1 Introduction

The majority of economists agree with the importance of “Productivity,” as it is one of the most frequently discussed topics in the construction industry (Yi and Chan 2013). Labor productivity is one of the key management factors to complete a project successfully. Hence, productivity has been creating remarkable interest in both the construction industry and academia. Furthermore, project managers have to improve construction productivity performance because of competitive business environments (Park et al. 2005).

Construction project management goals would be attainable by optimization of management, machinery, material, money and manpower (Allmon et al. 2000). Project management teams usually encounter significant challenges in order to estimate the manpower and productivity level (Hanna et al. 2005). According to Wilcox et al. (2000) improving productivity is a major concern for any profit-oriented organization, as it represents an effective and efficient conversion of resources into marketable products and determines business profitability. Consequently, considerable effort has been directed to understanding the productivity concept, with the different approaches taken by researchers resulting in a wide variety of definitions of productivity (Lema 1995; Oglesby et al. 1989; Pilcher 1992).

Due to the importance of the labor productivity in the construction sites, different studies about this issue have been found in western countries. A number of studies regarding the factors affecting labor productivity have been identified and mentioned in the next section. While numerous labor productivity research studies have been undertaken, only a few have addressed the construction labor productivity issue in Iran, such as (Ghoddousi and Hosseini 2012; Khanzadi et al. 2011; Zakeri et al. 1997, 1996).

Besides, most of the construction management research involves unobservable, or latent, variables. The Structural Equation Modeling (SEM) is a family of statistical models that seek to explain the relationship among multiple variables (Hair et al. 2010). When these variables can be measured through surrogate variables, SEM allows for a statistically defensible means of quantifying them (Molenaar et al. 2000). Therefore, numerous studies in construction management research have been conducted by SEM. However, there is no labor productivity model by SEM. On the other hand, the Analytical Hierarchy Process (AHP) is a flexible decision-making tool which was firstly introduced by Saaty (1980a). This study attempted to apply the AHP as a decision-making tool and SEM as a multivariate analysis technique and eventually compared the outcomes of both methods, in parallel for accuracy and reliability of findings.

2 Literature Review

Generally, there are various definitions of productivity and each company uses its own internal system to measure it (Thomas and Mathews 1986). There are two forms of productivity used in the previous studies; productivity = output/input, productivity = input/output. The other different definitions can be identified regarding the productivity in the construction activities, one refers to the productivity when the work is implemented, and the other one refers to the value of the work based on the cost (Knutson et al. 2009). On the other hand, productivity refers to the output or hours that each worker needs to do in order to complete the job. Usually different countries measure the productivity rate of their workers based on dollar production for each worker-hour or whole price per element of production (Knutson et al. 2009). Tenah (1985) believes that based on the theoretical definitions, productivity refers to the relationship between output and input. According to the Bureau of Labor Statistics of the USA, the amount of productivity is usually related to the physical or real amounts of things and facilities (productions), associated with the physical or actual amount of feedback (workload, energy, wealth).

2.1 Factors Affecting Labors Productivity in the Construction Industry

Several researchers are enthusiastic in the context of labor productivity. Due to the importance and vital role of labor in project enhancement, numerous studies have been done in various countries. Horner and Talhouni (1993) identified the most significant perceived factors influencing labor productivity in the UK as: Skill of labor, Build-ability, Quality of supervision and Method of working. Lim and Alum (1995) discovered seventeen issues that could affect construction productivity, and the greatest concerns are, namely, Difficulty in recruitment of supervisors, Difficulty in recruiting workers because of a high rate of labor turnover, absenteeism at work site, communication problems with foreign workers and inclement weather that requires work stoppage for 1 day or more. Dai et al. (2007) conducted a survey and identified eighty-three factors in the USA, and the most significant factors are as follows: Supervisor direction; Communication; Safety; Tools and consumables; and Materials. Durdyev and Mbachu (2011) discovered that internal constraints have a much higher impact on onsite productivity than the external factors. The internal constraints included: reworks, level of skill and experience of the workforce, adequacy of method of construction; build-ability issues and inadequate supervision and coordination. Dai et al. (2009) identified several factors affecting labor productivity, and it has been discovered through the principal factor analysis that ten latent variables have a negative impact on productivity in the following descending order: Construction Equipment, Materials, Tools and Consumables, Engineering Drawing Management, Direction and Coordination, Project Management, Training, Craft Worker Qualification, Superintendent Competency and Foreman Competency.

Subsequently, the studies concerning labor productivity performed in some developing countries are being compared to construction productivity problems with developed countries. Kaming et al. (1997b) realized that factors affecting the productivity of craftsmen in Indonesia comprise: lack of materials, rework, absenteeism of operatives and lack of suitable tools. Besides, Alwi (2003) further allocated the key factors impinging upon construction productivity in Indonesia into the following categories: (1) Characteristics of contractors; (2) Inadequate management strategy; (3) Organization’s focus. Makulsawatudom et al. (2004) identified five factors among twenty-three factors as the most critical factors, namely in Thailand: Lack of materials, Rework, Absenteeism of operatives, Lack of suitable tools and equipment and Crew interference. Kadir et al. (2005) discovered fifty productivity factors on Malaysian residential projects, and five of the most significant factors were: Shortage of material; Nonpayment to suppliers causing stoppage of materials delivery to sites; Change in orders by consultants; Late issuance of construction drawings by consultants; and The incapability of site management. In Uganda, Alinaitwe et al. (2007) ranked the following five factors as being the most significant: Incompetent supervisors; Lack of skills; Rework; Lack of tools/equipment; and Poor construction methods.

In addition, some researchers in Middle Eastern countries performed studies regarding labor productivity in order to evaluate the factors affecting labor productivity. Enshassi et al. (2007) identified forty-five factors affecting labor productivity within building projects in the Gaza Strip. The main factors negatively affecting labor productivity were: Material shortage, Lack of labor experience, Lack of labor surveillance, Misunderstandings between labor and superintendent and Alteration of drawings and specifications during execution. Jarkas and Bitar (2012) found that the most effective factors, out of forty-five discovered factors, were: clarity of technical specifications, extent of variation/change orders during execution, coordination level among various design disciplines, lack of labor supervision and proportion of work subcontracted. According to El-Gohary and Aziz (2013) the most significant factors in regard to the effects on construction labor productivity in Egypt comprised: labor experience and skills; incentive programs; availability of the material and ease of handling; leadership and competency of construction management; and competency of labor supervision. Zakeri et al. (1996) using the relative index ranking technique ranked the following five factors as major determinants of Iranian operatives’ efficiency: Materials shortage; Weather and site conditions; Equipment breakdown; Drawing deficiencies/change orders; and Lack of proper tools and equipment. Zakeri et al. (1997) identified five of the most important motivation factors of Iranian construction operatives, namely fairness of pay, incentive and financial rewards, on-time payment, good working facilities and safety. Ghoddousi and Hosseini (2012) determined and explored the most critical grounds affecting sub-contractors productivity in descending order as: Materials/Tools, Construction technology and method, Planning, Supervision system, Reworks, Weather and Jobsite condition. The potential factors affecting labor productivity from previous studies have been summarized in Table 1.

Table 1 Factors affecting labor productivity with descriptions
Full size table

2.2 Application of Analytical Hierarchy Process (AHP) in the Construction Management

The use of statistical analysis in order to identify critical features in construction engineering practice is reasonably widespread (Hanna et al. 2005; Iyer and Jha 2005). The Analytic Hierarchy Process (AHP) is a theory of measurement through pairwise comparisons which relies on the judgments of experts to derive priority scales (Saaty 2008). One of the advantages of AHP is that the analysis does not need a statistically significant sample size (Dias and Ioannou 1996). The simplicity of the AHP approach is that, unlike other “conjoint” methods, the qualities (or levels) of different attributes are not directly compared. The AHP approach thus removes the need for complex survey designs and can even be applied (in an extreme case) with only a single respondent (Saaty 1980b; Schot and Fischer 1993; Zahedi 1986).

Applications of AHP in construction management studies are pretty remarkable as many researchers and project managers apply this tool. Here, several construction management studies that implemented AHP are addressed briefly as follows: According to Al-Harbi (2001), AHP is a potential decision-making method in project management. Al-Harbi applied AHP for prequalification of contractors for a project. Doloi (2008) believed that poor construction labor productivity causes delay and cost overrun. Doloi discovered, by using AHP, that planning and programming has the highest impact on productivity. Cheng and Li (2002) examined a model by AHP regarding the construction partnering process and critical success factors. Skibniewski and Chao (1992) evaluated advanced construction technologies by applying AHP, in relation to the risk of traditional economic analysis techniques. Pan (2008) proposed a fuzzy AHP to select the most preferable bridge construction method in Taiwan. Al Khalil (2002) developed an AHP model to select the most suitable project delivery method. Chiang et al. (2017) applied AHP in order to prequalify and select the construction contractors. Raviv et al. (2017) implemented AHP to evaluate the risk potential of safety incidents for cranes. Tamošaitienė et al. (2017) used a hybrid multicriteria decision-making model by AHP in relation to supply chain management issues.

2.3 Application of Structural Equation Modeling (SEM) in the Construction Management

Structural Equation Modeling (SEM) is a methodology for representing, estimating and testing a network of relationships between variables. Through the SEM, the researcher could examine the direct and indirect interrelationships which exist between multiple dependent and independent variables (Gefen et al. 2000). SEM’s foundation lies on two familiar multivariate techniques: factor analysis and multiple regression analysis (Hair 2010). According to Hair (2010), the three distinguished characteristics of SEM models are: i) To estimate multiple and interrelated dependence relationships; ii) To represent unobserved concepts in these relationships and to account for measurement errors in the estimation process; and iii) To define a model to explain the entire set of relationships. It could be used as a more powerful alternative to path analysis, multiple regression, factor analysis, covariance analysis and time series analysis. In fact, SEM is a multivariate analysis which combines path analysis and confirmatory factor analysis simultaneously; through the path analysis, the regression weights will be discovered; and through the confirmatory factor analysis (CFA), the structure or group of factors or variables will be confirmed (Xiong et al. 2015).

Hence, SEM, as a statistical analysis tool, is being applied in construction engineering and management research these days. Molenaar (2000) mentioned that SEM is a statistical analysis tool that is underutilized in construction engineering and management research nowadays. Xiong et al. (2015) reviewed 84 articles which addressed construction problems and applied SEM. Xiong et al. (2015) discovered that SEM applications have been increasing over time. Moreover, it has been applied to a variety of issues and aspects in construction management such as trust in construction contracting by Wong et al. (2008), a composite model using SEM and fuzzy logic for supplier selection by Punniyamoorthy et al. (2011), feasibility and project success for public–private partnership (PPP) studied by Ng, Wong, and Wong (2010), construction contracting by Cheung et al. (2012), construction partnering assessed by Chen et al. (2012), relationship between an institution and its constituents studied by Oei and Ogunlana (2006), the implementation of enterprise resource planning software and the goal of competitive advantage performed by Ram et al. (2014) and contract disputes between owners and contractors investigated by Molenaar et al. (2000).

Furthermore, Xue et al. (2015) applied SEM to analyze the factors for measuring environmental and social influences of subway construction and their interrelationships. Deng et al. (2013) evaluated the capabilities of port logistics among five Chinese coastal port clusters by SEM. In construction safety management, Li and Xiang (2011) investigated the main causes of poor construction site safety using SEM in order to examine the importance of each aspect of the causes. Samee and Pongpeng (2016a) explored the causal relationships among components of construction equipment management, project performance and corporate performance. Samee and Pongpeng (2016b) also performed a survey of Construction Equipment Selection and Contractor Competitive Advantages and analyzed it through SEM. Waroonkun and Stewart (2008) proposed a conceptual model for International Technology Transfer in construction projects in Thailand.

3 Methodology

The aim of this study is to prioritize and highlight the factors that are most affecting construction labor productivity in Iran. To this aim, initially an intensive literature review has been done to identify the factors affecting labor productivity and a set of potential factors was collected. Based on the extracted factors from the literature review, a pretest questionnaire was designed. In order to validate the questionnaire, a pilot study was accomplished with seven experts who have more than 10 years of experience in the Iranian construction industry. The experts were asked to revise and critically comment on the questionnaire structure specifically in relation to the factors affecting labor productivity in the context of Iran. Some unnecessary factors were omitted, and some of the similar factors were combined together. Ultimately, the revised questionnaire with 33 potential factors was distributed to the construction project managers who are involved in the Iranian construction industry and have more than 5 years of experience. The participants were asked to assess the factors, based on the five-point Likert scale, from 1 (not applicable) to 5 (extremely effective). Out of 200 questionnaires, 157 questionnaires were fully completed and returned. Incomplete data were eliminated to ensure that the data set was suitable for statistical analysis. With 152 fully completed responses, we got an overall response rate of 78.5% which is quite reasonable.

Initially, the Cronbach’s alpha test had been done to determine the internal consistency of items in the survey to measure its reliability. Then, an Exploratory Factor Analysis (EFA) was conducted to reduce the number of variables and detect the structure in the relationships between variables in order to classify them. The Analytical Hierarchy Process (AHP) as a decision-making tool and the Structural Equation Model (SEM) as a multivariate analysis technique have been applied in parallel for accuracy and reliability of findings. Based on the EFA results, the two mentioned methods, AHP and SEM, were implemented and analyzed. A confirmatory factor analysis (CFA) was conducted to confirm the factor structure extracted from the EFA.

Four types of software have been used in this study; EFA analyzed by IBM SPSS Statistics 20 which is the most relevant statistics tool in research; AHP analyzed with Expert Choice 11 as a multicriteria decision-making software; SEM analyzed through IBM SPSS AMOS 22 and Microsoft Excel.

4 Data Analysis

4.1 Exploratory Factor Analysis (EFA)

Factor analysis is a class of multivariate procedures which aim to identify the underlying structure in a data matrix (Hair et al. 2010), and they also aim to reduce the number of variables and detect the structure in the relationships between variables to classify them.

Initially, Cronbach’s alpha test was done to determine the internal consistency of items in the survey to measure its reliability. The α is 0.898 which is 0.8 ≤ α < 0.9, and according to Field (2009), the reliability is good and it means that the test is 89% reliable (Table 2).

Table 2 Analyzing the reliability of the questionnaire
Full size table

According to Majid and McCaffer (1997), the factors with more than 3.5 mean index are classified in the “Extremely effective” rating group. Consequently, the factors with less than 3.5 score average mean index were removed from the potential factors list. Therefore, the Exploratory Factor Analysis was conducted with twenty factors.

According to Conway and Huffcutt (2003), EFA as an exploratory method has advantages of generating theories and arriving at a more parsimonious understanding of a set of measurement items (Fabrigar et al. 1999). Therefore, factor analysis is performed to analyze the latent relationship between the large numbers of success factors. The KMO and Bartlett’s test attempted to check whether the factor analysis is applicable or not. The KMO measure was 0.824 which should be higher than 0.6, and Bartlett’s test was less than 0.05 and, thus, extremely significant (Table 3). Therefore, the variables have a correlation and EFA is quite applicable.

Table 3 KMO and Bartlett’s test
Full size table

In the following steps of factor analysis, the Principal Component Method of extraction and the Varimax method of rotation have been applied in this study. From Table 4, it is considerable that six extracted factors have eigenvalues greater than 1.00 and these six components, by 61.56% variance, could be representative of 61.56 percent of data. Finally, Table 5 shows the factor loading for each variable.

Table 4 Total variance explained
Full size table
Table 5 Rotated component matrix
Full size table

4.2 Analytical Hierarchy Process (AHP)

The process of AHP has been followed according to Saaty (2008). First, describing and determining the objective of study, which in this study is factors affecting labor productivity. Secondly, the hierarchy from the top to the criteria and alternatives, as it is shown in Fig. 1. The factors have been categorized in a total of 6 components based on the Exploratory Factor Analysis findings and according to the rotated component matrix (Table 5). Moreover, an interrelated component name was selected for each set of factors. Thus, the hierarchy structure of factors affecting labor productivity (Fig. 1) was designed based on EFA components.

Fig. 1
figure 1

Hierarchy structure of factors affecting labors productivity

Full size image

The third step of AHP is constructing a set of pairwise comparison matrices; in this step an AHP questionnaire based on the hierarchy structure (Fig. 1) was designed and distributed among the experts who are project managers of the construction companies. The participants have been asked to rank the relative importance of each of the criteria and sub-factors (alternatives) on a scale of 1–9 in order to make the pairwise comparison, based on the nine-point scale by Saaty (1994) (Table 6). A total of 25 questionnaires were distributed, and 18 of them were returned. The feedback questionnaire from professionals was estimated by using the Consistency Index (CI) and Consistency Ratio (CR) to ensure their reliability and validity (Saaty 2008). Hence, the Inconsistency Ratio was calculated for each respondent. Six respondents were rejected because their CR was less than 10 percent. Therefore, the analysis was continued on the remaining 12 respondents.

Table 6 Nine-point scale by Saaty (1994)
Full size table

Only a Consistency Index (CI) and Consistency Ratio (CR) of less than 0.1 can be acceptable. If it is more than 0.1, it means there is inconsistency in pairwise comparison (Saaty 1994). The CI and CR would be calculated by the following equations:

$${\text{Consistency Index }}\left( {\text{CI}} \right) = \frac{{\lambda { \hbox{max} } - {\text{n}}}}{{{\text{n}} - 1}}$$
(1)
$${\text{Consistency}}\;{\text{Ratio}}\;({\text{CR}}) = \frac{\text{CI}}{\text{RI}}$$
(2)

where λmax: highest eigenvalue, n matrix size.

Judgment consistency could be checked by taking CR of CI with suitable value in Table 7.

Table 7 Average random consistency (RI) (Saaty 1980a, b, 1994)
Full size table

The judgments of several individuals should be combined to obtain a single judgment for the group. Judgments must be combined so that the reciprocal of the synthesized judgments is equal to the syntheses of the reciprocals of these judgments. It has been proved that the geometric mean, not the frequently used arithmetic mean, is the only way to achieve this (Saaty 2008). Accordingly, the geo mean of the twelve responses was calculated by Excel to synthesize them. Geo mean can be calculated by the following formula:

$$P_{i} = \sqrt[n]{{\mathop \prod \limits_{j = 1}^{n} a_{ij} }}$$
(3)

where aij: comparison between object i, P i : priority of object i, n matrix size.

Therefore, all individual judgments combined to a single synthesized judgment. The synthesized judgment was imported to the Expert Choice software in order to analyze and find out the priorities of the criteria with respect to the goal of the study. Additionally, the Inconsistency Ratio has been shown for the criteria and sub-criteria, which should be approximately 0.1 or less (less than 10 percent).

Table 8 shows the overall assessment for the criteria with respect to the aim of the study, which was to prioritize factors affecting labor productivity. According to this table, the “Labor Characteristics,” by 0.384 weights, is the most significant criteria, and then “Tools & Equipment” and “Management” with the same weights of 0.191 are the second and third dominant criteria. In addition, the Inconsistency Ratio for the criteria has been analyzed, which by 4 percent is quite reasonable and it is less than 10 percent. Moreover, Misunderstanding between labors, Delay and Safety and Communication are ranked as the fourth to sixth priority, respectively.

Table 8 Priority weights of criteria and sub-criteria
Full size table

Similarly, the pairwise comparisons were made for each of the five criteria’s sub-factors. The Inconsistency Ratio for each group has been checked as well. All of the CRs were less than ten percent and all were acceptable. In the “Tools and Equipment” group, “Lack of required tools and/or equipment” was the most significant sub-factor in this group by 0.444 weights. In the “Labor Characteristics” group, “Absenteeism,” with the weight of 0.388, was the dominant sub-factor. In the “Management” group, the “Control delay” with 0.297 weights was ranked as the significant sub-factor. Based on the findings, between the “Delay” sub-factors, the “Payment delays” with the weights of 0.691 was the most superior sub-factor. Finally, the “Ambiguity of project objective” was the most significant sub-factor among the Safety and Communication’s sub-factors (Table 8).

4.3 Structural Equation Modeling (SEM)

4.3.1 Confirmatory Factor Analysis (CFA)

In the Structural Equation Modeling, the first step is validating the measurement model and the second step is the assumed structural model testing. Confirmatory factor analysis (CFA) is a pure measurement model containing un-gauged covariance between each of the possible latent variable pairs. Hence, the CFA was conducted as a measurement model, and also to confirm the factor structure extracted in the EFA. According to Hair (2010), a single variable should be removed from the Structural Equation Model. Therefore, a single factor, namely “Misunderstanding between Labors,” was removed from the Sub-Structural Equation Model (Fig. 2). The modified measurement model is shown in Fig. 3.

Fig. 2
figure 2

Sub-Structural Equation Modeling of labor productivity

Full size image
Fig. 3
figure 3

Final measurement model pertaining to “labor productivity”

Full size image
4.3.1.1 Internal Consistency

The Internal Consistency of all latent variables is determined by Cronbach’s alpha (α), before the initiation of CFAs. Consequently, the Cronbach’s alpha (α) calculated for Tools and Equipment, Labor Characteristics, Management, Delay and Safety and Communication is shown in Table 9.

Table 9 Cronbach’s alpha (α) of each latent variable
Full size table

Cronbach’s alpha with a value of more than 0.7 is considered as “acceptable,” and the range from 0.6 to 0.7 is “questionable”; most of the values here are almost close to 0.7 or more than 0.7, and according to Loewenthal (2001), Cronbach’s alpha from 0.6 and 0.7 is not hopeless.

4.3.1.2 Discriminant Validity

One of the limitations of factor analysis is how to name the factors and it may be challenging. Factor names may not precisely reflect the variables within the factor, or “Split loading” which is known for interpretation difficulties of some variables, because they may load to more than one factor (Yong and Pearce 2013). These variables might be correlated with others to create a factor in spite of having underlying meaning to the factor (Tabchnick and Fidell 2006).

Discriminant validity discovers which factors are distinct and uncorrelated. In other words, variables should relate more strongly to their own factor than to another factor. One of the methods to examine discriminant validity is the Factor Correlation Matrix. Hence, the Factor Correlation Matrix is applied by Principal Axis Factoring for the extraction method and Promax is applied for the Rotation method. Promax is normally applicable when researchers are not certain. Correlations between factors should not exceed 0.7, and if it is greater than 0.7, it indicates a majority of shared variance (Gaskin 2012). As it is shown in Table 10, there is no correlation greater than 0.7, which suggests that the factors are not correlated and they are valid.

Table 10 Factor Correlation Matrix
Full size table
4.3.1.3 Model Fit

In order to improve the model fit, first, variables which had Standardized Regression Weights of less than 0.5 were removed. Those variables are, namely “Absenteeism” (abs) and “Ambiguity of Project Objective” (aop), which were eliminated from Sub-Structural Equation Modeling in Fig. 2 and modified in Fig. 3. The second step for improving model fit is adjusting the covariance (Modification indices). To this aim, an appropriate goodness-of-fit index of Structural Equation Modeling is used to confirm the model fit. Finally, the satisfactory structural model is identified and assessed by Modification indices. Model fit indicators are comprised of: p value, relative Chi-square (χ2/df), goodness-of-fit index (GFI), incremental fit index (IFI), comparative fit index (CFI), root mean square error of approximation (RMSEA) and Tucker–Lewis coefficient (TLI). The criterion values of goodness-of-fit and goodness-of-fit indices are shown and compared in Table 11. Model fit indicators are shown as the following formulas (Bentler 1990; Bentler and Bonett 1980; Bentler and Raykov 2000; Bollen 1989; Brown and Cudeck 1993; Hu and Bentler 1999; James 2011; Jöreskog and Sörbom 1984; Tanaka and Huba 1985):

Table 11 Goodness-of-fit criteria and goodness-of-fit indices for SEM
Full size table

Likelihood Ratio χ2 Chi-squared Test (baseline vs saturated models):

$$Xbs^{2} = 2\left\{ { log Ls - log Lb} \right\}$$
(4)

Likelihood Ratio χ2 Chi-squared Test (specified vs saturated models):

$$Xms^{2} = 2\left\{ { log Ls - log Lm} \right\}$$
(5)

where Lb: Log likelihood for the baseline model, Ls: Log likelihood for the saturated model, Lm: Log likelihood for the specified model.

dfbs = dfs − dfb

dfms = dfs − dfm

$${\text{CFI}} = 1 - \frac{{Xms^{2} - dfms}}{{Xbs^{2} - dfbs}}$$
(6)
$${\text{RMSEA}} = \sqrt {\frac{{\left( {Xms^{2} - dfms} \right) }}{{\left( {N - 1} \right)dfms}}}$$
(7)
$${\text{TLI}} = \frac{{\left( {\frac{{Xbs^{2} }}{dfbs}} \right) - \left( {\frac{{Xms^{2} }}{dfbs}} \right)}}{{\left( {\frac{{Xbs^{2} }}{dfbs}} \right) - 1}}$$
(8)
$${\text{IFI}} = \Delta2=\frac{{{\hat{\text{c}}{\text{b}}}-{\hat{\text{c}}}}}{{\hat{\text{c}}{\text{b}}} - {\text{d}}}$$
(9)

where Ĉ and d: discrepancy and the degrees of freedom for the model being evaluated, Ĉb and db: discrepancy and the degrees of freedom for the baseline model

$${\text{GFI}} = 1 - \frac{{F{{\hat{ }}}}}{{F{{\hat{ }}}b }}$$
(10)

\(F{{\hat{ }}}\) minimum value of the discrepancy function, \(F{{\hat{ }}}b\): evaluating F with ∑(g) = 0, g = 1,2,…,G.

4.3.2 Structural Equation Model of Labor Productivity

As mentioned above, the Structural Equation Model includes a measurement model and a structural model. The measurement model displays how latent variables are measured by observed variables (Fig. 3) and relationships between those latent variables are demonstrated by the structural model. In this step, the Structural Equation Model (Fig. 4) is examined to explore the causal relationship based on the five hypotheses as shown below:

Fig. 4
figure 4

Structural Equation Modeling of labor productivity

Full size image

  • H1: “Labor Productivity” has a positive relation with “Tools & Equipment”

  • H2: “Labor Productivity” has a positive relation with “Labor Characteristics”

  • H3: “Labor Productivity” has a positive relation with “Management”

  • H4: “Labor Productivity” has a positive relation with “Delay”

  • H5: “Labor Productivity” has a positive relation with “Safety & Communication”

In order to accept the alternative hypothesis, the p value should be less than 0.05. The Hypothesis test and Standardized Regression Weights of latent variables are presented in Table 12 as the overall final structural model. According to this table, all five hypotheses have a p value of less than 0.05 and were accepted. The Hypothesis test revealed that Sample data supported the hypotheses. Moreover, Standardized Regression Weights of latent variables are as follows, in descending order shown in Table 12: Delay (0.832), Tools and Equipment (0.822), Safety and Communication (0.726), Management (0.622) and Labor Characteristics (0.505).

Table 12 Hypothesis test and Standardized Regression Weights of latent variables
Full size table

5 Findings and Discussion

From the AHP findings, the authors discovered that “Labor Characteristics,” “Tools & Equipment” and “Management” are the most dominant group affecting labor productivity in the Iranian construction industry. Subsequently, the sub-factors are ranked as the following; “Lack of required tools and/or equipment,” “Absenteeism,” “Control Delays,” “Payment delays” and “Ambiguity of Project objective.”

According to Jay and Render (1993), Labor Characteristics include skills, experience, satisfaction and motivation, and they considered the Labor Characteristics to be one of the major productivity groups. An experienced management team with proper supervision and leadership has a direct critical impact on labor productivity. On the contrary, an unskillful manager leads the organization and project to loss of productivity. Therefore, “Management” is recognized as one of the main categories affecting labor productivity in previous studies (Abdul Kadir et al. 2005; Horner and Talhouni 1993; Jarkas and Bitar 2012). On the other hand, the selection of the appropriate type and size of construction equipment often affects the required amount of time for the project, so it is vital for site executives to be aware of the features of the main kinds of tools that are usually used in the construction sites. In order to increase the productivity rate of the construction sites, it is propitious to choose tools with good features and to ensure that their dimensions are the most appropriate for the work circumstances at a structure place. By providing imperfect tools, the productivity rate may be affected negatively (Lim and Alum 1995; Yates and Guhathakurta 1993).

Through the SEM analysis, five main latent variables and their sub-factors (observed variables) were analyzed through the path analysis to determine the relationships between variables. To this aim, a sub-structural model was designed based on the Exploratory Factors Analysis. The sub-factors were categorized into groups according to the EFA results. A Sub-Structural Equation Model was designed, and various indexes such as p value, χ2/df, GFI, IFI, CFI, RMSEA and TLI were compared with the standard criterions to check the goodness of fit between sampled data and the model. The results from Table 11 indicated that both the measurement model and the Structural Equation Model proved their goodness of fit satisfactorily and, therefore, the proposed framework is supported. Five proposed hypotheses were examined by the Hypothesis test, and all p values of hypotheses were less than 0.05 and were accepted (Table 12). From the SEM analysis, “Delay,” with 0.832 Standardized Regression Weights, is the most significant factor. Subsequently, the “Tools & Equipment,” “Safety & Communication,” “Management” and “Labor Characteristics” are the important factors with 0.822, 0.726, 0.622 and 0.505 Standardized Regression Weights, respectively.

Findings from AHP and SEM were compared and are revealed in Table 13. According to this table, “Tools and Equipment” has been selected as the most common significant factor in both AHP and SEM methods. From the AHP analysis findings, “Tools and Equipment” was discovered as the second most prioritized criteria in level 2. Additionally, “Lack of required tools and/or equipment” is the most significant sub-criteria in level 3. Similarly, from the SEM findings, “Tools & Equipment” was selected as the second most prominent latent variable and “Lack of required tools and/or equipment” as the second most significant observed variable as well. Hence, Tools and Equipment has a significant and direct impact on the construction labors productivity. Lack of proper tools or out of service equipment has a negative impact on the labor productivity. Dai et al. (2007) found that “misplaced tools,” “restrictive policy on consumables,” “poor tool quality” and “lack of extension cords” have a significant impact on construction productivity. Tools are mainly provided to the craftsmen who are involved on a full-time basis (Alinaitwe et al. 2007). Productivity depended on efficient usage of tools and equipment; hence, a lack of proper tools and equipment would have a critical impact on labor productivity (Mahmood Zakeri et al. 1996). Kaming et al. (1997a) discovered that lack of equipment and tools is one of the specific productivity problems in Indonesia.

Table 13 Comparative summary in descending order
Full size table

Moreover, in spite of the fact that “Control delay” was discovered as the significant sub-factor in the Management group in the AHP analysis, “Delay,” with the highest regression weights, is the most significant latent variable through the SEM analysis as well. Although “Delay” was not the most significant criteria in AHP, it has been chosen as the most significant latent variable in SEM. In addition, in both AHP and SEM, “Control Delays” was selected as the third significant factor (Table 13). Delay in construction could be contained: Project delay, Payment delays, Inspection delays, Supervision delays, Delay in responding to requests for information (RFI), etc. Enshassi et al. (2007) identified Payment delays as one of the most significant factors affecting labor productivity. Zakeri et al. (1996) discovered “Inspection delay” as one of the predominant factors influencing Iranian construction operative’s productivity. Kaming et al. (1997b) identified Supervision delays as one of the factors influencing craftsmen in Indonesia. Furthermore, Change order by consultants causes Project delay (Abdul Kadir et al. 2005).

6 Conclusion

Since the construction industry is labor intensive and improving labor productivity has a direct and effective impact on project time, cost and quality, the labor productivity issue is of remarkable interest in both the construction industry and academia. This study attempted to prioritize and highlight the factors most affecting construction labor productivity in Iran. Hence, the Analytical Hierarchy Process (AHP) and Structural Equation Model (SEM) were applied as the analytical tools. The results from both AHP and SEM were discovered and compared in parallel for accuracy and reliability of findings. Eventually, “Labor Characteristics” was selected as the most prioritized criteria. From the compared outcomes it was found that the most common significant factors influencing construction labor productivity in Iran are “Tools & Equipment” and “Delay.” There is a need to notice, inform and train our foremen and sub-contractors about the importance of productivity issues. It is also necessary for construction sites to be well equipped with the latest modern tools and equipment. Indeed, controlling and reducing delay has the ability to increase the labor productivity and the time–cost deduction as well. The results of this study would be useful for civil engineers, construction project managers, consultants, contractors and any parties who are involved in Iranian or Middle Eastern construction projects, based on the similar structure of construction sites of that area.