where D&D are a substantial part of the explanation for a
project’s late delivery and cost overruns.
A review of the literature confirmed there had not been much investigation thus far on developing
a model investigating the labor productivity impacts of change orders by employing the data from
the owner’s part incorporating the effect of overtime, overmanning, temperature, and learning
curves. Most productivity loss studies were financed from the contractor’s part and rely on the
contractor’s data. The need for such a model is apparent in the literature. This study aims to build
a model to fill the identified gap to help owners, contractors, and planners enhance the capability
of accepting or rejecting the responsibility of a change order for the occurrence of productivity
loss in the project.

8
METHODOLOGY
The goal of this research is to study the labor productivity impacts due to change orders in road
projects by utilizing the data from the owner’s part incorporating the effect of overtime,
overmanning, temperature, and learning curves. This study aims to address the following
objectives:
 Employing the developed system dynamics model by Ghazi Al-Kofahi et al. (2020) to test its
application
 Conducting a sensitivity analysis to evaluate the impact of overtime, overmanning,
temperature, and learning curve on the model's behavior. The effects of the overtime,
overmanning, temperature, and learning curve are simulated based on the gathered data
ranges from the minimum to the maximum captured historical data
 Testing the developed sensitivity analysis model, two different road construction projects
encountered due to owner-directed change orders issued for modification were analyzed
using both methods, namely, an MMA and an SD model. The data was derived from the
owner’s part and included daily work reports of the owner, contract documents, change orders,
and inspection reports
The majority of the models in the literature that examine the loss of productivity issues
recommend linear models (Cooper et al. 2002; Eden et al. 2000), while in the study by Ghazi
Al-Kofahi et al. (2020), the nonlinear relationships between productivity and various parameters
are considered throughout the model formulation. The available methods employed to monetize
and quantify in labor productivity loss believe that owner is responsible for all the losses during
the affected time (Ghazi Al-Kofahi et al. 2020), which is not a sound hypothesis for all the
scenarios. Thus, the inefficiencies on the contractor’s part can also be analyzed throughout the
computations. In the course of a change order in the project, several parameters are required

9
to be evaluated and examined. Since these parameters could influence the productivity level of
the project, it is critical to identify the leading factors.
To investigate the effects of Change Orders on Labor Productivity, a previously developed
Causal Loop Diagram (CLD) developed by Ghazi Al-Kofahi et al. (2020) was used in this study.
Therefore, an SD model was generated and validated by them to investigate the change orders
in labor productivity made by the owner part to solve the conflicts between the contractor and
owner. Their developed CLD is shown in Appendix A, and their SD model, the stock-flow
diagram (SFD) is shown in Appendix B. Parameters such as temperature, overmanning,
overtime, and learning curve effects generate feedback loops that influence the contractor’s
work rate flow and productivity. To evaluate the accurate implementation of the SD model, model
verification and validation tests were employed in Ghazi Al-Kofahi et al. (2020) research (more
details about their study are shown in the supplementary material).
System Dynamic’s Sensitivity Analysis
It is valuable to understand the range of behavior of an SD model, given the wide range of
extremes in the input. Performing sensitivity analysis helps the modeler to make the model more
robust by making any corrections recognized from the sensitivity analysis, which points out
possible weaknesses. Determining numerical sensitivity can be evaluated by varying any
constant that directly impacts productivity. For this research, a sensitivity analysis was
performed to evaluate the impact of overtime, overmanning, temperature, and learning on the
behavior of the model when their effects are simulated using a Monte Carlo multivariate
simulation. For each run of the sensitivity analysis, the model was simulated with all the
constants set to their baseline condition values, and then the impact of the factor under
consideration was analyzed using the random uniform distribution of the values from their high
to low thresholds.

10
Run #1 of the sensitivity analysis analyzed the effect of learning using the random uniform
distribution for learning rate from 0.8 to 1, where the value of 1 assuming no learning. Figure
1a (Productivity-Time) and Figure 1b (Work Really Done-Time) illustrate the confidence
boundaries or ranges for all potential output values of the variables, productivity, and work done,
respectively. The results of the sensitivity analysis illustrate that outer bounds of uncertainty (100
percent) show a 10 percent increase in the man hours required to accomplished one unit,
productivity, as the worst case, which indicates that no learning is acquired by the laborers while
doing the work, resulting in finishing only 92 percent of the work scope at the end of the
simulation time for the baseline conditions. Moreover, an improvement in labor productivity by
25 percent, as the best case, results in earlier delivery of the project when compared to the
baseline condition.
Figure 1. Sensitivity analysis, run #1
Run #2 of the sensitivity analysis analyzed the effect of percent crowding using the random
uniform distribution from 0% to 40%. Figure 2a (Productivity-Time) and Figure 2b (Work
Really Done-Time) illustrate the confidence boundaries or ranges for all potential output values
of the variables productivity and work done, respectively. The results of the sensitivity analysis
show that, with 50 percent confidence, in the worst case, the working hours required to
accomplished one unit, productivity, are increased by 27 percent. Compared to the baseline
conditions, it can be seen that the occurrence of crowding is limited, and this is expected in
roadway projects. For work done, the outer bounds of uncertainty (95 percent) show that the
project can be delivered two weeks earlier compared to baseline conditions as the best case,
compared to finishing only 84 percent of the work scope at the end of the simulation as the worst
case.
Figure 2. Sensitivity analysis, run #2

11
Run #3 of the sensitivity analysis analyzed the effect of learning, crowding and overtime.
Figure 3a (Productivity-Time) and Figure 3b (Work Really Done-Time) illustrate the
confidence boundaries or ranges for all potential output values of the variables productivity and
work done, respectively. The results of the sensitivity analysis illustrate that the outer bounds of
uncertainty (100 percent) show a 34 percent increase in the man hours required to accomplish
one unit, productivity, as the worst case and an improvement in the labor productivity by 46
percent as the best case. For work done, the outer bounds of uncertainty (100 percent) show
that the project can be delivered 16.5 weeks earlier compared to baseline condition as the best
case, in contrast to finishing only 80 percent of the work scope at the end of the simulation as
the worst case.
Figure 3. Sensitivity analysis, run #3
Run #4 of the sensitivity analysis analyzed the effect of crowding and overtime. Figure 4a
(Productivity-Time) and Figure 4b (Work Really Done-Time) illustrate the confidence
boundaries or ranges for all potential output values of the variables productivity and work done,
respectively. The results of the sensitivity analysis illustrate that the outer bounds of uncertainty
(95 percent) show a 25 percent increase in the man hours required to accomplish one unit,
productivity, as the worst case and an improvement in the labor productivity by 25 percent as
the best case. For work done, the outer bounds of uncertainty (100 percent) show that the project
can be delivered four weeks earlier compared to baseline conditions as the best case, in contrast
to finishing only 79 percent of the work scope at the end of the simulation as the worst case.
Figure 4. Sensitivity analysis, run #4
Run #5 of the sensitivity analysis analyzed the effect of learning and overtime. Figure 5a
(Productivity-Time) and Figure 5b (Work Really Done-Time) illustrate the confidence
boundaries or ranges for all potential output values of the variables productivity and work done,

12
respectively. The results of the sensitivity analysis illustrate that the outer bounds of uncertainty
(100 percent) show a 0.3 percent increase in the man hours required to accomplish one unit,
productivity, as the worst case and an improvement in the labor productivity by 44.8 percent as
the best case. For work done, the outer bounds of uncertainty (100 percent) show that the project
can be delivered 15.8 weeks earlier compared to baseline condition as the best case, in contrast
to finishing only 86.5 percent of the work scope at the end of the simulation as the worst case.
Figure 5. Sensitivity analysis, run #5
Run #6 of the sensitivity analysis analyzed the effect of learning and crowding. Figure 6a
(Productivity-Time) and Figure 6b (Work Really Done-Time) illustrate the confidence
boundaries or ranges for all potential output values of the variables productivity and work done,
respectively. The results of the sensitivity analysis illustrate that the outer bounds of uncertainty
(100 percent) show a 37 percent increase in the man hours required to accomplish one unit,
productivity, as the worst case and an improvement in the labor productivity by 40 percent as
the best case. For work done, the outer bounds of uncertainty (100 percent) show that the project
can be delivered 14.5 weeks earlier compared to baseline conditions as the best case, in
contrast to finishing only 76.8 percent of the work scope at the end of the simulation as the worst
case.
Figure 6. Sensitivity analysis, run #6
Run #7 of the sensitivity analysis analyzed the effect of Temperature using the random uniform
distribution from 54F to 91F; these values were chosen based on the maximum and minimum
temperatures recorded at the project location. Figure 7a (Productivity-Time) and Figure 7b
(Work Really Done-Time) illustrate the confidence boundaries or ranges for all potential output
values of the variables productivity and work really done, respectively. The results of the
sensitivity analysis illustrate that the outer bounds of uncertainty (100 percent) show an 18

13
percent increase in the man hours required to accomplish one unit, productivity, as the worst
case and an improvement in the labor productivity by 4.5 percent as the best case. For work
really done, the outer bounds of uncertainty (50 percent) show that the project can be delivered
three weeks earlier compared to baseline conditions as the best case, in contrast to finishing
only 84.4 percent of the work scope at the end of the simulation as the worst case.
Figure 7. Sensitivity analysis, run #7
Run #8 of the sensitivity analysis analyzed the effect of Temperature, learning, crowding, and
overtime. Figure 8a (Productivity-Time) and Figure 8b (Work Really Done-Time) illustrate
the confidence boundaries or ranges for all potential output values of the variables productivity
and work done, respectively. The results of the sensitivity analysis illustrate that the outer
bounds of uncertainty (100 percent) show an 18 percent increase in the man hours required to
accomplish one unit, productivity, as the worst case and an improvement in the labor productivity
by 28 percent as the best case. For work done, the outer bounds of uncertainty (100 percent)
show that the project can be delivered eighteen weeks earlier compared to baseline conditions
as the best case, in contrast to finishing only 71 percent of the work scope at the end of the
simulation as the worst case.
Figure 8. Sensitivity analysis, run #8
RESULTS AND DISCUSSION
One executed road project consists of the data of FDOT daily work reports, contract documents,
change orders, and inspection reports. To find the change order’s impact, the proposed model
defines and quantifies the influences of other contributing factors to the loss in labor productivity,
namely, overtime and fatigue, temperature, labor crowding, and learning curve effect.
Additionally, the working hours spent to accomplish the work during the period of a project

14
impacted by the issuance of the change order were assessed. The projects were analyzed using
both SD model and MMA approaches.
Case Study #1
FDOT has engaged a primary contractor to widen and add lanes for a road in the state project
“X.” This primary contractor hired a subcontractor to build the concrete pavement of this road.
The scope of work of this subcontract included placing 74781SY unreinforced concrete
pavement (twelve and a half inches). The cost of the work of $5,608,575.00- and the work
duration of 40-weeks duration and total number of work hours of 29,054 hours were agreed in
the beginning of the project. After twelve weeks of work, a change order (supplemental
agreement) was issued by the owner to increase the scope of work by 11 percent due to the
plan’s modification. The results of the model were then compared with the actual data and the
results obtained from the MMA approach. Referring to Figure 9, the distance between the black
lines presents the period in which the contractor’s work is impacted by the change order. Based
on the unaffected period, it can be verified that the contractor’s best performance on site during
the unaffected period is almost 0.3381 man hours per unit. This value is the MMA as well as the
baseline value used to start the model simulation.
Figure 9. Actual % cumulative quantity installed vs. actual cumulative man hours (Case #1)
The analysis shows an average loss of 11 percent along the project length due to placing the
available personnel on extended overtime. The loss, in this case, can be attributed to the fatigue
effect, whereas a 3 percent average loss in productivity occurred due to overmanning. The
crowding effect on productivity can explain this loss; as the number of laborers increased onsite,
so, too, did the site congestion increase, which negatively affects productivity. Less than 1
percent of loss in productivity is attributed to temperature effect. The results also show an

15
average increase of 20 percent to labor productivity due to the learning effect; this can be
attributed to the experience gained by the laborers each time they repeat the same activity.
Figure 10 demonstrates the change in laborers’ productivity along the project period for the
actual data, MMA approach, and SD modeling simulation. It can be seen that the contractor’s
laborers were working inefficiently even before the issuance of the change order when
comparing the actual productivity values to the MMA productivity values. Using SD simulation,
the reasons behind these inefficiency values can be explained by calculating the expected labor
productivity on site after introducing the factors that might affect productivity, namely, fatigue
due to the change order, overcrowding, temperature, and learning curve effects.
Figure 10. Time vs. productivity (Case #1)
The analysis shows an average loss of 11% along the project length due to placing the available
personnel on extended overtime. The loss, in this case, can be attributed to the fatigue effect,
whereas a 3% average loss in productivity occurred due to overmanning. This loss can be
explained by the crowding impact on productivity, as the number of laborers increased onsite,
so increased site congestion, which negatively affects productivity. Less than 1% of loss in
productivity is attributed to the temperature effect. The results also show an average increase
of 20% to labor productivity due to the learning effect; this can be attributed to the experience
gained by the labor each time they repeat the same activity.
Figure 11 presents the relationship between percent of cumulative amount installed and actual
cumulative man hours spent during installation. According to the figure, it took fewer actual
labors hours to install the first 15 percent of the project than the estimated labors hours from
both MMA and SD models. This can be attributed to the freshness of the laborers at the
beginning of the project during the first three weeks of the work. Considering the placement of
the available workers on overtime and crowding of laborers in some periods of the project, the

16
number of labor hours needed to install a certain number of units increased. This dynamic is
demonstrated by the increase in the slope of the actual data series. In the SD series, a steeper
slope can be seen in the beginning as a consequence of the effect of fatigue and crowding on
productivity. However, as the laborers repeat the same activity more and more, they gain
experience, which makes the tasks easier to perform. As a result, a decrease in time and effort
needed to perform the task can be varied, leading to an improvement in labor productivity. It can
also be seen, based on the SD model, that the project could be completed with a fewer total
number of work hours (29,054 hours), which were delivered earlier (40 weeks) than the actual
values of 32,747 hours and 44 weeks, respectively.
Figure 11. % Cumulative quantity installed vs. cumulative man hours (Case #1)
The calculations for the inefficient work hours during the impacted period of the project are
shown in Table 1.
Table 1. Calculation on inefficient work hours (Case #1)
Therefore, using the MMA approach, the contractor will request the owner to be compensated
for 2,191-man hours due to productivity loss. In contrast, when using the SD modeling approach,
it can be seen that the extra man hours used by the contractor for the impacted period (week 13
to week 20) of the case study #1 (Total work hours: 6526 – SD Work hours: 5927 = 599 man
hours) are not justified.
Case Study #2
FDOT has engaged a primary contractor to perform an 8-mile road resurfacing in the state
project number “XX.” The total bid amount of $3,396,600.00- and duration of 48-weeks and total
number of work hours of 29,054 were agreed in the beginning of the project. After 28 weeks of
work, a change order (supplemental agreement) was issued by the owner to increase the scope
of work by 3 percent due to the plan’s modification. Based on the SD model results, the project

17
can feasibly be completed with a lower total number of work hours (15,184.3 hours) and
delivered earlier (48 weeks), compared to the actual values of 22,338 hours and 60 weeks,
respectively. According to Figure 12, the distance between the black lines presents the period
in which the contractor’s work is impacted by the change order, based on the unaffected period.
It can be observed that the contractor’s best performance on site during the unaffected period
is almost 8,948.4 man hours/unit. This value is the MMA and the baseline value used to start
the model simulation.
Figure 12. Actual % cumulative quantity installed vs. actual cumulative man hours (Case #2)
Figure 13 shows the change in laborers’ productivity along the project period for the actual data,
MMA approach, and SD modeling simulation. It can be verified that the contractor’s laborers
were working inefficiently even before the change order issuance when comparing the actual
productivity values to the MMA productivity. Using SD simulation, the reasons behind these
inefficiency values can be explained through the results of expected labor productivity on site
after introducing the factors that might affect productivity, namely, fatigue due to the change
order, overcrowding, temperature, and learning curve effects. The analysis reveals an average
loss of 3 percent along the project length due to placing the available personnel on overtime.
The loss, in this case, can be attributed to the fatigue effect. Also, about 2 percent of the average
loss in productivity occurred due to overmanning. This loss can be explained by the crowding
effect on productivity since the number of laborers increased onsite, leading to site congestion,
and negatively affecting productivity. About 7 percent of loss in productivity is attributed to the
temperature effect. The results also indicate an average increase of 27 percent to labor
productivity due to the learning effect, which can be attributed to the experience gained by
repeating the same activity.
Figure 13. Time vs. productivity (Case #2)

18
Figure 14 displays the relationship between percent of cumulative amount installed and actual
cumulative man hours spent during installation. According to the figure, a significant number of
working hours is not justified when comparing the estimated laborers’ hours using SD simulation
and those spent on site. The SD series starts revealing a reduction in the slope after the
execution of about 63 percent of the project work. This implies that the effect of learning starts
to play an essential role in improving productivity as it eliminates the negative impact of fatigue,
crowding, and temperature on labor. Based on the SD model results, the project can feasibly be
completed with a lower total number of work hours (15,184.3 hours) and delivered earlier (48
weeks), compared to the actual values of 22,338 hours and 60 weeks, respectively.
Figure 14. % Cumulative quantity installed vs. cumulative man hours (Case #2)
The calculations for the inefficient work hours during the impacted period of the project are
shown in Table 2.
Table 2. Calculation on inefficient work hours (Case #2)
Therefore, using the MMA approach, the contractor will request the owner to be compensated
for 221.91-man hours due to productivity loss, whereas using the SD modeling approach, it can
be seen that the extra man hours used by the contractor for the impacted period (week 29 to
week 31) of the case study #2 (Total work hours: 987 – 789.89 = 197.11) are not justified.
CONCLUSIONS
This research highlighted the problem of productivity loss resulting from issuance of a change
order in construction projects. By utilizing the developed and validated SD model that analyzes
the changes in labor productivity due to the owner change orders to address the disputes
between the owner and contractor by Ghazi Al-Kofahi et al. (2020), this research analyzed

19
different road construction projects using two methods: the MMA and SD model. MMA assumes
that the difference between actual outcomes and the MMA prediction is all claimable by the
contractor, whereas the SD modeling approach directly addresses both constructive
acceleration and management’s actions made in effort of project compression. MMA only relies
on a linear extrapolation of the measure mile to estimate future progress; however, SD model
includes nonlinearities that emerge in real projects.
Most productivity loss studies were financed from the contractor’s part and rely on the
contractor’s data. In this research, however, the data used were derived from the owner’s part.
Moreover, using baseline productivity helps prove which portion of the productivity loss can be
attributed to the owner’s changes. This approach compares the best productivity achieved by
the contractor, which is the model’s baseline productivity, with actual hours.
In this study, the productivity loss is linked to its causes using SD modeling; thus, the
inefficiencies on the part of the contractor can be considered in the calculations. As the percent
of the scope of change increases, the difference between the MMA and the SD model
predictions become larger. It was observed that the quantity installed has a high impact on
productivity improvement, a phenomenon explained by the learning curve theory. Ultimately,
according to the Figures 1,2 and 7, it was found that compared to crowding and learning curve
impacts, temperature has more significant impact on labor productivity and delivery date. For
future studies, more related variables should be investigated to check their impact on
productivity loss. Moreover, the scholars could employ some advanced tools such as Machine
learning to solve the problem statement of this study. Ultimately, the positive impacts of the
change orders should also be monetized and be evaluated against the negative impacts.

20
REFERENCES
Eden, C., Howick, S., 2001. The impact of disruption and delay when compressing large
projects: going for incentives?, Journal of the Operational Research Society, 52:1, 26-34, DOI:
10.1057/palgrave.jors.2601045.
Eden, C., Williams, T., Ackermann, F. 2005. Analysing project cost overruns: Comparing the
"measured mile"; analysis and system dynamics modelling. International Journal of Project
Management, 23 (2) , pp. 135-139.
Eden, C., Williams, T., Ackermann, F., 2000. Howick S. On the nature of disruption and delay. J
Oper Res Soc; 51:291–300
Cooper, KG, Lyneis, JM, Bryant, BJ. 2002. Learning to learn, from past to future. Int J Project
Manage ;20:213–9.
Ghazi Al-Kofahi, Z., Mahdavian, A., Oloufa, A.A. 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
Gulezian, R., Samelian, F.Z. 2003. The Productivity Baseline. AACE International Transactions.
EST.02.
Habibi M, Kermanshachi S, Safapour E. 2018. Engineering, Procurement and Construction Cost
and Schedule Performance Leading Indicators: State-of-the-Art Review. Construction Research
Congress 2018 ASCE. https://doi.org/10.1061/9780784481271.037.
Hickson, B.G., Ellis, L. 2014. Factors affecting Construction Labour Productivity in Trinidad and
Tobago. Journal of the Association of Professional Engineers of Trinidad and Tobago 42 (1).
Kermanshachi, S., Dao, B., Shane, J., Anderson S. 2016. An Empirical Study into Identifying
Project Complexity Management Strategies. Procedia Engineering, 145 , pp. 603-610.
https://doi.org/10.1016/j.proeng.2016.04.050.
Nelson, D. 2011. The Analysis and Valuation of Disruption. Hill International, Inc.
Robles, G., Stifi, A., Ponz-Tienda, J. L., Gentes, S. 2014. Labor Productivity in the Construction
Industry-Factors Influencing the Spanish Construction Labor Productivity. Journal of Civil and
Environmental Engineering 8 (2014): 1061-1070.
Safapour E, Kermanshachi S, Ramaji I. 2018. Entity-Based Investigation of Project Complexity
Impact on Size and Frequency of Construction Phase Change Orders. Construction Research
Congress 2018 ASCE. https://doi.org/10.1061/9780784481271.066.
Society of Construction Law (SCL). 2017. Delay and disruption protocol. Oxford shire, UK: SCL.

21
Warhoe, S. P. 2013. Applying earned value management to design-bid-build projects to assess
productivity disruption: a system dynamics approach. Dissertation.Com. ISBN-10 : 1612334164.
Zhao, T., Dungan, M. 2014. Improved baseline method to calculate lost construction
productivity. Journal of Construction Engineering and Management. Vol. 140 (2).
https://doi.org/10.1061/(ASCE)CO.1943-7862.0000800.
Zhao, T., Dungan, M. 2018. Quantifying Lost Labor Productivity in Domestic and International
Claims. Journal of Legal Affairs and Dispute Resolution in Engineering and Construction. Vol.
10 (3). https://doi.org/10.1061/(ASCE)LA.1943-4170.0000269.

22
Appendix A
Causal Loop Diagram (CLD) of labor productivity by Ghazi et al. (2020)
Figure A1. CLD of labor productivity

23
Appendix B
Stock and Flow Diagram by Ghazi et al. (2020)
Figure B1. Stock and Flow Diagram

24
FIGURES
Figure 1. Sensitivity analysis, run #1
Figure 2. Sensitivity analysis, run #2

25
Figure 3. Sensitivity analysis, run #3
Figure 4. Sensitivity analysis, run #4
Figure 5. Sensitivity analysis, run #5

26
Figure 6. Sensitivity analysis, run #6
Figure 7. Sensitivity analysis, run #7

27
Figure 8. Sensitivity analysis, run #8
Figure 9. Actual % cumulative quantity installed vs. actual/span>cumulative man hours (Case #1)
0
5000
10000
15000
20000
25000
30000
35000
0 102030405060708090100
Cumulative Manhours
% Cumulative Quantity Installed
Impacted
period
Measured
Mile

28
Figure 10. Time vs. productivity (Case #1)
0
0.5
1
1.5
2
2.5
0 1020304050
Productivity (Man-hours/unit)
Time (Weeks)
Actual Productivity
Productivity-Measured Mile
Productivity-SD
Impacted
Period

29
Figure 11. % Cumulative quantity installed vs. cumulative man hours (Case #1)
0
5000
10000
15000
20000
25000
30000
35000
0 20406080100120
Cumulative Man-hours
% Cumulative Quantity Installed
Actual
Measured Mile
SD
Impacted
Period

30
Figure 12. Actual % cumulative quantity installed vs. actual cumulative man hours (Case #2)
0
5000
10000
15000
20000
25000
0 102030405060708090100
Cumulative Man-Hours
% Cumulative Quantity Installed
Impacted
Period
Measured Mile

31
Figure 13. Time vs. productivity (Case #2)
0
20000
40000
60000
80000
100000
120000
140000
160000
0 10203040506070
Productivity (Man-Hours/unit)
Time (Weeks)
Actual Productivity
Productivity- Measured mile
Productivity -SD
Impacted
Period

32
Figure 14. % Cumulative quantity installed vs. cumulative man hours (Case #2)
0
5000
10000
15000
20000
25000
0 20406080100
Cumulative Man-Hours
% of Cumulative Quantity Installed
Actual
Measured mile
SD
Impacted
Period

33
FIGURES CAPTION LIST
Figure 1. Sensitivity analysis, run #1
Figure 2. Sensitivity analysis, run #2
Figure 3. Sensitivity analysis, run #3
Figure 4. Sensitivity analysis, run #4
Figure 5. Sensitivity analysis, run #5
Figure 6. Sensitivity analysis, run #6
Figure 7. Sensitivity analysis, run #7
Figure 8. Sensitivity analysis, run #8
Figure 9. Actual % cumulative quantity installed vs. actual/span>cumulative man hours (Case #1)
Figure 10. Time vs. productivity (Case #1)
Figure 11. % Cumulative quantity installed vs. cumulative man hours (Case #1)
Figure 12. Actual % cumulative quantity installed vs. actual cumulative man hours (Case #2)
Figure 13. Time vs. productivity (Case #2)
Figure 14. % Cumulative quantity installed vs. cumulative man hours (Case #2)

34
TABLES
Table 1. Calculation on inefficient work hours (Case #1)
Week Work
hours
Quantity
installed
Should have
worked-MM
Inefficient
Work (MMA)
Work hours-
(SD)
13 745 1170.33 395.69 349.31 111
14 902 1885.92 637.63 264.37 745
15 970 1885.92 637.63 332.37 902
16 550 1885.92 637.63 -87.63 970
17 958 1885.92 637.63 320.37 550
18 704 1885.92 637.63 66.37 958
19 987 1110.8 375.56 611.44 704
20 710 1110.8 375.56 334.44 987
Total 6526 4334.96 2191.04 5927
Table 2. Calculation on inefficient work hours (Case #2)
Week Work
hours
Quantity
installed
Should have
worked- MMA
Inefficient
Work -MMA
Work hours-
SD
29 300 0.03 255.03 44.97 240
30 453 0.03 255.03 197.97 365.89
31 234 0.03 255.03 -21.03 184
Total 987 765.09 221.91 789.89