Modeling the Cumulative Impact of Change Orders
Iskandar, Karim Ashraf Sabry
MetadataShow full item record
Change orders occur in almost every construction project and regularly cause variations to the contractors’ anticipated working conditions, resources, and manner of work completion. Change orders are major source of additional congestion, change in sequence, and loss of momentum in the construction jobsite. They frequently cause unforeseen labor productivity loss, which forces contractors to extend their stays on projects. Contractors encounter a lot of resistance from owners when proving productivity loss attributable to change orders, which may lead to unresolved disputes and lengthy litigations. Previous researchers attempted to set standards and methods in order to quantify the cumulative impact of changes on labor productivity. Some of the previous studies were based on case studies of two or three projects, others included a larger number of projects and more reliable analysis. Generally, it is very difficult to conclusively determine the exact amount of productivity loss attributable to change orders. As a result, there is a continuous need to enhance and enrich the cumulative impact research field. This current research is based on a database of one hundred and forty-five mechanical and electrical projects, encompassing two project groups: projects impacted by changes, and projects unimpacted by changes. Using two-sample 𝑡-tests and Chi-squared tests, a series of numerical and categorical variables were found to be significant in distinguishing between impacted and unimpacted projects, thus revealing the underlying causes of productivity loss associated with change orders. Furthermore, sixty-eight impacted projects were used in order to quantify the cumulative impact of changes using linear regression analysis. A series of statistical model selection criteria were applied in order to carefully identify the best predictive models. Candidate models were statistically diagnosed and thoroughly tested to check their validity. Statistical tests and measures were used in order to check whether there are outlying or influential observations in the models. In addition to that, new projects were collected to verify the future predictive ability of the candidate models. The analysis identified the following six factors as best cumulative impact predictors: percent owner initiated change orders, overmanning, turnover, absenteeism, percent time spent by project manager on project, and productivity tracking. The models developed in this research provide the construction industry with means that could be used during dispute resolutions to support the contractors’ calculations and assertions for cumulative impact claims. Finally, this study incorporates a significant statistical component that highlights the most common challenges that analysts face when building linear regression models, such as multicollinearity and the presence of hidden extrapolations. The models developed in this research were extensively analyzed in full details through various statistical tests and measures in order to avoid misleading and deceptive results.