You are on page 1of 6

T ECHNICAL A RTICLE

Quantifying the Effect of change can result in a substantial loss in


efficiency. As a result, electrical contrac-
tors can be underpaid and often spend a
Change Orders on Electrical great deal of money and time document-
ing the true impact and cost of a change.
Construction Labor Efficiency Furthermore, some are facing litigation
and detrimental accusations of price in-
flating to cover anticipated or unantici-
Dr. Awad S. Hanna, PE; Dr. Jeffrey S. Russell, PE; and David Thomack pated losses. Electrical contractors have
not had industry-aggregated data to sub-
stantiate their position, whatever it may
many cases, contractors are not properly

T
he primary objective of this ar- be, because of the substantial effort re-
ticle is to quantify the effects compensated for the work they perform quired to quantify these effects.
that change orders have on and the results experienced on the esti-
project-level labor efficiency mated baseline productivity. At the same
in the electrical construction industry. time, owners have a tendency to feel they PRODUCTIVITY AND
Labor productivity is a risky aspect of a are being unfairly charged for work per- EFFICIENCY
project, and it affects a project’s prof- formed on directed changes, for reasons
itability. In addition, it is often difficult to ranging from mismanagement that results Efficiency is the ratio of actual per-
measure. Determining the exact cost of in poor productivity to an unreasonably formance to the theoretical maximum
change orders is a difficult, maybe even low initial bid. performance [2]. Depending on the
impossible, process. However, by tracking A change order can be defined as source, productivity can be defined by the
labor performance as it is affected by written authorization provided to a con- output divided by the input, or input by
change orders on a large number of past tractor that approves a change from the output. The input in most cases is labor
projects, a better explanation can be pro- original plans, specifications, or other hours, but also can be expressed in terms
vided by contractors hoping to obtain fair contract documents, as well as a change of capital. Upon collection of data from
compensation for work performed. The in the cost [2]. In the real world, change completed projects, the output variable
effects on efficiency are explored here at orders are not always authorized in writ- for productivity was found to be unattain-
the macroscopic level, and are not specif- ing. They can cause disruption in the able. The input, or labor hours expended,
ic to the scope, size, type, or timing of any planned work schedule and result in in- was readily available; for this reason, effi-
one specific change order on a given proj- creased costs through rework and de- ciency is used as the measurable in this
ect. Although there are many factors that creased productivity. Some problems re- study. Efficiency, for the purpose of this
reduce the efficiency and/or productivity lated to change orders include the research, is defined as the actual labor
of construction labor, this article explores following: performance compared to the estimated
the effects of change orders to better un- or standard performance. The variable
derstand their influence on labor produc- • changing the project scope, render- that will be used to determine efficiency
tivity. ing the original plan incomplete; is called “delta” (∆), and is defined as the
• loss of efficiency due to interruptions, difference between the base project labor
interference, and a lack of the tools, hours (actual hours - change order hours)
BACKGROUND labor, and materials needed to meet and the original estimated labor hours
the requirements of the new scope; when the contract was awarded. Figure 1
Owners have the right to initiate • difficulty in determining just com- illustrates this measure.
changes on a project. The only rule ap- pensation for the parties involved;
plying to such decisions is to provide di- • an increase in project management
rectives to the contractor doing the work. costs; LITERATURE REVIEW
It is to the advantage of the owner and the • an increase in the frequency of plan-
contractor that any notifications of ning and replanning efforts; and There have been few publications
change come as early as possible. The • loss of the learning curve effect due and/or studies dealing with the quantita-
contractor, in turn, has the right to equi- to interruption of work. tive effects of change orders on labor pro-
table adjustment for any and all services ductivity. This article focuses on the pro-
rendered. Conflicts can occur when the The connected nature of the electri- duction level of a construction project, in
cost of a change is not agreed upon by the cal trade magnifies the results of change part because productivity is the most im-
owner or his/her representative. The ef- orders during the construction process. portant determinant between a successful
fect of changes on the construction The electrical trade is a labor-intensive and unsuccessful project. The other rea-
process can be difficult to quantify. In field where, in many cases, a minor son for studying productivity is the mea-

36 Cost Engineering Vol. 40/No. 2 FEBRUARY 1998


projects. The results of the study showed
delta
can be positive or that while in many instances changes did
negative not affect labor productivity, on the aver-
age, a 30 percent loss in efficiency oc-
Total Labor curred during the performance of
Hours change-order work. To define efficiency, a
Labor Hours

∆ (base + change) performance ratio (PR) was first defined


as the actual productivity divided by the
Estimate Base baseline productivity. Efficiency was then
defined as the average PR, with a given
factor present divided by the average PR
under normal conditions. Thomas and
Napolitan point out that disruptions dur-
t = 0% t = 100% ing the work process were the major
cause of production loss, resulting in daily
Project Time Line efficiency losses in the range of 25 to 50
Figure 1—Graphic Illustration of ∆ percent.
The factor model [5] was used in this
surable, quantifiable nature of the pro- hypothesis was not possible. An investiga- study to analyze how change affected
duction. tion of hypothesis 2 showed that produc- labor productivity. Baseline productivity
Past research has indicated that there tivity rates lower than planned are was determined by monitoring and
is a direct correlation between an increase achieved when construction change is recording work performed and quantities
in change orders and a loss of productivi- greater than 5 percent [1]. This decrease installed on days when change-order
ty. One study [3] reported losses in pro- in total job productivity increases as the work was not present. The performance
ductivity associated with change orders, amount of change increases. This can be ratio (PR) was then determined by divid-
combined with other causes of disruption. attributed to the “ripple” effect of changes ing the actual productivity by the baseline
This relationship is more dramatic for the on all aspects of a construction project. productivity. Multivariate regression pro-
electrical and mechanical sectors than for The third hypothesis was that hidden duced the following model:
civil/architectural work due mostly to the change (an unquantified effect on the
interdependence of the trades. An inter- base contract) and project change are di- PR = 2.57 + 1.07 × changes indicator
esting finding is that losses in productivity rectly related. Statistical analysis revealed
as a result of change orders are not affect- that as the amount of project change in- where
ed by the type of construction, i.e., gener- creases, so does the amount of hidden the changes indicator is an indicator vari-
al building versus industrial projects. change, or the impact on baseline labor. able such that the presence of
A recent study conducted by the The major conclusion from the study was changes = 1 and the absence of
Construction Industry Institute (CII) at- that professionals must improve identifi- changes = 0.
tempted to quantify the effects of project cation and measurement techniques for (equation 1)
change. It focused upon three research project change to better measure the hid-
hypotheses: den change effects. With efficiency being the PR with
In a recent publication by Thomas the changes indicator present (2.57 +
• changes that occur late in a project and Napolitan [4], an attempt to quantify 1.07 = 3.64) over the PR under normal
are implemented less efficiently than the effects of construction changes on conditions (2.57), it was concluded that
those early-on; labor productivity was made. The authors the efficiency was 2.57/3.64 = 0.71, or an
• the greater the amount of change used daily productivity values from three average loss of efficiency of about 30 per-
there is on a project, the greater the projects: cent when changes are present.
negative effect on labor productivity;
and • a process plant—mechanical and
• the cumulative change effect increas- electrical trades were studied; METHODOLOGY
es proportionately to the amount of • a paper mill—electrical; and
change on a project [1]. • a refinery—mechanical, constructed A preliminary questionnaire was de-
over a period of 4 years. veloped and distributed to two electrical
The information necessary to investi- contractors to obtain data to quantify the
gate these hypotheses was difficult to ob- The project data included approximately loss of efficiency due to changes. Feed-
tain; many participants just didn’t have 522 working days, and represented a lim- back regarding the clarity of questions
the needed information. Only 19 percent ited sample (three projects). Multiple re- and the availability of the data collected
of the submitted projects included infor- gression techniques and an analysis of was obtained. The questionnaire was re-
mation needed for an analysis of hypothe- variance tests, among other methods, vised and distributed to six major electri-
sis 1; thus, a statistical evaluation of this were used to analyze data from the three cal contractors around the US. The ques-

Cost Engineering Vol. 40/No. 2 FEBRUARY 1998 37


tionnaire consisted of three sections: struction project types consisted of the fol- REGRESSION ANALYSIS
company-specific data, project-specific lowing categories:
data, and change order-specific data for Regression analysis is the statistical
the project. Upon completion of the ques- • industrial (manufacturing or process tool used to measure how well a group of
tionnaires, follow-up interviews were con- plants, paper mills, wastewater treat- independent variables “explains” a given
ducted to clarify and validate the data ment plants, etc.)—38 percent; dependent variable. The model presented
provided. Hypothesis testing and linear • commercial (banks, retail, schools, here consists of a two-term regression
regression were performed on the collect- office buildings, and others of the equation, with one of the terms being an
ed data using MINITAB, a statistical soft- type)—34 percent; indicator variable. The data were first ex-
ware package. • institutional (hospitals and correc- amined by graphical methods to show re-
tional facilities)—14 percent; and lationships between different parameters
• power plants—9 percent. that warranted further investigation and
ASSUMPTIONS possible inclusion in the model. In the
There were two responses of “other” (5 model formulation and fitting stages, the
Several assumptions had to be made percent—a theater and a museum proj- model was implemented, and fitted val-
for the calculations needed to determine ect). Industrial and commercial projects ues for the dependent variable were cal-
the effect of change orders on labor effi- constituted over 70 percent of all of the culated. These values were then plotted
ciency. The first assumption was that the project types. against their actual values in an effort to
project estimate was accurate and/or The project delivery approach was validate the model. Finally, diagnostic
valid. This assumption was made with the the next project-related variable collect- checking, or data and model criticism,
following factor in mind: the contractors ed. Three types were reported: was performed.
in this study had been successful in the The graphical distribution of estimat-
industry for a number of years, during • design-bid-build without a construc- ed labor hours was skewed to the left, giv-
which time they amassed large amounts tion manager—67 percent; ing unequal weight to projects with a
of data and knowledge used to arrive at • design-bid-build with construction lower number of labor hours. To normal-
accurate bids. Similarly, the second as- manager services—29 percent; and ize the data, they were plotted on a natur-
sumption was that the estimated labor • design-build approach with a con- al logarithmic scale. This distribution of
hours reported for change-order work struction manager—4 percent. the normalized data represents a much
were accurate. Over one-half of the proj- more symmetrical or normal observed
ects did not track the actual hours ex- distribution. This transformation is shown
pended on change orders. The reasons for Determination of Effects in figures 2 and 3.
not tracking these hours were insufficient The questionnaire asked contractors
time to collect data and difficulty in ob- to indicate if the job being considered
taining accurate numbers. were affected by change orders. This was Regression Model Selection
defined as a job experiencing more In an effort to be all-encompassing,
changes than past projects of similar size the initial regression model contained all
DATA COLLECTION and contract duration. The determina- 44 of the projects reported. The indepen-
AND CHARACTERISTICS tion of whether or not the job were affect- dent variables for the model were deter-
ed was made using the project manager’s mined to be estimated labor hours at proj-
Forty-four projects from five electri- experience. The question was subjective ect conception and if it were affected or
cal contractors were collected, spanning a for several reasons, the main one being not. The dependent variable was identi-
wide range of project types and sizes. The that in many instances, the numbers re- fied as ∆ measured as a percent of the
projects ranged from $50,000 to over ported in job cost reporting systems do total actual labor hours. The equation in-
$5,000,000 in total installed cost, and not adequately reflect the nature of the cludes the following:
covered a variety of project types, ranging project. More specifically, a job could be
from power plants to institutional and severely affected by change and still finish ∆%tot = 0.517 + .150 impact
commercial buildings. In the majority of ahead of the planned labor projections. - .0586 LnEstHR
these projects, the electrical contractor The projects were grouped into two out-
was a subcontractor. come categories: affected projects and un- where
affected projects. The projects were then ∆%tot = actual base labor hours minus es-
divided into two further categories: those timated labor hours divided by the
Company and Project Characteristics with a positive value for ∆, and those with total actual labor hours;
The average annual sales in 1995 for a negative value for ∆. Of the 21 positive impact = 1 if the project is affected and =
the participants were $62 million, with delta projects, 6 were characterized as un- 0 if unaffected;
the minimum being $18.5 million and affected; similarly, 6 projects were report- LnEstHR = original contract estimated
the maximum $123 million. The projects ed as affected and having delta values labor hours in natural logarithmic
primarily involved union labor. The con- below zero. units.
(equation 2)

38 Cost Engineering Vol. 40/No. 2 FEBRUARY 1998


tractor pursuing additional compensa-
25 tion.
The negative sign associated with the
20 natural-log of the original estimated labor
hours appears to be a cause for concern,
but can be partially explained by a num-
Frequency

15
ber of reasons. The main reason is that
small jobs, measured in direct field labor
10
hours, are susceptible to change orders of
any magnitude. This is true of jobs of any
5
size, but the issue here is that smaller
projects can have a larger delta percent-
0
age just because of their size and because
5 15 25 35 45 55 65 75 85 95 105 of how the change orders compare to the
Estimated Labor Hours (1,000 Hours) original size of the job, while larger jobs
have more hours to dilute the effect of
Figure 2—Distribution of Labor Hours these change-order hours. This issue of
scale is seen in the fact that it takes a
10 much larger delta, or difference between
the base and estimated labor hours, to af-
8 fect the delta measured as a percent of the
total labor hours on a large project versus
Frequency

6 a small one.
Another possible explanation is that
4 contractors react better or are forced to
give more attention to projects that con-
2 tain large numbers of change-order
hours. The converse of this is the possi-
0 bility that on jobs where change orders
6.5 7 7.5 8 8.5 9 9.5 10 10.5 11 11.5 account for only 5 to 15 percent of the
total labor, they become a nonissue for
Natural Log Estimated Labor Hours
managers and get lost in the everyday ac-
Figure 3—Natural Log Transformation Distribution of Labor Hours tivities of the project, and thus have more
effect on the baseline productivity of
workers.
∆ % Total Values Figure 4 illustrates the actual values
of delta as reported by contractors partici-
0.5- • pating in the study plotted against the val-
0.4- • ues of delta as calculated from equation
0.3- • • 2, or fitted values. It can be seen that the
0.2- •• • • prediction equation provides an accurate
• • • •
Actual Values

0.1- • result in the trend of the data. The t-val-


•• • • •• •
0.0- •• • • ues associated with each coefficient ex-
• •• •• • • ceed 3.5 in absolute value. The R2 value,
-0.1- • •
-0.2- • • or coefficient of determination, is 0.480,
R2 = 0.480 indicating that the model explains 48 per-
-0.3-
-0.4- cent of the variation in delta. The Ra2
value adjusted for the degrees of freedom
-0.2 -0.1 0.0 0.1 0.2 0.3 is 0.452, just slightly lower.
Fitted Values

Figure 4—Fitted Value Plot for ∆ % Total Values SAMPLE APPLICATION

The signs and magnitudes of the esti- expended. It is imperative that the owner, The delta is defined as the base con-
mated regression coefficients provide in- contractor, and specialty contractor ac- tract hours minus the original estimated
sight into the variables’ relationships. For knowledge the existence of this when hours, and given that the model returns a
a particular project, holding all other vari- changes occur. The responsibility to doc- value for delta in terms of the percentage
ables equal, an impacted project adds 15 ument the effects on labor as a result of of the total actual job labor, allows the
percent to delta in terms of the total labor change-order work belongs to the con-

Cost Engineering Vol. 40/No. 2 FEBRUARY 1998 39


computation of the affected base labor tion. The equation can be used to deter- Journal of Civil Engineering 8
hours. More specifically, mine the amount of lost efficiency on the (1991): 484-492.
base project resulting from change orders, 4. Thomas, H. Randolph, and Carmen
affected base labor hours = (∆%tot) but is just one measure and is not the end- L. Napolitan. Quantitative Effects of
× total actual labor hours all for future negotiations. It has a great Construction Changes on Labor Pro-
+ total estimated labor hours deal of merit in that it was derived from ductivity. Journal of Construction
(equation 3) actual electrical construction project Engineering and Management 121,
data, and the projects from which the no. 3 (1995): 290-296.
and, by definition data came included other factors that af- 5. Thomas, H. Randolph, and A.S.
fected productivity, such as weather con- Sakarcan. Forecasting Labor Produc-
total actual labor hoursaffected ditions, trade stacking, overtime, and tivity Using the Factor Model. Jour-
= base labor hoursaffected more. Statistically speaking, the model is nal of Construction Engineering
+ total change-order labor hours. a good predictor of delta, given the inde- and Management 120, no. 1 (Janu-
(equation 4) pendent variables in the equation. The ary 1994): 228-239.
model has a coefficient of determination,
To illustrate with a numerical exam- R2, higher than in previous studies per-
ple, data from a project submitted after formed on the topic, and is based on a siz- Dr. Awad S. Hanna,
the completion of the regression analysis able database composed of projects rang- PE, is an associate
was used in the regression equation. The ing in size from 1,700 to 147,000 total professor of construc-
estimated labor at the start of construction actual labor hours. However, the model tion at the University
was reported as 32,729 hours. The actual has limitations as well. It is an after-the- of Wisconsin-Madison,
number of change-order labor hours was fact tool, meaning that the nature of the in the department of
23,840, with an extra 6,753 hours of work variables allows the model to work only civil and environmen-
that were not accounted for; assuming when the project is complete. tal engineering. He
this project was affected, the estimated Perhaps the best application for the earned a master’s de-
impact on this project is 5.8 percent. model would be to prove or disprove an gree and a Ph.D. in civil engineering (with
owner’s belief that losses a contractor an emphasis in construction engineering
{∆%tot = 0.517 + 0.150 (1) - 0.0586 claims resulted from change orders actu- and management) from Pennsylvania
(ln 32729)}. ally resulted from mismanagement or a State University in 1986 and 1989, respec-
(equation 5) poor bid. For a contractor to get reim- tively. He worked as an assistant and then
bursed for additional labor hours per- as an associate professor at the Memorial
This is an estimate for ∆%tot. The actual formed as the result of effects to the base University of Newfoundland, Canada,
value for ∆%tot was 10.66 percent. hours, he/she has to first determine how from 1989 to 1993. He has authored over
many hours were added. By comparing 50 publications in journals and conference
the contractor’s desired or computed proceedings in the areas of construction
LIMITATIONS amount of additional labor hours to a cal- productivity, constructability, and meth-
culated value obtained with the model, ods improvement.
It should be noted that the data used the factor(s) of mismanagement and/or a
to formulate the model contained proj- bad bid drop out. This assumes that mis- Dr. Jeffrey S. Russell,
ects with a maximum value for change- management and/or poor bids were not PE, is an associate
order hours of 35 percent of the total common in the data used to formulate professor in the depart-
labor hours. This means that projects that the model—an unlikely scenario. ment of civil and envi-
include change orders to a greater degree ronmental engineering
than this cannot be used in the model. at the University of
Another limitation of the model is that REFERENCES Wisconsin-Madison,
many factors were not included during and he is the chair of
conception of the actual model. Factors 1. Ibbs, C.W., and Walter E. Allen. their construction en-
such as project type, project delivery ap- Quantitative Impacts of Project gineering and management program. Dr.
proach, schedule increase, frequency of Change. Source Document 108. Russell’s education includes a B.S. in civil
change orders, and others were not in- Austin, TX: The Construction Indus- engineering (1985) from the University of
cluded in an effort to simplify the model. try Institute, 1995. Cincinnati, as well as a master’s degree
2. Means Illustrated Construction (1986) and Ph.D. (1988) from Purdue
Dictionary. Kingston, MA: R.S. University. Dr. Russell’s research focuses

T
his article is a first step toward Means Company, Inc., 1991. on developing decision-support tools for
the quantification of the re- 3. Moselhi, O., C. Leonard, and P. the managerial aspects of construction, in-
sults of change. The model de- Fazio. Impact of Change Orders on cluding contractor prequalification, as
veloped from the regression Construction Productivity. Canadian well as human and robotic performance in
analysis should be inspected with cau- the construction industry. He is an author

40 Cost Engineering Vol. 40/No. 2 FEBRUARY 1998


and coauthor of numerous publications in these areas. He is a
member of AACE International. DON’T BE THE
David Thomack is a project engineer with ONLY ONE WHO
the M.A. Mortenson Construction Company
in Minneapolis, MN. He received his under-
graduate degree from the civil and environ-
DOESN’T
mental engineering department’s construc-
KNOW—
5
tion engineering and management program
in 1994 at the University of Wisconsin-Madi-
son. He graduated from the University of
Wisconsin with a master’s degree in civil en-
gineering in 1996. He has worked as a project estimator for
Mortenson and has experience in the mechanical construction GET YOUR COPY OF THE
industry as well.◆
BEST TECHNICAL
REFERENCE GUIDE
Available again AVAILABLE
for the first time FOR CONTRACTS &
in CLAIMS
several years . . .

Due to the demand from industry


and academia, this classic
textbook has been reprinted by
AACE International.

$35.00 for AACE International members This 2-volume reference is a collection of


papers from the AACE International
$41.50 for nonmembers Transactions, Cost Engineering, and other
plus shipping and handling sources. Edited by James G. Zack Jr., this is a must-
have for anyone who deals with contracts and
To order your copy of this softcover claims (and who doesn’t?)
book, call 800-858-COST
or 304-296-8444 $99.95 for AACE members (+S&H)
$195.95 for nonmembers (+S&H)

Cost Engineering Vol. 40/No. 2 FEBRUARY 1998 41

You might also like