A conceptual cost estimation model for the pre-design stage of road projects in New Zealand : a thesis presented in partial fulfilment of the requirements for the degree of Doctor of Philosophy (PhD) in Construction, School of Built Environment, Massey University, Auckland, New Zealand

Thumbnail Image
Open Access Location
Journal Title
Journal ISSN
Volume Title
Massey University
The Author
Cost overruns in construction projects have been a pervasive and challenging issue within the construction industry. Over the years, the researchers explored the causes, effects, and mitigation strategies for the cost overrun issue. However, there has been no major improvement, and projects still experience significant cost overruns. Due to the long project duration, scope, higher coverage of ground, and involvement of public funds, road construction faces more uncertainty compared to building projects. Thus, the cost overrun in road projects has become a crucial issue. Therefore, this study focused on mitigating the cost overrun issue by minimising the errors in conceptual cost estimation by developing a new cost model for the pre-design stage of road projects. This research followed a secondary data collection through systematic and bibliometric literature review and primary data collection from the New Zealand (NZ) Road construction industry to fill the above research gap and achieve the research aim. The multi-method quantitative research approach, including a questionnaire survey and multiple case study data, was used in this research. Initial cost overrun evaluation used one hundred and six cases from NZ road projects, while fifty-nine detailed cases were utilised to develop and validate the cost model. Firstly, the research identified ten crucial factors that affect the cost of NZ road projects: frequent design changes, poor planning and scheduling, poor and incomplete tender documentation, delays in design, mistakes/ errors in design and drawings, unforeseen ground conditions, inaccurate cost estimates, poor site management and supervision, poor project management, and inaccurate quantity take-off. Secondly, the research investigated the severity of the cost overruns in NZ road projects. According to the findings, NZ road projects experience approximately 20% of cost overruns, while the project size and duration significantly impact the magnitude of the cost overruns. Thirdly, using the case study data, the research developed three models to improve the conceptual cost estimation. Since the conceptual estimate is prepared during the pre-design stage, the variables should be calculated using less information. The performance of the models was evaluated as an error percentage comparing the estimated cost from the model with the actual project cost. The measure was the Mean Absolute Percentage Error (MAPE). The first model was developed using regression analysis (RA), which showed a MAPE of 21.35%. Then, another model was developed using artificial neural networks (ANN) with a MAPE of 11.82%. However, both models considered only the technical aspects of the projects. Therefore, the final model combined ANN with Monte Carlo simulation (MCS). The hybrid model demonstrated a minimum MAPE of 3.53%, significantly improved results compared to the continuous cost overruns experienced in the NZ road construction section. The findings of this research greatly support the NZ road construction sector to be aware of the severity of the cost overrun issue. Further, the model introduced in the research can be used in the industry to improve the current estimation practices. The study was conducted using the case study data from NZ road projects. However, the variables considered for the models were selected with the possibility of generalising the mode to other countries. Further, the findings and conclusions were limited to the modelling techniques identified in this research. Although other hybrid models are available, very little research was conducted on developing cost estimations for road projects. Most of the developed cost models considered technical characteristics in a project but did not consider any project risk factors. There were a few models considered, both using only one modelling technique but failing to produce reliable output. Therefore, hybrid models can be developed by combining several techniques without mixing the technical and risk variables. This research developed the first ANN and MCS hybrid model, considering technical characteristics and risk factors with a significantly lower error.
construction cost estimation, conceptual stage, road projects, artificial neural networks, Monte-Carlo simulation, cost overruns