Relationship between contractor satisfaction and project management performance

Despite the fast growth of project-based companies and industries, studies of the satisfaction-performance (S-P) nexus of project participants are lacking. This study aims to explore the role of contractor satisfaction in affecting contractor project management performance along with considering external effects from other key participants. Two broad dimensions of satisfaction toward noneconomic factors and economic factors were used to develop hypothesized models. Structural equation modelling techniques were applied with data collected from 117 projects, showing that it is insufficient to simply conclude that contractor satisfaction influences project managerial performance and vice versa , and that satisfaction disaggregation is necessary. Additionally, it was found that noneconomic satisfaction contributes to performance, which in turn contributes to economic satisfaction. The theoretical and practical implications are further discussed.

The notion of project participant satisfaction has been widely accepted, with the emergence of stakeholder theory (Freeman, 1984) and the increasing popularity of such project-based industries as construction, real-estate, offshore engineering, and software developmentone fifth of the world's economic activity is organized as a project (Esenduran, Hall and Liu, 2015). More stable owner-contractor relationships have been proposed to improve life cycle costing, contracting processes, risk management, teamwork, trust, and cooperation in construction work (Dozzi et al., 1996). In addition to the 'iron triangle' measures of cost, schedule, and quality, satisfaction has been recognized as another effective measure of project performance (Ibbs and Ashley, 1987;Toor and Ogunlana, 2010;Li, Ng and Skitmore, 2013;Ha et al., 2017). Studies of construction project participant satisfaction have become increasingly prominent over the past decade, with a positive change in focus from individual performance to a greater emphasis on stakeholder interests (Toor and Ogunlana, 2010;Hosseini, Chileshe and Zillante, 2014;Ibiyemi et al., 2014;El-Sawalhi and Salah, 2015;Forsythe, 2015;Ujene and Edike, 2015;Andriof et al., 2017;Mok, Shen and Yang, 2017;Xio, Liu and Pang, 2019). Additionally, project participant satisfaction provides an early sign of the likely outputs in executing complex projects (Williams et al., 2012;Xiong et al., 2014).
Despite the wide acknowledgment and study of the individual satisfaction-performance (S-P) nexus, its counterpart nexus between organizational satisfaction and performance has received little attention. Pearsall and Ellis's (2006) command-control simulation study, for example, examined the effects of critical team member assertiveness on team performance and team satisfaction. Chiu andNg's (2013, 2015) questionnaire surveys of quantity surveyors in Hong Kong found job satisfaction and organisational commitment improved through work group identification and working relationships. Panahi et al.'s (2016) questionnaire survey of professional construction consultants in Malaysia investigating the potential effect of value conflicts on the stakeholders' job satisfaction, revealed the importance of the interaction between personal and organizational values in construction organizations. Jaafar and Radzi's (2013) questionnaire survey of various construction industry players in Malaysia found levels of satisfaction to be related to the procurement systems used. Xio, Liu and Pang (2019) identified the competencies needed of Chinese real-estate project managers needed to guarantee the smooth implementation of projects. Similarly, Ingle and Mahesh's (2020) study of Indian construction projects identified the main factors affecting project performance, which include stakeholder satisfaction. However, none have specifically considered the S-P nexus. One approach to understanding the S-P nexus of project companies has been to follow the various studies that have been made at the individual level and assume that there is an overall S-P nexus with a complex relationship between performance and satisfaction divided into sub-dimensions. A few extant studies of contractor satisfaction, such as Xiong et al. (2014), have highlighted the different effects of external factors on contractor economic satisfaction (ES) and production-related satisfaction (PS). The nexus between contractor satisfaction and contractor performance is still unknown, however, which is identified here as the research gap.
In response to the research gap, this study aims to investigate the nature of the satisfactionperformance (S-P) nexus of construction contractors. The findings could be critical for researchers and practitioners to understand internal relationships and better decision-making. The results of a survey of 117 construction contractors' satisfaction and performance pertaining to their recent project work are analyzed through the development and comparison of two series of structural equation models. The first group of models treats contractor satisfaction as a holistic concept in examining the S-P nexus (i.e., combining economic-related satisfaction and production-related satisfaction). In carrying out a detailed analysis of contractor satisfaction, the second group of S-P models provides significantly improved explanatory power by dividing contractor satisfaction into two facets of (PS).

Literature review
Satisfaction has been defined as a response function of the discrepancy between 'How much is there?' and 'How much should there be?' (Wanous and Lawler, 1972). Early research in industrial settings mainly refers to satisfaction as job satisfaction. This became a popular term with the Hawthorne studies in emphasizing linkages between employees' job satisfaction and their performance -a relationship described as the 'Holy Grail' by industrial-organizational researchers ( Judge et al., 2001). However, the weak empirical evidence relating to the individual S-P nexus prompted some researchers (e.g. Judge et al., 2001) to argue that the unsatisfactory outcomes in S-P linkage research have been mainly caused by the ambiguity in defining satisfaction, although some measures of job satisfaction have been developed, such as the Job Descriptive Index ( JDI) (Smith, Kendall and Hulin, 1969;Onukwube, 2012) and Minnesota Satisfaction Questionnaire (MSQ) (Weiss, Dawis and England, 1967). Analysis should also treat economic satisfaction separately, since it is highly related to such pay factors as pay equity (Brown, 2001). Significantly, Lai (2007) divided satisfaction into social satisfaction and economic satisfaction when investigating the mediating effects of satisfaction in the Taiwanese motor industry, finding noneconomic satisfaction to be much more important than economic satisfaction in influencing performance.
A project is a natural and an ideal organizational form for developing products facing increasing product complexity, changing markets, cross-functional expertise cooperation, customer-oriented innovation, and technological uncertainty (Hobday, 2000). For many cross-functional projects, it is the collaboration of participating project organizations rather than the routine practices of workers in assembly lines that produce the final product (Turner, 2014). For projects, both external factors and perceived satisfaction affect the performance of participating companies (Soetanto and Proverbs, 2002;Mellado, Lou and Becerra, 2019;Nugraha and Putranto, 2019;Olugboyega and Windapo, 2019;Yakubu, Ogunsanmi and Yakubu, 2019).The best-value (BV) procurement process proposes to use other key factors as well as bid price in the evaluation and choice of the best-performing contractor (Elyamany and Abdelrahman, 2010). As Freeman's (1984) stakeholder theory holds, stakeholder satisfaction management is useful for solving corporate social-related issues effectively and generating good business performance (Porter and Kramer, 2006). In many cases, the perception of success for complex projects has little to do with whether they are completed on time, at cost, and with the desired quality, but the achievement of the desired objectives of different stakeholders (Turner and Zolin, 2012;El-Sawalhi and Salah, 2015;Ujene and Edike, 2015;Mok, Shen and Yang, 2017). Many accept participant satisfaction as a new dimension of project success (e.g. Serrador and Turner, 2015) and a new approach for use as an early warning sign (Xiong et al., 2014).
With the increasing popularity of project economics and stakeholder theory, project participant satisfaction has become an emergent area of study (Orlitzky and Swanson, 2012). Also studied by some researchers are the organizational satisfaction levels of companies taking part in construction projects, such as clients and contractors (e.g. Masrom, Skitmore and Bridge, 2013). For example, a criterion for defining success especially regarded by some researchers is client satisfaction (Munns and Bjeirmi, 1996). In exploring problems of the design and construction phases of the green-building project delivery process, dissatisfaction of the owner, and dissatisfaction of the general contractor due to "problems with material availability" can negatively affect the value and quality of the project in terms of cost and time of the project. (Seyis, Ergen and Pizzi, 2016). However, previous studies have focused on measuring and exploring the driving factors of participant satisfaction (Leung, Ng and Cheung, 2004;Li, Ng and Skitmore, 2013), contractor satisfaction (Soetanto and Proverbs, 2002;Masrom, Skitmore and Bridge, 2013;Xiong et al., 2014), and client satisfaction (Yang and Peng, 2008). Geyskens, Steenkamp and Kumar's (1999) meta-analysis of 71 studies of satisfaction in marketing relationships, for example, resonated with Lopez's (1982) results in demonstrating the need to distinguish between economic and noneconomic satisfaction as distinct constructs with different causes and consequences. Likewise, Rodríguez, Agudu and Gutierrez (2006) demonstrated the necessity of such a categorization by examining the determinants of economic and noneconomic satisfaction in manufacturer-distributor relationships. In addition to the supply-chain, such a separation of organizational satisfaction is also applicable to construction projects. For example, Xiong et al. (2014) obtained a similar outcome in examining the effects of other key project participants' performance on the economic satisfaction (ES) and production satisfaction (PS) of construction contractors. Therefore, this study aims to explore the nexus between satisfaction and performance of construction contractors by taking sub-dimensions of contractor satisfaction into account.

Conceptual models and hypotheses
The combined efforts of different project participants generate project outcomes, and contractors − as key performance assessors − have their own psychological interpretation of other participants' performance (Soetanto and Proverbs, 2002). Clients and designers are another two key participants in construction contracts (Lai and Lam, 2010). In examining the influence of other project participants, Wang and Huang's (2006) survey of construction supervising engineers in China found client performance to be significantly correlated with project success. A major contributing factor seems to be client-led changes in project scope, which cause up to 70% of poor schedule performance in Saudi Arabian projects for example (Assaf and Al-Hejji, 2006). Poor designer performance also contributes significantly to delayed government projects in Malaysia, where defective designs account for most quality problems (Sambasivan and Soon, 2007). Ling et al. (2008) explored the project management practices of foreign architectural, engineering, and construction (AEC) companies in China, finding that project management performance is critical in predicting project outcomes in terms of cost performance, time performance, quality performance, owner satisfaction, and profit margins. Design ability has been found to be important to ensure the speedy completion of projects (Ling et al., 2004). Consistent with previous studies (e.g. Soetanto and Proverbs, 2002;Assaf and Al-Hejji, 2006;Yang and Peng, 2008;Xiong et al., 2014), the contractor S-P nexus is explored by including the influences of the performance of clients and designers in terms of the client's clarity of objectives (OC), and designer performance (DP) in terms of carefulness in the design of products or services to meet those needs.
Puzzled by the weak empirical evidence concerning both directional S-P links at the individual level, Schwab and Cummings (1970) pointed out ambiguity in defining job satisfaction may be a cause. This study disaggregated satisfaction into two parts -economic satisfaction (ES) and production satisfaction (PS). The performance part was restricted to contractor project management performance. Although some researchers followed this in exploring the relationships between disaggregated satisfaction components and performance (e.g. Lai, 2007), these studies assumed that all disaggregated satisfaction facets share the same unidirectional relationship with performance. If such an assumption is correct and these studies obtained significant results, the link between holistic satisfaction and performance should also be significant, which would not be consistent with the conclusions reached to date. To test the S-P nexus of construction contractors, therefore, it is proposed that different directional relationships exist between performance and each satisfaction sub-dimension. Two conceptual models describing the relationship between contractor satisfaction (COS) and contractor project management performance (CPMP) are developed to test these assumptions as follows.

CONCEPTUAL MODEL 1
The development of conceptual model 1 assumed that COS is a holistic concept comprising economic-related satisfaction (ES) and production-related satisfaction (PS) as suggested by Xiong et al. (2014). In this context, ES refers to contractor satisfaction, with such economic issues as cost, profitability, and potential business opportunities arising from current business activities. In contrast, PS refers to contractor satisfaction with production and service quality (Xiong et al., 2014).
Consistent with earlier studies (e.g. Soetanto and Proverbs, 2002;Lai, 2007;Xiong et al., 2014), the following hypotheses were proposed: H1 concerns the relationships between the performance of customers and designers separate from the performance and satisfaction of the organization. H2 concerns the linkage between contractor satisfaction and performance, with H2A and H2B indicating that COS affects CPMP and CPMP affects COS respectively (H2A and H2B are contained in two separate, but otherwise identical, models named Model 1A and Model 1B).  Conceptual model 1 CONCEPTUAL MODEL 2 Similar to conceptual model 1, the influences of OC and DP are considered as antecedent variables in exploring the contractor S-P nexus. A difference here is that satisfaction is divided into ES and PS, prompted by Geyskens, Steenkamp and Kumar (1999), Rodríguez, Agudu and Gutierrez (2006), Lai (2007), and Xiong et al. (2014) as leading examples. Unlike these earlier studies, also included are the different directional relationships that ES and PS may have with performance. As suggested by the previous literature, this research proposes that PS has a positive effect on CPMP and CPMP has a positive effect on ES. The hypotheses are as follows:

Hypothesis 1A: OC has a positive direct effect on COS
H3 concerns the relationships between the performance of customers and designers separate from the performance and satisfaction of the organization. H4 concerns the linkage between the two facets of contractor satisfaction and performance.  The contributory factors of construction project situations to COS and their degree of influence were identified on a five-point Likert scale questionnaire survey by Masrom (2012). The questionnaire consists of general information concerning the construction contractors, respondents, and specific questions in relation to the COS dimensions and their contributory factors. The questionnaire was sent to 300 construction contractors registered with the Malaysian Construction Industry Development Board (CIDB) to provide feedback based on the company's most recent construction project. Of the 136 contractors' responses, 117 usable responses (i.e. a valid net response rate of 39%) were used. These contractors are evenly distributed in terms of size, with 53.0% being large companies (G7) and 47.0% small to medium companies (G1-G6), categorized by company size and permitted tendering capability (Masrom, 2012). Of the respondents (the company representatives), 93.2% have a diploma or higher degree and 82.9% have more than 5 years' working experience, indicating good ability to understand the questions and provide reasonable feedback. Table 1 summarizes the basic information concerning the construction projects involved. > 3 year 9 7.69% 100.00% The sample size here exceeds the recommended number of 100 cases suggested by Bagozzi and Yi (2012) for structural equation modelling (SEM) and is comparable with previous SEM studies in the construction industry (Xiong, Skitmore and Xia, 2015). For example, Cheung and Chow (2011) analyzed 103 responses to explore the underlying factors contributing to withdrawal in construction project dispute negotiations; and Wong and Lam (2011) used 107 responses to investigate the effect of organization learning and unlearning on the performance of construction organizations. In addition to the comparatively simple model structure, each construct contains at least three variables, which assures model identification of each measurement construct and allows a smaller sample size for fitting the model (e.g. Xiong, Skitmore and Xia, 2015).
To test the conceptual models and hypotheses, the measurement framework in Table 2 was built based on Kline's (2015) three-variable principle, where three observed variables are used to reflect a latent variable. To do this, the observed variables were extracted from Masrom's (2012) questionnaire. Company size was used as a control variable in model development, with 0 = small/medium and 1 = large contractors. SEM is widely accepted and used in exploring and testing relationships between different constructs in the social science disciplines -its evolution being regarded as the most important statistical progress in social sciences in recent decades (Hair et al., 2012). Its particular applications in recent construction/project management studies include Qureshi and Kang (2015) (Washington et al., 2020). Compared with such first-generation models as principle component analysis and linear regression, SEM is a second-generation multivariate analysis method (Fornell and Larcker, 1982). It has many strengths, such as enabling the use of one model to explore an entire set of complex relationships, or the use of several observable items to represent ambiguous constructs (Kline, 2015; Cho, Hong and Hyun, 2009). Additionally, the availability of many SEM software packages offering graphical interfaces for model development (Xiong, Skitmore and Xia, 2015) enhances the popularity of SEM.
There are two types of SEM approaches: covariance based (CB-SEM) and partial least squares based (PLS-SEM). CB-SEM is appropriate for confirming theoretical hypotheses, as it focuses on minimizing the difference between the model-implied covariance matrix and the sample covariance matrix and obtaining accurate parameter estimates. In contrast, PLS-SEM is preferred in prediction, as it focuses on maximizing the explained variance of targeted constructs (Hair, Ringle and Sarstedt, 2012). CB-SEM is more popular in theoretical studies for its stricter rules concerning data and sample size and accurate estimates of parameters, while PLS-SEM has recently increased in popularity due to its ability to provide accurate predictions of target variables with a comparatively small sample size. Additionally, PLS-SEM can accommodate both reflective measurement constructs and formative measurement constructs, while CB-SEM can only accommodate the reflective measurement construct (Ringle, Sarstedt and Straub, 2012). In reflective constructs, changes in latent variables lead to changes in observed variables but not vice versa, which is important for explaining the selected latent variable when deleting an observed variable. In formative constructs, the opposite is the case, with changes in observed variables leading to changes but not vice versa, which changes the theoretical meaning significantly when deleting an observed variable ( Jarvis, MacKenzie and Podsakoff, 2003).
In the current context, organizational satisfaction is represented by contractor satisfaction (COS), which is composed of economic-related satisfaction (ES) and production-related satisfaction (PS), with both ES and PS being reflected by several observed variables, which means that COS in conceptual model 1 is a second order reflective-formative model and therefore more suitable for PLS-SEM (Becker, Klein and Wetzel, 2012). However, as Hair, Ringle and Sarstedt (2012) argue, CB-SEM and PLS-SEM have different advantages and should be complementary rather than conflicting. In this case, both methods were used -PLS-SEM for developing conceptual model 1 and CB-SEM for developing conceptual model 2, with the software SmartPLS 2.0 (Ringle, Wende and Will, 2005) and SPSS AMOS 21.0 used for developing the models accordingly.

RELIABILITY TEST
Cronbach's alpha value is used to test the reliability of the hypothesized construct based on the data, where a value exceeding 0.7 is taken as indicating the received data is acceptable for meeting the consistency requirement (Cho, Hong and Hyun, 2009;Lai, 2007). As presented in Table 3, the items are measured in five variables and the overall constructs are sufficiently satisfied. PLS-SEM has become a popular SEM method in recent years for its ability to accommodate both reflective and formative measurement constructs (Becker, Klein and Wetzel, 2012;Ringle, Sarstedt and Straub, 2012). For this study, satisfaction was divided into economic-related satisfaction and noneconomic satisfaction that is specific to production-related satisfaction. That is, contractor satisfaction is a formative-reflective construct. To test H2A and H2B separately, two models were constructed. Each contains three latent variables (OC, DP, and CPMP) with corresponding reflective indicators, one second-order hierarchical latent variable (COS) and two first order (reflective) constructs that form the second order (formative) construct. The repeated indicator approach was used for the conceptual model due to its comparatively high reliability and wide applications in reflective-formative construct problems (Chin, 2010;Becker, Klein and Wetzel, 2012;Ringle, Sarstedt and Straub, 2012). For this approach, a higher-order latent variable comprising lower-order latent variables could be constructed by representing all the observed variables belonging to the lower-order latent variables. This approach can estimate scores of latent variables simultaneously instead of estimating different order constructs separately, and then uses the needed construct scores to test the proposed relationships as a separate step (Ringle, Sarstedt and Straub, 2012;Hair et al., 2017).

Model fit evaluation
To validate the measurement components, three types of validity were assessed: internal consistency reliability, convergent validity, and discriminant validity. Compared with Cronbach's alpha values presented in Table 4, values of composite reliability are preferred in the testing of internal consistency reliability, as such measures do not assume that the observed variables share the same out-loadings (Hair et al., 2017). The average variance extracted (AVE) was used to test convergent validity. As Table 4 indicates, the values of composite reliability and AVE are greater than the required 0.7 and 0.5 respectively (Hair et al., 2017); all factor loadings of the observed variables on latent variables are significant at the 0.01 level; the discriminant validity is satisfactory, as the square root of the average variance extracted for each construct is more than the maximum correlations with other constructs (Fornell and Larcker, 1981); and the loadings are greater than the cross loadings by 0.1 as required (Hair et al., 2017). In addition to validity assessment, multicollinearity issues were checked for relatively high correlations between some variables in

Model development
Having confirmed their validity, the results of developing conceptual model 1 are presented in Figs 3 and 4 respectively to test H2A and H2B. COS is a second order latent variable consisting of ES and PS, and the relationships between the three variables are calculated by the repeated indicator approach. The non-significant low R 2 results indicated the lack of a strong enough relationship should be considered further. Since other satisfactory results were achieved, only the links between the three latent variables of OC, DP, and CPMP, and the second order COS are presented. As the company size control variable has a weak and insignificant effect in both Models 1A and 1B, its results are not presented. These show that both COSCPMP and CPMPCOS are highly significant at the 0.1 level (t-statistic=1.64), but still far from significant at the 0.05 level (t-statistic=1.96). These results are generally consistent with many earlier studies of the S-P relationship, in that they are positive but weak. Table 5 summarizes the results of the hypothesis tests The results gained in conceptual model 1 indicate the potential benefits of satisfaction disaggregation. In developing Model 2, although PLS-SEM is used to obtain the significance of the links and coefficients, CB-SEM is preferred for its ability to provide more accurate parameter estimates for the first-order reflective constructs (Hair, Ringle and Sarstedt, 2012). A two-step modelling method is often used when applying CB-SEM, (Byrne, 2016;Anvuur and Kumaraswamy, 2012;Xiong, Skitmore and Xia, 2015). This involves firstly carrying out a confirmatory factor analysis (CFA) from the correlations of all the latent variables and then, if the model fit results of the CFA are acceptable, changing these to propose directional relationships for further analyses. Table 6 shows the standardized regression weights and squared multiple correlations (SMCs) for each observed item. All the regression weights (factor loadings) range from 0.627 to 0.894 and, being above 0.5, are therefore highly significant (Anvuur and Kumaraswamy, 2012). The SMCs range from 0.393 to 0.799 (mean=0.600, sd=0.113). The average SMCs of the items in the measurement models are the AVE of the latent variables and, ranging from 0.510 to 0.759, are all greater than the 0.5 threshold. In terms of goodness of fit as presented in Table 7, χ2 = 187.214 (df = 170, χ2/df = 1.101, p=0.174) for the CFA phase, the χ2/df value of less than 2 indicating a good fit (Xiong, Skitmore and Xia, 2015).

Structural equation modelling
As a good model fit was obtained in the CFA phase, the correlations between the latent variables are replaced by the hypothesized directional relationships of conceptual model 2. Fig.  5 shows the final model. Company size was excluded for being highly insignificant and weak. The observed variables Q1 to Q16, measurement errors, and connections of items are omitted due to space limitations.

Figure 5
Model 2 testing H4A and H4B     (2012), the bias-corrected bootstrap approach was used with 500 resamples and maximum likelihood estimation to provide a robust test of the significance of the indirect effects (see Awang, Afthanorhan an Asri, 2015; Hox and Bechger, 1998). This indicates the significance to range from 0.002 to 0.050. For example, the indirect effect of DP on CPMP via the mediation of PS is 0.337×0.388=0.131, with a 95% confidence interval of (0.000, 0.322), p=0.045. The total effects are the sum of the direct effects and indirect effects. Fig. 5 shows the standardized direct effects, and Table 9 provides the standardized indirect effects and total effects. To maintain consistency, and as the effects of the insignificant OC-CPMP link are small, Table 9 shows the influencing effects on the dependent variables without deleting this link.

RECOGNIZING SUB-DIMENSIONS OF CONTRACTOR SATISFACTION
This paper explored the project participant satisfaction-performance nexus with empirical evidence from Malaysian construction contractors. This was partially inspired by the question proposed by Judge et al. (2001) for future research of whether the S-P relationship will be stronger at group or organization level? With this concern in mind, two satisfaction subdimensions comprising economic-related satisfaction (ES) and production-related satisfaction (PS) were introduced in the context of construction projects. Two series models were developed to test the S-P nexus of construction contractors. When satisfaction was seen as the usual holistic concept, the relationship between satisfaction and performance was weaksimilar to previous studies at the individual level. The good fit of Model 2 demonstrated the benefit of satisfaction disaggregation and the new S-P relationships proposed in this study.

RECOGNIZING DIFFERENT DIRECTIONAL RELATIONSHIPS BETWEEN CONTRACTOR SATISFACTION SUB-DIMENSIONS AND CONTRACTOR PERFORMANCE
The development of the first conceptual model gave a similar result to previous research at the individual level, in that the linkage of satisfaction and performance was positive but weak. Although Organ (1988) found a high probability of the existence of a functional S-P relationship, and Judge et al. (2001) obtained a 0.30 mean true correlation by conducting a meta-analysis, significant evidence of both directional models was still lacking. Our findings for this model to test the S-P links of construction contractors were similarly insignificant. However, these results are sufficiently counterintuitive for research to continue in the measurement and estimation of project participant satisfaction in order to improve satisfaction (e.g. Soetanto and Proverbs, 2002;Leung, 2004;Li, Ng and Cheung, 2013;Masrom, 2013;Xiong et al., 2014). This suggests the solution to the S-P problem is more complex than previously realized. Therefore, unlike previous empirical work, which assumed that all satisfaction facets share common unidirectional relationships with performance (e.g. Lai, 2007), the Model 2 presented here hypothesizes that PS had a positive effect on CPMP and CPMP positively affected ES. In addition to providing a satisfactory R 2 explanatory ability of Model 2, the hypothesized paths were significant. The conflicting relationships of the two satisfaction dimensions on performance validated here may also explain some of the previous findings at the individual level. For example, Lai (2007) found noneconomic satisfaction to be much stronger than economic satisfaction in determining the influence of decision strategies on performance. This suggests that a key problem of the theoretical debate on the S-P nexus to be the conflicting dimensions of satisfaction, which has important implications for future research at both individual and organizational levels.

Conclusions
The nature of the nexus between satisfaction and performance has been debated for decades without a satisfactory resolution. This paper provided results of the empirical validation of the hypothesized conceptual models among construction project participants. The results demonstrated that the definitions of satisfaction and performance should be more specific when discussing the S-P nexus. The findings of the disaggregated sub-dimensions of contractor satisfaction and the different directional relationships between contractor satisfaction sub-dimensions and contractor performance offered an innovative and reasonable way to understand the S-P nexus among project participants, which has significant implications for future research and practice.
This study was consistent with findings of previous studies in that other participant performance factors are critical to contractor performance. In addition, production-related satisfaction (PS) was found to mediate the effects of the factors on contractor performance. Future similar studies are encouraged to identify common antecedent variables (e.g. OC and DP) of the S-P nexus, as well as critical factors to PS (e.g. safety climate). Contractor performance could be improved by improving these aspects.
A further contribution was the comprehensive use of CB-SEM and PLS-SEM in solving the research problem. Although CB-SEM and PLS-SEM are increasingly used in social science studies, they are rarely used together in the same research (Hair, Ringle and Sarstedt, 2012). This study combined the advantages of both approaches to provide a suitable model development procedure for testing the formative-reflective construct in the first step and then disaggregating the formative component in the second step. This can be seen as a good demonstration of using both methods simultaneously and provides a reference for similar future research.
Being a pioneer study of the organizational satisfaction-performance nexus, the research was limited by the measures for satisfaction and performance being less well established than such widely used scales as JDI and MSQ for individual satisfaction. Building a comprehensive analysis framework (e.g., considering organizational citizenship behaviour, further classifying project quality into project quality and process quality, and considering the dimension of relational satisfaction) will be beneficial for further studies, otherwise the inconsistency of measures may lead to a significant bias and difficulty in conducting comprehensive analyses (e.g. meta-analysis). The framework also needs to be sufficiently flexible since the items to measure noneconomic satisfaction may be different for different industries and working roles.
This study contributed to this aspect by demonstrating the necessity to distinguish between economic and noneconomic dimensions. Another limitation is that, although two antecedent variables and one control variable were considered here, some potential moderators may exist between satisfaction and performance. For example, for trust among project participants, one possible situation is that when trust is high, the positive link between noneconomic satisfaction and performance will be larger, and vice versa. Similarly, design errors are likely to have an influencing role. Additionally, public/private sector differences, industry differences, cultural differences, and country differences are also possible as well as changes over time. Future research will therefore benefit from identifying the influence of potential antecedents, mediators, and moderators. Although the hypotheses were verified by evidence at the time of the research from the Malaysian construction industry, which is typical of the construction industries in most market economies, the relevance of our results to other industries and other countries is uncertain: more empirical studies are needed of different industries, contexts and time periods to understand the bigger picture. Similarly, the effects of stakeholders other than owners and designers are in need of further investigation; it would be beneficial to test the hypothesis that noneconomic satisfaction positively affects performance, and performance positively affects economic satisfaction at the individual level; and it would be interesting to investigate the use of the PLS and CB models in a single research to see how they complement each other and which offers a more accurate predictive power.
Finally, this study only focused on the relationships between satisfaction and performance of project participating companies. Further studies are needed to explore such cross-level influences as the impact of perceived organizational economic/noneconomic satisfaction on individual performance, organizational performance on individual satisfaction, and vice versa. Additional work is also needed to understand on how these findings could be of practical benefit in creating harmonious relationships between participants. Despite this study being derived in the context of the Malaysia construction industry, the models developed, and findings provided are likely to be useful for other industries and countries in understanding and improving project planning, implementation, and outcome assessments.