Research for Risk Management of Construction Projects in Taiwan

Abstract

Risks inevitably exist in all stages of a project. In a construction project, which is highly dynamic and complex, risk factors affect the expected achievement rates of the three main performance goals, namely schedule, cost, and quality. A comprehensive risk management procedure requires three crucial steps: risk confirmation, analysis, and treatment. Risk analysis is the core of risk management. Through structural equation modeling, this study developed a risk analysis model that takes a different perspective and considered the occurrence probability of risk events and the extent to which these events affect a project. The contractor dimension was discovered to exert the strongest influence on an overall project, followed by the subcontractor and design dimensions. This paper proposes a novel construction project risk analysis model, which considers the entire project. The proposed model can be used as a reference for risk managers to make decisions about project risks, so as to achieve the ultimate goal of saving resources and the sustainable operation of the construction project.

1. Introduction

The rapid development of economic infrastructure in Taiwan and the complex content of, considerable investment in, high technical requirements of, and long construction periods in construction projects have posed various risks to units involved in such projects. These risks, if not mitigated properly, hinder the process of construction, increasing operating costs and undermining construction quality. Risk problems are critical variables in a construction project; the existence of risk leads to uncertainty in the final cost, duration of construction, and construction quality [1]. Therefore, the purpose of risk management is to reveal factors that could negatively influence the expected project cost or quality and to facilitate appropriate countermeasures for alleviating these risk factors [2].

The tendering for construction projects in Taiwan has so far been conducted through either turnkey or subcontracting contracts. Subcontracted construction projects are not usually delivered on time because design flaws can lead to design–construction interface problems, resulting in late deliveries by subcontractors. Additionally, workplace accidents that occur in construction projects account for more than 60% of all workplace accidents in Taiwan, and this figure shows no sign of decline. Such numerous workplace accidents are attributable to construction design flaws, incomplete health and safety facilities, improper inspection practices, and on-site workers having insufficient knowledge of workplace safety, leading to countless property losses, casualties, and deaths. More crucially, on-site workers usually lack knowledge of engineering ethics.

Accordingly, problems occurring during the risk management of Taiwanese construction projects must be reviewed. To identify the cause of these problems, a systematic risk management model should be established at the aggregate level of a project [3,4]. The model should involve appropriate risk identification, risk analysis, risk assessment, risk response strategies, and risk control [2,3,4] and should be adopted to continually improve the operating performance of businesses.

The present study hypothesized that a construction project is affected by various risk factors that can cause each other through circular causality; therefore, problems can snowball into intractable situations when appropriate risk management is not in place.

This study identified the most relevant risk factors for construction projects from a comprehensive perspective by examining both the actual on-site operations of, and research findings on, Taiwanese construction projects and categorized the risk factors into five dimensions. The relative importance and incidence of each risk factor were determined, and the factors were ranked according to the quantitative results. These rankings revealed the extent to which each risk factor affected an overall project and thus will enable decision makers to formulate preventive, control, and management measures in advance. Additionally, the findings may provide a reference for the industry as to how risk management procedures and response strategies can be employed appropriately throughout an entire construction project to achieve the project goals.

Risk-related terms are used in this article, and their definitions are as follows: Risk refers to the possibility of a certain loss occurring in a certain period of time under a certain environment. Event refers to a single event or matter. Factors refer to the reasons or conditions that determine the success or failure of things. The Risk category refers to grouping risks according to characteristics groups, and dimensions have the same meaning.

2. Literature Review of Risk in Construction Projects

The primary statement of construction project risk management theses from numerous scholars in the past is as follows: Construction projects are considered a highly risky line of business with a complex and dynamic nature, which has given rise to their association with high uncertainty and risk [3]. Risks are a critical factor affecting whether the primary goals of a construction project can be achieved, such as goals concerning construction duration, cost, quality, safety, and environmental sustainability [2,3,4]. Risks and uncertainty may lead to destructive consequences in a construction project [5]. Risks are measures of how likely unpredictable events or accidents are to lead to negative deviations from the expected length of construction or estimated cost; poor risk management is the most critical reason for a project’s failure [6]. According to the Project Management Institute, a project risk is defined as an uncertain event that, if it occurs, will at have a positive or negative outcome on project objectives such as scope, cost, time, and quality [7]. Therefore, effective risk management is strongly positively correlated with project success [8].

To achieve construction project goals relating to the construction duration, cost, quality, safety, and environmental sustainability, risk management has been considered a critical procedure within project management [3,9].

Risk management which involves the appropriate use of finances and the coordination of resources as well as the formulation of response measures after risk confirmation, assessment, and prioritization can reduce the likelihood that negative events will occur, mitigates the impact of such events, and maximizes the achievement of project goals [10]. The ultimate objective of risk management is to confirm which risk factors exist and to develop suitable risk response strategies [11].

Project risk management must involve major procedures such as risk identification, risk analysis, and the development of risk response strategies. In particular, risk response strategies involve risk confirmation, risk assessment, and the selection and execution of effective actions to reduce risks, alleviate the negative influence of negative events, and thus enhance the benefits of projects [3].

In Rehacek’s paper [12], the risks are divided into three dimensions, which also include the five dimensions referenced in this study. In this study, the risks that have frequently occurred in Taiwan today were analyzed according to the interviews with the senior engineers and owners of consulting companies related to construction engineering (e.g., design, manufacturing supervision, and project management), summarizing these risks into the following five dimensions: (1) client, (2) design, (3) contractor, (4) subcontractor, and (5) external risk factors.

Accordingly, the main steps in risk management are as follows: risk confirmation, risk analysis, risk response strategies, and risk control [7,13,14]. Risk control is the continual process of analyzing and assessing risk factors [4].

The main purpose of risk confirmation is to identify all risk factors that could influence a project by conducting qualitative analysis of the relevant risk factors; such analysis is usually performed by considering personal experience and using brainstorming, expert interviews, or a questionnaire [12,13]. Risk analysis mainly involves the assessment (i.e., quantitative analysis) of the confirmed risk factors and aims to determine the consequences of each risk factor and the extent to which the factor may affect a project in the future [9,14]. Risk confirmation and analysis help decision makers respond, take action, and make appropriate judgements when problems occur [8,9].

Risk response strategies are a crucial implementation stage that is dependent on the risk identification and assessment results, and the objective of using these strategies is to reduce project risks and improve project efficiency [15]. The implementation of risk response strategies through risk avoidance reduces the probability that negative events will occur and weakens the negative impact of these risks, thereby facilitating risk transfer, risk sharing, and risk acceptance [3] and improving a project’s value and implementation efficiency [11,15,16]. Two other strategic actions for risk response are risk prevention and risk adaptation [11,17].

According to the literature, construction project risk management (CPRM) is a systematic dynamic risk management model that—through a continual circular process of risk confirmation, analysis, assessment, control, and feedback—enables the making of appropriate decisions, the use of appropriate resources, and a favorable communication and coordination. The use of CPRM enables the development of preventive measures, continual improvement, and reduction of the influence of risk events on a project and is thus conducive to achieving project goals.

Focusing on risk factors, this study developed a systematic quantitative method for analyzing and evaluating potential losses and severe consequences caused by risk factors from a comprehensive perspective. This method may provide a reference for project risk managers in relevant decision making and create a new perspective on the continual improvement of project risk management.

The purpose of this study was to conduct risk factor identification, analysis, and assessment and to design response measures for CPRM in Taiwan; to bridge the gap in research related to risk assessment and evaluation models; and to continually improve CPRM by coordinating the inconsistent viewpoints of construction project risk managers.

3. Research Methodology

Studies related to risk management have predominantly employed general regression, mean, standard deviation, and relative importance as their research analysis methods [9,13,18]. Adeleke et al. [7] used partial least path modeling to discuss how external factors affect the risk management of construction projects in Nigeria. Studies have also employed the analytic hierarchy process [13,19,20] and Monte Carlo simulation [21,22] to explore problems in CPRM.

A construction project is typically implemented in a dynamic and risky environment throughout the entire project period and is thus affected by various factors associated with highly uncertain risks [2,3]. The risk factors encountered by a project are not independent from other factors, and a causal relationship may exist between two types of factors. Therefore, the present study used structural equation modeling (SEM) to analyze the probability of a negative event occurring and the extent to which these events affect a project; an assessment method was developed for determining the consequences of adverse events, thus providing a new perspective on CPRM.

SEM is a statistical analysis method combining multiple regression and factor analysis; the strength of SEM is mainly its ability to completely reflect the different levels of influence of observed variables (i.e., risk factors) on latent variables (i.e., risk factor dimensions) as well as the causal relationships between latent variables [23]. SEM with the optimal goodness of fit can be constructed by establishing a theoretical hypothetical model, conducting related statistical tests on this model [23], and then modifying the model accordingly [24].

SEM comprises a measurement model and structural model [25,26,27]. The measurement model is used to verify the relationship between observed and latent variables [23] and to determine the level of influence between these variables. The structural model is composed of various measurement models and is used to analyze the relationship between endogenous and exogenous latent variables [23], determining the causal influence between latent variables [24].

In SEM, factor loadings (FLs) are calculated, which are the measured direct effect of the particular observed variables (independent variables, questionnaire items) on latent variables (dependent variables) and also the regression coefficient weight [24]. The size of an FL reveals the extent to which the independent (observed) variable affects the dimension factor (latent variable).

On the basis of the literature review, risk analysis of Taiwanese construction projects, analysis of the questionnaire results, and discussion of construction workers’ experiences, this study established a CPRM model by using SEM. The SEM risk management procedure for construction projects is displayed in Figure 1.

The procedure as described in Figure 1 is based on [28], page 129. The description that “Risk management is the systematic process of identifying, analyzing, and responding to project risk” has been provided, and SEM has been conducted to establish an analysis model and provide a conclusion and relevant risk management implications.

In the proposed procedure, all steps were strongly positively correlated with the risk factor analysis performed in step 2 (assessment and evaluation of the influence level), because risk management is a continual process of improvement throughout an entire project period. To achieve project goals, risk assessment must be conducted continually, and the risks continually monitored and controlled.

Project success relies on the confirmation of the risk factors and predictions of the probability of an adverse event occurring [15]. The level of influence of a risk factor on a project has often been investigated according to the rule of the SLAC National Accelerator Laboratory [29], who suggested that a project risk mitigation plan must reduce not only the probability (P) of risk event occurrence but also the strength of influence of the risk event to an acceptable level (I) and proposed the following equation for evaluating risk severity (S): severity (S) = probability (P) × level of influence (I). Mills [14] believes that risk factors are a quantitative measurement method, which can be regarded as the product of the probability of future events and the degree of impact of the event; the formula is as follows: $RI=L\times C ,$ where RI is the influence of the risk event, L is the probability that the risk event will occur, and C is the consequence of the risk event. Rehacek [12] proposed a risk assessment method, which is a qualitative measurement method that divides PI areas into nine categories. The formula is as follows: PI factor = Probability of risk × Impact of risk. According to Vlăduț-Severian [30], the losses (R) caused by a potential risk event in a project are the product of the risk severity (G) and occurrence probability (P); the formula is as follows:$R=G\times P$ [30].

On the basis of previous research by the Stanford Linear Accelerator Center [12,14,29,30], this research proposes a method to integrate the two-stage new risk measurement formula. Firstly, calculate the direct impact of individual facet risk factors on the facet result (Individual dimension measurement mode), and secondly, it measures the final impact of the risk factors in the integrated SEM model dimension on the final risk assessment model of the entire construction project. The proposed evaluation method is as follows:

$$R_x=P\times I\times L \left(x=1,2\right)$$

(1)

$R_1=P\times I$ is the consequence caused by a risk factor to a particular dimension in the measurement model of risk management. $R_2=P\times I\times L$ is the consequence caused by a risk factor to a project overall in the structural model of risk management. P is the probability that a risk factor will occur. I is the FL of a risk factor in the measurement model, specifically the extent to which this risk factor affects the particular dimension. L is the path coefficient in the SEM model (Figure 2 and

Table 1. FL, influence-level ranking, and reliability for the five dimensions in the finalized measurement model for construction risks.

Item	Measurement Models and Risk Factors	FL * (I = FL) (2)	Influence Level and Ranking of Risk Factors ${\mathit{R}}_1=\mathit{P}\times \mathit{I}$			Influence Level and Ranking of Models ${\mathit{R}}_2={\mathit{R}}_1\times \mathit{L}$
			P(1)	${\mathit{R}}_1\left[\left(1\right)\times \left(2\right)\right]$ (3)	Rank	${\mathit{R}}_2\left[\left(3\right)\times \mathit{L}\right]$	Rank
I	Measurement model for the client dimension  L  L = 0.64 (coefficient of path)
O3	Change in decisions	0.71	0.31	0.22	3	0.141	3
O4	Poor…communication	0.79	0.32	0.25	1	0.162	2
O7	Difficulties in design–construction coordination	0.67	0.34	0.23	2	0.146	4
O8	Difficulties in quality control with low budget	0.65	0.31	0.20	4	0.129	5
O9	Late submission of remodified or approved documents	0.66	0.35	0.23	2	0.148	1
O10	Late completion due to design–construction interface problems	0.61	0.33	0.20	4	0.129	5
II	Measurement model for the design dimension  L  L = 0.83 (coefficient of path)
D1	Insufficient design time	0.68	0.27	0.18	8	0.149	9
D2	Insufficient design information	0.77	0.35	0.27	2	0.224	2
D3	Insufficient number of designers	0.69	0.34	0.23	5	0.195	6
D4	Lack of experience and knowledge	0.83	0.30	0.25	4	0.207	4
D5	Insufficient understanding of the actual construction procedures	0.73	0.36	0.26	3	0.218	3
D6	Lack of a design standard	0.66	0.35	0.23	6	0.192	7
D7	Lack of communication and coordination in the design–construction interface	0.71	0.40	0.28	1	0.236	1
D8	Incomprehensive coordination in the design–construction interface	0.71	0.35	0.25	4	0.206	5
D9	Insufficient budget	0.69	0.27	0.19	7	0.155	8
III	Measurement model for the contractor dimension  L  L = 0.96 (coefficient of path)
C4	Insufficient competencies of construction workers	0.76	0.34	0.26	5	0.249	5
C5	Failure to integrate construction equipment, materials, and techniques	0.71	0.41	0.29	2	0.279	3
C6	Work correction due to incorrect construction operations	0.78	0.36	0.28	3	0.269	4
C7	Inappropriate scheduling of construction operations	0.78	0.34	0.27	4	0.254	6
C8	Poor on-site supervision and management	0.83	0.39	0.32	1	0.311	1
C9	Incorrect construction operations due to careless reading of construction drawings	0.82	0.36	0.29	2	0.283	2
C10	Failure to provide a detailed construction diagram	0.78	0.33	0.26	5	0.247	7
C11	Perfunctory establishment of health and safety facilities	0.74	0.32	0.24	6	0.227	8
IV	Measurement model for the subcontractor dimension   L  L = 0.86 (coefficient of path)
SC1	Lack of specialized labor	0.61	0.35	0.21	5	0.184	7
SC3	Lack of correct concepts of health and safety	0.63	0.38	0.24	3	0.206	4
SC4	Poor communication and coordination with contractors	0.45	0.38	0.17	6	0.147	8
SC6	Work correction due to incorrect construction operations caused by careless reading of construction drawings	0.97	0.42	0.41	1	0.350	1
SC7	Lack of educational training for operators	0.88	0.34	0.30	2	0.257	2
SC8	Ignorance of construction work ethics	0.73	0.31	0.23	4	0.195	6
SC9	Poor on-site management	0.79	0.38	0.30	2	0.248	3
SC10	Construction operations conducted according to personal experience and in violation of relevant regulations	0.81	0.29	0.23	4	0.202	5
V	Measurement model for the external factor dimension L L = 0.47 (coefficient of path)
Ex6	Legal restrictions	0.73	0.36	0.26	2	0.123	2
Ex7	Manufacturing techniques	0.87	0.31	0.27	1	0.127	1
Ex8	Predatory pricing by competitors	0.69	0.25	0.17	4	0.081	4
Ex9	Creditor requirements	0.65	0.27	0.18	3	0.082	3

Note: L is the path coefficient (i.e., factor loading), from the center latent variable (CRF) to five measurement model factors in Figure 2. Figure 2 depicts the finalized SEM measurement and structural models after model modification and goodness-of-fit testing. The FL is the extent to which each risk factor affects the corresponding dimension. R_x denotes a risk factor’s potential influence on the corresponding dimension, whereas R_xy is a risk factor’s latent influence on the overall project risk (Table 1). (The arrow symbol means the flow of the process.)

)

4. Application of SEM to Risk Management

This study established an SEM for project risk management for assessing and analyzing the level of influence of various risk factors proposed in a risk management plan.

Generally, an SEM is established by using the following procedure: (1) the research topics and theories are confirmed, (2) the model is specified, (3) the measurement variables are confirmed and questionnaire data collected, (4) the model is estimated and modified, and (5) analysis and discussion are performed. During model establishment, research topics and theories are confirmed through a literature review. Specifically, the theoretical hypothetical model and correlations between variables and dimensions are confirmed.

5. Sample Data Analysis

By reviewing relevant recent papers [2,3,4,7,8,12,14,20,30,31,32,33,34] and analyzing interviews with senior construction workers to discover their work experiences, this study identified the 54 most common risk factors on construction sites in Taiwan. These factors were in five categories: (1) client, (2) design, (3) contractor, (4) subcontractor, and (5) external risk factors.

This study administered a questionnaire to different parties involved in Taiwanese construction projects to determine which construction risk factors these parties had encountered or learned about. The questionnaire was scored using a 5-point Likert scale (from 5 = extremely important to 1 = not at all important). The respondents were professionally related to construction projects in Taiwan. A total of 250 questionnaires were distributed, with 211 valid responses returned, achieving a valid response rate of 84.4%; this met the requirement proposed by Marsh [35] and Boomsma [36] that a valid sample size must be larger than 150 and should ideally be larger than 200. The valid sample had the following structure: (1) in terms of work roles in the construction industry, 16%, 42%, 27%, and 15% of the respondents were supervisors, field engineers, design engineers, and project management engineers, respectively; (2) in terms of work experience in the construction industry, 11%, 30%, 31%, and 28% of the respondents had 0–5, 6–10, 11–15, and 16 or more years of experience, respectively.

The initial statistical analysis of the valid responses revealed the following. (1) Regarding the multivariate normal distribution, the absolute value of the highest kurtosis was less than 7, and that of the highest skewness was less than 2, conforming to the properties of a multivariate normal distribution [37]. (2) In reliability analysis, the internal consistency coefficients (Cronbach’s α) of the dimensions were found to be greater than 0.8, which conformed to the internal consistency coefficient requirement for questionnaires [38]. The Cronbach’s α of the dimensions was as follows: client (0.877), design (0.907), contractor (0.917), subcontractor (0.893), and external factors (0.882), indicating the satisfactory internal consistency of the questionnaire (

Table 2. FL, influence-level ranking, and reliability for the five dimensions in the initial measurement model for construction risks.

Item	Measurement Models and Risk Factors	FL * (I = FL) (2)	Influence Level and Ranking of Risk Factors ${\mathit{R}}_1=\mathit{P}\times \mathit{I}$			Reliab-Ility (α)
			P(1)	${\mathit{R}}_1 [\left(1\right)\times \left(2\right)$	Rank
I	Measurement model for the client dimension
O1	Unclear definition of requirement	0.57	0.28	0.16	5	0.877
O2	Lack of personnel	0.36	0.31	0.11	8
O3	Change in decisions	0.67	0.31	0.21	3
O4	Poor departmental communication	0.72	0.32	0.23	2
O5	Delayed schedule and late payment	0.55	0.26	0.14	6
O6	Complex contracting procedures	0.61	0.33	0.20	4
O7	Difficulties in design–construction coordination	0.69	0.34	0.23	2
O8	Difficulties in quality control with low budget	0.68	0.31	0.21	3
O9	Late submission of re-modified or approved documents	0.77	0.35	0.27	1
O10	Late completion due to design–construction interface problems	0.70	0.33	0.23	2
O11	Insufficient information on health and safety facilities	0.43	0.31	0.13	7
II	Measurement model for the design dimension
D1	Insufficient design time	0.76	0.27	0.21	7	0.907
D2	Insufficient design information	0.83	0.35	0.29	1
D3	Insufficient number of designers	0.75	0.34	0.26	3
D4	Lack of experience and knowledge	0.81	0.30	0.24	5
D5	Insufficient understanding of the actual construction procedures	0.73	0.36	0.26	3
D6	Lack of a design standard	0.67	0.35	0.23	6
D7	Lack of communication and coordination in the design–construction interface	0.69	0.40	0.28	2
D8	Incomprehensive coordination in the design–construction interface	0.71	0.35	0.25	4
D9	Insufficient budget	0.70	0.27	0.19	8
D10	Interference from commissioned designers	0.62	0.30	0.19	9
D11	Un-quantified health and safety facilities	0.55	0.31	0.17	10
III	Measurement model for the contractor dimension
C1	Insufficient construction time	0.56	0.39	0.22	7	0.917
C2	Low-price contracting	0.58	0.42	0.24	6
C3	Lack of communication and coordination in the design–construction interface	0.62	0.41	0.25	5
C4	Insufficient competencies of construction workers	0.79	0.34	0.27	4
C5	Failure to integrate construction equipment, materials, and techniques	0.77	0.41	0.32	1
C6	Work correction due to incorrect construction operations	0.81	0.36	0.29	2
C7	Inappropriate scheduling of construction operations	0.80	0.34	0.27	4
C8	Poor on-site supervision and management	0.81	0.39	0.32	1
C9	Incorrect construction operations due to careless reading of construction drawings	0.79	0.36	0.28	3
C10	Failure to provide a detailed construction diagram	0.73	0.33	0.24	5
C11	Perfunctory establishment of health and safety facilities	0.72	0.32	0.23	6
C12	Insufficient finances	0.64	0.27	0.17	8
IV	Measurement model for the subcontractor dimension
SC1	Lack of specialized labor	0.70	0.35	0.25	5	0.893
SC2	Lack of on-site workers	0.57	0.34	0.19	7
SC3	Lack of correct concepts of health and safety	0.73	0.38	0.28	3
SC4	Poor communication and coordination with contractors	0.71	0.38	0.27	4
SC5	Late payment requests	0.45	0.29	0.13	8
SC6	Work correction due to incorrect construction operations caused by careless reading of construction drawings	0.75	0.42	0.32	1
SC7	Lack of educational training for operators	0.84	0.34	0.29	2
SC8	Ignorance of construction work ethics	0.70	0.31	0.22	6
SC9	Poor on-site management	0.77	0.38	0.29	2
SC10	Construction operations conducted according to personal experience and in violation of relevant regulations	0.75	0.29	0.22	6
V	Measurement model for the external factor dimension
Ex1	Political environment	0.52	0.29	0.15	6	0.882
Ex2	Economic environment	0.51	0.27	0.14	7
Ex3	Force majeure events	0.48	0.36	0.17	5
Ex4	Source of material supply	0.64	0.35	0.22	3
Ex5	Client requirements	0.59	0.32	0.19	5
Ex6	Legal restrictions	0.66	0.36	0.24	1
Ex7	Manufacturing techniques	0.75	0.31	0.23	2
Ex8	Predatory pricing by competitors	0.67	0.25	0.17	5
Ex9	Creditor requirements	0.73	0.27	0.20	4
Ex10	Contractors’ low cash flow caused by late payment	0.61	0.25	0.15	6

FL * is a statistical estimated value of the direct influence of an observed variable on a latent variable and is also a regression coefficient. The fourth column represents the initial estimated factor loading (FL); the numbers in bold are those greater than the threshold of 0.65 and are thus those of the main factors of influence selected later in the modified model.

). Accordingly, the questionnaire responses were valid. The modeling model of CPRM was then established and analyzed with reference to the results of the literature review, interviews, and questionnaire survey.

5.1. Application of SEM

In the SEM process, the measurement and structural models were first constructed; this was followed by model specification and measurement evaluation, evaluation of the goodness of fit for the model and model modification, and then finalization of SEM of CPRM.

The SEM proposed in this study is a combined model (a) that includes a measurement model (b) and structural model (c), which are detailed as follows [24,27]:

(a) η = Bη + Γξ

(2)

(b) x = Λ_x ξ + δ

(3)

(c) y = Λ_y η + ε

(4)

(d) η = Γξ+ ζ

(5)

Formula (2), where B is a mediated matrix consisting of Λ_y (the coefficient vector that relates y to η), Γ is a matrix consisting of Λ_x (the coefficient vector that relates x to ξ), η is a vector for endogenous variables, and ξ is a vector for exogenous variables expressing latent errors.

Formula (3), where x is the observed variable, δ is the error term associated with the observed variable (x), Λ_x is the coefficient vector (F.L.) that relates x to ξ, ξ is a vector for endogenous latent variables,

Formula (4), where y is the indicator (observed variables) and ε is the error term associated with the ith observed variable y (indicator). Λ_y is the coefficient (F.L.) vector that relates y to η, ξ is a second-order exogenous latent variable.

Formula (5), η represents the five endogenous latent variables, Γ is the coefficient matrix of the factors loadings (i.e., path coefficient) of the five endogenous latent variables (dimension factors) and the exogenous latent variable (CRF), and ζ is the measurement error of the five endogenous latent variables.

5.2. Establishment of the SEM Model of CPRM

5.2.1. SEM Model Specification

The initial SEM comprised measurements and structural models for five dimensions (i.e., client, design, contractor, subcontractor, and external factors; Figure 2). Table 2 presents the model specification and evaluation results and reveals that the risk factors all influenced the corresponding dimensions. Table 2 details the FLs of the risk factors in the initial SEM measurement models and ranks these factors according to their level of influence (R_xy).

FLs represent the extent to which the observed variables (risk factors) affect the different dimensions (latent variables). According to Table 2, the risk factor exerting the strongest influence on the client dimension was “late submission of remodified or approved documents” (O9), for which the score for the influence level (R₁) was 0.27. The most influential factor for the design dimension was “lack of design information” (D2; R₁ = 0.29). There were two most influential risk factors for the contractor dimension, namely “Failure to integrate construction equipment, materials, and techniques” (C5; R₁ = 0.32) and “poor on-site supervision and management” (C8; FL = 0.81; R₁ = 0.32). The risk factor with the strongest influence on the subcontractor dimension was “work correction due to incorrect construction operations caused by careless reading of construction drawings” (SC6; FL = 0.75; R₁ = 0.32). The risk factor with the highest influence level was “legal restrictions” (EX6; FL = 0.66; R₁ = 0.24).

The FLs and R_xy obtained from the initial SEM measurement models revealed the quantitative estimates of the extent to which each risk factor affected the corresponding dimension. This information could assist managers in the preliminary planning of construction projects as well as the development of measures for preventing potentially adverse events.

5.2.2. Goodness-of-Fit Verification and Modification of SEM

A stable and reliable SEM model can be obtained only when its goodness of fit is higher than a certain standard. This study employed the following seven most commonly used absolute and relative goodness-of-fit indices: χ²/df (chi-squared/degrees of freedom) [39]; goodness-of-fit index (GFI) [40]; comparative fit index (CFI) [41]; root-mean-square residual (RMR) [42]; standardized RMR (SRMR) [43]; composite reliability (CR) [42]; and discriminant validity (e.g., average variance extracted (AVE)) [42].

The initial SEM model, which comprised measurement and structural models for five dimensions, was tested for its goodness of fit and modified using an FL threshold of 0.65 for factor elimination [44]. Thus, the final SEM model of CPRM was obtained (Figure 2). Both the measurement and structural models met the optimal goodness-of-fit requirements (

Table 3. Goodness-of-fit results of the initial and finalized SEM measurement models and finalized SEM structural model.

Model Fit Index	Acceptable Range (No Fit to Perfect Fit)	SEM Measurement Models for the Five Dimensions										Structural Model for the Five Dimensions
		Client Dimension		Design Dimension		Contractor Dimension		Subcontractor Dimension		External Factor Dimension
		Initial Model	Final Model	Initial Model	Final Model	Initial Model	Final Model	Initial Model	Final Model	Initial Model	Final Model
χ2/df	5 to 1	4.454	1.579	7.491	2.358	4.214	2.985	6.365	1.351	6.308	2.502	2.216
GFI	0 to 1	0.853	0.986	0.768	0.953	0.831	0.942	0.787	0.977	0.834	0.988	0.760
CFI	0 to 1	0.828	0.993	0.801	0.978	0.887	0.968	0.838	0.995	0.769	0.990	0.879
RMR	1 to 0	0.090	0.030	0.095	0.046	0.080	0.040	0.091	0.028	0.113	0.037	0.095
SRMR	1 to 0	0.071	0.024	0.075	0.036	0.064	0.033	0.079	0.023	0.083	0.026	0.074
CR	~ ≧ 0.7	0.872	0.856	0.914	0.916	0.910	0.919	0.894	0.895	0.861	0.823	~
AVE	~ ≧ 0.5	0.413	0.503	0.532	0.548	0.508	0.588	0.462	0.520	0.387	0.543	~

Note: 1. This compares the goodness-of-fit results of the initial and finalized SEM measurement models and the structural model; the finalized models were established by eliminating factors in the initial model with FL less than 0.65. 2. Composite reliability (CR) tests the composite reliability of the measurement models; average variance extracted (AVE) refers to the average variance extracted and is used to test the discriminant validity of the measurement models.

). Table 1 presents the FLs of the risk factors and their influence-level rankings. Wan Mohamed Radzi et al. [26] established a firm sustainability performance index model by applying both classical and Bayesian structural equation modeling.

6. SEM Model of CPRM

According to the evaluation and analysis results for the measurement and structural models of the five dimensions, as detailed in Section 4 the SEM of CPRM was obtained (Figure 2, Table 1). The details of this model are as follows.

The dimension that had the most risky influence on a construction project was the contractor dimension, for which the influence level (path coefficient) was 0.96, followed by the subcontractor dimension with an influence level of 0.86, the design dimension with an influence level of 0.83, the client dimension with an influence level of 0.64, and the external factor dimension with an influence level of 0.47. (2) The main influential risk factor in the client dimension was “poor interdepartmental communication” (O4; FL = 0.79; P = 0.32; R₁₄ = 0.23), whose latent influence level on the overall project risk was R₂₄ = 0.147. However, “late submission of remodified or approved documents” (O9) was more likely to occur (P = 0.35; FL = 0.66; R₁₉ = 0.27; R₂₉ = 0.173), the influence level of O9 (R₂₉ = 0.173) on the client dimension was higher than that of O4 (R₂₄ = 0.147).

The main influential risk factor in the design dimension was “lack of experience and knowledge” (D4; FL = 0.83; P = 0.30); its influence level on the dimension was R₁₄ = 0.24, and its latent influence level on the overall project risk was R₂₄ = 0.199. “Lack of communication and coordination in the design–construction interface” (D7; FL = 0.71; R₁₇ = 0.28; R₂₇ = 0.232) had a higher probability of occurrence (P = 0.40) than did D4. Therefore, the influence level of D7 (R₁₇ = 0.28) on the design dimension was higher than that of D₄ (R₁₄ = 0.24); its latent influence level (R₂₇ = 0.232) on the overall project risk was also higher than that of D4 (R₂₄ = 0.199).

The main influential risk factor in the contractor dimension was “poor on-site supervision and management” (C8; FL = 0.83; probability [P] = 0.39), which had an influence level of R₁₈ = 0.32 on the dimension and a latent influence level of R₂₈ = 0.311 on the overall project risk. Although C8 had a lower probability of occurrence than “failure to integrate construction equipment, materials, and techniques” (C5; P = 0.41), these two items had the same influence level on the contractor dimension (R₁₈ = R₁₅ = 0.32) and the same latent influence level on the overall project risk (R₂₈ = R₂₅ = 0.307). (5) The primary influential risk factor in the subcontractor dimension was “work correction due to incorrect construction operations caused by careless reading of construction drawings” (SC6; FL = 0.97; P = 0.42); its influence level on the dimension was R₁₆ = 0.41, and its latent influence level on the overall project risk was R₂₆ = 0.35. (6) The primary influential risk factor in the external factor dimension was “manufacturing techniques” (EX7; FL = 0.87; P = 0.31; R₁₇ = 0.23), and its latent influence level on the overall project risk was R₂₇ = 0.108. Because “legal restrictions” (EX6; FL = 0.73; P = 0.36; R₁₆ = 0.24; R₂₆ = 0.113) had a higher probability of occurrence than did EX7, the influence level (R₁₆ = 0.24) on the external factor dimension and latent influence level (R₂₆ = 0.113) on the overall project risk were both higher for EX6 than EX7 (R₁₇ = 0.23; R₂₇ = 0.108).

According to these evaluation analysis and assessment results, this study recommends that risk reduction strategies in CPRM be developed with reference to the level of influence and relative occurrence probability of each risk factor. (1) A risk event with a low level of influence and low relative occurrence probability can be controlled and prevented in advance; therefore, once it occurs, such risk event should be addressed through retention or absorption measures by capitalizing on relevant resources. (2) A risk event with a low level of influence and high relative occurrence probability cannot be controlled in advance and thus should be addressed through avoidance or monitoring measures. (3) A risk event with a high level of influence and high relative occurrence probability cannot be controlled or avoided in advance and therefore should be addressed through risk transfer or diversification measures. (4) A risk event with a high level of influence and low relative occurrence probability can be controlled and prevented in advance and should therefore be addressed through risk absorption and transfer by using relevant resources.

These recommended risk reduction measures may help CPRM decision makers formulate project risk plans and use resources appropriately to develop risk preventive measures. They can enable decision makers to manage risks, control risks, complete tasks stage by stage, and ultimately meet goals regarding the duration of construction, cost, quality, and environmental sustainability.

7. Conclusions and Recommendations

This study used the SEM method to explore CPRM in Taiwan, and specifically the risk factors affecting five dimensions (i.e., client, design, contractor, subcontractor, and external factor) of risk problems. Each factor’s level of influence on the corresponding dimension and overall project outcome was determined. Using a literature review, questionnaire data, and data consistency testing, an SEM model was established, tested with goodness-of-fit tests, modified, and finalized to obtain the SEM model of CPCRM (Figure 2).

This study obtained the following findings. (1) The contractor dimension (FL = 0.96) had the strongest influence on the overall project risk, followed by the subcontractor dimension (FL = 0.86) and design dimension (FL = 0.83). (2) “Poor on-site supervision and management” (FL = 0.83) was the risk factor with the strongest influence on the contractor dimension; “incorrect construction operations caused by careless reading of construction drawings” (FL = 0.82) was the second most influential factor. (3) Regarding the subcontractor dimension, the most influential risk factor was “work correction due to incorrect construction operations caused by careless reading of construction drawings” (FL = 0.97), followed by “lack of educational training for operators” (FL = 0.88). (4) “Lack of experience and knowledge” (FL = 0.83) had the strongest influence on the design dimension, whereas “lack of design information” (FL = 0.77) had the second strongest influence.

According to the SEM of CPRM for construction projects in Taiwan (Figure 2), the risk factors that should be emphasized in each dimension were as follows: “poor on-site supervision and management” (FL = 0.83) and “incorrect construction operations due to careless reading of construction drawings” (FL = 0.82) in the contractor dimension (FL = 0.96); “work correction due to incorrect construction operations caused by careless reading of construction drawings” (FL = 0.97) and “lack of educational training for operators” (FL = 0.88) in the subcontractor dimension (FL = 0.86); and “lack of experience and knowledge” (FL = 0.83) and “lack of design information” (FL = 0.77) in the design dimension.

The SEM of CPRM (Figure 2) revealed that the client (FL = 0.64) and external factor (FL = 0.47) dimensions exerted relatively weak influences on the overall project risk. According to the questionnaire results, on-site workers were less strongly affected by the client and external factor dimensions; they were most influenced by the contractor dimension (FL = 0.96), followed by the subcontractor (FL = 0.86) and design (FL = 0.83) dimensions.

These research results may help project risk managers and decision makers select appropriate risk response strategies and develop preventive measures according to different factors’ level of influence and relative probability of occurrence. This could improve the quality of their decision making and overall project performance.

This study makes the following five risk management recommendations for parties involved in Taiwanese construction projects:

• Contractors should practice thorough on-site supervision and management.
• Subcontractors should enhance on-site workers’ ability to read construction drawings and improve the educational training for these operators.
• Designers’ professional knowledge should be respected and information updates be followed closely.
• Interdepartmental communication and coordination with clients should be enhanced to reduce unnecessary changes in decisions.
• Regarding on-site operations, construction techniques should be improved, and relevant legal restrictions and legal changes should be followed closely and responded to in a timely manner.

Future studies may use larger samples, recruit respondents from more regions or industries, confirm the validity of the model, use different model evaluation and analysis methods, enhance the applicability of risk management, and improve the energy conservation of construction projects through risk management to achieve the ultimate goal of sustainable operation.