You are on page 1of 9

Scholarly Paper

Effect of Change Orders on Cost and Schedule for


Small Low-Bid Highway Contracts
Pramen P. Shrestha, Ph.D., P.E., M.ASCE 1; and Ruiko Maharjan, Ph.D., S.M.ASCE 2

Abstract: For design-bid-build (DBB) projects, the assumption is that because the bid cost of the lowest bidder is usually below the esti-
Downloaded from ascelibrary.org by Nottingham Trent University on 08/25/19. Copyright ASCE. For personal use only; all rights reserved.

mated cost of the project, change orders during construction will increase. This study collected data on 615 small highway DBB projects (with
costs of less than $1 million) from the Florida DOT (FDOT) related to the construction of roads and structures, and determined the correlation
between these two variables. The study determined whether change orders affected cost and schedule performance; for projects that had no
change orders or negative change orders, cost and schedule performance were significantly better than for of projects that had positive change
orders. The results of this study demonstrate that projects that had positive change orders were those in which the contractors submitted bids
significantly higher than those submitted for projects that had no change orders or negative change orders, even though—in both groups—the
bids were lower than the estimated costs. An in-depth analysis showed that the projects that had no change orders or negative change orders
had significantly lower cost and schedule growth as well as higher construction intensity compared to those of projects that had positive
change orders. These results indicate that change orders have detrimental effects on cost and schedule growth, as well as construction intensity
in small highway projects. The researchers recommend using a change order management process in order to reduce the effects of change
orders on project costs and schedule performance. The researchers also recommend conducting a similar study with the data of all state DOTs
in order to test whether the effect of change orders on cost and schedule performance of highway projects is similar to this study. The major
contributions of this study to the body of knowledge are that the low bid had no effect on change orders, and positive change orders increased
the cost and schedule growth as well as decreased the construction intensity of the small highway projects. DOI: 10.1061/(ASCE)LA.1943-
4170.0000323. © 2019 American Society of Civil Engineers.
Author keywords: Lowest bidder; Change order percentage; Bid deviation; Cost growth; Schedule growth; Construction intensity.

Introduction bidding method in selecting the construction contractors in small


DBB highway projects is one of the reasons for change orders dur-
In any state DOTs, highways and bridges are generally built using ing construction.
a design-bid-build (DBB) delivery method. This method is also Change orders are defined as the changes made by the owner
called the traditional delivery method because this is the delivery and accepted by both parties (owner and contractor) during the con-
method state DOTs have used for decades. According to 2016 data, struction phase. Change orders can be generated for several causes,
the Florida DOT (FDOT) has total of 19,483 km (12,106 centerline e.g., owner initiated, unforeseen conditions, design changes, or
miles) in the state highway system, which is equivalent to changes due to state regulations. Whatever the sources of changes,
70,682 km (43,920 lane miles) (FDOT 2018b). On these state high- if the owner and the contractor agree with the changes and asso-
ways, a total of 518 million km (322,050,000 vehicle miles) are ciated cost and time, these changes can be officially recorded as
traveled daily. The total budget of the FDOT is about $10.8 billion, change orders. For FDOT, the process of change order management
out of which $4.1 billion is spent on the construction of highway is spelled out clearly in its construction documents. Some important
projects. The taxpayer money spent on the construction of high-
information required for change orders to process are an explana-
ways is a large amount; therefore, it is necessary to determine
tion of the reason for the change order, a detailed description of the
whether FDOT is completing their projects within budget and
proposed changes, the impact of the changes on cost and time, and
on schedule. Consequently, the main goals of this study are to de-
the signatures of all parties (FDOT 2018a).
termine the extra cost spent on change orders for the construction of
In DBB projects, construction contractors are generally selected
small highway projects and the effects of these changes on cost and
based on the lowest bid cost. State DOTs construct the majority of
schedule. The study will also investigate whether using the lowest
their highways based on this procurement process. However, re-
1
Professor, Dept. of Civil and Environmental Engineering and
cently there have been some alternative delivery methods used by
Construction, Howard R. Hughes College of Engineering, Univ. of state DOTs to design and build highway projects, such as construc-
Nevada, Las Vegas, NV 89154 (corresponding author). ORCID: https:// tion manager/general contractor (CM/GC), lump-sum design-build
orcid.org/0000-0001-6362-2315. Email: pramen.shrestha@unlv.edu (LSDB), progressive design-build (PDB), and public–private part-
2
Ph.D. Student, Dept. of Civil and Environmental Engineering and nership (PPP). The assumption is that when contractors bid low
Construction, Howard R. Hughes College of Engineering, Univ. of Nevada, on highway projects, they will try to make up the cost difference
Las Vegas, NV 89154. Email: maharaja3@unlv.nevada.edu by generating change orders during construction. Therefore, the
Note. This manuscript was submitted on August 17, 2018; approved on
general belief is that change orders on projects will be higher when
February 12, 2019; published online on July 24, 2019. Discussion period
open until December 24, 2019; separate discussions must be submitted for the bid deviation from the estimated cost is high. However, no stud-
individual papers. This paper is part of the Journal of Legal Affairs and ies have been conducted to test this hypothesis. It is also necessary
Dispute Resolution in Engineering and Construction, © ASCE, ISSN to determine the impact of these changes to the cost and schedule of
1943-4162. the projects. This area of study is critical for any state DOT because

© ASCE 04519025-1 J. Leg. Aff. Dispute Resolut. Eng. Constr.

J. Leg. Aff. Dispute Resolut. Eng. Constr., 2019, 11(4): 04519025


it can assist them in predicting whether their contractors will com- also found that more change orders were generated for large proj-
plete a project within time and budget. ects compared to small projects.

Change Orders in Highway and Infrastructure Projects


Scope and Objectives of the Study
There are limited studies to determine the correlation between bid
Change orders captured the cost data of the scope changes, which cost and construction cost growth for highway projects. Williams
reduced or increased the work volume of the originally signed con- (2002) used a neural network to determine the relationship between
tract. Therefore, the authors tracked the total cost added to or sub- the bid cost and the final construction cost of highway projects built
tracted from the original contract cost due to change orders. During by the New Jersey DOT. The study showed a high correlation be-
the data collection for this study, FDOT was asked to provide the tween low bid and final construction cost of projects (correlation
cost of change orders that reduced or increased the total contract coefficient ¼ 0.98). This showed that as the bid cost increases, the
cost of the project. Any changes that increased the total cost of the final construction cost increases. Shrestha and Pradhananga (2010)
Downloaded from ascelibrary.org by Nottingham Trent University on 08/25/19. Copyright ASCE. For personal use only; all rights reserved.

project compared to contract cost were labeled as positive change conducted a study to determine whether construction cost growth
orders. The changes that decreased the cost of the project compared will increase if a contractor bids the project with the lowest bid
to contract cost were labeled as negative change orders. amount. Data from roughly 113 public road projects showed that
The main goal of this study was to determine whether change there is no significant correlation between bid cost and the con-
orders have any relationship with cost and schedule performance in struction cost growth. This showed that in road projects, not only
small highway projects from FDOT. The data set used in this study did the lowest-bid projects have construction cost growth, but the
was from the construction of roads, pavements, structures, drains, highest-bid projects also had construction cost growth.
culverts, and so forth. The analysis focused on determining the cor- Some studies determined the impact of change orders on the
relation of change order percentage with bid deviation, total project cost and schedule growth of highway projects. Based on four case
cost, cost growth, schedule growth, and construction intensity. studies representing public works in Oman for water transmission,
Therefore, the main objectives of this study were as follows: roads, buildings, and ports, Alnuaimi et al. (2010) found that the
• Determine the change order percentage occurring in new small impacts of change orders were cost overruns, schedule delays, and
highway projects; disputes. Assaf and Al-Hejji (2006) also found that change orders
• Assess whether the change order percentages are correlated to increased the construction duration of public and private projects
the difference of lowest bid amount and the estimated cost of the (including infrastructure and building projects) in a study based on
project (bid deviation); a time-performance survey with owners, consultants, and contrac-
• Determine the correlation between the change order percen- tors in Saudi Arabia. Using the interview data of 57 contractors,
tages and project size as well as project performance data (cost consultants, and owners, they found that two-thirds of the respond-
growth, schedule growth, and construction intensity); and ents agreed that change orders increased the schedules of their proj-
• Determine whether the projects with no change orders or nega- ects by about 10%. Bordat et al. (2004) found that for the Indiana
tive change orders had lower cost and schedule growth and DOT (INDOT), cost and schedule overruns occurred due to change
higher construction intensity compared to those of projects with orders. By analyzing 2001 INDOT project data, the researchers
positive change orders. found that due to change orders, the cost overrun was higher for
bridge projects (8.1%) compared to road resurfacing (5.6%), traffic
projects (5.6%), and maintenance projects (7.5%). Choi et al.
Literature Review (2016) found that the impact of change orders on highway projects
depended on the type of contract. Change orders more severely im-
The researchers conducted a comprehensive literature review of pacted cost growth for incentive/disincentive (I/D) contracts than
change orders on infrastructure and building projects. The review A + B (cost + time) contracts. However, the opposite was true for
revealed that there is a limited amount of literature related to the schedule growth. A recent study by Shrestha and Maharjan (2018)
impact of change orders on the cost and schedule of construction found that there was detrimental effect of change orders on the
projects. Most of the literature was related to the causes of change cost and schedule growth of large Texas highway projects. The data
orders for highway and building projects. Some of the causes for from 185 large highway projects (projects costing more than
change orders were poor estimation, unforeseen site conditions, $10 million) showed that the correlation coefficients of change
poor site investigation, and owner-initiated changes (Halwatura and orders with cost and schedule growth were 0.57 and 0.44, respec-
Ranasinghe 2013). Some studies also found that design change gen- tively. Statistical tests showed significant increases in cost and
erates most of the change orders in building projects (Ndihokubway schedule growth as change orders increased. The study found that
and Haupt 2008; Alnuaimi et al. 2010; Günhan et al. 2007; Bordat when change orders increase more than 5%, the impact on cost and
et al. 2004). Other studies found that state regulations, client– schedule growth became more severe.
contractor relationships, and political pressure also contributed to
change orders (Ndihokubway and Haupt 2008; Alnuaimi et al.
2010; Jawad et al. 2009; Serag et al. 2010). Change Orders in Building Projects
Few researchers have studied the causes for change orders on The researchers also investigated the impacts of change orders on
highway projects. Shrestha et al. (2017) found that change orders cost and schedule growths of building projects. Jawad et al. (2009)
occurred at a significantly higher rate on mixed-surface road pave- found that change orders increased the cost of building projects
ment compared to earthen and gravel roads. They also found that in the range of 5%–10% of the original contract cost. Similarly,
change orders were more prevalent on large highway projects than the change orders delayed these projects by 10% of their original
on small highway projects. Similarly, a study by Anastasopoulos contract duration. Ibbs (2012) found that the change orders in in-
et al. (2010) showed that a higher number of change orders was dustrial building projects increased the chance of cost growth by
generated for resurfacing and traffic maintenance projects com- 42%. Another study by Ibbs (2008) determined that change orders
pared to earthwork and subsoil treatment projects. The researchers decreased productivity by 20% in 40% of the building projects

© ASCE 04519025-2 J. Leg. Aff. Dispute Resolut. Eng. Constr.

J. Leg. Aff. Dispute Resolut. Eng. Constr., 2019, 11(4): 04519025


analyzed, and change orders increased project duration by 16% prepared in Microsoft Excel in order to collect the required numeri-
in 50% of the projects. Similarly, many researchers found that the cal data, as well as information about the projects. The information
impact of change orders was loss of productivity (Thomas and collected related to the projects were the project IDs, notices to
Napolitan 1995; Ibbs 2008; Hanna et al. 1999; Vandenberg 1996). proceed (NTPs), types of work (e.g., highways, culverts, drains, re-
Most of the previous studies have found a detrimental effect of taining walls, pavement construction), lowest bidder amounts, and
change orders on the productivity of projects. However, in this study locations of the projects. The authors asked for data on the types of
the productivity was not measured, but the construction intensity, change orders, e.g., owner-initiated change orders and unforeseen
which is also a type of measure of productivity of an overall project, conditions change orders, as well as errors and omissions. However,
was measured and analyzed in order to determine the correlation FDOT did not provide these data, so the authors were unable to
with change orders. The US Department of Commerce developed determine whether the changes originated due to FDOT or for other
the construction productivity metrics used in the project. The project reasons.
productivity level can be created by using the value of construction The questionnaire was then sent to FDOT. The research team
in the numerator and the number of work hours in the denominator first contacted the public information officer (PIO) to determine
Downloaded from ascelibrary.org by Nottingham Trent University on 08/25/19. Copyright ASCE. For personal use only; all rights reserved.

(Huang et al. 2009). Therefore, in this study, the authors developed a how to proceed with data collection. The researchers provided the
construction intensity metric by using the total cost of the project as questionnaire to the PIO to help in collection of the data. The re-
the numerator and the total working days as the denominator. In this searchers asked for data from the year 2000 to the present. The
way, the construction intensity is also a project-level productivity researchers also focused on projects that were small, with project
metric of the construction projects. size limited to $1 million (without adjusting the cost to the base cost
A recent study conducted by Shrestha and Zeleke (2018) found of 2018). All data used were not received from FDOT; some data
that change orders for school building renovation projects increased were sent by FDOT, and the researchers found other data from
project costs by 3.56% of the total contracted cost. The impact of FDOT websites, including the amount of change orders, final pay-
change orders on schedule growth was significantly higher in proj- ment amounts, and final durations. These data were identified by
ects that have change orders (5.26%) compared to the projects that matching the contract ID numbers to the database provided by
had no change orders (2.51%). Another recent study on new school FDOT personnel. By means of these two methods, the researchers
building projects showed that change orders had a detrimental ef- prepared the completed data set required for the analysis. The re-
fect on schedule growth, but not on cost growth (Shrestha et al. searchers would have liked to collect data related to types of change
2018). A study of 95 new school building projects showed that as orders, but the data were not available.
the change orders increased by more than 5%, the impact on sched- The researchers checked all parameters of each project for any
ule growth was severe, increasing the schedule growth twofold. missing values, and the projects with missing values were removed
Overall, past research shows that change orders have negative ef- from the analysis. After the elimination process, the combined data-
fects on the costs and schedules of building projects.
base contained data from 615 projects completed between the years
Based on the literature, it is clear that a gap in the research exists
2000 and 2016, with cost ranges between $19,000 and $1 million
regarding the effects of change orders on cost and schedule growth
(after cost adjustment to 2018 base cost). The researchers calcu-
of highway projects, especially small projects. In addition to this,
lated these costs by converting the total completion costs into the
few studies have been conducted to determine whether low project
base cost of July 2018 using the Engineering News-Record (2018)
bids generate more change orders in small DBB highway projects.
cost index.
This study hopes to fill this knowledge gap by determining the re-
After obtaining the raw data from FDOT, the researchers devel-
lationship between bid deviation and change order percentage, as
oped change order, cost, and schedule metrics to form research hy-
well as the impact of change orders on cost growth, schedule growth,
potheses. One metric related to change order, two metrics related to
and construction intensity of small highway projects.
cost, and two metrics related to project schedule were calculated to
determine the relationship between change orders with costs and
Research Methods schedule performances in small highway projects. The metrics
and their detailed descriptions are shown in Table 1.
This study collected highway contract data from FDOT. To collect To achieve the objectives of this study, the researchers created
the data, the researchers prepared a questionnaire, which included research hypotheses and converted them to null hypotheses. We
questions regarding the amount of change orders, estimated costs, tested null hypotheses using statistical tests. The following gives
contract costs, final completion costs, estimated durations, contract the research hypotheses, null hypotheses, and statistical tests con-
durations, and final completion durations. The questionnaire was ducted for this study.

Table 1. Metrics for change order, cost, and schedule performance


Metric Equation to calculate the metric Unit
Total change order cost
Change order percentage × 100 %
Contract award cost
Bid cost-estimated cost
Bid deviation × 100 %
Estimated cost
Total completion cost-estimated cost
Cost growth × 100 %
Estimate cost
Total completion duration-contract duration
Schedule growth × 100 %
Contract duration
Total completion cost
Construction intensity × 100 $=day
Total completion duration

© ASCE 04519025-3 J. Leg. Aff. Dispute Resolut. Eng. Constr.

J. Leg. Aff. Dispute Resolut. Eng. Constr., 2019, 11(4): 04519025


Research Hypotheses be verified. There are three major assumptions to be tested to con-
duct a PLC test: level of measurement, related pairs, and normality
The research team developed six research hypotheses for this study.
of variables (Statistics Solutions 2018).
The first research hypothesis is related to the change orders’ cor-
The first assumption related to the PLC test is that each variable
relation with bid deviation: if the contractors bid the project below
should be continuous. In this data set, all of the data related to
the estimated cost, then the chances of change orders during con-
change order percentage, bid deviation, total project cost, cost
struction will be high. The second research hypothesis is the pos-
growth, schedule growth, and construction intensity were continu-
itive correlation between the change order and total project cost.
ous variables. The second assumption, related to having pairs of
The third research hypothesis states that as change orders increase,
data, was also true in this data set. To fulfill the third assumption,
they will have a detrimental effect on cost growth. The fourth re-
normality tests of change order percentage, bid deviation, total
search hypothesis states that as change orders increase, the sched-
project cost, cost growth, schedule growth, and construction inten-
ule growth increases. The fifth research hypothesis assumes that as
sity need to be conducted. To check the normality of the data, the
change orders increase, they impact the productivity of the con-
histograms of change order percentage, bid deviation, total project
Downloaded from ascelibrary.org by Nottingham Trent University on 08/25/19. Copyright ASCE. For personal use only; all rights reserved.

struction workers and the construction intensity decreases. The


cost, cost growth, schedule growth, and construction intensity were
final research hypothesis determines whether the projects that have
plotted to see whether the data are normally distributed. If the data
no change orders or negative change orders have less bid deviation,
were not normally distributed, then a nonparametric test was con-
lower cost and schedule growth, and higher construction intensity
ducted to verify whether there was any significant correlation be-
than the projects that have positive change orders. The researchers
tween these variables.
will convert all research hypotheses to null hypotheses to verify
The researchers conducted a t-test to determine whether the
whether they are statistically valid.
bid deviation, cost growth, schedule growth, and construction in-
tensity significantly increase for two types of projects (one with no
Null Hypotheses change orders or negative change orders, and the other with pos-
The first five null hypotheses state that the correlation coefficients itive change orders). Before conducting the t-test, the researchers
between the change order percentages and bid deviations, total conducted the assumption tests (Laerd Statistics n.d.). Generally,
project cost, cost growth, schedule growth, and construction inten- for the t-test, there should be one independent variable and it should
sity are not significantly different from zero. The researchers can have two levels, which was true in this data set. Similarly, the data
subdivide the final null hypothesis into four null hypotheses. These should be normally distributed, which was verified during the cor-
null hypotheses are related to determining whether there are any relation test. The third assumption is that the dependent variables
mean differences of bid deviation, cost growth, schedule growth, or should be in ratio scales, which was true because all of the values of
construction intensity in two types of projects: projects with no bid deviation, cost growth, schedule growth, and construction in-
change orders or negative change orders, and projects with positive tensity were in ratio scale. The fourth assumption is that the data
change orders. Therefore, the null hypotheses state that there are no should be randomly selected, which was also true for this data
significant differences in the mean values of bid deviation, cost set. The final assumption is about the equal homogeneity of var-
growth, schedule growth, and construction intensity significantly iances in the groups. To determine this, the researchers conducted
differ in two types of projects. Table 2 shows the mathematical rep- a Levene’s test to see whether the variances of the two groups were
resentation of these null hypotheses. If the p-value of these tests are equal or not. If the variances of the groups were not equal, then the
found to be equal to or less than 0.05, the null hypotheses will be researchers conducted a t-test with unequal variance. The results of
rejected, confirming the research hypotheses. all these assumption tests are described in the “Results” section.

Pearson Correlation Test


Statistical Analysis The PLC test shows the correlation coefficients among these var-
The researchers conducted two types of statistical tests to test the iables. According to Rumsey (n.d.), a correlation coefficient below
null hypotheses. To test the first five null hypotheses (related to the 0.30 is considered to be weak or show no relationship; a value be-
relationships of change order percentage with bid deviation, total tween 0.30 and 0.50 is considered to be a moderate relationship;
project cost, cost growth, schedule growth, and construction inten- values from 0.50 to 0.70 are considered as having a strong relation-
sity) the researchers conducted Pearson linear correlation (PLC) ship; and values above 0.70 are considered to show a very strong
tests. Before conducting these tests, assumption tests needed to relationship among the variables. The p-values were checked to
determine whether these correlations were significant.

t-Test
Table 2. Null hypotheses to conduct statistical tests A t-test was conducted to determine whether the values of bid
No. Null hypotheses deviation, cost growth, schedule growth, and construction intensity
were significantly different in two groups based on the change
1 β change order percentage versus bid deviation ¼ 0
2 β change order percentage versus total project cost ¼ 0 orders. The research hypotheses were confirmed if the null hypoth-
3 β change order percentage versus cost growth ¼ 0 eses was rejected after the statistical tests.
4 β change order percentage versus schedule growth ¼ 0
5 β change order percentage versus construction intensity ¼ 0
6 μbid deviation of projects with no change orders or negative change orders Results
¼ μbid deviation of projects with positive change orders
7 μcost growth of projects with no change orders or negative change orders The entire data set for the 615 projects was analyzed first to get
¼ μcost growth of projects with positive change orders descriptive statistics of the variables being considered. Then the
8 μschedule growth of projects with no change ordersor negative change orders researchers conducted the assumption tests for the PLC to deter-
¼ μschedule growth of projects with positive change orders mine whether the data set was suitable for a linear correlation test.
9 μconstruction intensity of projects with no change orders or negative change orders Finally, the researchers conducted the equal variance assumption
¼ μconstruction intensity of projects with positive change orders
test (of t-tests) to verify whether t-tests for difference in means

© ASCE 04519025-4 J. Leg. Aff. Dispute Resolut. Eng. Constr.

J. Leg. Aff. Dispute Resolut. Eng. Constr., 2019, 11(4): 04519025


Table 3. Descriptive statistics of variables
Sample
Variables Unit size Mean Median
Change order percentage % 615 2.22 0.18
Bid deviation % 615 −13.03 −13.50
Total project cost $ 615 $434,000 $388,000
Cost growth % 615 −16.90 −16.93
Schedule growth % 615 15.74 10.00
Construction intensity $=day 615 3,668 3,170

could be conducted to prove the research hypotheses. The results of


Downloaded from ascelibrary.org by Nottingham Trent University on 08/25/19. Copyright ASCE. For personal use only; all rights reserved.

these analyses are described in the following.

Descriptive Statistics Fig. 1. Change order percentage (N ¼ 615).

The mean and median values for change order percentage, bid
deviation, total project cost, cost growth, schedule growth, and con-
struction intensity are shown in Table 3. Most of these variables
have their mean and median values very near to each other, except
for schedule growth and construction intensity. This shows that the
data are quite homogenous, and the sample collected represented
the population of the data. The data analysis showed that the mean
value of change order percentage is 2.22% for the entire project.
The bid deviation data showed that the contractors are bidding,
on average, 13.03% below the estimated cost. The total project cost
data showed that the average size of a project is a little less than half
a million dollars. Similarly, the cost performance of these small
highway projects is very good because, on average, these small
highway projects were completed below the estimated cost of the
projects (16.9% below). However, the average schedule growth is
15.74%, showing the project took longer than the contracted dura- Fig. 2. Bid deviation (N ¼ 615).
tion. The construction intensity showed that the average cost spent
during construction was about $3,668=day. The PLC test was con-
ducted to determine the relationship of change order percentage
with other variables. The results of this test are described in the
following section.

Pearson Linear Correlation Test Results


Before conducting the PLC test, the researchers conducted assump-
tion tests. To check the normality of the data set, a histogram was
plotted for each of the variables showing normal distribution curves.
Figs. 1–6 show the histogram plots of change order percentage, bid
deviation, total project cost, cost growth, schedule growth, and con-
struction intensity. Checking these histograms, we can conclude that
most of the variables are normally distributed, except for change
order percentage and total project cost. However, the normality as-
sumption is very robust, and if the sample size increases by more
than 30 it will not cause a major problem (Ghasemi and Zahedias
2012). This is because the central limit theorem states that if a Fig. 3. Total project cost (N ¼ 615).
large sample size is randomly selected from the population, then
the mean of the sample will be equal to the mean of the population
(Investopedia 2018). In this study, we selected 615 data sets, which
can be considered a large sample. orders versus schedule growth, respectively. The scatterplots also
Because the data are normally distributed and the dependent and show that the strength of the correlation is very weak.
independent variables are in ratio scale, the PLC test was conducted Similarly, the results also show that there is no correlation of
to verify whether there are significant relationships between the change order percentage with bid deviation, total project cost, or
change order percentage and other independent variables. The re- construction intensity. Therefore, we can conclude that these perfor-
sults of the PLC test are shown in Table 4. The results show that mance metrics related to cost and schedule did not increase linearly
change order percentage had a significant positive correlation with as the change order percentage increased. Therefore, the research-
cost growth (0.33) and schedule growth (0.28). Figs. 7 and 8 give ers divided the projects into two groups: one with projects having
the scatterplots of change orders versus cost growth and change no change orders or negative change orders, and the other projects

© ASCE 04519025-5 J. Leg. Aff. Dispute Resolut. Eng. Constr.

J. Leg. Aff. Dispute Resolut. Eng. Constr., 2019, 11(4): 04519025


Table 4. Pearson correlation test results for change order percentage
Sample Correlation
No. Performance metrics size coefficient p-value
1 Bid deviation 615 0.01 0.83
2 Total project cost 615 0.004 0.92
3 Cost growth 615 0.33 0.00a
4 Schedule growth 615 0.28 0.00a
5 Construction intensity 615 −0.08 0.07
a
Significant at α level 0.05.
Downloaded from ascelibrary.org by Nottingham Trent University on 08/25/19. Copyright ASCE. For personal use only; all rights reserved.

Fig. 4. Cost growth (N ¼ 615).

Fig. 7. Change order versus cost growth (N ¼ 615).

Fig. 5. Schedule growth (N ¼ 615).

Fig. 8. Change order versus schedule growth (N ¼ 615).

projects higher than the projects that have positive change orders.
The other three t-tests were similar tests to determine whether the
Fig. 6. Construction intensity (N ¼ 615). cost growth, schedule growth, and construction intensity are better
in projects that have no change orders or negative change orders
compared to the projects that have positive change orders. The re-
sults of the t-tests for the case of bid deviation is described in the
having positive change orders in order to determine whether the per- following.
formance of these two groups of projects are significantly different.
To validate these findings, t-tests were conducted, and the results of Results of t-Tests on Bid Deviation
these tests are described in the following. Before conducting t-tests, the researchers need to check the
assumption tests. The main assumption test is normality, and the
normality of the data was checked by plotting histograms during
Results of t -Tests
the PLC test. Therefore, the researchers can assume that the data
The researchers conducted t-tests to check the four major hypoth- were normally distributed. However, another important assumption
eses. The first was to prove that for projects that have no change of the t-test is equal variance in both groups of projects. To test equal
orders or negative change orders, the contractors are bidding these variances among these groups, a Levene’s test was conducted.

© ASCE 04519025-6 J. Leg. Aff. Dispute Resolut. Eng. Constr.

J. Leg. Aff. Dispute Resolut. Eng. Constr., 2019, 11(4): 04519025


Table 5. Levene’s test results for bid deviation
Group Sample size Total project cost Variance F-statistic p-value
Projects with no change orders or negative change order percentage 299 $397,000 548.66 10.28 0.001a
Projects with positive change order percentage 316 $420,000 459.17
a
Significant at α level 0.05.

Table 6. t-test with unequal variance results for change order percentage Table 7. Levene’s test results for cost growth
versus bid deviation
Sample Levene
Sample Mean Group size Variance statistic p-value
Group size (%) t statistic p-value
Projects with no change orders or 299 593 3.28 0.07
−14.96 0.03a
Downloaded from ascelibrary.org by Nottingham Trent University on 08/25/19. Copyright ASCE. For personal use only; all rights reserved.

Projects with no or negative 299 2.19 negative change order percentage


change order percentage Projects with positive change 316 459
Projects with positive change 316 −11.21 order percentage
order percentage
a
Significant at α level 0.05.
Table 8. t-test results for change order percentage versus cost growth

The results of this test are shown in Table 5. The test results show Sample Mean
that the variances of these two groups were not equal. Therefore, the Group size (%) t-statistic p-value
researchers conducted a t-test with unequal variance to test this hy- Projects with no or negative 299 −21.32 −4.63 001a
pothesis. In addition, there was no large difference in the total change order percentage
project costs of these two groups of projects. An in-depth analysis Projects with positive 316 −12.76
of the types of projects showed that both groups had projects that change order percentage
involved the construction of roads, pavements, drains, culverts, re- a
Significant at α level 0.05.
taining walls, and other things.
The result of the t-test with unequal variances is shown in
Table 6. The test results show that there was a significant difference Table 9. Levene’s test results for schedule growth
in the mean values of bid deviation on these two types of projects:
Sample Levene
those with no change orders or negative change orders, and the Group size Variance statistic p-value
other with positive change orders. This shows that in the projects
that have no change orders or negative change orders, the winning Projects with no change orders or 299 2,675 0.50 0.82
contractors bid the projects, on average, 14.96% less than the esti- negative change order percentage
Projects with positive change 316 1,770
mated cost. However, the contractors bid, on average, 11.21% less
order percentage
than the estimated cost on projects that have positive change orders.
Therefore, it can be concluded that irrespective of how low the con-
tract bid is, change orders occur, and they occur at a higher rate on
the projects that have higher bids than on the projects that have Table 10. t-test results for change order percentage versus schedule growth
lower bids compared to the estimated costs of the projects. Sample Mean
Group size (%) t-statistic p-value
Results of a t-Test on Cost Growth
Projects with no or negative 299 7.32 −4.32 0.001a
The correlation test performed previously showed that there was
change order percentage
not a strong correlation between change order percentage and Projects with positive 316 23.71
the cost growth of small highway projects. Therefore, the research- change order percentage
ers conducted a t-test to determine whether the cost growth was
significantly lower on the projects that had no change orders or
a
Significant at α level 0.05.
negative change orders than that of the projects that had positive
change orders in order to show whether there is a significant det- cost performance when they had no change orders or negative
rimental effect of change orders on cost growth for small highway change orders.
projects. Before conducting the t-tests, the researchers conducted a
Levene’s equal variance test for these analyses. The results of the Results of t-Test on Schedule Growth
Levene’s test show that the variances were equal in these two The researchers conducted a similar t-test to determine whether the
groups of projects (Table 7). Therefore, the researchers conducted change orders had a detrimental effect on the schedule growth of
t-tests with equal variance. these projects. First, a Levene’s test was conducted to check
The t-tests with equal variance showed that there was a signifi- whether the variances were equal for these two groups of projects.
cant difference between the mean values of cost growth when the Table 9 shows the results of the Levene’s test; the variances of these
projects were divided into two groups: one with no change orders two groups were equal. Therefore, a t-test with equal variance was
or negative change orders, and the other with positive change or- conducted to test this hypothesis.
ders (Table 8). The mean values of cost growth showed that both The t-test results showed that the difference of schedule growth
groups of projects had negative cost growths; however, the cost for these two groups of projects was significant (Table 10). The
growth increased by about 40% when the projects had positive results showed that the schedule growth increased by threefold
change orders compared to the projects with no change orders in the projects that had positive change orders compared to the
or negative change orders. Therefore, the projects had excellent projects that had no change orders or negative change orders.

© ASCE 04519025-7 J. Leg. Aff. Dispute Resolut. Eng. Constr.

J. Leg. Aff. Dispute Resolut. Eng. Constr., 2019, 11(4): 04519025


Table 11. Levene’s test results for construction intensity the bid deviation and the change order percentage. However, when
Sample Levene the projects were divided into two groups, the analysis showed that,
Group size Variance statistic p-value on average, the contractors were bidding much lower than the esti-
mated cost on the projects that had no change orders or negative
Projects with no change orders or 299 8,398,404 34.61 0.01a
negative change order percentage change orders when compared to the projects that had positive
Projects with positive change 316 5,527,201 change orders. Therefore, based on the data of FDOT, the research-
order percentage ers cannot prove the assumption that low-bid contractors will com-
pensate their low bids by using construction change orders. The
a
Significant at α level 0.05.
change orders occurred randomly, independent of the bidding pat-
tern. However, data related to the types of change orders were not
available, so it is not clear whether the change orders occurring in
Table 12. t-test with unequal variance results for change order percentage these projects were due to scope changes by FDOT or due to con-
versus construction intensity
tractors asking for changes related to errors or omissions in design.
Downloaded from ascelibrary.org by Nottingham Trent University on 08/25/19. Copyright ASCE. For personal use only; all rights reserved.

Sample Mean Therefore, the authors would like to caution readers that it cannot be
Group size ($=day) t-statistic p-value determined whether the change orders generated in these projects
Projects with no or negative 299 3,858 1.93 0.05a were due to contractors’ intentions to increase project costs or due
change order percentage to scope changes from FDOT.
Projects with positive 316 3,488 Another important objective of this study was to determine the
change order percentage correlation between the change order percentage and cost growth,
a
Significant at α level 0.05. schedule growth, and construction intensity. The study could not
find a strong positive correlation of change order percentage with
cost growth, schedule growth, or construction intensity. However,
This showed that the change orders had a detrimental effect on the when the projects were divided into two groups of projects, one
schedule growth of small highway projects. with no change orders or negative change orders and the other with
positive change orders, the tests showed that change orders had a
Results of t-Test on Construction Intensity
detrimental effect on cost growth, schedule growth, and construc-
To check whether construction intensity was significantly different
tion intensity.
in the two types of projects, the researchers again conducted a
The analysis showed that cost growth increased by about 40% in
t-test. The results of the Levene’s test, to test the equal variances
projects that had positive change orders compared to the projects
in these two groups, are shown in Table 11. They showed that the
that had no change orders or negative change orders. However, both
variances of construction intensity for the two groups of projects
groups of projects had overall negative cost growth, showing that
were not equal. Therefore, a t-test of unequal variance was con-
the projects were completed below the estimated costs, which is
ducted to determine the effect of change orders on the construction
impressive.
intensity of small highway projects.
The analysis of the impact of change orders on schedule growth
The results of the t-test with unequal variances are shown in
of these two groups of projects showed a similar pattern as cost
Table 12. The results showed a similar pattern as the cases of cost
growth. The schedule growth increased significantly when positive
and schedule growth. When positive change orders occurred in the
change orders occurred on the projects. The analysis showed that
small highway projects, the construction intensity was significantly
reduced. The construction intensity decreased by about 10% in the the schedule growth increased by about 224% when the projects
projects that had positive change orders compared to the projects had positive change orders compared to the projects that had no
with no change orders or negative change orders. Therefore, the change orders or negative change orders.
change order effect is significant and detrimental on construction The construction intensity findings were also similar to cost
intensity. and schedule growth findings. There was a reduction of about 4%
of construction intensity on the projects that had positive change
orders compared to the projects that had no change orders or neg-
Discussions ative change orders.
The limitation of this study is that the project data were collected
The major objective of this study was to determine the mean values only from Florida DOT; therefore, the findings of this study cannot
of change order percentages, cost growth, schedule growth, and be generalized. In addition, only small highway project data were
construction intensity occurring in small highway projects. It was analyzed; therefore, it is recommended that investigation of the im-
found that the average change order percentages, cost growth, and pact of change orders on cost, schedule, and construction intensity
schedule growth were 2.22%, −16.90%, and 15.74%, respectively. should be conducted on large Florida highway projects to deter-
The mean construction intensity of these projects was found to be mine whether the size had any effect on these findings.
$3,668=day. The average total project cost of each project was
about $434,000. The data showed that in small highway projects,
even though change orders occurred, the projects were completed, Conclusions
on average, 17% below the estimated cost, which is very impressive.
Another objective of this study was to determine whether the This paper analyzed the change order percentage, bid deviation,
bidding pattern of lowest responsive bidder had any impact on cost growth, schedule growth, and construction intensity of small
the change orders in small highway projects. It is generally as- highway projects. The major finding of this study is that there is no
sumed that contractors generally bid a project low and try to make correlation between low bid and the change order percentage. The
up the difference of low bids by generating change orders during study could not prove the hypothesis that contractors will make up
construction. This study aimed to determine whether this pattern the bid difference with change orders during construction.
was seen in the small highway projects constructed by FDOT. The Another important finding of this study is about the negative
PLC test showed that there is no significant correlation between impact of change orders on cost growth, schedule growth, and

© ASCE 04519025-8 J. Leg. Aff. Dispute Resolut. Eng. Constr.

J. Leg. Aff. Dispute Resolut. Eng. Constr., 2019, 11(4): 04519025


construction intensity on small highway projects. It was found that Günhan, S., D. Arditi, and J. Doyle. 2007. “Avoiding change orders in
when projects had no change orders or negative change orders, the public school construction.” J. Prof. Issues Eng. Educ. Pract. 133 (1):
cost growth, schedule growth, and construction intensity were supe- 67–73. https://doi.org/10.1061/(ASCE)1052-3928(2007)133:1(67).
rior to those of the projects that had positive change orders. There- Halwatura, R. U., and N. P. N. P. Ranasinghe. 2013. “Causes of variation
orders in road construction projects in Sri Lanka.” ISRN Constr. Eng.
fore, it is necessary that the change orders be controlled and state
2013: 381670. https://doi.org/10.1155/2013/381670.
DOTs make an effort not to have positive change orders on their Hanna, A. S., J. S. Russell, E. V. Nordheim, and M. J. Bruggink. 1999.
projects so that cost and schedule performance can be improved. “Impact of change orders on labor efficiency for electrical construc-
The primary contribution of this study to the body of knowledge tion.” J. Constr. Eng. Manage. 125 (4): 224–232. https://doi.org/10
was to show the significant impact of change orders on cost growth, .1061/(ASCE)0733-9364(1999)125:4(224).
schedule growth, and construction intensity for small highway proj- Huang, A. L., R. E. Chapman, and D. T. Butry. 2009. Metrics and tools for
ects. Because this is the first comprehensive study on the impact measuring construction productivity: Technical and empirical consid-
of change order percentages on cost growth, schedule growth, and erations. Rep. No. NIST SP 1101. Gaithersburg, MD: US Dept. of
construction intensity for small FDOT highway projects, the re- Commerce, National Institute of Standards and Technology.
Downloaded from ascelibrary.org by Nottingham Trent University on 08/25/19. Copyright ASCE. For personal use only; all rights reserved.

searchers recommend using data from other state DOTs for further Ibbs, C. W. 2008. “The cumulative impact of change on construction labor
productivity.” Accessed October 13, 2017. http://constructionclaims
study. This effort could help engineers from state DOTs to deter-
.com/Content/THE-CUMULATIVE-IMPACT-OF-cpCCA14.aspx.
mine whether the impact of change orders on cost and schedule in Ibbs, C. W. 2012. “Construction change: Likelihood, severity, and impact
their states were similar to that of FDOT. on productivity.” J. Leg. Aff. Dispute Resolut. Eng. Constr. 4 (3): 67–73.
https://doi.org/10.1061/(ASCE)LA.1943-4170.0000089.
Investopedia. 2018. “Central limit theorem.” Accessed June 22, 2018.
Acknowledgments https://www.investopedia.com/terms/c/central_limit_theorem.asp.
Jawad, R. S., R. Abdulkader, and A. A. Abang Ali. 2009. “Variation orders
The researchers would like to acknowledge the support provided in construction projects.” J. Eng. Appl. Sci. 4 (3): 170–176.
by FDOT during data collection for this study. Without its support, Laerd Statistics. n.d. “Independent t-test for two samples.” Accessed
the data collection and analysis for this study would not have been June 22, 2018. https://statistics.laerd.com/statistical-guides/independent
-t-test-statistical-guide.php.
possible. Any views and findings provided here in the paper are
Ndihokubway, R., and T. C. Haupt. 2008. “Uncovering the origins of varia-
of solely those of the researchers, and have no input from FDOT tion orders.” In Proc., 5th Postgraduate Conf. Construction Industry
personnel. Development Board. Port Louis, Mauritius: Construction Industry
Development Board.
Rumsey, D. J. n.d. “How to interpret a correlation coefficient r.” Accessed
December 16, 2018. http://www.dummies.com/education/math/statistics
References
/how-to-interpret-a-correlation-coefficient-r/.
Serag, E., A. Oloufa, L. Malone, and E. Radwan. 2010. “Model for quan-
tifying the impact of change orders on project cost for US roadwork
Works Cited
construction.” J. Constr. Eng. Manage. 136 (9): 1015–1027. https://doi
Alnuaimi, A. S., R. A. Taha, M. Al Mohsin, and A. S. Al-Harthi. 2010. .org/10.1061/(ASCE)CO.1943-7862.0000206.
“Causes, effects, benefits, and remedies of change orders on public con- Shrestha, P. P., and R. Maharjan. 2018. “Effects of change orders on cost
struction projects in Oman.” J. Constr. Eng. Manage. 136 (5): 615–622. growth, schedule growth, and construction intensity of large highway
https://doi.org/10.1061/(ASCE)CO.1943-7862.0000154. projects.” J. Leg. Aff. Dispute Resolut. Eng. Constr. 10 (3): 04518012.
Anastasopoulos, P. C., S. Labi, A. Bhargava, C. Bordat, and F. L. https://doi.org/10.1061/(ASCE)LA.1943-4170.0000264.
Mannering. 2010. “Frequency of change orders in highway construc- Shrestha, P. P., and N. Pradhananga. 2010. “Correlating bid price with the
tion using alternate count-data modeling methods.” J. Constr. Eng. number of bidders and final construction cost of public street projects.”
Manage. 136 (8): 886–893. https://doi.org/10.1061/(ASCE)CO.1943 Transp. Res. Rec. 2151: 3–10. https://doi.org/10.3141/2151-01.
-7862.0000198. Shrestha, P. P., K. K. Shrestha, and T. K. Kandie. 2017. “Effects of change
Assaf, S. A., and S. Al-Hejji. 2006. “Causes of delay in large construction orders on the cost and schedule of rural road maintenance projects.”
projects.” Int. J. Project Manage. 24 (4): 349–357. https://doi.org/10 J. Leg. Aff. Dispute Resolut. Eng. Constr. 9 (3): 04517010. https://doi
.1016/j.ijproman.2005.11.010. .org/10.1061/(ASCE)LA.1943-4170.0000227.
Bordat, C., B. G. McCullouch, S. Labi, and K. Sinha. 2004. An analysis of Shrestha, P. P., K. K. Shrestha, and H. B. Zeleke. 2018. “Probability
cost overruns and time delays of Indiana Department of Transportation of change orders and the effect on cost and schedule for new public
projects. Rep. No. FHWA/IN/JTRP-2004/7, SPR-2811. Washington, school buildings.” J. Eng. Constr. Arch. Manage. https://doi.org/10
DC: Transportation Research Board. .1108/ECAM-01-2018-0017 26 (6): 1087–1104.
Choi, K., H. Lee, J. Bae, and D. Bilbo. 2016. “Time-cost performance effect Shrestha, P. P., and H. Zeleke. 2018. “Effect of change orders on cost and
of change orders from accelerated contract provisions.” J. Constr. Eng. schedule overruns of school building renovation projects.” J. Leg. Aff.
Manage. 142 (3): 04015085. https://doi.org/10.1061/(ASCE)CO.1943 Dispute Resolut. Des. Constr. 10 (4): 04518018. https://doi.org/10.1061
-7862.0001071. /(ASCE)LA.1943-4170.0000271.
Engineering News-Record. 2017. “Construction cost index history.” Statistics Solutions. 2018. “Pearson’s correlation coefficient.” Accessed
Accessed October 5, 2017. https://www.enr.com/economics/historical December 12, 2018. https://www.statisticssolutions.com/pearsons
_indices. -correlation-coefficient.
FDOT (Florida DOT). 2018a. “Change order procedure for local agency Thomas, H. R., and C. L. Napolitan. 1995. “Quantitative effects of construc-
projects.” Accessed July 25, 2018. https://www.fhwa.dot.gov/federal tion changes on labor productivity.” J. Constr. Eng. Manage. 121 (3):
-aidessentials/catmod.cfm?id=4. 290–296. https://doi.org/10.1061/(ASCE)0733-9364(1995)121:3(290).
FDOT (Florida DOT). 2018b. “Reports on highway mileage and travel.” Vandenberg, P. J. 1996. “The impact of change orders on mechanical con-
Accessed July 25, 2018. https://www.fdot.gov/statistics/mileage-rpts struction labor efficiency.” Master’s thesis, Dept. of Civil Engineering,
/default.shtm. Univ. of Wisconsin-Madison.
Ghasemi, A., and S. Zahedias. 2012. “Normality tests for statistical analy- Williams, T. P. 2002. “Predicting completed project cost using bidding
sis: A guide for non-statisticians.” Int. J. Endocrinol. Metab. 10 (2): data.” J. Constr. Manage. Econ. 20 (3): 225–235. https://doi.org/10
486–489. https://doi.org/10.5812/ijem.3505. .1080/01446190110112838.

© ASCE 04519025-9 J. Leg. Aff. Dispute Resolut. Eng. Constr.

J. Leg. Aff. Dispute Resolut. Eng. Constr., 2019, 11(4): 04519025

You might also like