Impact of Skilled Labor Availability on

Construction Project Cost Performance

Hossein Karimi, Ph.D., A.M.ASCE 1; Timothy R. B. Taylor, Ph.D., P.E., M.ASCE 2;
Gabriel B. Dadi, Ph.D., P.E., M.ASCE 3; Paul M. Goodrum, Ph.D., P.E., M.ASCE 4;
and Cidambi Srinivasan, Ph.D. 5
Downloaded from ascelibrary.org by La Trobe University on 05/03/18. Copyright ASCE. For personal use only; all rights reserved.

Abstract: The North American construction industry began to experience a shortage of skilled labor in the 1980s, which has continued as a
repetitive cyclic trend over the last three decades. Although this issue has received significant attention from researchers, less attention has
been given to quantifying the impact of skilled labor availability on project performance. This study contributes to the body of knowledge by
filling a gap in the existing literature by quantitatively modeling and elucidating the influence of skilled labor availability on construction
project cost performance. Data from 97 construction projects completed in the United States and Canada were collected from two industry
databases. The primary analysis shows that projects that experienced craft shortages underwent significantly higher growth in cost overruns
compared with projects that did not. Further analysis on both databases returned two robust multiple regression models that demonstrate
similar patterns of the risk that a skilled labor shortage poses to cost performance. By using data-combining techniques, two data sets were
combined to obtain more reliable informative models. The primary benefit of this risk estimation is to enable project management teams to
determine whether mitigation strategies are warranted to prevent potential shortfalls in project cost performance. Understanding the level
of impact that craft shortages have on project performance can serve as a primary step in developing motivation for industry leaders,
communities, and construction stakeholders to address the challenge of skilled labor shortage in construction industry. DOI: 10.1061/
(ASCE)CO.1943-7862.0001512. © 2018 American Society of Civil Engineers.
Author keywords: Craft labor; Cost; Skills shortage; Risk assessment; Data combining.

Introduction and Background
Construction is a labor-intensive industry. Labor costs account for
30–50% of the total cost of a project (McTague and Jergeas 2002;
Hanna 2001). It is considered an element that contains the most
risks in a project cost. The other major elements of a project costs
(materials and equipment) are significantly influenced by market
price and often beyond the control of a project manager. Consid-
ering the 2–3% profit margin of many construction projects, a con-
tractor’s management of labor cost and productivity is critical to
the financial success of a construction project (Hanna et al. 2005;
Ernzen and Schexnayder 2000).
The North American construction industry began to experi-
ence a shortage of skilled labor in the 1980s, which has continued
1
Postdoctoral Associate, School of Civil Engineering, College of
Engineering, Univ. of Tehran, 16 Azar Ave., P.O. BOX 14395-477, Tehran,
Iran (corresponding author). ORCID: https://orcid.org/0000-0001-7290-
7822. Email: karimi.hossein@ut.ac.ir; hossein.karimi@uky.edu
2 labor shortage.
Associate Professor and Terrell-McDowell Chair of Construction
Engineering and Project Management, Dept. of Civil Engineering, Univ.
of Kentucky, Lexington, KY 40506-0281.
3
Assistant Professor, Dept. of Civil Engineering, Univ. of Kentucky,
Lexington, KY 40506-0281.
4
Nicholas R. Petry Professor in Construction Engineering and Manage-
ment, Dept. of Civil, Environmental, and Architectural Engineering, Univ.
of Colorado, Boulder, CO 80309.
5
Professor, Dept. of Statistics, Univ. of Kentucky, Lexington, KY
40506-0281.
Note. This manuscript was submitted on July 24, 2017; approved on
January 15, 2018; published online on May 3, 2018. Discussion period
open until October 3, 2018; separate discussions must be submitted for
individual papers. This paper is part of the Journal of Construction En-
gineering and Management, © ASCE, ISSN 0733-9364.
with a repetitive cyclic trend over the last three decades. The inevi-
table consequences of this shortage are difficulties in recruiting
and retaining skilled labor for projects, which puts the financial
success of projects in a precarious situation. The evidence of this
shortage between the 1980s and 2008 Great Recession has been
discussed frequently in previous literature (Business Roundtable
1983, 1997; Construction Industry Institute 1990; Abdol et al.
1999; Rosenbaum 2001; Goodrum 2004; Sawyer and Rubin 2007).
The slowdown in construction work during the 2008 Great Reces-
sion in the United States began to temporarily ameliorate the craft
shortage for a period of time. However, the postrecession rapid eco-
nomic recovery again initiated a craft shortage in some segments
of the industry. The Construction Industry Institute (CII) (2015)
revealed that the severe shortages among key craft trades, including
welders, pipefitters, and electricians, appeared, particularly in the
US southeast and southwest regions. Taylor et al. (2016) conducted
a survey among the North American construction companies and
revealed that 52% of surveyed projects were impacted by skilled
Projects encountering craft labor shortages endure difficulties
in recruiting and retaining a required level of skills and/or quantity
of craft labors. When the required quality level of skills cannot
be met, a project will be executed with less-skilled workers. When
a craft labor quantity issue arises, a project cannot meet its basic
labor demand. A project executed under either of these two con-
ditions is highly likely to experience cost escalation. Previous
works provide a wealth of information on how a shortage of skills
can affect project cost performance. To propose testable hypoth-
eses and models built on previous research findings, the authors
reviewed and examined the body of literature that discusses the
impact of skilled labor availability on construction projects’ cost
performance.

© ASCE 04018057-1 J. Constr. Eng. Manage.

J. Constr. Eng. Manage., 2018, 144(7): 04018057

 Impact of Skilled Labor Availability on Cost
Performance: Evidence from Previous Researches
Cost performance of a construction project is a complex function of
many factors. Over the last two decades, there have been numerous
studies about predicting the accuracy of construction project actual
costs. Doloi (2013) summarized the highlighted key factors in the
past research into seven groups as follows:
• project-related factors (such as scope, size, and type of project);
• contract-related factors (such as contract management and form
of procurement);
• project management factors (capability of construction team);
• quality-related factors (such as inspection and testing of work,
method, and techniques);
Downloaded from ascelibrary.org by La Trobe University on 05/03/18. Copyright ASCE. For personal use only; all rights reserved.
• planning-related factors (such as effective monitoring process
and construction control);
• market-related factors (such as availability of labors and short-
age of material); and
• contractor-related factors (such as experience and communica-
tion between client and contractor).
Doloi (2013) also conducted a survey on 160 Australian con-
struction clients, consultants, and contractors. The contractors and
clients ranked “availability and supplies of labor and material” as
the 21st and 15th important factors (total of 48 factors) contributing
to the cost performance, respectively. Akintoye (2000) conducted a
comparative study on 84 British contractors ranging from very
small to large firms about factors influencing contractor cost esti-
mating practices. Overall, “availability and supply of labor and
material” was ranked 10th among 24 identified factors. In addition,
by performing factor analysis, this variable was also recognized as
one of the significant contributors to project cost performance.
Gharaibeh (2014) conducted a Delphi study on project manage-
ment teams of two megaprojects in Canada. “Lack of contingency
and escalation for material and craft labor costs in the initial esti-
mate” was identified as one of the top 10 most important problems
in managing project cost. In addition, “applying quantity-tracking
concept to monitor changes of material and craft labor quantities”
was mentioned as one of the top 10 most important lessons learned
in managing project cost performance.
In a quantitative analysis and to establish a model for early cost
estimation, Trost and Oberlender (2003) performed a factor analy-
sis and multivariate regression on 45 identified cost drivers on 67
completed construction projects across the world. The Bidding and
Labor Climate factor, recognized as one of the five variables that
significantly impacts cost estimate accuracy, accounted for 14.5%
of the prediction in the model. This factor includes the effects of
bidding climate, labor productivity, contract type, project schedule,
and logistics for engineering and construction. RSMeans (2016)
suggested a cost allowance of 10% of the total construction cost
for building projects encountering a shortage of skilled labor. This
allowance rises to 11% for heavy industrial projects.
These previous works emphasize the significance of the direct
impact of skilled labor availability on cost performance. Further
exploration of previous literature indicates that mechanism of the
impact of skilled labor availability on project cost is complex and
can be through various processes. For instance, cost escalation can
result from increasing total craftwork hours due to the reduction of
labor productivity in a project that has experienced a skills short-
age. Another example is the increase in safety incidents due to the
lack of experienced/skilled craft labor (Karimi et al. 2016), which
results in imposing additional direct costs (i.e., clean up and repair
or equipment damage) and indirect costs (i.e., disrupted schedules)
on a project (Improving Construction Safety Performance 1982).
In summary, loss of productivity, increases in extended overtime,
© ASCE 04018057-2
J. Constr. Eng. Manage., 2018, 144(7): 04018057
increases in hourly wage, escalation in safety incidents, increases in
amount of rework, and escalation of schedule overruns are major
processes that ultimately result in further cost overruns in a project
encountering a skilled labor shortage.
Impact through Loss of Productivity
Project productivity can be significantly affected by a shortage
of skilled labor. Dai et al. (2009) identified 10 groups of factors
that represent the underlying structure of productivity. Among
these groups, four factors were related to labor issues, namely train-
ing, craft worker qualification, superintendent competency, and
foreperson competency. They also found that the qualification of
workers is an area that has the greatest possibility for project pro-
ductivity improvement. Wang et al. (2008) showed that the effec-
tiveness of craft training on project productivity is significant. They
also showed that higher percent of certified workers in a project
was statistically associated with higher project productivity. Heravi
and Eslamdoost (2015) developed a predictive model for estimating
construction labor productivity and found that by improving “labor
competence,” labor productivity could be increased by 13–18.7%,
which makes it one of the top five influential factors that can highly
improve labor productivity. Karimi et al. (2017) analyzed 97 con-
struction projects completed in the United States and Canada and
found a significant decline in productivity of projects that experi-
enced a craft shortage compared with those that did not. By con-
ducting a regression analysis, they also demonstrated that there is a
significant relationship between increased skilled labor recruiting
difficulty and the decline in project productivity.
Impact through Use of Overtime and Growth in
Hourly Wage
Projects experiencing a shortage of skilled craft workers usually
have a tight schedule to meet a project deadline. Hanna et al. (2005)
argued that overtime scheduling has become the prevalent option in
this situation because it accelerates a project schedule. They ex-
plained that an associated premium pay with overtime can attract
the required workforce as well. Karimi et al. (2017) demonstrated
that there is a significant relationship between an increased craft
recruiting difficulty and increased use of overtime hours in con-
struction projects. The extended overtime can significantly dimin-
ish a project’s productivity (Li et al. 2017; Lyneis and Ford 2007;
Hanna et al. 2005), which ultimately results in growth of total craft
workhours and final cost of a project.
This strategy is usually associated with premium pay, which,
along with other approaches such as bonuses, loyalty rewards, and
promotions, is a common short-term solution that project managers
use for attracting and retaining a workforce (Abdol et al. 1999;
Hanna et al. 2005). Therefore, the shortage of skilled labor often
encompasses increases in craft labor’s hourly wage. CII (2015)
reported that trades with the highest level of shortage in the
United States also had the highest wage growth in recently com-
pleted projects. The top three trades with the highest level of shortage
were pipe welders, pipefitters, and structural welders, with average
actual wage growth of 6.0, 5.4, and 3.1% in a project, respectively.
Impact through More Frequent Safety Incidents
Karimi et al. (2016) analyzed the influence of a skilled labor short-
age on project safety performance and demonstrated that there is
a strong relationship between increased skilled labor staffing diffi-
culty and increased total recordable incident rate (TRIR). The
impact of construction injuries on project cost performance with
regard to the indirect cost of injuries is substantial. Hinze and
J. Constr. Eng. Manage.
 Appelgate (1991) demonstrated that indirect costs, which are sub-
stantially greater than direct costs, make the cost of even a minor
injury considerable. Indirect costs are but not limited to the loss of
productivity, disrupted schedules, administrative time for investiga-
tions and reports, training of replacement personnel, wages paid to
the injured workers and other workers for time not worked, clean
up and repair, adverse publicity, third-party liability claims against
the owner, and equipment damage (Improving Construction Safety
Performance 1982). Everett and Frank (1996) found that injuries
account for 7.9–15% of the cost of nonresidential new construction
projects. That indicates the significant contribution of safety acci-
dents to cost overruns, particularly when projects experience a craft
labor shortage.
Downloaded from ascelibrary.org by La Trobe University on 05/03/18. Copyright ASCE. For personal use only; all rights reserved.
Impact through Increasing Schedule Overrun
Karimi et al. (2017) presented a series of quantitative analyses on a
total of 97 construction project completed in the United States and
Canada and demonstrated that there was a significant difference be-
tween the average time overrun of projects that experienced a craft
shortage and ones that did not. The study also showed that there is a
statistically significant relationship between craft recruiting difficul-
ties and time overruns in construction projects. Delays in a project
cause direct and indirect costs. Direct costs are costs related to
additional labor, equipment, and material to complete the job, and
indirect costs are often, but not limited to, site and home office over-
heads (Chester and Hendrickson 2005). The evidence of the impact
of project duration variation on cost performance has been demon-
strated in previous studies (Akintoye 2000; Flyvbjerg et al. 2004).
Impact through Decline in Quality Performance
Shortages of skills on a project causes construction field rework.
Fayek et al. (2003) quantified the cost of each cause of rework
in a mining expansion venture project. The root-cause analysis
on 125 field rework incidents (accounting for 0.87% of the total
cost of construction) showed that inadequate technical knowledge
of supervisors, lack of training and experience of supervisors, inad-
equate supervisor/foreperson/tradesperson ratios, lack of adherence
to procedure, and shortage of skilled labor and supervision ac-
counted for a total of 13.12% of total construction field rework
costs. With much research effort focusing on the impact of rework
on the cost of projects, it is now well-known that rework is one of
the key contributors to a project’s cost overrun. Researchers found
rework could account for 2.4–12.4% of the total cost of a project
(Hwang et al. 2009; Love and Li 2000).
Departure Point
The body of literature provides strong quantitative and qualitative
support for various processes of the impacts of skilled labor short-
age on cost performance. In addition, it provides strong qualitative
support for the influence of skilled labor availability on project cost
performance. However, no studies have yet examined an empirical
overall impact of craft labor shortage on project cost performance.
This study contributes to the body of knowledge by collecting and
analyzing empirical data of projects recently completed in the
United States and Canada to quantitatively model and elucidate the
influence of skilled labor availability on construction project cost
performance.
Besides regarding empirical evidence provided by the previous
studies for various influential processes of the impacts of skilled
labor shortage on cost performance, remaining processes that have
not been proved quantitatively will be examined. These processes
include the impact of craft shortages on cost performance through
© ASCE 04018057-3
J. Constr. Eng. Manage., 2018, 144(7): 04018057
the impact on hourly wages and quality performance. This study
also examines the influence of time overruns on cost overruns
as a part of the process of a craft shortage’s influence on cost per-
formance through impacts on time overruns. The importance of
addressing these knowledge gaps lies in the fact that doing so will
explicate and clarify the significance of skilled labor availability as
one of the main causes of cost overruns in construction projects.
To accomplish the previously mentioned aims, the following re-
search objectives were defined:
• Examine whether there is a significant relationship between
craft staffing difficulty and hourly wage as well as quality per-
formance. In addition, it will be examined if there is a significant
influence of time overrun on cost overrun. Fulfilling this objec-
tive, along with results of recent aforementioned studies men-
tioned, will complete the quantitative support for all processes
of influence of skilled labor shortage on cost performance.
• Identify whether there is a significant difference in cost overrun
of projects that experienced a craft labor shortage versus the
ones that did not.
• Develop a model that quantifiably links skilled labor variability
to project actual cost overrun.
Data Collection
The data used in this research were obtained from two different sour-
ces to enhance the validity and reliability of the study. The first
source was a primary data collection effort through a CII Research
Team 318 (RT-318) survey (Taylor et al. 2016). This survey collected
project performance and workforce demographic data on completed
construction projects in the United States and Canada. The survey
was developed, pilot-tested, and distributed to CII and non-CII
member construction organizations. There were 29 total responses
to the survey (30% response rate), with 26 projects from the United
States and three from Canada. Most of the survey responses were
industrial projects (25 out of 29), whereas the remaining projects
were a building (one project), light industrial (one project), and infra-
structure (two projects). The projects were distributed across North
America covering 18 states in the United States and three Canadian
provinces. Except for two projects performed between 2009 and
2011, all of the other 27 projects were performed and completed
between 2011 and 2014. The results of this survey were used to con-
struct a database of project characteristics. Additional details of the
survey effort were described by Taylor et al. (2016).
The second data source was obtained through the CII Bench-
marking and Metrics (CII BM&M) database. The CII BM&M data-
base was designed to get comprehensive data of construction
projects performed by CII member companies. For this research,
all projects in this database that reported data of availability of
skilled labor were selected. This subset consisted of 68 completed
projects, of which 59 were performed in the United States and nine
in Canada. Of these 68 projects, 31 projects (45%) were heavy
industrial, 24 projects (35%) were commercial building, seven
projects (10%) were light industrial, and six projects (9%) were in-
frastructure projects. All projects in this database were executed be-
tween 2001 and 2013. Additional details of the CII BM&M survey
and the subset of data used in this study can be found in Taylor et al.
(2016) and Karimi (2017). Table 1 provides the average, median,
and range of the size of these projects in terms of actual cost, actual
time, and actual craft direct work hours in both databases.
Skilled Labor Availability Measurement
In both databases, an estimate of the level of craft shortages in proj-
ects relied on subjective evaluations by project management teams.
J. Constr. Eng. Manage.
 Table 1. Summary of projects size in RT-318 survey and CII BM&M
database
Average Minimum
Database Project size (median) and maximum
RT-318 survey Actual cost ($ millions) 455.2 (45) 3.6 and 8,549
(29 projects) Actual schedule (days) 554.7 (533) 134 and 1,648
Craft work hours 610.6 (321) 13.3 and 3,777.9
(1,000 h)
CII BM&M Actual cost ($ millions) 142.5 (40.1) 0.5 and 1,799.3
(68 projects) Actual schedule (days) 1,054.5 (678) 46 and 3,131
Craft work hour 732.5 (110) 2.5 and 8,870.6
(1,000 h)
Downloaded from ascelibrary.org by La Trobe University on 05/03/18. Copyright ASCE. For personal use only; all rights reserved.
Table 2. Levels of craft staffing difficulties in the RT-318 survey
Level Definition Score
No Difficulty There was no shortage. Able to staff the
project with no delay on construction
Slight Recruiting difficulties led to consumption of
schedule float and/or contingency
Moderate Recruiting difficulties led to delay of
completing project activities on time
Severe Recruiting difficulties led to delay of
completing project milestones
Very Severe Recruiting difficulties led to project delay
That is because collecting quantitative data on labor availability (sup-
ply and demand) over the whole period of a project and for different
trades is a complex task. It requires collecting the number of required
craft labor (demand) as well as actual hired craft labor (supply) in
each week or month. In addition, the level of skills of the hired craft
workers needs to be taken into consideration, which increases the
complexity of this measurement. Therefore, it is justified to choose
subjective measurements by the project management team as an ap-
propriate way to measure skilled labor availability in a project.
In the RT-318 survey, respondents were asked to indicate the
level of craft staffing difficulty they experienced on their project
for 13 craft labor trades, which included carpenters, pipefitters,
electricians, boilermakers, sheet-metal workers, ironworkers, pipe
welders, structural welders, equipment operators, crane operators,
millwright workers, instrument fitters, and supervisors. There were
five levels of staffing difficulty defined in the survey, ranging from
No Difficulty to Very Severe (Table 2). To provide an overall level
of craft staffing difficulty for each project, the authors calculated an
aggregate average of craft staffing difficulty across these 13 trades
for each project as follows:
Craft staffing difficulty score of a project
ð0 × AÞ þ ð1 × BÞ þ ð2 × CÞ þ ð3 × DÞ þ ð4 × EÞ
¼ hypotheses. The null hypothesis for each analysis is that no rela-
13
in which A, B, C, D, and E = numbers of trades in each level of
staffing difficulty from No Difficulty to Very Severe.
In the CII BM&M database, respondents indicated the level of
availability of skilled labor across all trades compared with what
had been specified during the planning stage of their projects.
These levels ranged from Extremely Negative (−5) to Extremely
Positive (þ5), with zero representing the as-planned situation.
Cost and Schedule Performance
The cost and schedule performances in both databases were mea-
sured by the percentage of cost and schedule overrun relative to the
planned ones as follows:
© ASCE
J. Constr. Eng. Manage., 2018, 144(7): 04018057
Cost overrunð%Þ
Actual construction cost − Budgeted construction cost
¼ × 100
Budgeted construction cost
Schedule overrunð%Þ
Actual construction schedule − Planned construction schedule
¼
Planned construction schedule
× 100
Quality Performance
The quality performance in CII BM&M projects was measured
and obtained by a subjective evaluation of project management
team. The respondents of the CII BM&M questionnaire were asked
to indicate the level of success of their project’s quality perfor-
mance on the scale of 1 as Not At All Successful to 7 as Extremely
0 Successful. The RT-318 survey did not collect any data related to
quality performance.
1
2 Hourly Wage Change
3 The RT-318 survey asked respondents to provide an estimated and
actual raw hourly wage for each of the aforementioned 13 trades in
4 their projects. Therefore, an hourly wage change in a project for
each trade was measured by a percentage of change in wage relative
to the estimated one as follows:
Hourly wage changeð%Þ
Actual hourly wages − Estimated hourly wages
¼ × 100
Estimated hourly wages
The CII BM&M database had no data related to the hourly wage
of craft labor in projects.
Data Analysis
Preliminary Hypothesis Testing
The purpose of preliminary hypothesis testing is to examine the
influential processes of the impact of craft shortage on cost perfor-
mance that have not yet been examined empirically in previous
works. These processes include the impact of craft labor shortages
on quality performance decrease as well as craft hourly wage in-
creases. In addition, the impact of time overruns on cost overruns
will be examined as a part of the impact process of a craft shortage
on cost performance through time overruns.
Simple linear regression analysis was selected to examine the
tionship exists between the two variables. The alternative hypoth-
esis is that there is a significant relationship between the two
variables, which was determined by obtaining a p-value of less than
0.05. For all regression analyses, outliers were detected and re-
moved from the analysis using the Cook’s distance as suggested by
Agresti (2015). The two variables used in the regression analyses to
assess the level of craft labor availability were the craft staffing dif-
ficulty (RT-318 survey) and skilled labor availability (CII BM&M).
Both variables are categorical with natural ordering; hence, they can
be referred to as ordinal variables. To use more powerful methods
available for quantitative variables such as regression, it is possible
to assign numerical scores to categories of ordinal variables and treat
them as an interval variable (Agresti and Finlay 2009). Both vari-
ables had been assigned numerical scores and were suitable to be
04018057-4 J. Constr. Eng. Manage.
 Table 3. Summary of preliminary hypothesis development
Hypothesis Database (unit) Dependent variable Independent variable Null hypothesis Alternative hypothesis
H1 RT-318 survey (trade) Hourly wage change Craft staffing difficulty There is no relationship There is a relationship between
between variables hourly wage change and craft
staffing difficulty
H2 RT-318 survey & Cost overrun Time overrun There is no relationship There is a relationship between
CII BM&M (project) between variables cost overrun and time overrun
H3 CII BM&M (project) Meeting quality Skilled labor availability There is no relationship There is a relationship between
performance between variables meeting quality performance
expectation expectation and skilled labor
availability
Downloaded from ascelibrary.org by La Trobe University on 05/03/18. Copyright ASCE. For personal use only; all rights reserved.

Table 4. Summary of the results of preliminary hypothesis testing Hinze & Appelgate (1991)
Karimi et al. (2016) Safety Everett & Frank (1996)
Number of Performance
Model data points Equationa R2 P-value
H1 96 trades Y ¼ −1.03 þ 1.83X 0.24 <0.0001 Karimi et al. (2017)
Overtime
H2 72 projects Y ¼ −5.38 þ 0.45X 0.32 <0.0001 Hanna et al.
H3 55 projects Y ¼ 5.86 þ 0.16X 0.13 0.0076 Karimi et al. (2017) (2005)
Heravi & Eslamdoost (2015) Labor
a
X = independent variable; and Y = dependent variable as defined in Table 3. Productivity Project Cost
Skilled Labor
Availability H1 Performance
Hourly Wage

used in regression analysis. In addition, the Meeting Quality Expect- Hwang et al. (2009)
Fayek et al. (2003) Quality Love & Li (2000)
ation variable in BM&M database also has a similar status as skilled H3 Performance
labor availability variable. All hypotheses are described in detail in
Table 3. Karimi et al. (2017) Schedule Flyvbjerg et al. (2004)

The regression analysis for Hypothesis 1 (H1) shows that there Performance H2

is a significant relationship between increased craft staffing diffi-

culty and increased hourly wages in construction projects. The
p-value of less than 0.0001 resulted in rejection of the null hypoth-
esis and acceptance of the alternative hypothesis. In Hypothesis 2,
because the type and method of measuring the two variables were
the same in both data sets, two databases were combined. The
analysis shows that the impact of time overruns, which can be par-
tially due to a shortage of skilled labor (Karimi et al. 2017), on cost
overruns is considerable. The p-values of less than 0.0001 indicates
the significance of the relationships.
The last analysis examines the influence of skilled labor avail-
ability on quality performance. The p-value of the model was
0.0076, which supports a conclusion that there is a significant
relationship between these two variables. Table 4 presents the de-
tailed results of all hypothesis testing. Fig. 1 illustrates the proc-
esses of the influence of skilled labor shortage on cost performance.
It also lists studies that demonstrated the significance of the impact
empirically.
The analyses of the preliminary hypotheses testing along with
various quantitative studies explained previously explicate the proc-
esses of impact of a skilled labor shortage on projects cost perfor-
mance. The next analysis examines the significance of the overall
influence of a skilled labor shortage on projects cost performance.
Main Hypothesis Testing
The purpose of the main hypothesis testing is to examine the overall
impact of skilled labor shortages on project cost performance.
To conduct more reliable hypothesis testing, the authors decided
to define a single hypothesis for all projects in both databases.
Because the two databases used in this study had different scales
to measure skilled labor availability, it was required to convert one
database’s scale to that of the other and create a single unified meas-
urement across all projects in the two data sets. As will be explained
Fig. 1. Processes of the impact of skilled labor availability on project
cost performance. The studies are those that quantitatively demonstrate
the significance of the impact. H1–H3 are the hypotheses tested in this
study and show the significance of the impact.
subsequently, this also helps to have comparable regression models
derived from each of these data sets.
The scale in the RT-318 survey has no measure for the condition
of surplus of craft labor like the CII BM&M scale (þ1 to þ5);
therefore, the only option was to convert the CII BM&M’s scale
to the RT-318’s scale. Each number between zero (As Planned)
to þ5 (Extremely Positive) in the CII BM&M database was con-
verted to the No Difficulty level (0) in the RT-318 survey. The CII
BM&M questionnaire defined the zero score as a situation where
actual skilled labor availability was similar to what was expected
during project planning (i.e., skilled labor availability did not pos-
itively or negatively affect project performance when compared
with the project plan) and scores greater than zero as the condition
where the availability of craft workers had a positive impact on
project performance. Therefore, it can be reasonably assumed that
when a project in this database has a score of equal or greater than
zero, the project experienced no craft shortage whereas projects
with scores between −1 and −5 experienced some level of skilled
labor shortage. There were 17 projects among the total of 68 proj-
ects in this database that had a score greater than zero. The scores
between −5 (Extremely Negative) to −1 in the CII BM&M were
scaled proportionally to the number between 1 (Slight) to 4 (Very
Severe), as defined in the RT-318 survey. This can be done by
multiplying any score between −1 and −5 by −4=5. The final
skilled labor availability score for projects in both databases can
be any number between 0 and 4. Fig. 2 illustrates this scale con-
version process.

© ASCE 04018057-5 J. Constr. Eng. Manage.

J. Constr. Eng. Manage., 2018, 144(7): 04018057

 CII BM&M Survey Tool
Extremely Extremely
Table 6. Hypothesis testing result of cost overrun comparison between
Negative As Planned projects that experienced a skilled labor shortage and projects that did not
Positive
-5 -4 -3 -2 -1 0 +1 +2 +3 +4 +5
Availability of Skilled Hypothesis analysis Parameter Value
Labor
Projects not impacted by Mean cost overrun −6.7%
(-4/5)
craft shortage (score ¼ 0) N 48
W 0.96a
Projects impacted by Mean cost overrun 6.3%
Craft Staffing
Difficulty
Very Severe (4) Severe (3) Moderate (2) Slight (1) No Difficulty (0) craft shortage (0 < score ≤ 4) N 35
W 0.89b
RT-318 Survey Tool
Levene’s F-value 1.43a
T-value — 3.14
Fig. 2. Process of converting the BM&M scale of availability of craft Degree of freedom — 81
workers to RT-318 survey’s scale. P-value — 0.001
Note: N = Number; and W = Shapiro-Wilk statistics.
Downloaded from ascelibrary.org by La Trobe University on 05/03/18. Copyright ASCE. For personal use only; all rights reserved.

a
Not significant at 0.05 level.
b
Table 5. Summary of main hypothesis development Significant at 0.01 level.

Hypothesis section Description

Projects classification Projects impacted by craft shortage
(0 < score ≤ 4) versus not impacted
(score ¼ 0)
Database RT-318 survey and CII BM&M
Null hypothesis (Ho ) There is no difference in mean cost overrun
Alternative hypothesis (Ha ) The mean cost overrun is higher in projects
impacted by a craft shortage
After scale conversion, all projects in both databases were di-
vided into two groups: (1) projects with craft staffing difficulty
score equal to zero (score ¼ 0), which were classified as projects
that experienced no shortage of skilled labor; and (2) projects with
a craft staffing difficulty score greater than zero (0 < score ≤ 4),
which were classified as projects that experienced some level of
shortage of skilled labor. Table 5 provides details of this hypothesis
testing.
A t-test was conducted to determine if the difference in average
cost overrun between two groups of projects was statistically sig-
nificant. The underlying assumption of the test is to have a nor-
mally distributed population. This assumption was assessed using
Shapiro-Wilk test. The test returned a p-value greater than 0.05
for the first group (score ¼ 0), suggesting that the assumption of
normality is justifiable for this group. The test in second group
(0 < score ≤ 4) returned a p-value less than 0.05, indicating the
deviation from normality in this data set.
Agresti and Finlay (2009) stated that the t-test is robust to the
violation of the normal population assumption, particularly when
the data points exceed 30 according to Central Limit Theorem.
They, however, stated that caution needs to be exercised if there
is an extreme outlier or extreme skew in the data points. Because
there are more than 30 data points in the second group (n ¼ 35) and
they do not contain any extreme outliers and have no extreme
deviation from normality, the t-test can be adopted to test the differ-
ence in average cost overrun between the two groups. Levene’s test
was performed to examine the assumption of equality of variances.
The test returned a p-value greater than 0.05, suggesting that the
assumption of equality of variances is justifiable. As indicated in
Table 6, the average cost overrun in projects affected by a skilled
labor shortage was 6.3% compared with −6.7% for projects not
affected by a skilled labor shortage. This difference is significant
at the 0.001 level, which allows the null hypothesis (Ho ) to be re-
jected and supports a conclusion that cost performance in projects
is substantially affected by shortages of skilled labor.
The analyses of main hypothesis testing demonstrate the sub-
stantial influence of a skilled labor staffing difficulty on project cost
performance. The results justify a deeper exploration of these in-
fluential relationships in construction projects.
Regression Analysis
Multiple linear regression analysis was performed to further exam-
ine the relationship between a craft labor staffing difficulty and
project cost overruns. Two predictor variables of the analysis were
actual construction cost and craft staffing difficulty scores. In de-
veloping a predictive labor demand model for construction proj-
ects, Wong et al. (2008) found that construction cost is the most
significant determinant of labor demand. Although the craft staffing
difficulty variable inherently includes both demand and supply of
craft labor in a project, it is reasonable to examine its influence on
cost overrun with consideration of actual cost of a project. It is ex-
pected that a labor shortage on a larger project results in a higher
cost overrun because the execution of a project relies on more num-
ber of craft workers. The other important and independent predic-
tors related to the analysis, such as project type, actual project
schedule, project complexity, and total site work hours, were also
examined and found to have no statistically significant impact on
project cost overrun when considering the impact of craft staffing
difficulties.
Analysis of CII BM&M Database
A total of 64 data points were available for this analysis. The analy-
sis returned the regression model containing both aforementioned
variables. The p-value of the model was less than 0.0001, which
indicates the adequacy of the model. The R2 value of the model is
0.59 with an adjusted R2 of 0.58. Table 7 presents the details of the
statistical analysis of the model.
Analysis of RT-318 Survey Database
The total available data points for this analysis were 17. The analy-
sis again returned the regression model containing both variables.
The p-value of the model was 0.0019, which indicates the signifi-
cance of the model. The R2 value of the model is 0.59 with an ad-
justed R2 of 0.53. The details of the statistical analysis of model are
given in Table 7.
Analysis of the Combined Data Set
Because the models derived from two available databases with
different temporal contexts [BM&M projects (2001–2013) and
RT-318 survey (2011–2014)] showed a similar pattern of the

© ASCE 04018057-6 J. Constr. Eng. Manage.

J. Constr. Eng. Manage., 2018, 144(7): 04018057

 Table 7. Regression model of the impact of craft staffing difficulty on construction cost overrun
Number projects Actual Craft staffing Adjusted
(database) Dependent variable Constant construction cost difficulty F-value P-value R2 R2
64 (BM&M) Construction cost overrun −15.84a (−5.79) 0.077a (6.54) 10.699b (3.80) 43.91 <0.0001 0.59 0.58
17 (RT-318) Construction cost overrun −9.167c (−1.63) 0.0058d (3.25) 11.132e (2.61) 10.08 0.0019 0.59 0.53
81 (combined) Construction cost overrun −10.74b (−3.76) 0.008d (3.06) 15.04b (5.56) 23.89 <0.0001 0.38 0.36
Note: t-values shown in parentheses.
a
t-statistics significant at 0.0001 level.
b
t-statistics significant at 0.001 level.
c
t-statistics significant at 0.15 level.
d
t-statistics significant at 0.01 level.
e
t-statistics significant at 0.05 level.
Downloaded from ascelibrary.org by La Trobe University on 05/03/18. Copyright ASCE. For personal use only; all rights reserved.

impact of skilled labor availability on cost performance, it was de- transfer estimator (CTE). Introduced by Atherton and Ben-Akiva
cided to examine this impact across all available data. The most (1976), the Bayesian updating method is based on classical
common method of data combining used in past literature is the Bayesian analysis and estimates updated parameters of model with
joint context estimation. This method combines two databases and combining the parameters of models derived from two data sources.
uses them simultaneously in determining the parameters of a new Based on a normality assumption, the method is expressed math-
model. The basic theory of the method was discussed by Bradley ematically as follows (Karasmaa 2007):
and Daly (1991) and Ben-Akiva and Morikawa (1990). The under- !−1 !
X−1 X−1 X−1 −1
X
lying idea of this method is to adjust the random variation in the
βd
BU ¼ þ βbi þ βbj
utility function of the different data sets to be equal (Karasmaa
i j i j
2007). Under the assumption of equality of parameter vectors of
models from two databases, the least-square estimator of combined in which βBU = updated vector of parameters of the final model;
model is calculated as follows: βi and βj = estimated
P P vectors of parameters of the first and second
β ¼ ð½XT ½XÞ½XT ½Y models; and i and j = covariance matrixes of the first and sec-

 ond
P models.
P−1 The covariance of final model can be calculated as
Y1 ð −1i þ
−1
j Þ .
in which ½Y ¼ = vector of response variable form two
Y2 The combined transfer estimator method is a generalization


X1 of the Bayesian updating method because it explicitly takes into con-
databases; and ½X ¼ = explanatory variables from two
X2 sideration of the bias transfer (Shyr 1993). This procedure calculates
databases. In order to make the variance of error in two models the weighted average of models’ parameters and assigns the weight
equal, the second model needs to be multiplied by correction factor to each parameter in such a way that the mean square error (MSE) of
(μ), which is defined as μ2 ¼ Varðε1 Þ=Varðε2 Þ (Karasmaa 2007), the updated parameters is minimized (Ben-Akiva and Bolduc 1987).
in which ε1 and ε2 are the errors of Models 1 and 2, respectively. In The model can be expressed as follows (Karasmaa 2007):
this analysis, the correction factor (μ) can be reasonably assumed to !−1 !−1
X−1 X−1
equal to 1 because the test shows no significant difference between βd
CTE ¼ þΔΔ T
þ
Var (ε1 ) and Var (ε2 ) (F ¼ 1.53 < Fα¼0.05 ¼ 2.22). The null hy- i j
pothesis in the test is Ho ∶Varðε1 Þ ¼ Varðε2 Þ, and the alternative −1
!−1
−1
!
hypothesis is Ha ∶Varðε1 Þ ≠ Varðε2 Þ. The details of test are pro- X X
þ þΔΔ T
βbi þ βbj
vided in Table 8. i j
The regression analysis on total available data points (81 proj-
ects) returned the model containing both variables. The p-value of in which βCTE = updated vector parameter of the final model;Pβi and
the model was less than 0.0001, which indicates the significance β
P j = estimated parameters of the first and second models; i and
of the model. The R2 value of the model is 0.38 with an adjusted j = covariance matrixes of the first and second models; Δ ¼ ðβi−
R2 of 0.36. The details of the statistical analysis of the model are βj Þ is the transfer bias; and ΔT = transpose of the matrix Δ. The
given in Table 7. covariance matrix of final model can be calculated as
P2
P 0 −1
i P
CTE ¼ 2 . Table 9 shows the final models derived
Other Methods of Data Combining 0 j

The two other common data-combining procedures used in pre- from all of the data (both databases) with three different common
vious literature are the Bayesian updating method and combined methods of data combining. All three models demonstrate similar
patterns of the impact of skilled labor availability, which enhance
the reliability of the result.
Table 8. Comparison of variance of error in two models
Database Parameter Value
Discussion
CII BM&M MSE 330.26
Degree of freedom 61 Cost performance of a construction project is a complex function of
RT-318 MSE 216.26 many factors. The models presented in Table 9 should be regarded
Degree of freedom 14 as informative models that illustrate the pattern of the risk that a
— F-value 1.53 skilled labor shortage poses to cost performance. The primary ben-
— F-value (α ¼ 0.05) 2.22
efit of this risk estimation is to enable project management teams to

© ASCE 04018057-7 J. Constr. Eng. Manage.

J. Constr. Eng. Manage., 2018, 144(7): 04018057

 Table 9. Regression model of the impact of craft staffing difficulty on construction cost overruns using common data-combining methods (CII BM&M and
RT-318 survey)
Data-combining method
Joint context estimation Construction cost overrun ¼ −10.74 þ 0.008 × actual construction cost þ 15.04 × craft staffing difficulty
Bayesian updating method Construction cost overrun ¼ −10.64 þ 0.007 × actual construction cost þ 14.62 × craft staffing difficulty
Combined transfer estimator Construction cost overrun ¼ −10.63 þ 0.006 × actual construction cost þ 11.06 × craft staffing difficulty
determine whether mitigation strategies are warranted to prevent
potential shortfalls in project cost performance.
The strength of these informative risk models lies in the fact that
two primary models derived from two databases with two different
temporal contexts show similar patterns within the two robust re-
Downloaded from ascelibrary.org by La Trobe University on 05/03/18. Copyright ASCE. For personal use only; all rights reserved.
gression models. This affirms the reliability and consistency of the
results because they externally validate each other. To take the most
of these two available data sets and enhance the model reliability,
the two models were combined using three major data-combining
methods. The first model was obtained by performing a regression
analysis on the combined data set. The other two models were cal-
culated by two statistical methods of combining the models derived
from two data sets.
Because the final model is intended to be used as an informative
model not predictive one and all three final models show similar
patterns, the available tests used for assessing the performance
of models (predictability) were not conducted. All models provide
a proper risk perception of a shortage of skilled labor with regard to
the size of the project.
Limitations of the Study
Although the presented models provide beneficial and statistically
valid results, it is recognized that they are subject to the following
limitations:
• The analysis was based heavily on industrial projects (90% of
projects in RT-318 survey and 56% in CII BM&M database
were industrial projects).
• Although regression models are statistically significance with
adequate R2 value for prediction purpose, because they contain
only two influential variables of cost performance, they can only
be used as an informative risk model instead of predictive tool.
Conclusions and Recommendations
The main purpose of this research was to quantitatively model
the empirical influence of skilled labor availability on construction
project cost performance. Data from 97 construction projects com-
pleted in the United States and Canada between 2001 and 2014
were collected from two data sources. Preliminary hypothesis test-
ing evidenced the processes of the impact of craft shortage on cost
performance that have not yet been examined quantitatively. These
processes include the effects of skilled labor shortage on increasing
craft hourly wages and decreasing the quality performance. In ad-
dition, the impact of time overruns on cost overruns was examined
and shown to be significant as a part of influence of skilled labor
shortages on cost performance through schedule performance. The
impact of skilled labor shortages on time overruns has been evi-
denced in the recent quantitative study. The main hypothesis testing
examined the overall impact of a skilled labor shortage on cost per-
formance and demonstrated that projects experiencing a labor
shortage experienced a higher cost overrun compared with projects
that experienced no shortages. Further analysis by means of multi-
ple regression analysis resulted in two robust models derived from
© ASCE 04018057-8
J. Constr. Eng. Manage., 2018, 144(7): 04018057
Final model
two databases that demonstrated similar patterns of the significant
impact of skilled labor staffing difficulties on project cost perfor-
mance. Finally, by using the three main valid methods of data com-
bining, the two data sets were combined to obtain more reliable
models. The final models are presented as risk informative models
that illustrate the pattern of skilled labor shortage risk on project
cost overruns with regard to the size of a project.
The main contribution of this work to the body of knowledge is
to fill the gap in the existing literature by quantitatively modeling
and elucidating the influence of craft labor availability on construc-
tion project cost performance. This study supports and validates
previous qualitative studies that used opinion-based data to anec-
dotally link a craft labor shortage to a project’s cost growth.
The theoretical implications of this study will be on construction
industry workforce development system as one of the major US
industry sectors. Considering the results of this study, which shows
the adverse impact of a skilled labor shortage on project perfor-
mance, and on the other hand, studies that evidenced the recent
structural change in the US construction industry workforce, makes
the risk of a skilled labor shortage alarming. Taylor et al. (2016)
showed that the US construction workforce is aging faster than that
of all other private industries. They also revealed that there is a sig-
nificant shift in craft workers’ preferences from work satisfaction to
higher income and job security while there are shrinking real wage
gaps between construction craft labor and all other private indus-
tries. CII (2015) cautioned that there are national shortages in key
industrial trades (e.g., welders, pipefitters, and electricians). These
challenges represent substantial changes in the construction industry
workforce, which makes the problem of skilled labor availability
critical. This fundamental problem cannot be expected to be amelio-
rated unless these challenges are addressed not only within the con-
struction industry but also in K–12 education and societal
perceptions toward construction. However, understanding the level
of impact that craft shortages have on project performance through
robust statistical analyses can serve as a primary step in developing
motivation for industry leaders, communities, and construction
stakeholders to address this challenge.
The practical implications of this study lie in identifying,
through a quantitative approach, the shortage of skilled labor as
one of the main root causes of cost overruns in construction proj-
ects. The study provides a new realization for practitioners that
the impact of craft shortages on cost performance is significant
and can arise through various indirect influential processes. These
processes include impacts through loss of productivity, higher
safety incidents, lower quality performance, higher schedule over-
runs, and increases in hourly wage and overtime usage. Therefore,
addressing the availability of skilled labor in a project can be re-
garded as a multipurpose mitigation strategy with substantial ben-
efits and gains for a project. In addition to controlling the overall
cost of a project, project managers can improve productivity, safety,
quality, and schedule performance by investing in hiring, training,
and retaining of skilled labors.
The presented risk models can be used by project management
teams at the planning stage of projects to estimate the risk of ex-
pected level of skilled labor availability to the cost of a project with
J. Constr. Eng. Manage.
 regard to its size. This risk estimation will enable them to determine
the best possible mitigation strategies to address a skilled labor
shortage in a project.
Data Availability Statement
Data generated or analyzed during the study (RT-318 Survey) are
available at http://uknowledge.uky.edu/cgi/viewcontent.cgi?article=
1059&context=ce_etds. Other data analyzed during the study were
provided by a third party. Requests for data should be directed
to the provider indicated in the Acknowledgements (CII BM&M).
Information about the Journal’s data sharing policy can be found
here: http://ascelibrary.org/doi/10.1061/%28ASCE%29CO.1943
Downloaded from ascelibrary.org by La Trobe University on 05/03/18. Copyright ASCE. For personal use only; all rights reserved.
-7862.0001263.
Acknowledgments
The authors would like to thank the Construction Industry Institute
for its support of this research through Research Team (RT) 318.
In addition, the authors thank the members of CII RT 318 and 335
for their substantial contributions to this research. The views and
opinions expressed herein are of the authors and do not necessarily
represent the views and opinions of the Construction Industry
Institute or the industry team members of RT-318 and RT-335.
References
Abdol, R. C., H. B. Brisbane, and G. D. Eric. 1999. “Causes of the
construction skilled labor shortage and proposed solutions.” In Proc.,
ASC 35th Annual Conf., 187–196. San Luis Obispo, CA: California
Polytechnic State Univ.
Agresti, A. 2015. Foundations of linear and generalized linear models.
East Orange, NJ: Wiley.
Agresti, A., and B. Finlay. 2009. Statistical methods for the social sciences.
4th ed. Upper Saddle River, NJ: Pearson Prentice Hall.
Akintoye, A. 2000. “Analysis of factors influencing project cost estimating
practice.” Constr. Manage. Econ. 18 (1): 77–89. https://doi.org/10.1080
/014461900370979.
Atherton, T., and M. E. Ben-Akiva. 1976. “Transferability and updating of
disaggregate travel demand models.” Transp. Res. Rec. 610: 12–18.
Ben-Akiva, M., and D. Bolduc. 1987. “Approaches to model transferability
and updating: The combined transfer estimator.” Transp. Res. Rec.
1139: 1–7.
Ben-Akiva, M. E., and T. Morikawa 1990. “Estimation of travel demand
models from multiple data sources.” In Proc., 11th Int. Symp. on Trans-
portation and Traffic Theory, Yokohama, Japan.
Bradley, M., and A. Daly. 1991. “Estimation of logit choice models using
mixed stated preference and revealed preference information.” In Proc.,
6th Int. Conf. on Travel Behaviour, edited by P. Stopher and M. Lee-
Gosselin, 209–232. New York: Elsevier.
Business Roundtable. 1983. More construction for the money, construction
industry cost effectiveness project. Summary Rep. Washington, DC:
Business Roundtable.
Business Roundtable. 1997. Confronting the skilled construction workforce
shortage. Washington, DC: Business Roundtable.
Chester, M., and C. Hendrickson. 2005. “Cost impacts, scheduling impacts,
and the claims process during construction.” J. Constr. Eng. Manage.
131 (1): 102–107. https://doi.org/10.1061/(ASCE)0733-9364(2005)
131:1(102).
Construction Industry Institute. 1990. An assessment of education and train-
ing needs among construction personnel. Research Rep. 158-11. Austin,
TX: Univ. of Texas at Austin.
Construction Industry Institute. 2015. Is there a demographic craft labor
cliff that will affect project performance? Research Summary 318-1.
Austin, TX: Univ. of Texas at Austin.
© ASCE 04018057-9
J. Constr. Eng. Manage., 2018, 144(7): 04018057
Dai, J., P. M. Goodrum, and W. F. Maloney. 2009. “Construction craft
workers’ perceptions of the factors affecting their productivity.” J.
Constr. Eng. Manage. 135 (3): 217–226. https://doi.org/10.1061
/(ASCE)0733-9364(2009)135:3(217).
Doloi, H. 2013. “Cost overruns and failure in project management:
Understanding the roles of key stakeholders in construction projects.”
J. Constr. Eng. Manage. 139 (3): 267–279. https://doi.org/10.1061
/(ASCE)CO.1943-7862.0000621.
Ernzen, J., and C. Schexnayder. 2000. “One company’s experience with
design/build: Labor cost risk and profit potential.” J. Constr. Eng.
Manage. 126 (1): 10–14. https://doi.org/10.1061/(ASCE)0733-9364
(2000)126:1(10).
Everett, J. G., and P. B. Frank, Jr. 1996. “Costs of accidents and injuries to
the construction industry.” J. Constr. Eng. Manage. 122 (2): 158–164.
https://doi.org/10.1061/(ASCE)0733-9364(1996)122:2(158).
Fayek, A. R., M. Dissanayake, and O. Campero. 2003. Measuring and clas-
sifying construction field rework: A pilot study. Edmonton, AB, Canada:
Dept. of Civil and Environmental Engineering, Univ. of Alberta.
Flyvbjerg, B., M. K. Skamris Holm, and S. L. Buhl. 2004. “What causes
cost overrun in transport infrastructure projects?” Transp. Rev. 24 (1):
3–18. https://doi.org/10.1080/0144164032000080494a.
Gharaibeh, H. M. 2014. “Cost control in mega projects using the Delphi
method.” J. Constr. Eng. Manage. 30 (5): 04014024. https://doi.org/10
.1061/(ASCE)ME.1943-5479.0000218.
Goodrum, P. M. 2004. “Hispanic and non-Hispanic wage differentials:
Implications for United States construction industry.” J. Constr.
Eng. Manage. 130 (4): 552–559. https://doi.org/10.1061/(ASCE)0733
-9364(2004)130:4(552).
Hanna, A. S. 2001. Quantifying the impact of change orders on electrical
and mechanical labor productivity. Research Rep. No. 158-11. Austin,
TX: Construction Industry Institute.
Hanna, A. S., C. S. Taylor, and K. T. Sullivan. 2005. “Impact of extended
overtime on construction labor productivity.” J. Constr. Eng. Manage.
131 (6): 734–739. https://doi.org/10.1061/(ASCE)0733-9364(2005)
131:6(734).
Heravi, G., and E. Eslamdoost. 2015. “Applying artificial neural networks
for measuring and predicting construction-labor productivity.” J.
Constr. Eng. Manage. 141 (10): 04015032. https://doi.org/10.1061
/(ASCE)CO.1943-7862.0001006.
Hinze, J., and L. L. Appelgate. 1991. “Costs of construction injuries.”
J. Constr. Eng. Manage. 117 (3): 537–550. https://doi.org/10.1061
/(ASCE)0733-9364(1991)117:3(537).
Hwang, B. G., S. R. Thomas, C. T. Haas, and C. H. Caldas. 2009. “Meas-
uring the impact of rework on construction cost performance.” J.
Constr. Eng. Manage. 135 (3): 187–198. https://doi.org/10.1061
/(ASCE)0733-9364(2009)135:3(187).
Improving Construction Safety Performance. 1982. Construction industry
cost effectiveness (CICE) project. Rep. A-3. New York: Business
Roundtable.
Karasmaa, N. 2007. “Evaluation of transfer methods for spatial travel de-
mand models.” J. Transp. Res. Part A 41 (5): 411–427. https://doi.org
/10.1016/j.tra.2006.09.009.
Karimi, H. 2017. “Quantitative analysis of the impact of craft labor avail-
ability on construction project performance.” Theses and Dissertations–
Civil Engineering, University of Kentucky. https://uknowledge.uky.edu
/ce_etds/56.
Karimi, H., T. R. Taylor, and P. M. Goodrum. 2017. “Analysis of the impact
of craft labour availability on North American construction project pro-
ductivity and schedule performance.” Constr. Manage. Econ. 35 (6):
368–380. https://doi.org/10.1080/01446193.2017.1294257.
Karimi, H., T. R. Taylor, P. M. Goodrum, and C. Srinivasan. 2016. “Quan-
titative analysis of the impact of craft worker availability on construc-
tion project safety performance.” J. Constr. Innovation 16 (3): 307–322.
https://doi.org/10.1108/CI-10-2015-0050.
Li, Y., L. Jiang, T. R. B. Taylor, and D. N. Ford. 2017. “Impact of labor
controls on tipping point dynamics in large complex projects.” Syst.
Res. Behav. Sci. in press.
Love, P. E., and H. Li. 2000. “Quantifying the causes and costs of rework in
construction.” Constr. Manage. Econ. 18 (4): 479–490. https://doi.org
/10.1080/01446190050024897.
J. Constr. Eng. Manage.
 Lyneis, J. M., and D. N. Ford. 2007. “System dynamics applied to project
management: A survey, assessment, and directions for future research.”
Syst. Dyn. Rev. 23 (2/3): 157–189. https://doi.org/10.1002/sdr.377.
McTague, B., and G. Jergeas. 2002. Productivity improvement on Alberta
major construction projects. Construction Productivity Improvement
Rep. Edmonton, AB, Canada: Alberta Economic Development.
Rosenbaum, D. 2001. “Craft labor shortage provokes more studies of pay
and safety.” Engineering news record, 11. New York: McGraw-Hill.
RSMeans. 2016. Building construction cost data. Norwell, MA: Construc-
tion Publishers & Consultants.
Sawyer, T., and D. Rubin. 2007. “Leaders probe new solutions for indus-
try’s labor shortfall.” Engineering News Record, 15. New York:
McGraw-Hill.
Shyr, F.-Y. 1993. “Combining laboratory and field data in rain fatigue
analysis.” Ph.D. dissertation, Massachusetts Institute of Technology.
Downloaded from ascelibrary.org by La Trobe University on 05/03/18. Copyright ASCE. For personal use only; all rights reserved.
© ASCE 04018057-10
J. Constr. Eng. Manage., 2018, 144(7): 04018057
Taylor, T. R. B., H. Karimi, P. M. Goodrum, and M. Albattah. 2016. Is there
a demographic craft labor cliff that will affect project performance?
Construction Industry Institute Research Rep. 318-2. Austin, TX: Univ.
of Texas-Austin.
Trost, S. M., and G. D. Oberlender. 2003. “Predicting accuracy of early cost
estimates using factor analysis and multivariate regression.” J. Constr.
Eng. Manage. 129 (2): 198–204. https://doi.org/10.1061/(ASCE)0733
-9364(2003)129:2(198).
Wang, Y., P. M. Goodrum, C. T. Haas, and R. W. Glover. 2008. “Craft
training issues in American industrial and commercial construction.”
J. Constr. Eng. Manage. 134 (10): 795–803. https://doi.org/10.1061
/(ASCE)0733-9364(2008)134:10(795).
Wong, J. M. W., A. P. C. Chan, and Y. H. Chiang. 2008. “Modeling
and forecasting construction labor demand: Multivariate analysis.”
J. Constr. Eng. Manage. 134 (9): 664–672. https://doi.org/10.1061
/(ASCE)0733-9364(2008)134:9(664).
J. Constr. Eng. Manage.