ArticlePDF Available

Impact of Dynamic Workforce and Workplace Variables on the Productivity of the Construction Industry: New Gross Construction Productivity Indicator

September 2020
Journal of Management in Engineering 37(1):04020092

September 2020
37(1):04020092

DOI:10.1061/(ASCE)ME.1943-5479.0000862

Authors:

Rayan H. Assaad

New Jersey Institute of Technology

Islam El-adaway

Missouri University of Science and Technology

Construction productivity is the industry's predominant determinant of performance. Although the construction industry periodically provides large amount of data, existing studies have not fully exploited such data sets, especially relating to the overall productivity of the construction industry rather than labor productivity. This paper addresses this critical knowledge gap by statistically examining and modeling the causalities between different dynamic workforce and workplace variables and the productivity of the entire construction industry. Multivariate time-series data between 2006 and 2019 were collected for the productivity of the construction industry and 11 dynamic workforce and workplace variables: job openings, job hires, turnover or job separations, total compensation, gross job gains, gross job losses, average hourly earnings, fatalities, occupational injuries and illnesses, gross domestic product, and unemployment rate. Statistically significant relationships and causalities were examined between the response variable-productivity of the construction industry-and these 11 variables. A vector autoregression (VAR) framework was developed to model the temporal variations in the productivity of the construction industry. The developed VAR model was validated by predicting the construction productivity for the 2016-2019 period an acceptable mean average percentage error of 5.13%. Based on the findings, the paper concludes that (1) all considered dynamic workforce and workplace variables, except job openings, statistically cause fluctuations in the construction productivity; (2) the new concept of gross construction productivity is justified statistically and should be implemented in the construction industry; (3) the gross construction productivity is an additional valuable information that construction companies should consider to make different insightful and well-educated industry-related decisions; (4) the health of the construction industry needs to be studied based on the productivity of the industry as a whole rather than based on labor productivity alone; and (5) the construction industry should move toward the development of a notion of gross construction productivity indicator used to measure, evaluate, and predict the performance of the entire industry. Ultimately, this paper proposes a new indicator or index for gross construction productivity. The outcomes of this paper add to the body of knowledge by providing a better understanding of the impact of different dynamic workforce and workplace variables on the construction productivity and by offering a new concept called gross construction productivity.

shows the collected time series data with monthly time interval. After collecting and visualizing the data, the cointegration

…

Variables and their sources

…

Results of stationary test

…

Results of cointegration test

…

Fitted VAR statistical model for construction productivity

…

Figures - uploaded by Rayan H. Assaad

Content may be subject to copyright.

Content uploaded by Rayan H. Assaad

Content may be subject to copyright.

Impact of Dynamic Workforce and Workplace

Variables on the Productivity of the

Construction Industry: New Gross

Construction Productivity Indicator

Rayan Assaad, S.M.ASCE1; and Islam H. El-adaway, F.ASCE2

Abstract: Construction productivity is the industry’s predominant determinant of performance. Although the construction industry peri-

odically provides large amount of data, existing studies have not fully exploited such data sets, especially relating to the overall productivity

of the construction industry rather than labor productivity. This paper addresses this critical knowledge gap by statistically examining

and modeling the causalities between different dynamic workforce and workplace variables and the productivity of the entire construction

industry. Multivariate time-series data between 2006 and 2019 were collected for the productivity of the construction industry and

11 dynamic workforce and workplace variables: job openings, job hires, turnover or job separations, total compensation, gross job gains,

gross job losses, average hourly earnings, fatalities, occupational injuries and illnesses, gross domestic product, and unemployment rate.

Statistically significant relationships and causalities were examined between the response variable—productivity of the construction

industry—and these 11 variables. A vector autoregression (VAR) framework was developed to model the temporal variations in the pro-

ductivity of the construction industry. The developed VAR model was validated by predicting the construction productivity for the 2016–2019

period an acceptable mean average percentage error of 5.13%. Based on the findings, the paper concludes that (1) all considered dynamic

workforce and workplace variables, except job openings, statistically cause fluctuations in the construction productivity; (2) the new concept

of gross construction productivity is justified statistically and should be implemented in the construction industry; (3) the gross construction

productivity is an additional valuable information that construction companies should consider to make different insightful and well-educated

industry-related decisions; (4) the health of the construction industry needs to be studied based on the productivity of the industry as a whole

rather than based on labor productivity alone; and (5) the construction industry should move toward the development of a notion of gross

construction productivity indicator used to measure, evaluate, and predict the performance of the entire industry. Ultimately, this paper

proposes a new indicator or index for gross construction productivity. The outcomes of this paper add to the body of knowledge by providing

a better understanding of the impact of different dynamic workforce and workplace variables on the construction productivity and by offering

Engineers.

Introduction

The construction industry is one of the key contributors to the

growth of economies. For instance, the construction industry con-

tributes to about 4.4% of the total gross output in the US (US

Bureau of Economic Analysis 2018). In addition, the construction

industry employs 6%–10% of the workforce (Shohet et al. 2019).

Moreover, the construction industry plays an important role in driv-

ing the activities of other industries such as manufacturing, mining

and agriculture, transportation (Donkor 2011), and infrastructure

(Assaad and El-adaway 2020g,2020e), among others. In addition,

construction is a complex business (Assaad and Abdul-Malak

2020a), and different construction-related uncertainties can impact

productivity (Choy and Ruwanpura 2005), such as unseasonable

weather conditions (Sexton et al 2020).

Construction productivity is a fundamental piece of information

for different essential construction activities such as estimating,

budgeting, and scheduling (El-Gohary et al. 2017). In addition,

construction productivity is an important metric that provides feed-

back about the industry trends and improvements (Vereen et al.

2016). In fact, construction productivity is considered the indus-

try’s predominant determinant of performance (Jarkas 2016).

Consequently, it is important to understand the fluctuations or

movements of construction productivity.

The construction industry is prone to many disruptions (Assaad

et al. 2020c), and the dynamics of the workforce and workplace

play an important role in the fluctuations of construction produc-

tivity (Chaturvedi et al. 2018;Durdyev et al. 2018). However, this

does not mean that the role of these variables is understood and

quantifiable. In other words, although construction productivity is

affected by different dynamic workforce and workplace variables,

the effects of these variables cannot be understood or quantified

unless they are researched. Dynamic variables refer to variables

1Ph.D. Candidate, Dept. of Civil, Architectural, and Environmental

Engineering, Missouri Univ. of Science and Technology, Rolla, MO 65409.

Email: rayan.assaad@mst.edu

2Hurst-McCarthy Professor of Construction Engineering and Manage-

ment, Professor of Civil Engineering, and Founding Director of Missouri

Consortium for Construction Innovation, Dept. of Civil, Architectural,

and Environmental Engineering/Dept. of Engineering Management and

Systems Engineering, Missouri Univ. of Science and Technology, Rolla,

MO 65409 (corresponding author). ORCID: https://orcid.org/0000-0002

-7306-6380. Email: eladaway@mst.edu

Note. This manuscript was submitted on March 19, 2020; approved

on July 29, 2020; published online on September 28, 2020. Discussion

period open until February 28, 2021; separate discussions must be sub-

mitted for individual papers. This paper is part of the Journal of Man-

agement in Engineering, © ASCE, ISSN 0742-597X.

J. Manage. Eng., 2021, 37(1): 04020092

with values changing from one period to another or that are linked

to other variables depending on time (Abbas and Mosallamy 2016).

In simple terms, dynamic variables can take different values over

time.

Although the construction industry produces a large amount of

data on a periodic basis, existing data sets have not been exploited

fully (Cao and Goh 2019). To the best knowledge of the authors,

no previous research work exploited publicly available workforce

and workplace variables to understand and model the fluctuations

in the productivity of the construction industry as a whole. In other

words, although the productivity’s body of knowledge has produc-

tivity has many previous research works that provided important

information on construction productivity, including but not limited

to Durdyev et al. (2018), Gurmu and Ongkowijoyo (2020), Gupta

et al. (2018), and El-Gohary et al. (2017), these efforts focused on

labor productivity rather than on the productivity of the construc-

tion industry as a whole. Thus, construction productivity in this

paper is defined as, and refers to, the productivity of the overall

construction industry; that is, the monthly generated output (in

USD) per worker-hour. Thus, this paper takes a high-level perspec-

tive by focusing on the productivity of the overall construction

industry.

Goal and Objectives

The goal of this paper is to study and model the impacts of

dynamic workforce and workplace variables on the productivity of

the construction industry. The associated research objectives are to

(1) examine the statistical relationships and causalities between nu-

merous workforce and workplace variables on one hand and

the productivity of the construction industry on the other hand;

(2) associate the changes in construction productivity with its past

fluctuations and the past movements in different dynamic work-

force and workplace variables; and (3) develop and validate a stat-

istical framework that models and predicts the productivity of the

construction industry.

Background Information

The construction industry is subject to schedule overruns, numer-

ous uncertainties (Assaad et al. 2020a,Assaad and Abdul-Malak

2020b), risks, and ever-changing conditions (Assaad and El-

adaway 2020b). One of the most critical uncertainties or risks is

productivity. Thus, this section provides the needed background

information related to construction productivity. This section also

includes information related to time series analysis.

Previous Studies Related to Construction Productivity

This subsection details the relevant previous research work on

construction productivity. Johari and Jha (2020b) explored the re-

lationships between the work motivation of workers and their

productivity based on 116 construction workers from 4 construc-

tion sites in India. Florez et al. (2020) proposed a new metric to

measure compatibility of personality among workers in a crew

and revealed how personality factors affect productivity based

on rigorous methods to analyze correlations for construction ex-

periments. Ghodrati et al. (2018) quantified the effectiveness of a

set of implemented management strategies in improving produc-

tivity in construction projects in New Zealand. Pan et al. (2019)

examined the nature of the constraints on productivity advance-

ment and explored the rationale underpinning the productivity

enhancement strategies within the construction industries of

Singapore, Hong Kong, and the United Kingdom. Johari and Jha

(2020a) established the relationship between construction workers’

aptitudes and their productivity based on data collected on 112

workers. Gurmu and Ongkowijoyo (2020) attempted to predict

construction labor productivity based on implementation levels of

human resource management (HRM) practices by performing a

correlation and associations analysis between productivity, HRM

practices, company profiles, and project properties based on data

collected from 39 contractors. Durdyev et al. (2018) developed

a structural equation model of the factors affecting construction

labor productivity in the Malaysian construction industry. Gupta

et al. (2018) determined different site amenities and worker’s

welfare factors that impact the workforce productivity in Indian

construction projects. El-Gohary et al. (2017) introduced an en-

gineering concept to document, control, predict, and improve

contractors’labor productivity for formwork and reinforcing steel

fixing crafts, including factors of the project management and

administration level and of the activity level. Bonham et al.

(2017) applied data mining techniques to quantify the relative in-

fluence of design and installation attributes on labor productivity.

Kisi et al. (2018) tested and evaluated the validity of the traditional

two-prong strategy on a complex and labor-intensive operation and

provided a framework for estimating the optimal productivity for

the fabrication activity of sheet metal ducts. Zhao and Dungan

(2018) reviewed the methods used to quantify lost labor produc-

tivity in the US construction industry, along with relevant cases,

to examine the practical considerations in selecting the proper pro-

ductivity quantification method. Gurmu (2019) developed a tool

for scoring materials management practices for building projects

and built a tool for predicting productivity based on a question-

naire distributed to construction experts and contractors.

Summary of Existing Literature on Construction

Productivity

To provide a general description of the existing knowledge on

construction productivity, this subsection summarizes the previous

literature at three levels: high level, medium level, and detailed

level. The high-level overview provides the broad research areas

on construction productivity. The medium-level summary offers

more-elaborate information on the existing knowledge on con-

struction productivity. The detailed-level review provides a more

thorough summary of the existing literature on construction

productivity.

At the high level, Durdyev et al. (2018) demonstrated that the

main thread in all contextual interpretations or definitions of pro-

ductivity is related to efficiency and effectiveness. Efficiency is re-

lated to answering the question of how efficiently scarce resources

are used throughout the implementation process to attain the in-

tended objectives (Durdyev et al. 2018). Effectiveness is related

to answering the question of how effectively the resources are uti-

lized to fulfill the set targets (Durdyev et al. 2018). At the medium

level, Yi and Chan (2014) conducted a systematic literature review

of labor productivity in the construction industry and found that the

existing knowledge could be classified into the following research

areas: the effect of variations on construction labor productivity,

methods and technology for productivity improvement, factors af-

fecting labor productivity, modeling and evaluation of construction

productivity, productivity’s trends and comparisons, and baseline/

benchmarking construction labor productivity. At the detailed level,

the existing knowledge on labor productivity involves numerous

productivity aspects, including loss of labor productivity; the im-

pact of work changes on labor productivity; the effect of engineer-

ing aspects on productivity; the influence of labor force skills

J. Manage. Eng., 2021, 37(1): 04020092

and experience on productivity; and productivity factors related to

management and control practices on construction sites, project

financing elements, availability and use of the needed material

and equipment, the properties of the project, external components,

manpower-related characteristics, technical factors, and many

others.

Combining the aforementioned high, medium, and detailed lev-

els with the literature review conducted in the previous subsection,

a conclusive summary was made of the existing knowledge on

construction productivity. The knowledge is considered to fall

under the following main categories: (1) prediction, modeling, and

evaluation of labor productivity, because it is an important piece

of information used in numerous construction-related aspects;

(2) identification of the factors that affect construction labor pro-

ductivity and its variations, to better understand how to improve

the productivity of the construction workforce; (3) development of

practical strategies and best practices, to enhance the productivity

on construction sites as well as in the construction industry; and

(4) examination of the effect of different methods, technologies,

and constrains on productivity enhancements, efficiency, and ef-

fectiveness in different construction markets and countries. The

previous literature on construction productivity includes qualita-

tive, quantitative, and mixed-methods approaches (Kisi et al. 2018)

that address the aforementioned diverse aspects of construction

productivity.

Knowledge Gap

This subsection identifies the knowledge gap to provide adequate

rationale to support the study undertaken. Previous research studies

mainly focused on studying labor productivity in the construction

industry. Although previous research works provided important

knowledge on labor productivity, no previous research work, to

the best knowledge of the authors, attempted to study the causalities

and relationships between the dynamic workforce and workplace

variables and the productivity of the industry as a whole. Conse-

quently, there is a knowledge gap in the literature in terms of the

quantification of the impacts of different workforce and workplace

variables on the productivity of the entire construction industry.

This paper proposes a new view of construction productivity

by approaching it from the perspective of the entire construction

output generated by the US construction industry with respect to

total construction employment rather than the individual labor pro-

ductivity of each construction worker or construction workforce

occupation (such as electricians, carpenters, and plumbers, among

others). Although the construction industry produces a large

amount of data on a periodic basis, existing data sets have not been

exploited fully (Cao and Goh 2019), and there is always a need for

more advanced and holistic modeling techniques (Assaad et al.

2020b). To this end, this paper addresses this critical knowledge

gap by developing a statistical framework that can model the

causalities and the relationships between different dynamic varia-

bles and construction productivity. Consequently, this paper takes

the previous research directions a step further by rigorously inves-

tigating the relationship between dynamic workforce and work-

place variables and the overall productivity of the construction

industry.

Determination of Workforce and Workplace Variables

Different workforce and workplace variables related to construc-

tion productivity are mentioned in the literature. The expression

workforce and workplace refers to the variables related to the

construction labor force and the workplace’s working conditions.

This section identifies and describes these variables. Durdyev et al.

(2018) considered turnover an influencing factor on construction

productivity. Turnover refers to the number of construction work-

ers who leave a construction company. Koch (2017) considered

job openings to play a role in construction productivity and hiring

practices. Job openings refers to the total number of open job

vacancies that need to be filled by a construction worker. Rojas

and Aramvareekul (2003) considered construction employment to

investigate whether construction labor productivity is declining

and compared it with manufacturing labor productivity. Con-

struction employment refers to the total number of construction

workers in the construction industry. Sveikauskas et al. (2016)

considered average weekly hours worked in the construction in-

dustry to measure productivity growth. Average weekly hours re-

fers to the average hours per construction worker for which pay

was received. Ozturk et al. (2020) considered job losses to cause

changes in productivity. Job losses refers to the total number of

jobs lost by construction workers in the construction industry.

Sveikauskas et al. (2018) considered the value of construction put

in place (VIP) to assess productivity in the construction industry

because this measure provides good information about the total

construction output. VIP is a measure of the value of construction

work installed or erected at construction sites. Vereen (2013) con-

sidered job gains to affect labor demand in the US construction

industry, which in turn affects productivity. Job gains refers to the

total number of jobs gained by construction workers in the con-

struction industry. Also, the construction industry has been noted

for its high incident rates and poor safety performance (Abdul

Nabi et al. 2020b). In relation to that, Abrey and Smallwood

(2014) considered injuries and illnesses as unsatisfactory working

conditions that impact productivity in the construction industry.

Injuries and illnesses refers to the total number of nonfatal work-

place injuries and illnesses that construction workers experience.

Mirhadi (2018) considered hires to affect the efficiency of the

construction workforce, and thus also construction productivity.

Hires refers to the total number of employees that were hired in

the construction industry. Setiani and Abd Majid (2019) consid-

ered fatalities and safety to be a factor influencing productivity

in construction projects. Fatalities refers to the total number of

workplace accidents that led to the death of construction workers.

Kuznetsova et al. (2019) considered that there is a relationship

between productivity and unemployment. Unemployment is the

share of the construction labor force that is jobless. Hendrickson

(2005) considered changes in average hourly earnings to affect

the productivity measure in the construction industry. Average

hourly earnings refers to the average dollar amount that construc-

tion workers earn per hour. In addition, the size and contribution

of the construction industry to the economy usually is assessed as

a percentage of gross domestic product (GDP) (Mahamid 2013).

Higher GDP could reflect better labor well-being, which could

result in a better productivity. GDP is the monetary value of all

finished goods and services made within the US. Allmon et al.

(2000) considered total compensation rates as a factor that affects

construction productivity. Total compensation is the cost paid by

the construction employer for construction employee compensa-

tion per hour worked.

The previous workforce and workplace variables were consid-

ered for analysis in this paper. These variables were used because

they were mentioned in the literature and because they are available

from reliable sources such as the US Bureau of Labor Statistics

(BLS), the US Census Bureau, and the US Bureau of Economic

Analysis. These reliable sources provide consistent, periodically

updated, and well-maintained information and records on these

different variables.

J. Manage. Eng., 2021, 37(1): 04020092

Time-Series Analysis and Vector Autoregression

This paper studied and modeled the impacts of different workforce

and workplace variables on the productivity of the construction in-

dustry. Because the collected data for this paper were time series

showing the fluctuations of the workforce and workplace variables

with time, the collected data were multivariate time series. Vector

autoregression (VAR) is the most commonly used statistical tech-

nique for modeling and predicting multivariate time series (Singh

2018). Therefore, time series analysis and VAR were deemed to be

feasible and suitable for the paper’s research objectives and data set.

This subsection provides the needed background information on

time series analysis and VAR.

Overview

Time series represent the changes in the values of dynamic varia-

bles over time. In other words, time series are the simplest form of

temporal data, and are a sequence of real numbers collected regu-

larly in time (Gunopulos et al. 2001). Studies utilizing time-series

analysis are growing at a very fast rate due to its wide applications

in a large variety of research fields (Gao et al. 2017). Time-series

analysis is a useful tool for better understanding the cause-and-

effect relationships between different dynamic variables (Kadilar

and Kadilar 2017). In simple terms, time-series analysis is a stat-

istical method to model and predict the future values of a variable

based on previously observed values of the same variable or other

relevant dynamic variables (Yang and Liu 2019). Time-series stat-

istical analysis methods are divided into two main types: univariate

time-series analysis, and multivariate time-series analysis. Univari-

ate time-series analysis includes studying a single sequence of

values for a particular variable. Multivariate time series are more

common in real life and real-world applications due to the inherent

complexities of dealing with two or more dynamic variables (Wang

et al. 2017). Although many methods exist to analyze/predict time-

series data, such as moving average, exponential smoothing, and

autoregressive moving average, among others (Brownlee 2018),

VAR is the most commonly used statistical technique for modeling

and predicting multivariate time series (Singh 2018). VAR is a stat-

istical modeling technique that expresses the terms in a multivariate

time series of Kdynamic variables as a linear combination of the

previous pvalues of the Kvariables (Abdu-Aguye and Gomaa

2018).

Previous Studies

Due to the numerous benefits of VAR, such as good forecasting

capabilities, simplicity of implementation, and ease of estima-

tion (Anggraeni 2016), it has been widely used in a variety of

applications (Wang and Ding 2018) such as economic, finance,

construction management, and engineering, among others. In rela-

tion to previous studies that used time series and temporal data

analysis in the construction engineering and management area,

Lingard et al. (2017) examined the temporal relationship between

the safety performance indicators to uncover time-dependent causal

relationships. Cao and Goh (2019) used time-series analysis to

identify the leading indicators or predictors of construction acci-

dents, and they developed three different time-series models for

predicting accidents’occurrences. Xu and Lin (2016) used VAR

to analyze the influencing factors leading to changes in carbon

dioxide emissions in the construction industry to develop appropri-

ate energy policy and planning for the iron and steel industry.

Faghih and Kashani (2018) relied on time-series analysis to fore-

cast the short- and long-term prices of construction materials based

on a set of relevant explanatory variables. Lee et al. (2016) developed

a vector error correction model to perform an empirical analysis of

the impact of diversification on construction companies’insolvency.

Ilbeigi et al. (2017) used time-series analysis to forecast the asphalt-

cement price and examine whether and how time-series forecasting

models can predict future prices of asphalt-cement with higher ac-

curacy than the existing approaches. Lin et al. (2018) utilized time-

series to find correlations between intellectual capital and business

performance to enable corporations to shape policy decisions that

benefit business performance. Swei et al. (2017) employed univariate

time-series models to project future costs and prices of concrete and

asphalt based on a probabilistic approach. Vereen (2013) developed

a VAR model to forecast the labor demand and found that 5.3–

6.3 million skilled workers will be in demand by 2022. Because the

construction industry is a project-based industry, with projects taking

several months and years to complete, the monthly productivity of

the construction industry generally is considered to be dependent on

its previous lagged values. Therefore, time series and VAR were

employed in this paper to model and predict the productivity of the

construction industry.

Methodology

As shown in Fig. 1, the authors followed a methodology composed

of different steps (Fig. 1). Details on each one of these steps are

provided in the next subsections.

Data Collection and Description

The collected data included the following 14 dynamic workforce-

and workplace-related variables (mentioned in the section “Back-

ground Information”): VIP, total construction employment, average

weekly hours, GDP, job openings, job hires, turnover or job

separations, total compensation [includes wages and salaries, and

total benefits (insurance and retirement)], gross job gains, gross

job losses, average hourly earnings, work-related fatalities, work-

related occupational injuries and illnesses, and unemployment rate.

The VIP data were automatically collected from the US Census

Bureau through Python’s Scrapy package version 1.8.0, which is

used for data scraping. Data scraping is “an automated process of

compiling data from multiple Web pages in a systematic way”

(Bonifacio et al. 2015). The data for total construction employment,

average weekly hours, total compensation, unemployment rate, job

openings, job hires, turnover, gross job gains, gross job losses,

average hourly earnings, fatalities, and occupational injuries and

illnesses were collected from the US BLS. GDP was collected from

the Federal Reserve Economic Data (FRED) at the Federal Reserve

Bank of St. Louis (data from US Bureau of Economic Analysis).

All collected data were converted to a monthly time-step by inter-

polation and the addition of variability or noise whenever needed.

Table 1summarizes the variables used in this paper and their as-

sociated sources.

According to Sveikauskas et al. (2018), productivity can be cal-

culated as the ratio of total construction output to total hours

worked. The productivity of the construction industry, which is the

response variable in this paper, was obtained from the collected

data using Eq. (1)

Construction productivity ¼Total value of construction put in place ð$=monthÞ

Total construction empoyment ðemployee=monthÞ×total average weekly hours ðhours=weeksÞ×4ðweekÞ

ð1Þ

J. Manage. Eng., 2021, 37(1): 04020092

Three of the collected variables were used to calculate the pro-

ductivity of the construction industry: VIP, total construction em-

ployment, and average weekly hours. Therefore, the final set of

the variables used in the statistical analysis in this paper included

the calculated productivity of the construction industry using

Eq. (1) and the other 11 dynamic workforce and workplace vari-

ables. Although these variables have different scales of measure-

ment, the statistical causalities and relationships between them

can be examined. For instance, McGowan (2019) found that a

1% increase in the labor unemployment rate decreases construction

value by $12 billion. Also, Shahandashti and Ashuri (2016) used

time-series analysis and vector correction models to predict the

national highway construction cost index (NHCCI) based on many

potential leading indicators/variables such as total employment in

the construction industry, average weekly hours, building permits,

housing starts, and unemployment rate, among others; these vari-

ables have different scales of measurement.

A study period from 2006 to 2019, inclusive, was selected for

the collected data/variables in this paper. This range was selected

for the study period because the variable average weekly hours was

not recorded before 2006. Because this variable was used to cal-

culate the output variable total productivity of the construction in-

dustry [Eq. (1)], no values for the model’s output variable could be

obtained before 2006. This means that no model could be devel-

oped before the 2006 year. In addition, the variable average hourly

earnings was not recoded before 2006. Furthermore, at the time this

study was performed, no data were available for the entire year

2020 because this year had not concluded. Therefore, a 2006–2019

study period was chosen in this research. Henceforth, construction

productivity refers to the productivity of the construction industry

as a whole.

Statistical Analysis and Modeling

The main statistical methods used in this research were cointegra-

tion testing, unit root testing, Granger causality testing, and VAR.

Cointegration Relationship Testing

The cointegration between variables should be examined to study the

significant relationships between the different variables. The purpose

of the cointegration test is to study the long-term relationships

Fig. 1. Research methodology.

Table 1. Variables and their sources

Variable Source

Total compensation US Bureau of Labor Statistics

Average hourly earnings US Bureau of Labor Statistics

Unemployment US Bureau of Labor Statistics

Fatalities US Bureau of Labor Statistics

GDP US Bureau of Economic Analysis

Hires US Bureau of Labor Statistics

Occupational injuries and illnesses US Bureau of Labor Statistics

Gross job gains US Bureau of Labor Statistics

Gross job losses US Bureau of Labor Statistics

Job openings US Bureau of Labor Statistics

Turnover US Bureau of Labor Statistics

Total construction employment US Bureau of Labor Statistics

Value of construction put in place US Census Bureau

Average weekly hours US Bureau of Labor Statistics

Construction productivityaUS Bureau of Labor Statistics and

US Census Bureau

aCalculated using Eq. (1).

J. Manage. Eng., 2021, 37(1): 04020092

between the variables (Liu et al. 2019). The long-term relationship is

investigated by the cointegration test, which identifies whether two

or more time series are integrated together in a way that they cannot

deviate from equilibrium in the long term (CFI 2020). A long-term

relationship exists between variables when they share a similar trend

by being associated together over time rather than being correlated at

a specific or instantaneous time. In simple terms, even if the variables

deviate from each other in the short-term, they tend to return to the

trend in the long-term if they are cointegrated (Wei 201 6). Erica

(2020) provides more details on the cointegration test and the

long-term relationship between variables. In addition, the cointegra-

tion test analyzes whether there is a stable combination of the differ-

ent time-series data (Lee et al. 2019). This paper uses the Johansen’s

cointegration test due to its wide use (Mahadevan and Asafu-Adjaye

2007). When two or more time series are cointegrated, this indicates

they have a long-term statistically significant relationship. The null

hypothesis H0of the Johansen test is that there is no cointegration

between the variables (Elliott and Pesavento 2009). The trace statistic

is calculated for the Johansen cointegration test, and it is compared

with the critical trace statistic value at 95% (Rachev et al. 2007). The

trace test rejects the null hypothesis if the trace statistic exceeds the

critical value. In other words, if the null hypothesis is rejected, there

is a statistically significant relationship between the corresponding

two variables.

Unit Root Test

It is important to check the stationarity of the time series before

constructing the VAR model and before conducting the Granger

causality. In other words, it is critical to identify whether the var-

iables are stationary before developing the VAR model, because the

model can be applied only to stationary time series (Shahandashti

and Ashuri 2016). Stationary time series are those in which the

means, variances, and autocorrelation structures do not change over

time (Faghih and Kashani 2018). Stationarity can be assessed using

a unit root test (Lee et al. 2019). If the time-series data are not

stationary, they should be made stationary before developing the

VAR model by differencing, which eliminates trends’changes

by subtracting an earlier value from a later value. This research

implemented the widely used augmented Dickey–Fuller (ADF) test

(Dickey and Fuller 1979). The formulation of the ADF test is

shown in Eq. (2)

Δyt¼αþβtþγyt−1þX

p−1

i¼1

δiΔyt−iþutð2Þ

where t= time index; yt= time-series value corresponding to time t;

Δyt= lagged first differences (that is, yt−yt−1); α= intercept con-

stant (drift term); β= coefficient of time trend; and δcoefficient to

test if the data need to be differentiated to make them stationary.

The null hypothesis H0of the ADF test is that the time-series

data have a unit root; that is, they are nonstationary. The ADF test

generates a p-value that should be compared with a significance

level of 0.05 to determine whether the null hypothesis is rejected.

Differencing should be applied to all time series until all of them

become stationary. The order of integration represents the number

of differencings required to make a nonstationary time series sta-

tionary. Once all the time series are rendered stationary, the Granger

causality test can be applied to study the causalities between the

different variables and determine the explanatory variables causing

fluctuations in the response variable (Abediniangerabi et al. 2017).

Determination of Relevant Variables using Granger

Causality Testing

Before proceeding with developing a VAR model, it is quite

common to conduct the Granger causality test developed by

Granger (1969). The Granger causality test is a statistical technique

used in multivariate time-series analysis to examine whether the

lagged values of one variable helps to predict another variable

(Swei 2020). In other words, the Granger causality test is used

to identify the leading indicators or the relevant variables that affect

the response variable (Shiha et al. 2020). In simple terms, if a multi-

variate time series comprises two variables y1tand y2t, then Eq. (3)

is used to investigate the one lagged causality between these two

variables

Fðy1t;y2tÞ¼a1×y1t−1þa2×y2t−1þbð3Þ

The null hypothesis H0of the Granger test is that there is no

causality between y1tand y2t, that is, a2¼0[Eq. (3)]. In other

words, the null hypothesis states that y2tdoes not Granger-cause

y1t. Failing to reject H0means that y2tdoes not Granger-cause

y1t; therefore, there is no causality between them. In this case, the

variable y2tshould not be considered in the set of variables used to

develop the VAR statistical model. For a lag order p, the general

statistical formulation of the Granger causality test for two variables

can be represented by Eq. (4)(Lütkepohl et al. 2004)

y1t

y2t¼X

i¼1α11iα12i

α21iα22iy1t−1

y2t−1þutð4Þ

where i¼1;2;:::;p; and α12i¼0if y2tdoes not Granger-

cause y1t.

The Granger causality test was conducted to establish the cau-

salities between the response variable productivity of the construc-

tion industry and the other 11 potentially related variables with a

significance level of 0.05.

Vector Autoregression

The lag order pfor the VAR model can be iteratively determined by

fitting increasing orders of the VAR model and choosing the lag

order pthat gives a model with the lowest Bayesian information

criterion (BIC) (Swei et al. 2017;Ayhan and Tokdemir 2020).

When the lag order pis determined, the VAR statistical model is

fitted to quantitatively examine the relationships between the differ-

ent time series.

For two time series y1tand y2t, the associated VAR statistical

model for a lag order of 2 is shown in Eq. (5)

Fðy1t;y2tÞ¼a11 ×y1t−1þa12 ×y1t−2þa21 ×y2t−1

þa22 ×y2t−2þbð5Þ

If the VAR model is applicable to multivariate time series with K

variables and a lag order of p, the general representation of a VAR

model is shown in Eq. (6)(Lütkepohl et al. 2004)

Yt¼A1Yt−1þ··· þApYt−pþUtð6Þ

where Yt¼ðy1t; :::;yKtÞis a set of Ktime-series variables; Ap¼

ðkxkÞcoefficient matrix; and Ut= set of unobservable error terms

(Lütkepohl et al. 2004).

After fitting the VAR model based on the differenced time-series

data, the examination of the serial correlation of residuals (errors)

should be conducted (Prabhakaran 2019). Serial correlation is

important to check if there is any leftover pattern in the obtained

residuals (errors) from the VAR model. Checking for serial corre-

lation ensures that the fitted VAR model is sufficiently able to

explain the variances and patterns in the time series (Prabhakaran

2019). The Durbin–Watson (DW) statistic was calculated using

Eq. (7) to examine the serial correlation of residuals. Values of DW

J. Manage. Eng., 2021, 37(1): 04020092

test statistic in the range 1–3 are relatively normal, and any value

outside this range is a cause for concern (Field 2013)

DW ¼PT

t¼2ðet−et−1Þ2

t¼1e2

ð7Þ

where DW = Durbin–Watson statistic; and et= residuals at time t.

After fitting the VAR model, it was tested and validated on the

4-year period 2016–2019 inclusive. This 4-year period was selected

because it constituted about 30% (4 years divided by 14 years) of

the entire 14-year study period from 2006 to 2019, inclusive. This

is a recommended percentage for the testing data (Shahin et al.

2004) because it keeps aside enough data to give a good indication

of the model’s performance on unseen data. Because the VAR

model was fitted on the stationary differenced data series, the final

predictions were obtained by inverting (that is, integrating) the pre-

dicted differenced data series. The prediction accuracy of the fitted

VAR statistical model was determined based on the mean absolute

percentage error (MAPE) shown in Eq. (8)

MAPE ¼100

t¼1



yt−^

yt



ð8Þ

where MAPE = mean absolute percentage error; n= number of

observations during prediction period; yt= actual value; and ^

yt=

predicted value using developed VAR model.

The Python programming language was used in this paper for

data management, statistical analysis, and visualization. The fol-

lowing libraries were used: pandas version 1.0.0, which generally

is used for data manipulation and management of numerical tables

and data structures; numpy version 1.18.1, which generally is used

to add support for large and multidimensional arrays and matrices

and to provide high-level mathematical functions; matplotlib

version 3.1.2, which generally is used for data plotting and visu-

alization; and statsmodels version 0.10.2, which generally is

used to conduct statistical tests, statistical data exploration, and es-

timation of different statistical models.

Results and Analysis

Statistical Examination

Fig. 2shows the collected time series data with monthly time

interval. After collecting and visualizing the data, the cointegration

Fig. 2. Monthly time-series data.

J. Manage. Eng., 2021, 37(1): 04020092

relationship between the variables was examined using the

Johansen’s test statistic, and the obtained results are given in

Table 2.

There were significant relationships between the different var-

iables, because the obtained test statistic was greater than the criti-

cal test statistic at 95% for all variables (Table 2). Hence, the null

hypothesis H0that there is no cointegration between the variables

was rejected for all variables considered in this paper.

After examining the cointegration between the different varia-

bles, the stationarity of the data series was investigated. The ob-

tained results of the unit root test are listed in Table 3.

Only four of the original collected multivariate time-series data

were stationary (Table 3, Column 2): unemployment, fatalities,

occupational injuries and illness, and gross job gains. Therefore,

the data should be differenced so that all time series are stationary

before constructing the VAR model. After the first differencing

(Table 3, Column 3), six variables were stationary: construction

productivity, total compensation, hires, gross job losses, job open-

ings, and turnover. Because not all time series were differenced,

a second differencing of the data was needed. After the second dif-

ferencing of the data, all the time series were rendered stationary

(Table 3, column 4). Therefore, the VAR model could be developed

on the differenced stationary time-series data.

Before building the VAR statistical model, the Granger causality

test was conducted. The obtained results are listed in Table 4. The

null hypothesis was rejected for all variables except job openings.

In other words, all variables, except job openings, were considered

to Granger-cause the variable productivity of the construction in-

dustry. That is, the obtained p-value for the job openings variable

was higher than the 0.05 significant level, which indicates strong

evidence for the null hypothesis (Table 4). Because the null hypoth-

esis means that the input variable does not Granger-cause the out-

put variable (Shahandashti and Ashuri 2016), it can be concluded

that the job openings variable is the only variable that does not

Granger-cause the variable productivity of the construction indus-

try. Therefore, the variable job openings was not included in the

VAR statistical model developed for modeling and predicting the

productivity of the construction industry.

Vector Autoregression Model and Prediction

The VAR statistical framework was developed between the re-

sponse variable productivity of the construction industry, its lagged

values, and the following dynamic workforce and workplace var-

iables determined by the Granger causality test: total compensation,

average hourly earnings, unemployment, fatalities, GDP, hires, oc-

cupational injuries and illnesses, gross job gains, gross job losses,

and turnover. The VAR model was fitted on the stationary differ-

enced data series. The lag order pfor the VAR model was deter-

mined by fitting increasing orders of VAR models and choosing the

lag order pthat gave the lowest BIC. The obtained results for the

different lag orders and the associated values of BIC are listed in

Table 5.

The minimum BIC corresponded to a lag order pof 2 (Table 5).

Therefore, a lag order of 2 was selected for developing the VAR

statistical model for the construction productivity variable. A VAR

model was fitted using the collected data and the obtained results

are listed in Table 6.

Table 3. Results of stationary test

Variable

p-value

for no

differencing

p-value

for first

differencing

p-value

for second

differencing

Construction productivity 0.8011 <0.0001a<0.0001a

Total compensation 0.9661 0.0001a<0.0001a

Average hourly earnings 0.9893 0.0634 <0.0001a

Unemployment 0.0389a0.3958 <0.0001a

Fatalities 0.0047a0.4905 <0.0001a

GDP 0.9916 0.2423 0.0003a

Hires 0.5471 0.0273a<0.0001a

Occupational injuries

and illnesses

0.0015a0.5494 <0.0001a

Gross job gains 0.0024a0.0975 <0.0001a

Gross job losses 0.5229 0.0037a<0.0001a

Job openings 0.8779 0.0118a<0.0001a

Turnover 0.6412 <0.0001a<0.0001a

ap-value < 0.05 significance, meaning that the null hypothesis is rejected;

therefore, the series is stationary.

Table 2. Results of cointegration test

Variable

Obtained

test

statistic

Critical

test statistic

at 95%

Significant

relationship?a

Construction productivity 594.52 311.13 True

Total compensation 472.56 263.26 True

Average hourly earnings 376.89 219.41 True

Unemployment 285.20 179.52 True

Fatalities 222.36 143.67 True

GDP 164.36 111.78 True

Hires 118.82 83.94 True

Occupational injuries

and illnesses

88.62 60.06 True

Gross job gains 58.82 40.17 True

Gross job losses 38.57 24.28 True

Job openings 20.83 12.32 True

Turnover 6.84 4.13 True

aIf the obtained test statistic is greater than the critical test statistic at

95%, then there is a significant relationship (True); otherwise, there is no

significant relationship (False).

Table 4. Results of Granger causality test between productivity of con-

struction industry and other dynamic workforce and workplace variables

Null hypothesis (H0)p-value

Total compensation does not Granger-cause construction

productivity

0.0064a

Average hourly earnings do not Granger-cause

construction productivity

0.0173a

Unemployment does not Granger-cause construction

productivity

0.0440a

Fatalities do not Granger-cause construction productivity 0.0323a

GDP does not Granger-cause construction productivity 0.0139a

Hires do not Granger-cause construction productivity 0.0121a

Occupational injuries and illnesses do not Granger-cause

construction productivity

0.0484a

Gross job gains do not Granger-cause construction

productivity

0.0264a

Gross job losses do not Granger-cause construction

productivity

0.0408a

Job openings do not Granger-cause construction

productivity

0.1246

Turnover does not Granger-cause construction

productivity

0.0238a

ap-value < 0.05 significance, meaning that the null hypothesis is rejected;

therefore, there is causality with the response variable productivity of the

construction industry.

J. Manage. Eng., 2021, 37(1): 04020092

Table 6gives the VAR statistical/mathematical model developed

in this paper. The VAR model was developed by obtaining the co-

efficient associated with each lagged variable. For example, the

coefficient for the one-lagged differenced productivity variable was

−0.8137, the coefficient for the one-lagged differenced total com-

pensation variable was 21.1465, and so on (Table 6). The highest

coefficient in absolute value was that related to the one-lagged dif-

ferenced average hourly earnings variable, which means that this

lagged variable contributed the most to the fluctuations in construc-

tion productivity. The lowest coefficient in absolute value was that

related to the two-lagged differenced hires variable, which indicates

that this lagged variable contributed the least to the fluctuations in

construction productivity.

The serial correlation of the errors was checked using DW

statistic. This is important to examine if there is any leftover

pattern in the residuals (errors) obtained from the VAR model.

The DW statistic obtained was 2.12, which is between 1 and 3;

therefore, the developed VAR model is sufficiently able to explain

the variances and patterns in the time series (Prabhakaran 2019;

Field 2013).

Consequently, the developed VAR model was tested and vali-

dated by predicting the construction productivity between 2016 and

2019. The predictions were generated after inverting the predicted

differenced data series. To assess and validate the developed VAR

statistical model, the MAPE was calculated to evaluate the model’s

predictions. Fig. 3shows the actual and predicted values for the

productivity of the construction industry using the developed VAR

model. The obtained associated residuals are shown in Fig. 4.

The developed VAR statistical model for the productivity of the

construction industry had a MAPE of 5.13% (Fig. 3). In addition,

the residuals indicate that the developed VAR model slightly under-

estimated the productivity between 2016 and the end of 2018, and it

slightly overestimated the productivity for the period from the end

of 2018 to the end of 2019 (Fig. 4). Nevertheless, the overall ob-

tained residuals are considered acceptable because they are within a

satisfactory range of error. The obtained MAPE of 5.13% is accept-

able because it is less than 10%, which is the generally accepted

Table 5. BIC results for different VAR lag orders

Lag order pBIC

0 38.61

1 36.18

2 32.71a

3 34.71

4 36.47

5 37.45

6 39.61

7 41.49

8 43.6

9 44.08

10 44.41

11 42.65

12 39.19

aLowest BIC; therefore, a lag order of 2 was selected for the VAR model.

Table 6. Fitted VAR statistical model for construction productivity

Lagged variable Coefficient

L1. Productivity −0.8137

L1. Total compensation 21.1465

L1. Average hourly earnings −30.8967

L1. Unemployment −0.1639

L1. Fatalities 0.1980

L1. GDP −0.0428

L1. Hires 0.0105

L1. Occupational injuries and illnesses −0.1234

L1. Gross job gains 0.3613

L1. Gross job losses −0.0802

L1. Turnover 0.0079

L2. Productivity −0.2874

L2. Total compensation −6.2199

L2. Average hourly earnings −11.0521

L2. Unemployment −1.1089

L2. Fatalities 0.1471

L2. GDP −0.0182

L2. Hires 0.0027

L2. Occupational injuries and illnesses −0.1431

L2. Gross job gains −0.2235

L2. Gross job losses 0.2468

L2. Turnover 0.0135

Constant 0.4081

Note: Coefficients were obtained for the differenced (with an order of 2)

data series, and thus they reflect more the nature of change (decrease or

increase) in the productivity rather than the direct impact on the value of

the construction productivity itself.

Fig. 3. Prediction of the construction productivity using the fitted VAR

statistical model.

Fig. 4. Obtained prediction residuals for the developed VAR model.

J. Manage. Eng., 2021, 37(1): 04020092

MAPE for a robust prediction model (Fan et al. 2010). Hence, the

developed VAR statistical model is perceived to be a good frame-

work that can be used to model and predict the productivity of the

entire construction industry.

Analysis of Findings

Cointegration occurs naturally in economics and finance applica-

tions in which there are relationships between different variables.

For example, in the well-known permanent income model, there

is cointegration between consumption and income. Nevertheless,

cointegration might not be present naturally in engineering and

management applications. However, the findings of the cointegra-

tion test conducted in this paper indicate that there were equilibrium

relationships between the construction productivity variable and the

other dynamic workforce and workplace variables considered in

this paper. This shows that the collected time-series data sets were

linked to each other. Specifically, the paper investigated the long-

term equilibrium relationships between the variables because dif-

ferent forces, which act in response to deviations from equilibrium,

may take a long time to restore equilibrium (Zivot and Wang 2007).

This concludes that there could be a stationary linear combination

between construction productivity and the different workforce and

workplace variables, which cannot vary too much from an equilib-

rium (Johansen 2009,Dickey 2007).

The findings of the unit root test indicated that when the col-

lected times-series data were differenced twice, all the time-series

variables became stationary. This means that there was no stochas-

tic trend in the differenced time series, which often is referred to as

random walk with drift (Shifera 2019). Therefore, the differenced

construction productivity variable and the dynamic workforce and

workplace variables did not possess a unit root, and thus they did

not have a pattern that is unpredictable (Sodiq Olawale 2019).

According to Glen (2016b), this is a desirable property because the

existence of a unit root can create serious issues, including spurious

regressions (a high coefficient of determination even if the data

are uncorrelated) and errant behavior or invalid inferences (due

to assumptions for analysis not being valid).

The findings of the Granger test reflected on the structures of

the causal relationships between the different variables and the re-

sponse variable construction productivity. Specifically, the findings

highlighted that the majority of the variables (10 of the 11 dynamic

workforce and workplace variables) are useful for predicting con-

struction productivity. Therefore, it is inferred that the fluctuations

in construction productivity are associated with corresponding

changes in the 10 predictor variables, and hence have a robust pat-

tern of Granger causality. In other words, construction productivity

is affected by 10 of the dynamic workforce- and workplace-related

variables considered in this paper. All obtained results justified

the rationale behind conducting this research and have statistically

proved the concept of productivity for the entire construction

industry.

Nevertheless, because the focus of this study was to understand

the impacts of the dynamic workforce and workplace variables on

construction productivity, the individual impacts of the dynamic

variables were quantified by developing a VAR statistical model.

The obtained findings from the VAR model provided insights on

the particular relationships between the variables. More specifi-

cally, the dynamic nature of the variables was determined by inves-

tigating different orders for the leads and lags in the time series

data. It was found that a lag order of 2 best modeled the fluctuations

in construction productivity based on the changes in the 10 predic-

tor variables. This indicates that the changes in the workforce- and

workplace-related variables for two periods or steps have a relevant

impact on the construction productivity. These variables were con-

sidered dynamic because their values change from one period to the

other, and their values are linked to other variables depending on

time (Abbas and Mosallamy 2016). Because it is difficult to inter-

pret the large number of coefficients in a VAR model (Stanslaus

2017), the authors did not focus on the individual interpretation of

the obtained coefficients, however the coefficients could still be of

value to better understand the nature of the impact of each dynamic

variable on the construction productivity. To check if the developed

VAR model is sufficiently able to explain the variances and patterns

in the time series, the serial correlation of the error terms or resid-

uals was investigated using the DW statistic. Serial correlation oc-

curs when the error terms of a time series are transferred from one

period to another; that is, when the error in a period is correlated

with the error in a subsequent period (Glen 2016a). For example, an

overestimate of the construction productivity for one month would

result in an overestimate of the construction productivity for the

subsequent month. This would lead to biases, and thus serial cor-

relation of error terms is not desired because it can lead to myriad

problems, including but not limited to inefficient estimation of the

coefficients, underestimation of the error variance, and underesti-

mation of the variance of the coefficients, among many other issues

(NCSS 2006). The findings of the paper indicate that the obtained

DW statistic was 2.12, which is a desirable value because it lies

between 1 and 3 (Field 2013). Thus, the findings concluded that the

developed VAR statistical framework is sufficiently able to explain

the variances and patterns in the time series. Finally, the prediction

accuracy of the developed VAR model was acceptable, with a

MAPE of 5.13%.

Discussion and Limitations

Although labor productivity is not equal to the productivity of the

construction industry, the obtained results could still be examined

by comparing them with those of previous studies related to labor

productivity. This is because labor productivity is believed to influ-

ence the productivity of the entire industry since the labor force is

performing the construction work and generating the total construc-

tion output.

This paper showed that all considered workforce and work-

place variables have statistical causalities with the construction

industry productivity, except for job openings. These results could

be positioned with respect to previous studies. For instance, total

compensation is one of the most important workforce variables

that construction owners, contractors, and unions track for differ-

ent reasons, such as project planning, risk management, and esti-

mating (CLMA 2018). In addition, Brill et al. (2017) established

that productivity and compensation are correlated. Moreover,

Hendrickson (2005) considered changes in average hourly earnings

to affect the productivity measure in the construction industry. This

could be due to the fact that the construction productivity increases

when workers are monetarily motivated by good total compensa-

tion (including wages, salaries, and benefits such as insurance and

retirement) or high hourly earnings. Additionally, Ayodele et al.

(2020) stated that high labor turnover or job separations can affect

the construction productivity and the business performance of con-

struction firms. This could be attributed to the deterioration of

the learning curve due to the replacement workers and the loss of

knowledge and skill (Allen et al. 2010;Parrotta and Pozzoli 2012).

Moreover, McGowan (2019) argued that the unemployment rate

affects the value of construction put in place, and in turn, construc-

tion productivity. Furthermore, workplace fatalities and occupa-

tional injuries substantially affect the productivity on construction

J. Manage. Eng., 2021, 37(1): 04020092

sites because construction projects might be closed, slowed, or sus-

pended due to such inefficiencies (Walter 2008). In fact, the con-

struction industry has poor safety performance and high incident

rates compared with other industries (Abdul Nabi et al. 2020a).

In addition, job losses and job gains can cause changes in produc-

tivity (Ozturk et al. 2020;Vereen 2013). This can be explained by a

fundamental principle in productivity theory in which productivity

is expected to be maximized when there is neither overstaffing

(more workers than needed) nor understaffing (fewer workers than

needed). Moreover, the number of hires in the construction industry

can affect the efficiency of the construction workforce (Mirhadi

2018), and in turn the construction productivity. In fact, to over-

come the difficulties related to construction productivity and to find

enough experienced construction workers, construction companies

try to hire more people to match the current and expected construc-

tion demand (AGC 2018;Semler 2020). On the other hand, one

reason for the absence of causality between job openings and pro-

ductivity can be that, although construction jobs are available, there

could be a shortage of labor to fill the open positions. Thus, job

openings do not necessary have a significant impact on productivity

since labor was not necessarily employed for the available posi-

tions. Staffing difficulties are very common in the construction in-

dustry, and they have become more pronounced in recent years

because the construction industry failed to seize the opportunity

to replenish its supply of skilled laborers (CLMA 2013).

One of the limitations of this research is the unidirectional re-

lationship between the different dynamic workforce and workplace

variables on the one hand and the productivity of the construction

industry on the other hand. That is, this research was limited to

studying how the response variable (construction productivity)

was influenced by the predictor variables (workforce and work-

place variables), but not vice versa. Another limitation was the pa-

per’s focus on the workforce and workplace variables that can affect

construction productivity rather than other industry-, government-,

or economic-related variables. Future research work could try to

address this limitation by studying the effects of other variables

on construction productivity. Moreover, this paper was limited

to be a proof-of-concept for a new notion of gross construction pro-

ductivity, and thus it acts as a step toward the development of a new

gross construction productivity indicator.

Theoretical and Practical Implications

From the theoretical perspective, this paper proposes a new theo-

retical concept of gross construction productivity. The theoretical

implications of this research include the opportunities that this pa-

per could generate in relation to studying the health of the construc-

tion industry based on the productivity of the industry as a whole

rather than based on labor productivity alone. Therefore, this re-

search should help in moving toward the development of a theo-

retical notion of gross construction productivity. This notion was

inspired by the use of the gross domestic product to study and

evaluate the health of economies and countries, as well as by the

use of the word ‘gross’by the US Bureau of Labor Statistics for

other variables such as gross job gains and gross job losses in the

entire construction industry (US Bureau of Labor Statistics 2020).

The notion of gross construction productivity is expected to result

in an indicator or index that models and incorporates the relation-

ship between different dynamic variables and the broad economic

output. Many previous research efforts reflected the health or the

importance of the construction industry by examining its share of

or percentage contribution to GDP. Nevertheless, because the

construction industry is one of the largest industries nationwide,

the health of the industry is better examined in terms of gross con-

struction productivity, instead of being represented as a simple per-

centage or share of the total GDP. Thus, the theoretical implications

of this paper include motivating future scholars to delve into this

new theoretical area or notion of gross construction productivity

more deliberately.

The findings of this paper could open practical opportunities in

relation to using the productivity of the entire construction industry

to anticipate market volatility in the construction industry, as well

as to identify potential improvements in the overall productivity of

the industry. This is because construction productivity is an impor-

tant metric that provides feedback about the industry trends and

improvements (Vereen et al. 2016). This also could lead to a better

execution of construction projects, because productivity affects the

performance of projects in the construction industry (Soekiman

et al. 2011). Moreover, the practical implications of this paper are

reflected in the importance of the research to the different entities

that supply information to US construction companies. This paper

provides them with an additional and important piece of informa-

tion (i.e., the gross construction productivity) that could be used to

make different decisions. For example, the gross construction pro-

ductivity could be added as an additional piece of information to

the value of construction spending that the Information Handling

Services (IHS) closely observes to enlighten construction firms

about business risks in the construction industry (IHS 2020).

Another example is the Associated Builders and Contractors

(ABC), which can provide its members with the value of gross con-

struction productivity in addition to residential and nonresidential

construction spending, which ABC carefully tracks and reports

to its members on a monthly basis (ABC 2020). In fact, construc-

tion firms utilize information about the future trends of the overall

construction industry indicators in making their strategic business

decisions, such as market entry, consolidation (mergers and ac-

quisitions), business expansion and contraction, and staffing

(Abediniangerabi et al. 2017). Therefore, one of the practical con-

tributions of this paper is to add gross construction productivity

to these industrywide indicators. In summary, gross construction

productivity is additional valuable information that construction

companies can use to make different insightful and well-educated

industry-related decisions.

Conclusion

The paper identified a critical gap in the body of knowledge in terms

of construction productivity. Specifically, previous research works

concentrated on labor productivity, without studying whether the

productivity could be assessed for the entire construction industry.

This research examined and quantified the impacts of different

workforce and workplace variables on the productivity of the entire

construction industry. The authors collected empirical multivariate

time-series data for the period from 2006 to 2019, inclusive, for con-

struction productivity and for 11 dynamic workforce and workplace

variables: job openings, job hires, turnover or job separations, total

compensation, gross job gains, gross job losses, average hourly

earnings, fatalities, occupational injuries and illnesses, gross domes-

tic product, and unemployment rate. Causalities and statistically sig-

nificant relationships were examined. Relying on statistical test for

the collected data series, a VAR framework was developed based on

all dynamic workforce and workplace variables, except job open-

ings, to model the temporal variations in the construction produc-

tivity. The developed VAR model was validated by predicting the

construction productivity for the period 2016–2019, inclusive, with

an acceptable MAPE of 5.13%.

J. Manage. Eng., 2021, 37(1): 04020092

Based on the findings of the paper, it is concluded that (1) the

new concept of gross construction productivity is statistically jus-

tified and should be implemented in the construction industry, be-

cause the findings indicated that statistically significant long-term

relationships had a predictable (nonstochastic) statistical pattern,

and reflected statistical significant causalities between the produc-

tivity of the construction industry and different dynamic workforce

and workplace variables; (2) gross construction productivity is ad-

ditional valuable information that construction companies should

consider to make different insightful and well-educated industry-

related decisions; (3) the health of the construction industry needs

to be studied based on the productivity of the industry as a whole

rather than based on labor productivity alone; and (4) the construc-

tion industry should move toward the development of a notion of a

gross construction productivity indicator used to measure, evaluate,

and predict the performance of the entire industry. Ultimately, this

paper proposes a new indicator or index for gross construction pro-

ductivity. The outcomes of this paper add to the body of knowledge

by providing a better understanding of the impact of different dy-

namic workforce and workplace variables on the construction pro-

ductivity and by offering a new concept called gross construction

productivity.

Data Availability Statement

All data generated or analyzed during the study are included in the

published article.

Acknowledgments

The authors appreciate the comments, suggestions, and recommen-

dations provided by the anonymous reviewers, because they collec-

tively helped hone and strengthen the quality of this manuscript

during the blind peer-review process.

References

Abbas, S., and D. Mosallamy. 2016. “Determinants of FDI flows to devel-

oping countries: An empirical study on the MENA region.”J. Finance

Econ. 4 (1): 30–38.

ABC (Associated Builders and Contractors). 2020. “Construction

economic update.”Accessed May 7, 2020. https://www.abc.org

/NewsMedia/ConstructionEconomics/ConstructionEconomicUpdate

/tabid/270/categoryid/46/.

Abdu-Aguye, M. G., and W. Gomaa. 2018. “Novel approaches to activity

recognition based on vector auto regression and wavelet transforms.”

In Proc., 2018 17th IEEE Int. Conf. on Machine Learning and

Applications, 951–954. New York: IEEE.

Abdul Nabi, M., I. H. El-adaway, and C. Dagli. 2020a. “A system dynamics

model for construction safety behavior.”Procedia Comput. Sci.

168 (Jan): 249–256. https://doi.org/10.1016/j.procs.2020.02.254.

Abdul Nabi, M., I. H. El-adaway, S. Fayek, C. Howell, and J. Gambatese.

2020b. “Contractual guidelines for construction safety–related issues

under design–build standard forms of contract.”J. Constr. Eng. Man-

age. 146 (7): 04020074. https://doi.org/10.1061/(ASCE)CO.1943-7862

.0001855.

Abediniangerabi, B., S. M. Shahandashti, N. Ahmadi, and B. Ashuri. 2017.

“Empirical investigation of temporal association between architecture

billings index and construction spending using time-series methods.”

J. Constr. Eng. Manage. 143 (10): 04017080. https://doi.org/10.1061

/(ASCE)CO.1943-7862.0001391.

Abrey, M., and J. J. Smallwood. 2014. “The effects of unsatisfactory

working conditions on productivity in the construction industry.”

Procedia Eng. 85 (2014): 3–9. https://doi.org/10.1016/j.proeng.2014

.10.522.

AGC (Associated General Contractors of America). 2018. “Eighty per-

cent of contractors report difficulty finding qualified craft workers to

hire as association calls for measures to rebuild workforce.”Accessed

May 6, 2020. https://www.agc.org/news/2018/08/29/eighty-percent

-contractors-report-difficulty-finding-qualified-craft-workers-hire.

Allen, D. G., P. C. Bryant, and J. M. Vardaman. 2010. “Retaining talent:

Replacing misconceptions with evidence-based strategies.”Acad. Man-

age. Perspect. 24 (2): 48–64. https://doi.org/10.5465/amp.24.2.48.

Allmon, E., C. T. Haas, J. D. Borcherding, and P. M. Goodrum. 2000.

“US construction labor productivity trends, 1970–1998.”J. Constr.

Eng. Manage. 126 (2): 97–104. https://doi.org/10.1061/(ASCE)0733

-9364(2000)126:2(97).

Anggraeni, P. W. 2016. “Transmission mechanism of monetary policy

through asset price in Indonesia in the period 2002–2011.”J. Indone-

sian Appl. Econ. 6 (1): 123–141. https://doi.org/10.21776/ub.jiae.2016

.006.01.7.

Assaad, R., and M. A. Abdul-Malak. 2020a. “Legal perspective on treat-

ment of delay liquidated damages and penalty clauses by different

jurisdictions: Comparative analysis.”J. Leg. Aff. Dispute Resolut. Eng.

Constr. 12 (2): 04520013. https://doi.org/10.1061/(ASCE)LA.1943

-4170.0000387.

Assaad, R., and M. A. Abdul-Malak. 2020b. “Timing of liquidated

damages recovery and related liability issues.”J. Leg. Aff. Dispute

Resolut. Eng. Constr. 12 (2): 04520015. https://doi.org/10.1061

/(ASCE)LA.1943-4170.0000390.

Assaad, R., C. Dagli, and I. H. El-adaway. 2020c. “A system-of-

systems model to simulate the complex emergent behavior of vehicle

traffic on an urban transportation infrastructure network.”Procedia

Comput. Sci. 168 (Jan): 139–146. https://doi.org/10.1016/j.procs

.2020.02.277.

Assaad, R., I. El-adaway, and I. Abotaleb. 2020d. “Predicting project perfor-

mance in the construction industry.”J. Constr. Eng. Manage. 146 (5):

04020030. https://doi.org/10.1061/(ASCE)CO.1943-7862.0001797.

Assaad, R., and I. H. El-adaway. 2020e. “Bridge infrastructure asset

management system: Comparative computational machine learning

approach for evaluating and predicting deck deterioration conditions.”

J. Infrastruct. Syst. 26 (3): 04020032.

Assaad, R., and I. H. El-adaway. 2020f. “Enhancing the knowledge of

construction business failure: A social network analysis approach.”

J. Constr. Eng. Manage. 146 (6): 04020052. https://doi.org/10.1061

/(ASCE)CO.1943-7862.0001831.

Assaad, R., and I. H. El-adaway. 2020g. “Evaluation and prediction

of the hazard potential level of dam infrastructures using computa-

tional artificial intelligence algorithms.”J. Manage. Eng. 36 (5):

04020051. https://doi.org/10.1061/(ASCE)ME.1943-5479.0000810.

Assaad, R., I. H. El-Adaway, A. H. El Hakea, M. J. Parker, T. I. Henderson,

C. R. Salvo, and M. O. Ahmed. 2020h. “A contractual perspective

for BIM utilization in US construction projects.”J. Constr. Eng.

Manage. 31 (May): 204–214.

Ayhan, B. U., and O. B. Tokdemir. 2020. “Accident analysis for construc-

tion safety using latent class clustering and artificial neural networks.”

J. Constr. Eng. Manage. 146 (3): 04019114. https://doi.org/10.1061

/(ASCE)CO.1943-7862.0001762.

Ayodele, O. A., A. Chang-Richards, and V. González. 2020. “Factors

affecting workforce turnover in the construction sector: A systematic

review.”J. Constr. Eng. Manage. 146 (2): 03119010. https://doi.org/10

.1061/(ASCE)CO.1943-7862.0001725.

Bonham, D. R., P. M. Goodrum, R. Littlejohn, and M. A. Albattah.

2017. “Application of data mining techniques to quantify the relative

influence of design and installation characteristics on labor productiv-

ity.”J. Constr. Eng. Manage. 143 (8): 04017052. https://doi.org/10

.1061/(ASCE)CO.1943-7862.0001347.

Bonifacio, C., T. E. Barchyn, C. H. Hugenholtz, and S. W. Kienzle.

2015. “CCDST: A free Canadian climate data scraping tool.”

Comput. Geosci. 75 (Feb): 13–16. https://doi.org/10.1016/j.cageo.2014

.10.010.

Brill, M., C. Holman, C. Morris, R. Raichoudhary, and N. Yosif.

2017. “Understanding the labor productivity and compensation gap.”

J. Manage. Eng., 2021, 37(1): 04020092

Accessed February 7, 2020. https://www.bls.gov/opub/btn/volume-6

/understanding-the-labor-productivity-and-compensation-gap.htm?view

_full.

Brownlee, J. 2018. “11 classical time series forecasting methods in Python

(cheat sheet).”Accessed May 3, 2020. https://machinelearningmastery

.com/time-series-forecasting-methods-in-python-cheat-sheet/.

Cao, H., and Y. M. Goh. 2019. “Analyzing construction safety through time

series methods.”Front. Eng. Manage. 6 (2): 262–274. https://doi.org/10

.1007/s42524-019-0015-6.

CFI (Corporate Finance Institute). 2020. “Cointegration.”Accessed May 5,

2020. https://corporatefinanceinstitute.com/resources/knowledge/other

/cointegration/.

Chaturvedi, S., J. J. Thakkar, and R. Shankar. 2018. “Labor productivity in

the construction industry: An evaluation framework for causal relation-

ships.”Benchmarking 25 (1): 334–356. https://doi.org/10.1108/BIJ-11

-2016-0171.

Choy, E., and J. Y. Ruwanpura. 2005. “Situation based modeling for

construction productivity.”In Proc., Construction Research Congress

2005: Broadening Perspectives,1–10. Reston, VA: ASCE.

CLMA (Construction Labor Market Analyzer). 2013. “20/20 foresight

report.”Accessed February 17, 2020. http://www.myclma.com/clma

-tools-services/skilled-labor-market-analytics/.

CLMA (Construction Labor Market Analyzer). 2018. “Skilled labor com-

pensation data.”Accessed February 18, 2020. http://www.myclma.com

/clma-tools-services/wage-surveys/.

David, D. 2007. “Cointegration—General discussion.”Accessed June 14,

2020. https://www4.stat.ncsu.edu/∼dickey/Spain/pdf_Notes/Cointegration

.pdf.

Dickey, D. A., and W. A. Fuller. 1979. “Distribution of the estimators for

autoregressive time series with a unit root.”J. Am. Stat. Assoc. 74 (366):

427–431. https://doi.org/10.1080/01621459.1979.10482531.

Donkor, S. 2011. “Determinants of business failure: The perspective of

SMEs building contractors in the Ghanaian construction industry.”

Doctoral dissertation, Dept. of Building Technology, College of Archi-

tecture and Planning.

Durdyev, S., S. Ismail, and N. Kandymov. 2018. “Structural equation

model of the factors affecting construction labor productivity.”

J. Constr. Eng. Manage. 144 (4): 04018007. https://doi.org/10.1061

/(ASCE)CO.1943-7862.0001452.

El-Gohary, K. M., R. F. Aziz, and H. A. Abdel-Khalek. 2017. “Engineering

approach using ANN to improve and predict construction labor produc-

tivity under different influences.”J. Constr. Eng. Manage. 143 (8):

04017045. https://doi.org/10.1061/(ASCE)CO.1943-7862.0001340.

Elliott, G., and E. Pesavento. 2009. “Testing the null of no cointegration

when covariates are known to have a unit root.”Econom. Theory 25 (6):

1829–1850. https://doi.org/10.1017/S026646660999034X.

Erica. 2020. “A guide to conducting cointegration tests.”Accessed May 5,

2020. https://www.aptech.com/blog/a-guide-to-conducting-cointegration

-tests/.

Faghih, S. A. M., and H. Kashani. 2018. “Forecasting construction material

prices using vector error correction model.”J. Constr. Eng. Manage.

144 (8): 04018075. https://doi.org/10.1061/(ASCE)CO.1943-7862

.0001528.

Fan, R., T. Ng, and J. Wong. 2010. “Reliability of the Box–Jenkins model

for forecasting construction demand covering times of economic

austerity.”Constr. Manage. Econ. 28 (3): 241–254. https://doi.org/10

.1080/01446190903369899.

Field, A. 2013. Discovering statistics using IBM SPSS statistics. New York:

SAGE.

Florez, L., P. Armstrong, and J. C. Cortissoz. 2020. “Does compatibility of

personality affect productivity? Exploratory study with construction

crews.”J. Manage. Eng. 36 (5): 04020049. https://doi.org/10.1061

/(ASCE)ME.1943-5479.0000807.

Gao, Z. K., M. Small, and J. Kurths. 2017. “Complex network analysis of

time series.”Europhys. Lett. 116 (5): 50001. https://doi.org/10.1209

/0295-5075/116/50001.

Ghodrati, N., T. Wing Yiu, S. Wilkinson, and M. Shahbazpour. 2018. “Role

of management strategies in improving labor productivity in general

construction projects in New Zealand: Managerial perspective.”

J. Manage. Eng. 34 (6): 04018035. https://doi.org/10.1061/(ASCE)ME

.1943-5479.0000641.

Glen, S. 2016a. “Serial correlation/autocorrelation: Definition, tests.”

Accessed June 14, 2020. https://www.statisticshowto.com/serial

-correlation-autocorrelation/.

Glen, S. 2016b. “Unit root: Simple definition, unit root tests.”Accessed

June 14, 2020. https://www.statisticshowto.com/unit-root/.

Granger, C. W. 1969. “Investigating causal relations by econometric mod-

els and cross-spectral methods.”Econometrica: J. Econ. Soc. 37 (3):

424–438. https://doi.org/10.2307/1912791.

Gunopulos, D., G. Das, and G. Das. 2001. “Time series similarity measures

and time series indexing.”ACM SIGMOD Rec. 30 (2): 624. https://doi

.org/10.1145/376284.375808.

Gupta, M., A. Hasan, A. K. Jain, and K. N. Jha. 2018. “Site amenities and

workers’welfare factors affecting workforce productivity in Indian

construction projects.”J. Constr. Eng. Manage. 144 (11): 04018101.

https://doi.org/10.1061/(ASCE)CO.1943-7862.0001566.

Gurmu, A. T. 2019. “Tools for measuring construction materials manage-

ment practices and predicting labor productivity in multistory building

projects.”J. Constr. Eng. Manage. 145 (2): 04018139. https://doi.org

/10.1061/(ASCE)CO.1943-7862.0001611.

Gurmu, A. T., and C. S. Ongkowijoyo. 2020. “Predicting construction labor

productivity based on implementation levels of human resource man-

agement practices.”J. Constr. Eng. Manage. 146 (3): 04019115. https://

doi.org/10.1061/(ASCE)CO.1943-7862.0001775.

Hendrickson, C. 2005. “Discussion of ‘Is construction labor productivity

really declining?’by Eddy M. Rojas and Peerapong Aramvareekul.”

J. Constr. Eng. Manage. 131 (2): 269–270. https://doi.org/10.1061

/(ASCE)0733-9364(2005)131:2(269).

IHS (Information Handling Services). 2020. “Economics and country risk.”

Accessed May 7, 2020. https://www.ihs.com/industry/economics

-country-risk.html.

Ilbeigi, M., B. Ashuri, and A. Joukar. 2017. “Time-series analysis for fore-

casting asphalt-cement price.”J. Manage. Eng. 33 (1): 04016030.

https://doi.org/10.1061/(ASCE)ME.1943-5479.0000477.

Jarkas, A. M. 2016. “Effect of buildability on labor productivity: A prac-

tical quantification approach.”J. Constr. Eng. Manage. 142 (2):

06015002. https://doi.org/10.1061/(ASCE)CO.1943-7862.0001062.

Johansen, S. 2009. “Cointegration: Overview and development.”In Hand-

book of financial time series, 671–693. Berlin: Springer.

Johari, S., and K. N. Jha. 2020a. “How the aptitude of workers affects con-

struction labor productivity.”J. Manage. Eng. 36 (5): 04020055. https://

doi.org/10.1061/(ASCE)ME.1943-5479.0000826.

Johari, S., and K. N. Jha. 2020b. “Impact of work motivation on construc-

tion labor productivity.”J. Manage. Eng. 36 (5): 04020052. https://doi

.org/10.1061/(ASCE)ME.1943-5479.0000824.

Kadilar, G. Ö., and C. Kadilar. 2017. “Assessing air quality in Aksaray with

time series analysis.”In Proc., AIP Conf., 020112. College Park, MD:

American Institute of Physics.

Kisi, K. P., N. Mani, E. M. Rojas, and E. T. Foster. 2018. “Estimation

of optimal productivity in labor-intensive construction operations:

Advanced study.”J. Constr. Eng. Manage. 144 (10): 04018097. https://

doi.org/10.1061/(ASCE)CO.1943-7862.0001551.

Koch, M. J. 2017. Hiring practices and labor productivity. Abingdon, UK:

Taylor & Francis.

Kuznetsova, A., A. Askarov, R. Gusmanov, A. Askarova, and P. Pypłacz.

2019. “Differentiation of labor productivity level and wages as a

basis for changes in labor market.”Pol. J. Manage. Stud. 2019 (20):

345–357.

Lee, C., J. Won, and E. B. Lee. 2019. “Method for predicting raw material

prices for product production over long periods.”J. Constr. Eng. Man-

age. 145 (1): 05018017. https://doi.org/10.1061/(ASCE)CO.1943-7862

.0001586.

Lee, S., S. Tae, S. Yoo, and S. Shin. 2016. “Impact of business portfolio

diversification on construction company insolvency in Korea.”J. Man-

age. Eng. 32 (3): 05016003. https://doi.org/10.1061/(ASCE)ME.1943

-5479.0000413.

Lin, D.-J., W.-D. Yu, C.-M. Wu, and T-M. Cheng. 2018. “Correlation

between intellectual capital and business performance of construction

J. Manage. Eng., 2021, 37(1): 04020092

industry–an empirical study in Taiwan.”Int. J. Constr. Manage. 18 (3):

232–246. https://doi.org/10.1080/15623599.2017.1315528.

Lingard, H., M. Hallowell, R. Salas, and P. Pirzadeh. 2017. “Leading or

lagging? Temporal analysis of safety indicators on a large infrastructure

construction project.”Saf. Sci. 91 (Jan): 206–220. https://doi.org/10

.1016/j.ssci.2016.08.020.

Liu, H. J., P. E. Love, M. C. Sing, and J. Smith. 2019. “Ex post evaluation

of economic infrastructure assets: Significance of regional hetero-

geneities in Australia.”J. Infrastruct. Syst. 25 (2): 05019005. https://doi

.org/10.1061/(ASCE)IS.1943-555X.0000485.

Lütkepohl, H., M. Krätzig, and P. C. B. Phillips. 2004. Applied time series

econometrics. Cambridge, UK: Cambridge University Press.

Mahadevan, R., and J. Asafu-Adjaye. 2007. “Energy consumption, eco-

nomic growth and prices: A reassessment using panel VECM for de-

veloped and developing countries.”Energy Policy 35 (4): 2481–2490.

https://doi.org/10.1016/j.enpol.2006.08.019.

Mahamid, I. 2013. “Contractors perspective toward factors affecting labor

productivity in building construction.”Eng. Constr. Archit. Manage.

2013 (Sep): 2. https://doi.org/10.1108/ECAM-08-2011-0074.

McGowan, J. 2019. “How has the growth of E-commerce sales affected

retail real estate?”CMC Senior theses, The Robert Day School of

Economics and Finance, Claremont McKenna College.

Mirhadi, M. 2018. “Adverse effects of shift work on labor productivity.”

Accessed May 5, 2020. https://www.adroitprojectconsultants.com/2018

/03/24/adverse-effects-shiftwork-labor-productivity/.

NCSS. 2006. “Multiple regression with serial correlation.”Accessed June

14, 2020. https://ncss-wpengine.netdna-ssl.com/wp-content/themes/ncss

/pdf/Procedures/NCSS/Multiple_Regression_with_Serial_Correlation

.pdf.

Ozturk, M., S. Durdyev, O. N. Aras, S. Ismail, and N. Banaitienė. 2020.

“How effective are labor wages on labor productivity?: An empirical

investigation on the construction industry of New Zealand.”Technol.

Econ. Dev. Economy. 26 (1): 258–270. https://doi.org/10.3846/tede

.2020.11917.

Pan, W., L. Chen, and W. Zhan. 2019. “PESTEL analysis of construction

productivity enhancement strategies: A case study of three economies.”

J. Manage. Eng. 35 (1): 05018013. https://doi.org/10.1061/(ASCE)ME

.1943-5479.0000662.

Parrotta, P., and D. Pozzoli. 2012. “The effect of learning by hiring on

productivity.”RAND J. Econ. 43 (1): 167–185. https://doi.org/10.1111

/j.1756-2171.2012.00161.x.

Prabhakaran, S. 2019. “Vector autoregression (VAR)—Comprehensive

guide with examples in Python.”Accessed February 16, 2020.

https://www.machinelearningplus.com/time-series/vector-autoregression

-examples-python/.

Rachev, S. T., S. Mittnik, F. J. Fabozzi, and S. M. Focardi. 2007. Financial

econometrics: From basics to advanced modeling techniques.

Hoboken, NJ: Wiley.

Rojas, E. M., and P. Aramvareekul. 2003. “Is construction labor produc-

tivity really declining?”J. Constr. Eng. Manage. 129 (1): 41–46.

https://doi.org/10.1061/(ASCE)0733-9364(2003)129:1(41).

Semler, K. 2020. “The construction labor shortages direct impact on home

buyers.”Accessed November 17, 2020. https://expressmodular.com/the

-construction-labor-shortages-direct-impact-on-home-buyers/.

Setiani, Y., and M. Z. Abd Majid. 2019. “Safety practices and labour

productivity in construction projects.”Covenant J. Res. Built Environ.

7 (1): 84–95.

Sexton, D., I. H. El-adaway, M. Abdul Nabi, and A. H. El Hakea. 2020.

“Using the simplified acquisition of base engineer requirements in

construction projects: Contract administration guidelines.”J. Leg. Aff.

Dispute Resolut. Eng. Constr. 12 (3): 04520018. https://doi.org/10

.1061/(ASCE)LA.1943-4170.0000394.

Shahandashti, S. M., and B. Ashuri. 2016. “Highway construction cost

forecasting using vector error correction models.”J. Manage. Eng.

32 (2): 04015040. https://doi.org/10.1061/(ASCE)ME.1943-5479

.0000404.

Shahin, M. A., H. R. Maier, and M. B. Jaksa. 2004. “Data division for

developing neural networks applied to geotechnical engineering.”

J. Comput. Civ. Eng. 18 (2): 105–114. https://doi.org/10.1061/(ASCE)

0887-3801(2004)18:2(105).

Shifera, A. 2019. Modeling and forecasting volatility of coffee and gold

export price in Ethiopia using GARCH model. Nekemte, Ethiopia:

Wollega Univ.

Shiha, A., E. M. Dorra, and K. Nassar. 2020. “Neural networks model for

prediction of construction material prices in Egypt using macroeco-

nomic indicators.”J. Constr. Eng. Manage. 146 (3): 04020010. https://

doi.org/10.1061/(ASCE)CO.1943-7862.0001785.

Shohet, I. M., H. H. Wei, M. J. Skibniewski, B. Tak, and M. Revivi. 2019.

“Integrated communication, control, and command of construction

safety and quality.”J. Constr. Eng. Manage. 145 (9): 04019051. https://

doi.org/10.1061/(ASCE)CO.1943-7862.0001679.

Singh, A. 2018. “A multivariate time series guide to forecasting and

modeling (with Python codes).”Accessed February 17, 2020. https://

www.analyticsvidhya.com/blog/2018/09/multivariate-time-series-guide

-forecasting-modeling-python-codes/.

Sodiq Olawale, A. 2019. Impact of corruption on economic growth in

Nigeria. Lagos, Nigeria: Lagos State Univ.

Soekiman, A., K. S. Pribadi, B. W. Soemardi, and R. D. Wirahadikusumah.

2011. “Factors relating to labor productivity affecting the project

schedule performance in Indonesia.”Procedia Eng. 14 (Jan): 865–873.

https://doi.org/10.1016/j.proeng.2011.07.110.

Stanslaus, V. 2017. “The causality effects of macroeconomic factors on eco-

nomic growth in tanzania.”Doctoral dissertation, Dept. of Economics,

Open Univ. of Tanzania.

Sveikauskas, L., S. Rowe, J. Mildenberger, J. Price, and A. Young.

2016. “Productivity growth in construction.”J. Constr. Eng. Manage.

142 (10): 04016045. https://doi.org/10.1061/(ASCE)CO.1943-7862

.0001138.

Sveikauskas, L., S. Rowe, J. D. Mildenberger,J. Price, and A. Young. 2018.

“Measuring productivity growth in construction.”Mon. Lab. Rev.

Accessed May 5, 2020. https://www.bls.gov/opub/mlr/2018/article

/measuring-productivity-growth-in-construction.htm.

Swei, O. 2020. “Forecasting infidelity: Why current methods for predicting

costs miss the mark.”J. Constr. Eng. Manage. 146 (2): 04019100.

https://doi.org/10.1061/(ASCE)CO.1943-7862.0001756.

Swei, O., J. Gregory, and R. Kirchain. 2017. “Probabilistic approach

for long-run price projections: Case study of concrete and asphalt.”

J. Constr. Eng. Manage. 143 (1): 05016018. https://doi.org/10.1061

/(ASCE)CO.1943-7862.0001211.

US Bureau of Economic Analysis. 2018. “Interactive access to industry

economic accounts data: GDP by industry.”Accessed November 22,

2019. https://apps.bea.gov/iTable/iTable.cfm?ReqID=51&hx0026%3

Bstep=1#reqid=51&hx0026;step=51&hx0026;isuri=1&hx0026;5114=

a&hx0026;5102=1.

US Bureau of Labor Statistics. 2020. “Industries at a glance: Construction:

NAICS 23.”Accessed March 18, 2020. https://www.bls.gov/iag/tgs

/iag23.htm.

Vereen, S. C. 2013. “Forecasting skilled labor demand in the US construc-

tion industry.”Ph.D. dissertation, Dept. of Civil, Construction, and

Environmental Engineering, North Carolina State Univ.

Vereen, S. C., W. Rasdorf, and J. E. Hummer. 2016. “Development

and comparative analysis of construction industry labor productivity

metrics.”J. Constr. Eng. Manage. 142 (7): 04016020. https://doi.org/10

.1061/(ASCE)CO.1943-7862.0001112.

Walter, L. 2008. “Workplace fatalities: The impact on coworkers.”

Accessed February 18, 2020. https://www.ehstoday.com/safety/article

/21905011/workplace-fatalities-the-impact-on-coworkers.

Wang, L., and S. Ding. 2018. “Vector auto regression and envelope model.”

Stat 7 (1): e203. https://doi.org/10.1002/sta4.203.

Wang, S., G. Hua, G. Hao, and C. Xie. 2017. “A cycle deep belief network

model for multivariate time series classification.”Math. Prob. Eng.

2017 (Jan): 1–7.

Wei, W. 2016. Achieving inclusive growth in China through vertical

specialization. Cambridge, UK: Chandos.

Xu, B., and B. Lin. 2016. “Assessing CO2emissions in China’s iron

and steel industry: A dynamic vector auto regression model.”Appl.

Energy 161 (Jan): 375–386. https://doi.org/10.1016/j.apenergy.2015

.10.039.

J. Manage. Eng., 2021, 37(1): 04020092

Yang, X., and B. Liu. 2019. “Uncertain time series analysis with imprecise

observations.”Fuzzy Optim. Decis. Making 18 (3): 263–278. https://doi

.org/10.1007/s10700-018-9298-z.

Yi, W., and A. P. Chan. 2014. “Critical review of labor productivity research

in construction journals.”J. Manage. Eng. 30 (2): 214–225. https://doi

.org/10.1061/(ASCE)ME.1943-5479.0000194.

Zhao, T., and J. M. Dungan. 2018. “Quantifying lost labor productivity

in domestic and international claims.”J. Leg. Aff. Dispute Resolut.

Eng. Constr. 10 (3): 04518013. https://doi.org/10.1061/(ASCE)LA

.1943-4170.0000269.

Zivot, E., and J. Wang. 2007. Modeling financial time series with S-Plus®.

Berlin: Springer.

J. Manage. Eng., 2021, 37(1): 04020092

Impact of COVID-19 on Social and Psychological Well-Being of Workers in Construction Projects: A Comparative Analysis of Managers and Laborers

Article

Full-text available

Jul 2023

The COVID-19 pandemic brought challenges such as social distancing, health fears, reduced interaction, and increased stress for construction workers. Understanding their changing social and psychological states is crucial for effective management and performance. This study investigated the impact of the pandemic on the managers’ and laborers’ social and psychological well-being states and identified the changes in their social and psychological well-being states affecting project performance before and after the pandemic. Construction professionals, including construction managers, superintendents, and laborers, participated in a survey exploring thirteen social and psychological well-being variables and three performance variables. Data analysis involved paired t-tests and multiple regression. The findings revealed increased levels of anxiety and depression among both managers and laborers after the pandemic, with laborers more severely affected. Managers considered a broader range of variables, while laborers primarily focused on social factors influencing project performance. These disparities suggested that managers should prioritize health and safety measures, fair compensation, team cohesion, and stress management, while laborers’ motivation, work environment, knowledge acquisition, and sense of belonging should receive priority attention. This study contributes to providing managerial implications and guidance for improving the construction workforce, including managers’ and site laborers’ performance in the post-pandemic period.

Integrating early assessment of robotic constructability into design optimization of a standalone classroom

Article

Nov 2023
AUTOMAT CONSTR

Multi-objective optimization can enhance design quality through performance simulation. However, ease or efficiency of construction is also important, and optimization may lead to difficult-to-build designs. Early quantification of constructability would allow designers to balance performance and construction issues. While it is difficult to quantify all factors that influence constructability, robotic construction simulations offer rich datasets to compare potential outcomes. This paper integrates constructability knowledge into early-stage design and examines the impact on multi-objective optimization. It evaluates robotic material delivery systems in constructing a standalone classroom optimized for structural, daylighting, and energy goals. When considering robotic pick and place time, the optimized designs differ, offering the opportunity to change design directions based on construction knowledge. Broader implications also become observable, as incorporating robots with higher carrying capacities reduces the structural elements and embodied carbon of optimal designs. This paper thus demonstrates benefits of incorporating robotic constructability simulation into early design optimization.

System Dynamic Modeling to Study the Impact of Construction Industry Characteristics and Associated Macroeconomic Indicators on Workforce Size and Labor Retention Rate

Article

Oct 2023

Assessing construction workforce diversity of skills and education with the probabilistic linguistic fuzzy Petri net

Article

Jun 2024
INFORM SCIENCES

Modeling Inflation Transmission among Different Construction Materials

Article

May 2024

Cost estimating in the construction industry is challenging due to the high uncertainty associated with the supply chain, especially after the COVID-19 pandemic. Some research studies have addressed such problems by developing models that predict material cost. In fact, all material can be interconnected and interrelated with lead-lag relationships such that any inflation in one material’s price can be associated with inflation in other materials’ prices, referred to as transmission of inflation. Despite the latter, none of the existing studies have investigated inflation transmission among all construction materials. This paper fills this knowledge gap. The authors used a multistep research methodology. First, Producer Price Index (PPI) data for 16 construction materials were collected and sorted. Second, the vector autoregression technique was used to model the relationships between each pair of material and subsequently validate the associations using the Granger causality test. Third, network analysis was performed to identify the inflation transmission capacity (out-degree centrality), inflation susceptibility (in-degree centrality), and inflation intermediatory capacity (betweenness centrality) for each material. Finally, modularity-based clustering was conducted to categorize the materials based on their price indices’ interconnections and examine inflation transmission path among different sectors of construction-related material. The results show that metals and plastic products are generally the highest transmitters of inflation to other construction material including “Fabricated structural metal products” and “Plastic construction products.” Furthermore, the results show that “Concrete products,” “Flat glass,” “Brick and structural clay tile,” and “Architectural coatings” are also high transmitters of inflation and thus can be key indicators of increase in the overall construction cost. This paper provides the industry stakeholders with leading indicators and early warning signs for the inflated material prices. Contractors and owners can utilize those warning signs to enhance their procurement plans and budgeting decisions.

An Accurate-Pricing Estimate Game-Theoretic Model for Determining Price Escalations in Construction Projects during Economic Uncertainties

Conference Paper

Mar 2024

A Classification Model Using Personal Biometric Characteristics to Identify Individuals Vulnerable to an Extremely Hot Environment

Article

Mar 2024

Enhanced planning and scheduling in building construction projects: an innovative approach to overcome scheduling challenges

Article

Full-text available

Dec 2023

This article introduces an Enhanced Planning and Scheduling (EPS) approach for complex construction projects with multiple sections, offering an innovative solution to tackle existing planning and scheduling challenges. This research presents a systematic and computational approach to optimize project duration and enhance schedule reliability and productivity. The goal is to overcome the inherent limitations of traditional scheduling methods. The proposed methods have the potential to be integrated with emerging cross-industrial project management strategies to enhance planning and scheduling processes in construction. The EPS method is executed in three stages: the enhanced Milestone, Phase, and Production schedules. Fundamental elements of this methodology encompass work areas, activities categories, labor hours estimation, and planning indicators. Drawing on combinatorial optimization concepts, this study offers industry practitioners an effective, easily implementable solution. A case study was conducted to assess the application and effectiveness of the Milestone schedule. The results showed a 40-workday reduction in project duration compared to the traditional scheduling methods. https://www.tandfonline.com/doi/full/10.1080/15623599.2023.2286888 Free ePrint: https://www.tandfonline.com/eprint/GIHQA6Q6PWBMHFXUS6VK/full?target=10.1080/15623599.2023.2286888

Differing approaches to embedding low energy construction and climate literacy into vocational education and training

Chapter

Nov 2023

A Literature Study on Digital Technology Applications in China Construction Industry

Conference Paper

Aug 2023

Contractual Perspective for BIM Utilization in US Construction Projects

Article

Full-text available

Sep 2020

Although building information modeling (BIM) is experiencing substantial development and use in the construction industry, its legal and business structures lag far behind. While many research efforts have been devoted to promoting the implementation of BIM in the construction industry, very few of them have actually attempted to explore its contractual environment. In fact, previous studies reported the lack of, and need for, a holistic understanding of BIM contractual requirements. To fill this gap, this paper assists the various project parties in better understanding BIM contractual integration and utilization in the construction industry at different key levels: contractual incorporation of BIM addenda/exhibits, priority of contract documents, roles and obligations, copyright ownership and intellectual property rights, standard of care, discrepancies in or between models, compensation and information establishment, BIM management and collaboration, use of models, and indemnity. Within the context of documents developed by the American Institute of Architects and ConsensusDocs, the authors used a three-step research methodology that comprised (1) reviewing and analyzing essential clauses, protocols, and addenda; (2) highlighting law cases that are related to BIM contracting in practice; and (3) comparing the different BIM implementation procedures and processes. To this end, the paper presents 20 contractual recommendations to help project parties better manage BIM-based projects from a contractual point of view. To verify/validate the usefulness and benefits of the performed analysis, the authors shared the paper’s recommendations with leading experts in BIM and digital/virtual construction. Some of the specific research recommendations offered in this paper include: preparation of a list of project participants that have incorporated BIM contractual documents into their agreements to provide reasonable evidence that BIM provisions have been included in the main and affiliated contracts, and the prohibition on disclosing any received BIM data to any third party except as required by law or a court order to avoid issues with copyright ownership and intellectual property rights. This research adds to the body of knowledge by providing an enhanced understanding of various BIM contractual aspects on construction projects.

Bridge Infrastructure Asset Management System: Comparative Computational Machine Learning Approach for Evaluating and Predicting Deck Deterioration Conditions

Article

Full-text available

Sep 2020

Bridge Infrastructure Asset Management System: Comparative Computational Machine Learning Approach for Evaluating and Predicting Deck Deterioration Conditions

Article

Full-text available

Jun 2020

Bridge infrastructure asset management system is a prevailing approach toward having an effective and efficient procedure for monitoring bridges through their different development phases including construction, operation, and maintenance. Damage to any structural component of a bridge will negatively affect its safety, integrity, and longevity. Bridge decks are more susceptible to severe deterioration because they are exposed to harsh conditions including heavy traffic, varying temperatures, road salts, and abrasive forces. The ability to forecast the conditions of bridges in an accurate way has been a great challenge to transportation agencies. Many previous research studies highlighted the need to have a data-driven approach in predicting and evaluating the deterioration conditions of bridges. As such, this paper develops a computational data-driven asset management system to evaluate and predict bridge deck deterioration conditions. A multistep interdependent research methodology was utilized. First, the best set of variables affecting the conditions of bridge decks was identified. Second, two computational machine learning models were developed for the prediction of deck conditions using artificial neural networks (ANNs) and k-nearest neighbors (KNNs). Third, a comparison between the developed models is conducted to select the ultimate model with the highest accuracy. The result is a framework that is able to evaluate and predict deck conditions with a prediction accuracy of 91.44%. While this research is applied to bridges in Missouri, the technique can be used on any similarly available data set nationwide. This study adds to the body of knowledge by devising a computational data-driven framework that is valuable for transportation agencies as it allows them to evaluate and predict deck conditions with high accuracy. Consequently, this will help in ensuring proper and effective distribution of funds allocated for the maintenance, rehabilitation, and repair of bridges. Ultimately, this will result in minimizing efforts, time, and costs associated with site inspection of bridge decks.

Does Compatibility of Personality Affect Productivity? Exploratory Study with Construction Crews

Article

Full-text available

Jun 2020

Crew productivity is a function of how efficiently labor is utilized in the construction process. However, previous research in construction has not comprehensively investigated the relationship between personality and crew productivity. This paper uses personality profiles to investigate a new fundamental concept, the relationship between compatibility of personality and crew productivity at the task level. Twenty-eight masons completed a revised questionnaire of the Big Five to indicate their personality. Personality scores were used to calculate compatibility in each of the 20 participating two-mason crews working on eight projects. Regression analysis was performed to establish the relationship between compatibility and crew productivity. Results show that that there is a high positive correlation between compatibility and crew productivity. Compatibility accounts for more than half of the predictable variance in productivity. This paper makes four major contributions: it proposes a new metric to measure compatibility of personality among workers in a crew; it reveals how personality factors affect productivity; it provides rigorous methods to analyze correlations (using confidence intervals and Bayesian inference) for construction experiments; and it provides theoretical contributions to advancing the theory of personality and productivity in construction projects.

DIFFERENTIATION OF LABOR PRODUCTIVITY LEVEL AND WAGES AS A BASIS FOR CHANGES IN LABOR MARKET

Article

Dec 2019

Hiring Practices and Labor Productivity

Book

Jun 2017

Marianne Koch

How the Aptitude of Workers Affects Construction Labor Productivity

Article

Jun 2020

Aptitude is the potential to demonstrate the ability to perform a certain kind of work at a certain level. While several studies have highlighted the relationship between aptitude and employee performance, very few have tested such a relationship using empirical data, especially in the construction domain. Therefore, to address this gap, this study attempts to establish the relationship between construction workers’ aptitudes and their productivity. To achieve this aim, first the aptitude of the workers was measured using a general aptitude test battery (GATB). Second, the productivity of the same workers was measured via direct observation at construction sites. Subsequently, by applying various regression models to the data collected on 112 workers, best-fit curves were developed between aptitude and observed productivity. The results revealed that the physical aptitude of the workers impacts performance more than mental aptitude. Moreover, spatial and clerical perception have a negative impact on performance. The study concludes that aptitude is an important performance indicator of any construction worker, and thus, to enhance labor productivity, contractors must ensure the availability of high-aptitude workers on site.

Evaluation and Prediction of the Hazard Potential Level of Dam Infrastructures Using Computational Artificial Intelligence Algorithms

Article

Jun 2020

Failures of dams cause immense property and environmental damages and take thousands of lives. As such, the goal of this paper is to evaluate and predict the hazard potential level of dams in the US using a comparative approach based on computational artificial intelligence (AI) algorithms. The research methodology comprised data collection from the National Inventory of Dams (NID); data pre-processing; data processing; and model selection and evaluation. To this end, the authors: (1) identified the best subset of variables that affect the prediction of the hazard potential level of dams in the US; (2) investigated the performance of two AI computational algorithms: artificial neural networks (ANNs) and k-nearest neighbors (KNNs) for the evaluation and prediction of the hazard potential levels of US dams; and (3) developed a decision support tool that could be used by the agencies responsible for the management of dams in the US with the capability to predict the hazard potential with good accuracy. The obtained results reflected that the ANN algorithm yielded better accuracy compared to the KNN algorithm. In addition, the conclusions indicated that 19 variables pertaining to dams in the US could affect the hazard potential level of dams. The output is a decision support system that is able to evaluate the hazard potential of dams with a prediction accuracy of 85.70%. This study contributes to the management in engineering's body of knowledge by devising a data-driven framework that is valuable for dams' owners and authorities. Ultimately, the developed computational AI algorithm could be used to evaluate and predict the hazard potential level of US dams with good accuracy while minimizing the efforts, time, and costs associated with formal inspection of the dams.

Impact of Work Motivation on Construction Labor Productivity

Article

Jun 2020

Poor performance of workers is a perennial problem for the construction industry worldwide. While there have been many studies reporting the influence of motivation on construction labor productivity (CLP), much of the evidence has been anecdotal in nature. Therefore, to address this gap, the objective of this study is to explore the relationships between the work motivation of workers and their productivity. The work-motivation scale was adopted for measuring the work motivation of 116 construction workers from 4 construction sites in India. Then, using nonlinear regression analysis, the relationship among their observed productivity and their motivation was established. The study reveals that amotivation, extrinsic regulation (social), introjected regulation, and identified regulation affect CLP. On the other hand, extrinsic regulation (material) and intrinsic motivation do not influence CLP substantially. The study also concludes that by increasing the motivation of workers, their productivity and their probability of staying and working at one particular site increase substantially. In this way, the findings can help training providers focus on and improve the significant motivational attributes of workers influencing CLP at construction sites.

IMPACT OF CORRUPTION ON ECONOMIC GROWTH IN NIGERIA

Research

May 2020

Sodiq Abdulazeez

The study investigated the impact of corruption on economic growth in Nigeria for a 26 years period ranging from 1992 - 2018. The study specifically examined the impact of corruption level, government expenditure and unemployment on economic growth in Nigeria. Corruption perception index (CPI) proxies level of corruption, unemployment rate (UN) proxies unemployment, level of government spending proxied by government expenditure (GEX) and growth rate of GDP (GGDP) represents economic growth.

Impact of Dynamic Workforce and Workplace Variables on the Productivity of the Construction Industry: New Gross Construction Productivity Indicator

Abstract and Figures

Recommended publications

Stock Prices of Architectural, Engineering, and Construction Firms as Leading Economic Indicator: A...

Key Factors Affecting Labor Productivity in Offsite Construction Projects

Modeling Inflation Transmission among Different Construction Materials

System Dynamic Modeling to Study the Impact of Construction Industry Characteristics and Associated...