Examining factors affecting budget overrun of construction projects undertaken through management procurement method using PLS-SEM approach

Available online at www.sciencedirect.com ScienceDirect Procedia - Social and Behavioral Sciences 107 ( 2013 ) 120 128 ELPIIC 2013 Evaluation of Learning for Performance Improvement International Conference, Malaysia, 25 26 February, 2013 Examining factors affecting budget overrun of construction projects undertaken through management procurement method using PLS-SEM approach Ismail Abdul Rahman a, Aftab Hameed Memon b*, Ahmad Tarmizi Abd Karim c a,b,c Faculty of Civil Engineering, University Tun Hussein Onn Malaysia, 86400 Parit Raja, Johor, Malaysia Abstract This paper focused on examining the effect of various factors on budget overrun in construction projects undertaken using management procurement method in Malaysia. It adopted a quantitative method for data collection using structured questionnaire survey amongst contractors, consultant and clients. A total of 118 samples were collected against 200 questionnaires that had been distributed nationwide. Gathered data was analyzed with an advanced multivariate method of structural equation modeling with PLS approach using SmartPLS software. The major finding showed all the constructs in model contributed significantly to budget overrun with R 2 value of 0.623. Also, the developed model has substantial explaining power with GoF value of 0.62. This indicates that the model was able to be generalized in representing the budget overrun factors occurring in construction projects nationwide. By identifying these factors, it will help the construction community to take measures in improving the cost performance of the projects. 2013 The Published Authors. Published by Elsevier by Elsevier Ltd. Selection Ltd. and peer-review under responsibility of Universiti Malaysia Selection Kelantan, and Malaysia peer-review under responsibility of Universiti Malaysia Kelantan, Malaysia Keywords: Management Procurement Method; Budget overrun; PLS-SEM * Aftab Hameed Memon. Tel.: +60-14-2725620 E-mail address: aftabm78@hotmail.com. 1877-0428 2013 The Authors. Published by Elsevier Ltd. Selection and peer-review under responsibility of Universiti Malaysia Kelantan, Malaysia doi:10.1016/j.sbspro.2013.12.407

Ismail Abdul Rahman et al. / Procedia - Social and Behavioral Sciences 107 ( 2013 ) 120 128 121 1. Introduction In achieving successful completion of construction projects in Malaysia, traditional procurement method has been replaced with strategic and integrated approaches such as management procurement method. This method improves construction performance through a better process (Morledge, Smith, & Kashiwagi, 2006) where client appoints contractor as management consultant for managing design and construction of a project (Rashid et al., 2006). It allows practicing value-engineering at the early stage because of the participation of the contractor at the design stage (Abdullah, 2010). In spite of it, construction industry is still facing major challenges in completing the construction projects within the estimated budget (Abdullah, 2010; Abdullah, Aziz, & Rahman, 2009; Sambasivan & Soon, 2007) where more than 50% of projects faced overrun (Endut, Akintoye, & Kelly, 2009). Research on budget overrun appears to be limited in the Malaysian context and in particular among construction industries (Ibrahim, Roy, Ahmed, & Imtiaz, 2010); thus it is important to identify and evaluate the causes of budget performance in construction projects. However, this paper aims to identify significant and dominant factors which affect the budget overrun faced by construction projects that employed management procurement method. 2. Research method This study adopted quantitative mode of research where data collection was carried out through structured questionnaire survey amongst the practitioners (i.e. client, consultants and contractors) who involved in handling construction projects that employed management procurement method. A total of 200 questionnaire sets were distributed and 118 of the completed questionnaire sets were collected. Of these, majority of the respondents with 50.8% belong to contractor s organization while 29.7% and 19.5% of the respondents were from consultants and clients organizations respectively. Most of the respondents (i.e. 81.4%) had handled large projects and only 18.6% of them were involved in small scale projects. These respondents have working experience for more than 15 years in construction. In terms of academic ability most of respondents (i.e. 78%) have a minimum of civil engineering degree. The gathered data was then analyzed with PLS-SEM approach using SmartPLS v.20 (Ringle, Wende, & Will, 2005) software package for simulation and modeling process in determining the significant and dominant factors. 3. PLS Structural Equation Modeling (PLS-SEM) PLS-SEM is a regression based modeling approach which uses a component-based (similar to principal components factor analysis) technique in analyzing path models (Vinzi, Trinchera, & Amato, 2010). PLS path models comprises of two sets of linear equations: the outer model also referred as measurement model and the inner model also referred as structural model. The inner model specifies the relationships between unobserved or latent variables, whereas the outer model specifies the relationships between a latent variable and its observed or manifest variables (Ringle, Sarstedt, & Mooi, 2010). PLS-SEM technique follows a systematic sequential procedure in analyzing and assessing the theoretical model as described in Appendix A. 3.1. Develop Theoretical Model and Data Input A complete theoretical model for evaluation is developed based on 35 factors affecting budget overrun. Theoretical model defines the causal relationships between the factors of budget overrun to assess their effect on cost overrun. This model is drawn in SmartPLS v2.0 software for simulation

122 Ismail Abdul Rahman et al. / Procedia - Social and Behavioral Sciences 107 ( 2013 ) 120 128 process. Input data to this model was done through comma delimited (*.csv) file format which contains the respondent s input from the questionnaire survey. Appendix B describes PLS-model together with input variables in SmartPLS software where exogenous and endogenous latent variables are presented with blue coloured oval shapes and indicators representing manifest/known variable represented by yellow coloured rectangles in the path diagram. The figure shows that 35 of the manifest/measured variables of budget overrun factors are categorized in 7 exogenous latent variables/groups which are assigned as (i) CSM, (ii) DDF, (iii) FIN, (iv) ICT, (v) LAB, (vi) MMF and (vii) PMCA; while there is only one endogenous latent variable i.e. budget overrun, where, CSM represents contractor s site management group which consists of 8 manifest variables (CSM01: poor site management and supervision, CSM02: incompetent subcontractors, CSM03: schedule delay, CSM04: inadequate planning and scheduling, CSM05: lack of experience, CSM06: inaccurate time and cost estimates, CSM07: mistakes during construction, and CSM08: inadequate monitoring and control). DDF represents design and documentation group which consists of 5 manifest variables (DDF01: frequent design changes, DDF02: mistakes and errors in design, DDF03: incomplete design at the time of tender, DDF04: poor design and delays in design, and DDF05: delay in preparation and approval of drawings). FIN represent Financial Management group which consists of 6 manifest variables (FIN01: cash flow and financial difficulties faced by contractors, FIN02: poor financial control on site, FIN03: financial difficulties of owner, FIN04: delay in progress payment by owner, FIN05: delay payment to supplier /subcontractor, and FIN06: contractual claims, such as, extension of time with cost claims). ICT represents information and communication group which consists of 3 manifest variables (ICT01: lack of coordination between parties, ICT02: slow information flow between parties, and ICT03: lack of communication between parties). LAB represents labour group which consists of 5 manifest variables (LAB01: labour productivity, LAB02: shortage of site workers, LAB03: shortage of technical personnel (skilled labour), LAB04: high cost of labour, and LAB05: severe overtime). MMF represents material and machinery group which consists of 4 manifest variables (MMF01: fluctuation of prices of materials, MMF02: shortages of materials, MMF03: late delivery of materials and equipment, and MMF04: equipment availability and failure). PMCA represents project management and contract administration group which consists of manifest variables (PMCA01: poor project management, PMCA02: change in the scope of the project, PMCA03: delays in decisions making, and PMCA04: inaccurate quantity take-off). 3.2. Evaluation of Outer Model Evaluation of the outer model (measurement model) is to examine the reliability and validity of the constructs of the model (Hulland, 1999). It determines how well the indicators (specific questions) load on the theoretically defined constructs. This can be carried out in two stages as: A) Convergent validity of the measures (Hulland, 1999). B) Discriminant validity of the research instruments (Gefen, Straub & Boudreau, 2000) 3.2.1. Convergent validity of constructs Convergent validity is the measure of the internal consistency. It is estimated to ensure that the items assumed to measure each latent variable measures them and not measuring another latent variable (Fornell & Larcker, 1981; Hulland, 1999). Convergent validity of the construct can be determined by

Ismail Abdul Rahman et al. / Procedia - Social and Behavioral Sciences 107 ( 2013 ) 120 128 123 calculating individual item reliability, Cronbach's alpha, Composite reliability (CR) and Average Variance Extracted (AVE) as suggested by (Aibinu, Ling & Ofori, 2011). Individual Item Reliability is the extent to which measurements of the latent variables measured with multiple-item scale reflects mostly the true score of the latent variables relative to the error. It is assessed by calculating standardized loadings of each variable where items with loadings of less than 0.4 should be dropped (Hulland, 1999) while (Chin, 1998) suggested that item with loading lower than 0.5 should be dropped. Cronbach's Alpha is the coefficient of reliability (or consistency). It measures how well a set of items (or variables) measures a single one dimensional latent construct. (Litwin, 1995) suggested that value of cronbach alpha should be higher than 0.7. Composite Reliability (CR) measure is used to check how well a construct is measured by its assigned indicators. However, the interpretation of composite reliability score and Cronbach's Alpha is same. (Chin, 1998; Hair, Ringle, & Sarstedt, 2011) suggested 0.7 as a benchmark for modest composite reliability. Average Variance Extracted (AVE) test is used to assess internal consistency of the construct by measuring the amount of variance that a latent variable captures from its measurement items relative to the amount of variance due to measurement errors (Fornell & Larcker, 1981). A basic assumption is that the average covariance among indicators has to be positive. Barclay, Thompson, and Higgins (1995) and Hair et al., (2011) stated that AVE should be higher than 0.5. This means that at least 50% of measurement variance is captured by the latent variables. 3.2.2. Discriminant validity of constructs Discriminant validity indicates the extent to which a given construct is different from other constructs (Hulland, 1999). It is tested through analysis of average variance extracted by using the criteria that a construct should share more variance with its measures than it shares with other constructs in the model (Fornell & Larcker, 1981). This is examined by comparing the AVE of construct shared on itself and other constructs. For valid discriminant of construct, AVE shared on itself should be higher than variance shared with other constructs (Chin, 1998). 3.3. Evaluation of Inner Model Inner model (structural model) evaluation is carried out to assess the relationship between exogenous and endogenous latent variables in respect of variance accounted (Hulland, 1999). It also determines the explanatory power of the model by evaluating squared multiple correlations (R 2 ) and path co-efficient (β) values, where R 2 indicates the percentage of a construct s variance in the model, whilst the path coefficients indicate the strengths of relationships between constructs (Chin, 1998). According to (Cohen, 1988; Cohen, Cohen, West, & Aiken, 2003) R² of endogenous can be assessed as substantial =0.26, moderate =0.13 and weak=0.02 while for path co-efficient assessment, β value of all structural paths is compared, highest β value indicates that effect of the construct is most significant and lowest value shows the lowest effect of construct on endogenous latent variable. 3.4. Model Representation Model representation was aimed to assess the power of developed model to generalize for construction industry of Malaysia in representing the effect of budget overrun factors. This was tested by assessing Goodness of Fit (GoF) index value. GoF is defined as the geometric mean of the average communality and average R 2 for all endogenous constructs (Akter, Ambra, & Ray, 2011). It can be used to determine

124 Ismail Abdul Rahman et al. / Procedia - Social and Behavioral Sciences 107 ( 2013 ) 120 128 the overall prediction power of the large complex model by accounting for the performance of both measurement and structural parameters. According to (Chin, 2010), the intent of GoF is to account for the PLS model performance at both the measurement and the structural model with a focus on overall prediction performance of the model. 4. Performance of Model Performance of the developed model was assessed with two-step process as (i) outer model evaluation to examine the reliability and validity of the construct, and (ii) inner model evaluation to assess the relationship between exogenous and endogenous latent variables (independent latent variables and dependent variable) in respect of variance accounted for (Hulland, 1999). 4.1. Convergent Validity of Outer Model Convergent validity assessment and modification of the PLS developed model was carried out simultaneously by adopting iterative process. A total of 2 iterations were run for achieving the optimum values of parameters as presented in Appendix C. Table in Appendix C shows that 5 constructs have achieved satisfactory convergent validity (factor loading, CR, Alpha, AVE values are all above the cut off values) in iteration 1. However, the other 2 constructs (FIN and MMF) managed to achieve cut-off values for factor loading, CR and Alpha but AVE value is lower that required value of 0.5. Hence, these two constructs were modified by omission of the manifest that has the lowest factor loading value from each of the construct. The 2 nd iteration has improved the convergent validity for all the constructs after omitting FIN04 and MMF01 manifest variables. 4.2. Discriminant Validity of Outer Model Discriminant validity of construct was evaluated by assessing correlation matrix generated from simulation of the model. The diagonals values of the matrix are replaced with the value of square root of the AVE. For adequate discriminant validity, the diagonal values should be greater than the off-diagonal values in the corresponding rows and columns (Hulland, 1999). This testifies that the outer model has achieved the discriminant validity needed for further analysis. 4.3. Inner Model Assessment Inner model was assessed by examining the R² and path co-efficient (β) values of all the paths. The R² value of endogenous latent variable (budget overrun) of the inner model was found as 0.623 which indicates that the all the exogenous latent variables (constructs) are significantly contributing effect to the endogenous latent variable. For path co-efficient, β values of each path was found as 0.435 for CSM, 0.089 for DDF, 0.288 for FIN, 0.026 for ICT, 0.131 for LAB, 0.044 for MMF and 0.407 for PMCA. These path co-efficient values of the model show that CSM group has the highest co-efficient value of 0.425. This means the CSM is the most significant group in causing budget overrun. 4.4. Model Representation Model representation is assessed by calculating GoF index value which is bounded between 0 and 1. Because of the descriptive nature of GoF index, there is no inference based criteria to assess its statistical significance (Vinzi et al., 2010). GoF for cut-off values were calculated using the guidelines suggested by

Ismail Abdul Rahman et al. / Procedia - Social and Behavioral Sciences 107 ( 2013 ) 120 128 125 (Wetzels, Schroder, & Oppen, 2009) and it was found as GoF small = 0.10, GoF medium = 0.25 and GoF large = 0.36. The model GoF value was calculated by using the following equation (Akter et al., 2011): GoF AVE 2 R GoF 0.615 0.623 0.62 The GoF value of the model is exceeding the large cut-off point. This indicates that the model has substantial explaining power in generalizing to represent the budget overrun factors occurring in construction projects nationwide. 5. Conclusion This study examined 35 factors that contribute to budget overrun in Malaysian construction industry using PLS-SEM approach. These factors are grouped into seven latent variables for modelling in SmartPLS software. These factors and variables were used to develop a PLS model of budget overrun. The developed model was tested in two stages: the outer model and inner model assessment. It was found that the outer model all the manifest in the model is reliable and valid. For inner model, it was found that all the constructs contributed significantly to the budget overrun and CSM group of factors is the most significant contributor to budget overrun. Finally, it was found that the overall model has high explaining power of the ability to generalize the model for nationwide representation. References Abdullah, M. R. 2010. Significant causes and effects of construction delay. Master thesis, University Tun Hussein Onn Malaysia. Abdullah, M. R., Aziz, A. A. A., & Rahman, I. A. 2009. Causes of delay and its effects in large MARA construction project. International Journal of Integrated Engineering (Issue on Mechanical, Materials and Manufacturing Engineering). Aibinu, A. A., Ling, F. Y. Y., & Ofori, G. 2011. Structural equation modeling of organizational justice and cooperative behaviour in the construction project claims process: contractors' perspectives. Construction Management and Economics, 29(5), 463-481. Akter, S., Ambra, J. D., & Ray, P. 2011. Trustworthiness in mhealth Information Services: An Assessment of a Hierarchical Model with Mediating and Moderating Effects Using Partial Least Squares (PLS). Journal Of The American Society For Information Science And Technology, 62(1), 100 116 Barclay, D., Thompson, R., & Higgins, C. 1995. The Partial Least Squares (PLS) approach to causal modeling: personal computer adoption and use as an illustration. Technology Studies, 2(2), 285 309. Chin, W. W. 1998. The partial least squares approach to structural equation modelling. In G. A. Marcoulides (Ed.), Modern Methods for Business Research (pp. 295 336): Erlbaum, Mahwah, NJ. Chin, W. W. 2010. How to Write Up and Report PLS Analyses. In V. Esposito Vinzi et al. (eds.) (Ed.), Handbook of Partial Least Squares, Springer Handbooks of Computational Statistics: Springer- Verlag Berlin Heidelberg. Cohen, J. 1988. Statistical power analysis for the behavioral sciences (2nd ed.). Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. 2003. Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences (Third ed.): Lawrence Erlbaum Associates, Publishers, Mahwah, New Jersey, London.

126 Ismail Abdul Rahman et al. / Procedia - Social and Behavioral Sciences 107 ( 2013 ) 120 128 Endut, I. R., Akintoye, A., & Kelly, J. 2009. Cost and time overruns of projects in Malaysia. retrieved on August 21, 2009, from http://www.irbnet.de/daten/iconda/cib10633.pdf, 243-252. Fornell, C., & Larcker, D. F. 1981. Evaluating structural equation models with unobservable variables and measurement error. Journal of Marketing Research, 18, 39 50. Gefen, D., Straub, D. W., & Boudreau, M. C. 2000. Structural equation modelling and regression: Guidelines for research practice. Communications of the Association for Information Systems, 4(5), 1 78. Hair, J. F., Ringle, C. M., & Sarstedt, M. 2011. PLS-SEM: Indeed a Silver Bullet. Journal of Marketing Theory and Practice, 19(2), 139 151. Hulland, J. 1999. Use of partial least squares (PLS) in strategic management research: a review of four recent studies. Strategic Management Journal, 20, 195 204. Ibrahim, A. R., Roy, M. H., Ahmed, Z., & Imtiaz, G. 2010. An investigation of the status of the Malaysian construction industry. Benchmarking: An International Journal, 17(2), 294-308. Litwin, M. S. 1995. How to Measure Survey Reliability and Validity: Sage, Thousand Oaks, CA. Morledge, R., Smith, A., & Kashiwagi, D. T. 2006. Building Procurement: Blackwell Publishing, UK Rashid, R. A., Taib, I. M., Ahmad, W. B. W., Nasid, M. A., Ali, W. N. W., & Zainordin, Z. M. 2006. Effect of procurement systems on the performance of construction projects. Paper presented at the Padang on 21-24 June 2006, Retrieved from http://eprints.utm.my/790/ dated: January 09, 2013. Ringle, C. M., Sarstedt, M., & Mooi, E. A. 2010. Response-Based Segmentation Using Finite Mixture Partial Least Squares Theoretical Foundations and an Application to American Customer Satisfaction Index Data. In Data Mining, Annals of Information Systems (Vol. 8, pp. 19-49): R. Stahlbock et al. (eds.). Ringle, C. M., Wende, S., & Will, S. 2005. SmartPLS 2.0 (M3) Beta, Hamburg 2005, http://www.smartpls.de. Sambasivan, M., & Soon, Y. W. 2007. Causes and effects of delays in Malaysian construction industry. International Journal of Project Management, 25, 517 526. Vinzi, V. E., Trinchera, L., & Amato, S. 2010. PLS Path Modeling: From Foundations to Recent Developments and Open Issues for Model Assessment and Improvement. In V. Esposito Vinzi et al. (eds.), Handbook of Partial Least Squares, Springer Handbooks of Computational Statistics (pp. 47-82). Wetzels, M., Schroder, G. O., & Oppen, V. C. 2009. Using PLS path modeling for assessing hierarchical construct models: Guidelines and empirical illustration. MIS Quarterly, 33(1), 177 195.

Ismail Abdul Rahman et al. / Procedia - Social and Behavioral Sciences 107 ( 2013 ) 120 128 127 Appendix A. Schematic Diagram of PLS-SEM Analysis Modify PLS model Not ok Develop Theoretical Model Data Input Assessing Outer Model Convergent validity Discriminant validity ok Assessing Inner Model R 2 < 0.02 Model is weak Model Moderate/Strong R 2 > 0.13/0.26 Model Representation Global Fit Index (GoF)

128 Ismail Abdul Rahman et al. / Procedia - Social and Behavioral Sciences 107 ( 2013 ) 120 128 Appendix B. Convergent Validity Results Iteration 1 Iteration 2 Loading AVE CR Alpha Loading AVE CR Alpha CSM01 0.841 0.613 0.926 0.911 0.841 0.613 0.926 0.911 CSM02 0.837 0.837 CSM03 0.685 0.685 CSM04 0.819 0.819 CSM05 0.798 0.798 CSM06 0.794 0.794 CSM07 0.731 0.731 CSM08 0.743 0.743 DDF01 0.802 0.676 0.913 0.881 0.802 0.676 0.913 0.881 DDF02 0.822 0.822 DDF03 0.814 0.814 DDF04 0.841 0.841 DDF05 0.831 0.831 FIN01 0.718 0.467 0.838 0.789 0.719 0.504 0.835 0.758 FIN02 0.774 0.775 FIN03 0.758 0.758 FIN04 0.531 Omitted FIN05 0.674 0.674 FIN06 0.613 0.613 ICT01 0.887 0.824 0.933 0.893 0.887 0.824 0.933 0.893 ICT02 0.937 0.937 ICT03 0.899 0.899 LAB01 0.827 0.510 0.837 0.755 0.827 0.510 0.837 0.755 LAB02 0.708 0.708 LAB03 0.753 0.753 LAB04 0.558 0.558 LAB05 0.697 0.697 MMF01 0.631 0.465 0.774 0.621 0.777 0.632 0.838 0.716 MMF02 0.679 0.792 MMF03 0.803 0.816 MMF04 0.597 Omitted PMCA01 0.766 0.544 0.826 0.720 0.766 0.544 0.826 0.720 PMCA02 0.749 0.749 PMCA03 0.722 0.722 PMCA04 0.711 0.711 Appendix C: Discriminant Validity Results CSM DDF FIN ICT LAB MMF PMCA CSM 0.783* DDF 0.612 0.822* FIN 0.638 0.536 0.710* ICT 0.755 0.524 0.455 0.908* LAB 0.603 0.464 0.627 0.501 0.714* MMF 0.525 0.426 0.517 0.497 0.533 0.795* PMCA 0.714 0.725 0.523 0.703 0.545 0.467 0.737* Note: * square root of the AVE