You are on page 1of 6

Change Orders Impact on Labor Productivity

Osama Moselhi, F.ASCE1; Ihab Assem, M.ASCE2; and Khaled El-Rayes, M.ASCE3

Abstract: This paper describes a study conducted to investigate the impact of change orders on construction productivity and introduces
a new neural network model for quantifying this impact. The study is based on a comprehensive literature review and a field investigation
of projects constructed in Canada and the USA. The field investigation was carried out over a 6-month period and encompassed 33 actual
Downloaded from ascelibrary.org by New York University on 05/13/15. Copyright ASCE. For personal use only; all rights reserved.

cases of work packages and contracts. Factors contributing to the adverse effects of change orders on labor productivity are identified and
a model presented earlier is expanded to account primarily for the timing of change orders, among other factors. The developed model,
as well as four models developed by others, have been incorporated in a prototype software system to estimate the loss of labor
productivity due to change orders. A numerical example is presented to demonstrate the use of the developed model, and illustrate its
capabilities.
DOI: 10.1061/共ASCE兲0733-9364共2005兲131:3共354兲
CE Database subject headings: Change orders; Neural networks; Productivity; Construction industry; Claims; Dispute resolution;
Computer software; Claims.

Introduction Bruggink 1997; Coffman 1997; Hanna et al. 1999a,b; Assem


2000兲. Reported studies can be grouped into two categories: 共1兲
Change orders have long been identified to have a negative im- General, covering construction stage 共Leonard 1988; Moselhi et
pact on construction productivity, leading to a decline in labor al. 1991a; Moselhi 1998; Abdo 1999兲 or both design and con-
efficiency and, in some cases, sizeable loss of man hours 共Barrie struction stages 共Ibbs 1997兲; and 共2兲 trade specific, covering me-
and Paulson 1996; Moselhi 1998兲. Change orders continue to chanical 共Hanna et al. 1999a兲 and electrical 共Bruggink 1997;
pose serious challenge to owners and contractors alike. On one Hanna et al. 1999b兲 works. Except for the recent work of Hanna
hand, they provide an essential mechanism for: 共1兲 Satisfying et al. 共1999a兲, none considered the impact of the timing of change
owners construction needs throughout the project delivery pro- orders over the project duration.
cess; and 共2兲 responding effectively to errors and/or omissions in This paper presents a study, conducted primarily to extend the
design, construction methods, and contract documents. This is model presented earlier by Moselhi et al. 共1991a兲 to include the
particularly true in fast-track construction, where construction timing effect of change orders. The study encompasses: 共1兲 A
starts prior to design completion and the project scope of work is field investigation for data collection; and 共2兲 development of a
adjusted along the way. On the other hand, change orders fre- new method and a prototype software system for estimating the
quently pose serious problems to owners and contractors, leading impact of change orders on labor productivity. The field investi-
to cost overruns and costly disputes. This could be attributed, at gation was conducted over a 6-month period at the offices of
least in part, to inadequate understanding and lack of appreciation Revay and Associates Limited in Montreal, Canada. Thirty-three
of the impact of these changes on project performance. work packages, referred to here as cases, extracted from projects
Quantifying the impact of change orders on labor productivity constructed in Canada and the USA between 1990 and 1998, were
remains to be a challenging task, despite the reported findings of analyzed and used in the developments made in this study. These
many studies and documented cases 共Revay 1985; Brunies 1988; cases were used to develop a neural network model to estimate
Leonard 1988; Moselhi et al. 1991a; Ibbs and Allen 1995; Tho- the impact of change orders on labor productivity. The developed
mas and Napolitan 1995; Thomack 1996; Vanderberg 1996; neural network model was incorporated in a software system
called ChangeOrders.E. The system also incorporates models de-
1
veloped by others, and provides an automated user-friendly envi-
Professor and Chair, Dept. of Building, Civil, and Environmental ronment that permits objective and timely evaluation of the ad-
Engineering, Concordia Univ., Montreal PQ, Canada H3G 1M8.
2 verse impact of change orders on labor productivity. A numerical
Research Assistant, Dept. of Building, Civil, and Environmental En-
gineering, Concordia Univ., Montreal PQ, Canada H3G 1M8. example is worked out to illustrate the salient features of the
3
Assistant Professor, Dept. of Civil and Environmental Engineering, developed method and its capabilities.
Univ. of Illinois at Urbana—Champaign, Urbana, Ill. 61801.
Note. Discussion open until August 1, 2005. Separate discussions
must be submitted for individual papers. To extend the closing date by Data Collection
one month, a written request must be filed with the ASCE Managing
Editor. The manuscript for this paper was submitted for review and pos-
sible publication on December 19, 2002; approved on February 13, 2004. The data used in this research were collected through a field in-
This paper is part of the Journal of Construction Engineering and Man- vestigation conducted over a period of 6 months at the offices of
agement, Vol. 131, No. 3, March 1, 2005. ©ASCE, ISSN 0733-9364/ Revay and Associates Limited in Montreal, Canada 共Assem
2005/3-354–359/$25.00. 2000兲. One hundred and seventeen 共117兲 projects, constructed in

354 / JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT © ASCE / MARCH 2005

J. Constr. Eng. Manage. 2005.131:354-359.


Canada and the USA between 1990 and 1998, were initially ex-
amined for possible use in this study. Only 33 work packages,
extracted from the 117 projects, were found to have sufficient
data that can be used for the intended developments. Of these
work packages, 3 were architectural, 9 electrical, and 21 mechani-
cal, respectively. These work packages, referred to here as cases,
have an original total value of more than $110 million, over a
million 共1,023,583兲 planned direct hours for the original scope of
work and 166,002 actual direct hours assigned to change orders.

Model Development
Downloaded from ascelibrary.org by New York University on 05/13/15. Copyright ASCE. For personal use only; all rights reserved.

The collected data were used to develop a neural network model


to estimate the impact of change orders on labor productivity. The
model was developed in three main stages: 共1兲 Identifying change Fig. 1. Direct manpower loading for analyzed cases
orders factors that affect labor productivity; 共2兲 modeling the tim-
ing impact; and 共3兲 developing a neural network model.
Modeling the Timing Impact
To model the timing impact of change orders, an effort was made
Identifying Change Orders Factors that Affect Labor
to account for the buildup and rundown of labor hours spent to
Productivity
perform the work. This was deemed essential in modeling the
Based on the field investigation conducted and a comprehensive ripple effect of the introduced changes. It should be noted here
literature review, the following factors were found to influence that the hypothesis introduced by Hanna et al. 共1999a兲 for mod-
the impact of change orders on labor productivity: eling the timing impact assumes that the timing impact increases
1. Intensity: This factor can be expressed in terms of: 共1兲 Num- from project inception to completion in a linear manner. This
ber of change orders; 共2兲 their frequency; and/or 共3兲 ratio of hypothesis was not supported by the findings of several recent
change orders hours to contract hours. The ratio of change studies. For example, Bruggink 共1997兲 and Coffman 共1997兲 re-
orders hours to planned or actual contract hours is commonly ported that the highest impact of change orders occur in the third
used to represent this factor 共Leonard 1988; Moselhi et al. quarter of the project duration. Furthermore, Ibbs and Allen
1991a; Ibbs 1997; Vandenberg 1997; Hanna et al. 1999a,b兲. 共1995兲 did not find that changes which occur late in a project are
2. Timing in relation to project duration: Coffman 共1997兲 con- implemented less efficiently than changes that occur early.
firmed the significance of the timing factor stating: “When In order to improve the modeling of the timing impact of
evaluating change orders, regardless of their cause, the most change orders, this study performed an in-depth analysis of the
significant factor is when the change occurs.” The timing buildup and rundown of man hours for each of the 33 considered
factor has recently been introduced by Hanna et al. 共1999b兲, cases. This buildup and rundown for the 33 cases can be analyzed
assuming that the timing impact increases from project in- by plotting the direct manpower loading curves that depict the
ception to completion in a linear manner. This assumes utilization of man hours over the construction duration, as shown
higher labor productivity losses to occur toward the end of in Fig. 1. To allow for a comparison with the method proposed by
project duration, and therefore it does not consider the ripple Hanna et al. 共1999a兲, the construction duration, in all cases con-
effect of change orders on the remaining unchanged work. sidered, was divided into 5 equal periods as shown in Fig. 1 and
3. Work type: The type of work 共i.e., civil, architectural, elec- Table 1. For each of these periods 共i = 1 to 5兲, a direct manpower
trical, or mechanical兲 affects the degree of impact of change loading ratio 共ai / A兲 was calculated to represent the scope of work
orders on labor productivity 共Leonard 1988兲. This is mainly expressed in man hours in period i 共ai兲 in relation to those of the
due to: 共1兲 The differences in the level of skill required to entire project 共A兲. These direct manpower loading ratios 共ai / A兲
perform the work and its level of complexity; and 共2兲 the were averaged for the 33 considered cases as shown in Table 1
interdependency which varies from one type of work to an- and Fig. 1, providing a useful representation of the buildup and
other and among work types 共Coffman 1997; Leonard 1988兲. rundown of man hours in these cases.
4. Type of impact: This variable illustrates whether the type of
impact on construction productivity is limited to change or-
Table 1. Direct Manpower Loading Ratios 共ai / A兲
ders only or is a combined impact of change orders and
additional productivity-related factors, such as overtime, Project National Electrical
congestion, and weather. The number of additional factors Project duration Weighted Contractors
that may affect construction productivity varies from one period covered average of Industry Association
project to another, and can be represented by 0, 1, or 2 ad- 共i兲 共%兲 considered cases average average
ditional factors, as explained in the model of Moselhi et al. 1 0–20 0.09 0.06 0.06
共1991a兲. 2 20–40 0.21 0.19 0.16
5. Project phase: This factor differentiates between changes in- 3 40–60 0.29 0.30 0.27
troduced during the design phase and the construction phase 4 60–80 0.26 0.31 0.32
as presented in the model of Ibbs 共1997兲. 5 80–100 0.15 0.13 0.19
6. On-site management: This factor accounts for the Project
Total 100 1.00 1.00
Manager’s years of experience 共Hanna et al. 1999b兲.

JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT © ASCE / MARCH 2005 / 355

J. Constr. Eng. Manage. 2005.131:354-359.


共PHi兲 in each of the five construction periods as shown in Eq. 共1兲.
As such, TPi is capable of considering the combined impact of:
共1兲 The timing of the change order; and 共2兲 the intensity of change
orders in each period

HCOi
TPi = 共1兲
PHi

where TPi = timing impact of change order in period i; HCOi


= actual change order hours during period i; PHi = planned hours
during period i; and i = period when change order occurs, where
i = 1 to 5.
The newly introduced parameter of TPi is used to account for
Downloaded from ascelibrary.org by New York University on 05/13/15. Copyright ASCE. For personal use only; all rights reserved.

the timing impact of change orders and their intensity, among


Fig. 2. Comparison of direct manpower loading curves other relevant factors on productivity loss using a back propaga-
tion neural network model of the type shown in Fig. 3. Neural
networks are utilized in view of their powerful capabilities in
These direct manpower loading ratios were compared to in- modeling pattern recognition problems 共Moselhi et al. 1991b;
dustry trends reported by others 共NECA 1983; Bent and Thuman NeuroShell 1996; Moselhi 1998; Abdo 1999兲. It should be noted
1988兲, as shown in Fig. 2 in order to examine the feasibility of that neural networks of the type used in this study are well suited
utilizing them in modeling the timing impact of change orders. for the development of parametric cost estimating models, similar
Bent and Thuman 共1988兲 and the AACE Education Board 共1989兲 to the model developed in this study, and are generally recognized
suggested a typical trapezoidal shape to model the direct man- to outperform regression models.
power loading in construction. The National Electrical Contrac-
tors Organization 共NECA 1983兲 developed an industry-average
Developing Neural Network Model
type curve, representing the buildup and rundown of manpower
consumption for electrical construction. In view of the good A neural network model was developed to estimate the impact of
agreement between the results obtained in this study and those change orders on labor productivity. In order to identify and in-
representing industry-average for general construction and for corporate the most relevant factors in the model, the impact of the
electrical work 共see Fig. 2 and Table 1兲, it was decided to use aforementioned six factors on productivity loss 共PL兲 was statisti-
NECA distribution for electrical work and the trapezoidal distri- cally examined 共Assem 2000兲. Productivity loss in this model is
bution of Bent and Thuman 共1988兲 for other types of work. defined as the unproductive labor hours attributable to the causes
In this model, the aforementioned NECA and trapezoidal dis- of impact under examination, expressed as percentage of the labor
tributions are used to subdivide the total planned man hours at the hours spent on original contract work. The results obtained from
project level into its subcomponents of planned man hours 共PHi兲 scatter plots, correlation, and regression analyses indicated a
in each of the five construction periods 共i = 1 to 5兲 in order to strong correlation between the percentage of PL and each of the
enable the consideration of the timing impact of change orders. following four factors: 共1兲 Timing impact of change orders as
PHi is then used to develop a new parameter 共TPi兲 that considers represented by the newly developed parameter 共TPi兲; 共2兲 change
the buildup and rundown of man hours shown in Figs. 1 and 2. orders intensity which is also represented by TPi that provides a
TPi is introduced in this model to consider the direct impact of a ratio of actual change order hours to planned hours in each of the
change order on period i when it occurs, and its ripple effect on five construction periods 共i = 1 to 5兲; 共3兲 work type; and 共4兲 type
the remaining periods. TPi is designed to track the ratio of actual of impact 共TI兲 which can be change orders only or change orders
change order hours 共HCOi兲 to the original planned hours of work plus 1 or 2 additional causes of productivity related impact.

Fig. 3. Developed neural network model

356 / JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT © ASCE / MARCH 2005

J. Constr. Eng. Manage. 2005.131:354-359.


Downloaded from ascelibrary.org by New York University on 05/13/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 4. System graphical user interface

The remaining two factors described earlier 共i.e., project phase the developed neural network based on a definition file. The sys-
and on-site management兲 were not statistically examined in this tem was implemented using Microsoft Visual Basic as a stand-
study due to the lack of sufficient data in the 33 cases utilized in alone application that runs on Microsoft Windows 1998, 2000,
the present development. All change orders in the 33 cases were and NT. To facilitate data entry and reporting, a set of user-
issued during the construction phase and they did not include friendly interactive screens were developed using the Multiple
explicit data on project management experience. Consequently, Document Interface 共MDI兲 tool available in Microsoft Visual
the above four factors were used in the development of the pro- Basic. Fig. 4 depicts the developed MDI and a set of interactive
posed neural network model, as shown in Figs. 3 and 4. The screens for the developed system.
detailed procedure and algorithm used in the development of the
neural network model can be found in Assem 共2000兲. The model
was developed using NeuroShell2 共1996兲, which operates in the Numerical Example
MS Windows environment and offers a number of ready-to-use
neural network paradigms. To demonstrate the use of ChangeOrders.E and illustrate its main
features, a numerical example is presented using a set of eight

Prototype Software System Table 2. Parameters of the Cases Used 共data from Bruggink 1997兲
Case Work hours
A prototype software system was developed to provide a tool for No. Productivity loss
quantifying the adverse impact of change orders on labor produc- 共j兲 Planned Actual Total 共h兲
tivity. The software was named ChangeOrders.E 共Change Orders 1 2,461 3,461 4,262 801
Estimator兲. It provides a user-friendly interface and incorporates: 2 1,769 2,119 3,192 1,073
共1兲 The newly developed neural network model; and 共2兲 previ- 3 4,419 5,419 7,012 1,593
ously developed models for general construction 共Moselhi et al.
4 5,332 8,532 10,273 1,741
1991a; Ibbs 1997兲, mechanical construction 共Hanna et al. 1999a兲,
5 10,213 11,733 14,393 2,660
and electrical construction 共Hanna et al. 1999b兲.
6 8,394 9,404 14,543 5,139
The neural network model was incorporated in
ChangeOrders.E using the “Runtime Facilities” feature of 7 46,269 60,460 74,173 13,713
NeuroShell2 共1996兲. This feature provides a direct interface with 8 18,170 25,595 36,170 10,575

JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT © ASCE / MARCH 2005 / 357

J. Constr. Eng. Manage. 2005.131:354-359.


Table 3. Change Orders Data 共data from Bruggink 1997兲 Table 4. Planned Hours 共Based on Distribution of National Electrical
Contractors Association 1983兲
Case Change Change orders hours for period
No. orders Case Planned hours for period
共j兲 hours i=1 i=2 i=3 i=4 i=5 No.
共j兲 Total i=1 i=2 i=3 i=4 i=5
1 1,000 200 200 200 200 200
2 350 0 105 140 105 0 1 2,461 148 394 664 788 468
3 1,000 240 210 200 190 160 2 1,769 106 283 478 566 336
4 3,200 256 544 960 1,056 384 3 4,419 265 707 1,193 1,414 840
5 1,520 85 67 49 286 1,034 4 5,332 320 853 1,440 1,706 1,013
6 1,010 0 303 303 202 202 5 10,213 613 1,634 2,758 3,268 1,940
7 14,191 2,271 2,271 2,271 2,838 4,541 6 8,394 504 1,343 2,266 2,686 1,595
8 7,425 297 965 2,079 2,599 1,485 7 46,269 2,776 7,403 12,493 14,806 8,791
8 18,170 1,090 2,907 4,906 5,814 3,452
Downloaded from ascelibrary.org by New York University on 05/13/15. Copyright ASCE. For personal use only; all rights reserved.

actual electrical cases. These cases were selected from literature


共Bruggink 1997兲 as they have adequate data which are essential respectively, as shown in Table 5. This demonstrates that the pro-
for the application of the three considered models in this analysis. posed consideration of the timing impact in the developed neural
The data pertaining to those eight cases are summarized in Tables network model was useful in improving the accuracy and reliabil-
2 and 3. It should be noted that the distribution of the planned ity of estimating the impact of change orders on productivity
hours for the cases considered 共see Table 4兲 was generated to losses.
follow the previously mentioned distribution of NECA 共1983兲.
The cases were analyzed using the models included in ChangeOr-
ders.E, i.e., the general regression model of Moselhi et al. Summary and Concluding Remarks
共1991a兲, the electrical regression model of Hanna et al. 共1999b兲,
and the developed neural network model for electrical work. Es- The findings of a study conducted to examine the impact of
timated PLs due to change orders were obtained using the three change orders on construction productivity have been presented.
models, and were compared to the actual PLs reported by Brug- The study is based on a comprehensive literature review and a
gink 共1997兲, as shown in Table 5. A comparison of the results field investigation of projects constructed in Canada and the USA.
shows that the proposed model outperforms available models in The field investigation was carried out over a 6-month period to
estimating the impact of change orders on productivity losses. collect data needed for the study. Factors contributing to the ad-
The average estimating error of the proposed model for the ana- verse effects of change orders on labor productivity are identified,
lyzed eight cases was 17.8%, which is significantly lower than and the model presented earlier by Leonard 共1988兲 and Moselhi et
those 共30.5% and 40.5%兲 associated with the general regression al. 共1991a兲 is extended to account for the timing effect. A proto-
model 共Moselhi 1991a兲 and the electrical regression model type software system is developed to estimate the percentage loss
共Hanna 1999b兲. The average absolute estimating error of the pro- of labor productivity due to change orders. The developed soft-
posed model, the general regression model, and the electrical re- ware provides a user-friendly interface to facilitate data entry and
gression model was also calculated to be 19.4%, 25.3%, and 30%, to assist users in generating a response to a number of what-if

Table 5. Actual versus Estimated Productivity Loss


Moselhi et al. 共1991a兲 Hanna et al. 共1999b兲 Proposed model

Actual Estimated Estimated Estimated


Case productivity productivity Estimating productivity Estimating productivity Estimating
No. loss loss errorb loss errorb loss errorb
共j兲 共h兲 共h兲 共%兲 共h兲 共%兲 共h兲 共%兲
1 801 731 8.74 650 18.85 940 −17.35
2 1,073 1,050a 2.14 713 33.55 1,245a −16.03
3 1,593 884 44.51 1,467 7.91 1,240 22.16
4 1,741 2,070 −18.90 2,041 −17.23 2,674 −53.59
5 2,660 1,744 34.44 2,966 −11.50 2,699 −1.47
6 5,139 N/A N/A 2,485 51.64 5,078a 1.19
7 13,713 11,469 16.36 10,644 22.38 15,169 −10.62
8 10,575 5,073 52.03 2,438 76.95 7,077 33.08
Average errorc 共%兲 30.45 40.50 17.83
d
Average absolute error 共%兲 25.30 30.00 19.44
Note: N / A = not applicable 共i.e., off model range兲.
a
Two additional types of impact are considered in the estimate 共i.e., TI= 3兲.
b
Estimating error j = Actual loss j − Estimated loss j / Actual loss j, where j represents case 1 to 8.
c
Average error= 兺8j=1兩Actual loss j − Estimated loss j兩 / 兺8j=1Actual loss j.
d
Average absolute error= 兺8j=1兩Estimating error j兩 / 8.

358 / JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT © ASCE / MARCH 2005

J. Constr. Eng. Manage. 2005.131:354-359.


scenarios. In addition to the developed neural network model, the Barrie, D., and Paulson, B. 共1996兲. Professional construction manage-
software incorporates four models developed by others. A numeri- ment, 3rd Ed., McGraw–Hill, New York.
cal example is analyzed to demonstrate the use of the developed Bent, J., and Thuman, A. 共1988兲. “Project management for engineering
model, and illustrate its capabilities. The results of the analyses and construction.” Prentice–Hall, Englewood Cliffs, N.J.
indicated that the present neural network model provides, in com- Bruggink, M. 共1997兲. “An investigation into the impacts of change orders
parison to the other models, more accurate estimates of the impact on labor efficiency in the electrical construction industry.” MSc the-
sis, Univ. of Wisconsin-Madison, Wisc.
of change orders on productivity.
Brunies, R. 共1988兲. “Impact costs: What are they and how to quantify
them.” Proc., PMI Seminar/Symposium, PMI, 386–393.
Coffman, G. 共1997兲. “Effect of change orders on labor productivity.”
Acknowledgments
Proc., 5th Construction Congress V, ASCE, Reston, Va., 141–148.
Hanna, A., Russel, J., Gotzion, T., and Nordheim, E. 共1999a兲. “Impact of
The writers would like to acknowledge Revay and Associates change orders on labor efficiency for mechanical construction.” J.
Limited for providing access to the data used in this study and to Constr. Eng. Manage., 125共3兲, 176–184.
thank Ms. Regula Brunies and Mr. Stephen Revay for their coop- Hanna, A., Russel, J., Nordheim, E., and Bruggink, M. 共1999b兲. “Impact
Downloaded from ascelibrary.org by New York University on 05/13/15. Copyright ASCE. For personal use only; all rights reserved.

eration. The financial support provided by the Construction Indus- of change orders on labor efficiency for electrical construction.” J.
try Research Institute of Canada 共CRIC兲 is highly appreciated. Constr. Eng. Manage., 125共4兲, 224–232.
Ibbs, C. 共1997兲. “Quantitative impacts of project change: Size issues.” J.
Constr. Eng. Manage., 123共3兲, 308–311.
Notation Ibbs, C., and Allen, W. 共1995兲. “Quantitative effects of project change.”
Source Document 108, Construction Industry Institute, Austin, Tex.
The following symbols are used in this paper: Leonard, C. 共1988兲. “The effect of change orders on productivity.” MS
thesis, Concordia Univ., Montreal, Quebec.
A ⫽ total area under the normalized direct manpower Moselhi, O. 共1998兲. “Estimating the cost of change orders.” Trans. Am.
curve; Assn. Cost. Eng., EST.06.1–EST.06.5.
ai ⫽ area under the normalized direct manpower curve for Moselhi, O., Charles, L., and Fazio, P. 共1991a兲. “Impact of change orders
project period i; on construction productivity.” Can. J. Civ. Eng., 18共4兲, 484–492.
HCOi⫽ hours of change order during period i; Moselhi, O., Hegazy, T., and Fazio, P. 共1991b兲. “Neural networks as tools
i ⫽ project period, where i = 1 to 5; in construction.” J. Constr. Eng. Manage., 117共4兲, 606–625.
PHi ⫽ planned hours during period i; and National Electrical Contractors Association 共NECA兲. 共1983兲. “Rate of
TPi ⫽ timing impact of change order in period i. manpower consumption in electrical construction.” Rep., NECA,
Washington, D.C.
NeuroShell2. 共1996兲. “NeuroShell2 user’s manual.” Ward Systems Group,
Md.
References Revay, S. 共1985兲. “Impact costs.” Construction Law Reports, Carswell
Co. Ltd., 11, pp. 15–26.
Abdo, D. 共1999兲. “A neural network model for quantifying the impact of Thomack, D. 共1996兲. “The impact of change orders on electrical con-
change orders on construction productivity.” M. Eng. Major Technical struction labor efficiency.” MSc thesis, Univ. of Wisconsin—
Rep., Concordia Univ., Montreal. Madison, Wisc.
American Association of Cost Engineers 共AACE兲 Education Board. Thomas, R., and Napolitan, C. 共1995兲. “Quantitative effects of construc-
共1989兲. “Practical project control.” Proc., 33rd Annual Meeting, tion changes on labor productivity.” J. Constr. Eng. Manage., 121共3兲,
AACE, Skills and Knowledge of Cost Engineers, Section 4.6, San 290–296.
Diego. Vandenberg, P. 共1996兲. “The impact of change orders on mechanical con-
Assem, I. 共2000兲. “Estimating productivity losses due to change orders.” struction labor efficiency.” MSc thesis, Univ. of Wisconsin—
MSc thesis, Concordia Univ., Montreal. Madison, Wisc.

JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT © ASCE / MARCH 2005 / 359

J. Constr. Eng. Manage. 2005.131:354-359.

You might also like