Professional Documents
Culture Documents
Effects of Change Orders On Cost Growth
Effects of Change Orders On Cost Growth
Abstract: Previous research have shown that change orders have a detrimental effect on cost growth and schedule growth of construction
projects. These studies mainly concentrated on the number of change orders for building or maintenance projects. However, correlation of the
number and change order growth with cost growth, schedule growth, and construction intensity of new highway projects has not been studied.
This study collected data from 185 projects from the state of Texas, whose costs amounted to more than $10M. The results showed that the
average change order growth for these projects was 7.0%, and also found a significant positive correlation in change order growth with cost
growth and schedule growth of these projects. The correlation coefficients of change order growth with cost growth and schedule growth were
found to be 0.57 and 0.44, respectively. When the projects were divided into two groups based on the change order growth, the results showed
that the cost growth and schedule growth significantly increased as the change order growth increased to greater than 5%. The number of
change orders were significantly and positively correlated with cost growth (0.23), schedule growth (0.30), and construction intensity (0.46).
When the number of change orders increased by more than 20, then the cost and schedule growth increased significantly. It is recommended
that Texas Department of Transportation engineers use an effective change management process to control the change orders in their highway
projects so that cost and schedule overruns could be minimized. DOI: 10.1061/(ASCE)LA.1943-4170.0000264. © 2018 American Society of
Civil Engineers.
Author keywords: Change order growth; Number of change orders; Texas Department of Transportation; Correlation.
Anastasopoulos et al. 2010; Ibbs 2008, 2012). Therefore, the main poor estimation, unforeseen site conditions, political pressure dur-
objectives of this study were: ing construction stage, poor soil investigation, and client-initiated
1. Determine the CO growth occurring in new highway projects. variation (Halwatura and Ranasinghe 2013). The findings of this
2. Determine the correlation between CO growth and project study were based on surveys of 50 respondents who had experience
performance data (i.e., cost growth, schedule growth, and con- in road construction projects in Sri Lanka. Dickson et al. (2015)
struction intensity). conducted a questionnaire survey to determine the factors that
3. Determine the correlation between the number of COs and cause COs to occur in civil engineering construction projects in
project performance data (i.e., cost growth, schedule growth, Kenya. There were 12 clients, 32 consultants, and 51 contractors
and construction intensity). who responded to this survey. They ranked delay in land acquis-
4. Determine the optimum values of COs for a project over which ition, differing site conditions, change in scope, change in schedule
there would be a detrimental effect on the cost, schedule growth, by client, and lack of coordination as the top five important causes
and construction intensity of the project. of COs. Rowland (1981) found that the rate of COs increased as the
project size and complexity increased. These findings were based
Definition of Change Order and Cost and Schedule on a study of 18 public works construction projects in the southern
Metrics Used in the Study United States.
Past research found that COs mostly were generated due to de-
To test the hypothesis, a total of five metrics were developed, two sign changes in the projects (Ndihokubway and Haupt 2008;
metrics related to COs, one related to cost, and two related to sched- Alnuaimi et al. 2010; Gunhan et al. 2007; Bordat et al. 2004). Some
ule. The aim of this study was to determine the correlation between of the research found that factors that contributed to changes in the
CO growth and number of COs with the cost and schedule perfor- projects included clients, consultants, contractors, state regulation
mance of new highway projects. Therefore, the COs were repre- statutes, adverse weather conditions, political pressure, poor estima-
sented by two metrics, one for CO growth, expressed in terms of tion, poor site investigation, and unforeseen site conditions
percentage growth against the total contract cost, and the other (Ndihokubway and Haupt 2008; Alnuaimi et al. 2010; Jawad et al.
for the number of COs. One cost-performance metric was cost 2009; Serag et al. 2010). Serag et al. (2010) developed models to
growth and the two schedule-performance metrics were schedule predict CO growth based on 11 variables found in 16 Florida
growth and construction intensity. Table 1 shows the equations used DOT projects. The contract values of these projects ranged of
to calculate these metrics. $10 million–$25 million. The authors found that CO growth was
significantly correlated to the timing of the COs and unforeseen
conditions.
Literature Review Other research found that CO growths occurring in road main-
tenance projects had a relationship with road surface types and
Review of past studies mainly focused on the causes and the effects project types (Shrestha et al. 2017; Anastasopoulos et al. 2010).
of COs on the cost, schedule, and construction intensity of trans- Shrestha et al. found that mixed-surface road pavements (14.7%)
portation and other types of construction projects. COs affect the had significantly higher amounts of COs compared to gravel surfaces
overall cost and schedule of a project, resulting in claims and (13.1%) and earthen surfaces (14.7%) in Kenyan highway projects.
disputes between the owner and the contractor. Therefore, it is an Anastasopoulos et al. (2010) found that resurfacing and traffic main-
important parameter, and requires close attention to complete a tenance work had a significantly greater number of COs than earth-
project successfully. work and subsoil treatment projects. Both of these studies found that
β change order growth vs:cost growth ¼ 0 ð1Þ p-values were found to be less than 0.05, then the null hypothesis
would be rejected, confirming the research hypothesis that the
β change order growth vs:schedule growth ¼ 0 ð2Þ values of these groups were significantly different. If p-values
were found to be less than 0.01, the difference in values would
β change order growth vs:construction intensity ¼ 0 ð3Þ be considered as highly significant. In this case also, whichever test
yielded strongest difference, those results were reported.
β number of change order vs:cost growth ¼ 0 ð4Þ
β number of change order vs:construction intensity ¼ 0 ð6Þ First, the entire data set for 185 projects was plotted based on the
CO growth, the number of COs, cost growth, schedule growth, and
μcost growth below optimum change order growth value construction intensity. The box plots in Figs. 1–5 indicate that there
were outliers in the data set based on these five variables. However,
¼ μcost growth above optimum change order gowth value ð7Þ
outliers were not removed because the authors could not prove that
the outliers were due to a nonrepetitive error. Therefore, all the data
μcost growth below optimum number of change order
for 185 projects were used to determine the mean, the median, and
¼ μcost growth above optimum number of change order ð8Þ the standard deviation values for these variables.
Descriptive Statistics
Statistical Analysis The mean values for CO growth, cost growth, and schedule growth
To prove or reject the null hypotheses, statistical tests need to be was plotted, as shown in Fig. 6. Results indicate that the average
conducted. For the correlation test, the appropriate test was the values of CO growth for these projects were found to be 7%;
Pearson linear correlation test because the variables are in a ratio furthermore, these COs increased the cost by 6.6%, on average,
scale. However, before conducting this test, it should be verified and schedule by 21%.
whether the data distributions related to the COs, cost, and schedule Table 2 shows a summary of (1) the descriptive statistics for the
metrics are normally distributed (Real Statistics 2017). To check CO growth, (2) the number of COs, and (3) the cost growth, sched-
the normality of the data, the Shapiro-Wilko test was conducted; ule growth, and construction intensity of these projects. This table
it was found that all the data set were not normally distributed. indicates that schedule growth was found to be higher than cost
(The results of this test are shown in the “Results” section). How- growth and CO growth. Data analysis showed that 35 COs occurred
ever, the normal distribution test is a robust test. As the size of the in highway projects, on average. The mean value for construction
sample was more than 30 in this case, this parametric test could be intensity was found to be about $38,809=day. By comparing the
conducted to determine the correlation between these variables.
Also, a nonparametric test, i.e., Spearman’s correlation test, was
conducted to determine the correlation between these two variables
to determine the differences in the results (Leard Statistics 2017b).
Similarly, both the t-test and the Mann-Whitney U test were con-
ducted to determine whether there was a difference among the
values for cost growth, schedule growth, and construction intensity
based on an optimum value of CO growth and the number of COs
(Leard Statistics 2017a).
Fig. 6. Mean values of change order growth, cost growth, and schedule
growth.
Fig. 3. Boxplot of cost growth.
mean and median values of all these variables, it can be seen that
there are some outliers in this data set. However, the outliers were
not removed because the data were collected randomly from the
TxDOT, and it could not be proved that the outliers were due to
a nonrepetitive error.
and construction intensity. Similarly, in order to determine the The data also was analyzed using the nonparametric Spearman’s
optimum values for CO growth and the number of COs—after correlation test, and the results are shown in Appendix I. The results
which cost growth, schedule growth, and construction intensity showed that correlation coefficients for all these relationships were
significantly increases—the t-test was conducted. However, to lower than for the results from the Pearson correlation test.
use the t-test, the data needed to be checked for normality. There- Pearson and Spearman’s Correlation Test Results for the
fore, the Shapiro-Wilk test was conducted to determine whether the Number of Change Orders. Results from the parametric correla-
data were normally distributed. tion test showed that the number of COs was significantly corre-
Table 3, which shows the results of Shapiro-Wilk test, indicates lated to cost growth, schedule growth, and construction intensity
that none of the data were normally distributed. In this test, the null (Table 5). The correlation coefficient between the number of
hypothesis stated that the population is normally distributed, which COs and cost growth (0.23) and the number of COs and schedule
would be rejected if the p-value was less than 0.05. Therefore, for growth (0.30) were less than that obtained for CO growth. The
CO growth, the number of COs, cost growth, schedule growth, strength of the correlation between these variables was not strong;
and construction intensity, the p-values were found to be less than however, in both cases, the correlations were found to be highly
0.05, which showed that the data were not normally distributed. significant at alpha level 0.05. This test also showed that the
However, normal distribution tests are very robust, and conceiv- number of COs had a significant correlation with construction in-
ably can be applied to nonnormally distributed data. Also, if the tensity (0.46), and correlation was found to be moderately strong.
sample size is more than 30, the data can be considered to be It showed that if the number of COs increased, then the speed of
normally distributed, which was true in this case. Therefore, the construction would increase. The correlation test was conducted
Pearson correlation test and the t-test could be performed instead once again using the nonparametric test, and the results are shown
of nonparametric tests, Spearman’s correlation test, and the Mann- in Appendix II. The results showed the similar findings as in the
Whitney U test. However, the authors conducted both the paramet- Pearson correlation test.
ric and nonparametric tests in order to determine the differences
in the results. Results of the t-Test and Mann-Whitney U Test
The t-test and the Mann-Whitney U test were used to determine the
Correlation Test Results difference between the performance metrics for cost and schedule,
Pearson and Spearman’s correlation tests were conducted to based on amount of CO growth and the number of COs. This analy-
determine the association of CO growth and the number of COs sis determined the optimum values for CO growth and the number
with cost growth, schedule growth, and construction intensity.
of COs above which cost growth, schedule growth, and construc-
The results of the Pearson test for the correlation of CO growth
tion intensity would increase significantly. These tests were
and the number of COs and cost, schedule growth, and construc-
conducted for CO growth as well as the number of COs, described
tion intensity are described as follows. Also, the results of
as follows.
Spearman’s correlation test among these variables is shown in
t-Test and Mann-Whitney U Test Results for Chang Order
Appendixes I and II.
Growth. Before conducting this test, data for cost, schedule,
Pearson and Spearman’s Correlation Test Results for Change
and construction intensity were divided into two groups, based
Order Growth. Table 4 shows the results of Pearson correlation
on CO growth. It was found that the median value for CO growth
test for the coefficients values of CO growth with cost growth,
for this dataset was 3.9%. Therefore, the analysis was conducted
schedule growth, and construction intensity. Test results showed
that the CO growth had a positive correlation with cost growth four times by dividing the project data into two groups based
(0.57) and schedule growth (0.44). Both of these correlation coef- on 3–6% CO growth. When the cost growth was divided into
ficients were moderately strong and highly significant at alpha level two groups based on 3–6% CO growth, the t-statistics for the t-tests
0.05. In addition, the results showed that the COs increased the cost were found to be 4.42, 5.57, 5.96, 5.70, respectively (Table 6).
growth and the schedule growth of highway projects. However, Therefore, it can be concluded that the difference in mean values
there was no significant correlation between CO growth and con- for cost growth in the two groups was significantly the largest at
struction intensity. This finding showed that the speed of the alpha level 0.05 when the groups were divided based on 5% CO
growth. This showed that if the CO growth increased by more 5%,
then it would have a significantly adverse effect on cost growth of
Table 4. Pearson correlation test results for change order growth
the projects. The difference between the values of cost growth was
Performance Sample Correlation 19.8% when the projects were divided based on 5% cutoff point for
Number metrics size coefficient p-value CO growth. Table 6 also showed that in the case of schedule
1 Cost growth 185 0.57 0.00a growth, the difference was highest and significant when the proj-
2 Schedule growth 185 0.44 0.00a ects were divided based on a 5% cutoff point for CO growth. The
3 Construction intensity 185 0.04 0.54 difference between the values of schedule growth was found to
a
Significant at alpha level 0.01. be 19.1%.
a
Significant at alpha level 0.01.
The results of t-test regarding construction intensity was not • As the CO growth increased beyond 5%, the effects of this
shown in Table 6 because the difference was not found to be sig- increase on cost growth and schedule growth was highly signif-
nificant. Results showed that the amount of CO growth did not have icant and severe.
any adverse effect on the speed with which construction in highway • As the number of COs increased, cost growth, schedule
projects were completed for those projects whose data were growth, and construction intensity increased for these highway
collected for this study. The nonparametric test, the Mann-Whitney projects.
U test, was conducted to determine whether the results were • As the number of COs increased beyond 20, the effect on cost
similar, which they were. The results of this test for two groups of and schedule growth was highly significant and severe.
projects, divided based on a 5% cutoff point of CO growth, is This study found that the mean CO growth for new highway
shown in Appendix III. projects constructed by TxDOT that cost more than $10 M was
Mann-Whitney U Test Results for Number of Change Order. 7.0%. This CO growth for new highways was less than for bridge
The t-test and the Mann-Whitney U test were conducted once again, projects (8.1%) and maintenance projects (7.5%), but higher than
based on the number of COs. The median value for the number of for road resurfacing (5.6%) and traffic projects (5.6%), as reported
COs was found to be 20. Therefore, the projects were divided into by Bordat et al. (2004). When compared to the findings from
two groups, one consisting of projects that had less than 20 COs and Shrestha et al. (2017) for road maintenance projects of Kenya,
the other that had more than 20 COs. The results of the t-test, shown new highway projects of TxDOT had significantly fewer COs.
in Table 7, indicated that as the number of COs increased by Therefore, compared to previous studies for new highways and
more than 20, the cost growth and schedule growth increased sig- maintenance projects, the CO growth for TxDOT new highway
nificantly. The cost growth and schedule growth of projects that projects were comparatively lower.
had more than 20 COs (11.5 and 28.5%, respectively) were The correlation between CO growth and cost growth for these
significantly higher than for cost growth and schedule growth of projects was similar to those found for other infrastructure and main-
the projects that had less than 20 COs (0.2 and 11.6%, respectively). tenance projects as well as for building projects. This study found a
Similar results were found for construction intensity. Therefore, as significant and moderately strong positive correlation of CO growth
the number of COs increased to over 20, the cost growth, schedule with cost growth of the projects. The cost growth was measured as
growth, and construction intensity increased significantly for new the difference between the final project cost and the engineer’s
highway projects. It showed that the engineers need to limit the num- estimate divided by the engineer’s estimate. Similarly, other studies
ber of change orders within 20, so that the effect on cost and sched- (e.g., Shrestha et al. 2017; Ibbs 2008, 2012) found that COs nega-
ule growth will be minimum. In addition, the Mann-Whitney tively affected cost growth. This study also found that the limit for
test was conducted to determine whether the results would be differ- CO growth was 5% of the total project cost. When the percentage
ent; however, it was found that this nonparametric test yielded exceeded this limit, then the effect of COs on cost growth was highly
similar results. The results of the Mann-Whitney test are shown significant. According to this data set, a percentage of 5% of COs is
in Appendix IV. the maximum that the state DOTs should limit; if the COs exceeded
this percentage, then cost growth significantly increases.
This study also determined that the number of COs had a detri-
Discussions mental effect on cost growth. The results showed a significant
positive correlation between these two variables, validating the
The major findings of this study are as follows: hypothesis that as the number of CO increases, the cost growth
• As the CO growth increased, the cost growth and schedule increases. This study found that the optimum number of COs is
growth increased for these new highway projects. 20; if the number of COs increased more than this number, then
also was found, which validated the hypothesis that as the number whether the effects of COs in their state were similar to that
of COs increased, the schedule growth will be negatively affected. for TxDOT.
Finally, Anastasopoulos et al. (2010) found a nonlinear correlation
of the number of COs with contract duration. This current study
showed that when the number of COs increased more than 20, then Appendix I. Spearman’s Correlation Test Results for
the schedule growth will increase significantly. Change Order Growth
This study did not find any significant correlation between CO
growth and construction intensity. However, when the number of Performance Sample Correlation
COs increased over 20, then the construction intensity increased Number metrics size coefficient p-value
significantly. This result is counterintuitive in that the increase 1 Cost growth 185 0.46 0.00a
in the number of COs had a negative effect on productivity and 2 Schedule growth 185 0.32 0.00a
construction speed. The authors thought that the correlation found 3 Construction intensity 185 0.03 0.71
between these variables was not due to causation; therefore, further a
Significant at alpha level 0.01.
investigation should occur on whether the construction intensity
increased due to an increase in the number of COs. Past researchers
had not studied the correlation between the number of COs and
construction intensity for any type of project. Appendix II. Spearman’s Correlation Test Results
for Number of Change Order
This study investigated the correlation between CO growth and 1 Cost growth 185 0.28 0.00a
2 Schedule growth 185 0.31 0.00a
number of COs with cost growth, schedule growth, and construc-
3 Construction intensity 185 0.35 0.00a
tion intensity. This study found that in new highway construction
a
built by TxDOT, the average CO growth was 7% and the number of Significant at alpha level 0.01.
COs was 35. The results showed that CO growth and the number of
COs negatively affected cost growth. This study also found that
up to a 5% growth in COs, the effects on cost growth was not Appendix III. Results of Mann-Whitney U Test for
severe. However, COs significantly affected cost growth once Change Order Growth
the limit of 5% was crossed. Similarly, when the number of
COs increased to greater than 20, the effect on cost growth became Mean values of
highly significant. performance metrics
Performance Chi-square
The findings regarding the effects of CO growth and the number metrics ≤5% change >5% change value p-value
of COs on schedule growth was similar to findings for cost growth,
Cost growth −1.3% 18.3% 31.6 0.00a
showing a negative effect on schedule growth. Similar to cost Schedule growth 12.8% 31.9% 14.0 0.01a
growth, for schedule growth, a 5% growth in COs was the optimum Construction $38,254=day $39,553=day 1.4 0.24
value beyond which schedule growth would be affected signifi- intensity
cantly. In addition, the number of COs beyond which schedule a
Significant at alpha level 0.01.
growth would be affected significantly was 20. To the best knowl-
edge of the authors, this study was the first to consider COs for new
highway construction; the results were similar to those for studies
involving building and highway maintenance projects. Appendix IV. Results of Mann-Whitney U Test for
The findings from this study could be a great value for TxDOT the Number of Change Orders
planning engineers, planners, and administrators, because they can
Mean values of
use the findings to reduce the amount and number of COs in their performance metrics
projects so that there would be minimal effect on cost and schedule Performance Chi-square
performance during the construction phase. The policy makers and metrics ≤20 changes >20 changes value p-value
engineers should scope and design the projects more effectively in Cost growth 0.2% 11.5% 12.1 0.00a
order to reduce the number and amount of COs during new high- Schedule growth 11.6% 28.5% 13.0 0.00b
way construction. It is recommended that TxDOT engineers use Construction $30,917=day $45,091=day 12.4 0.00b
more effective change management practices in order to reduce intensity
the effect of COs on cost and schedule overruns in their projects a
Significant at alpha level 0.05.
b
during construction phase. Significant at alpha level 0.01.