Quantifying and modeling the cumulative impact of change orders

The modern construction industry in the United States generated 5% of the country's gross domestic product (GDP) in 2016. In an industry this large, it is inevitable that changes will detrimentally effect projects. Yet, existing studies on the impact of change orders suffer from insufficient sample size, lack of rigor in validation, or both. This study uses extensive data from 68 electrical and mechanical construction projects affected by changes to develop a rigorous statistical regression model that predicts the cumulative impact of changes. The input variables of the model developed in this paper include the percent of change orders initiated by the owner, productivity tracking, turnover, percent of time spent by the project manager on the project, and overmanning. This paper uses a process that differs from previous studies in that it supplements linear regression with multiple-variable selection criteria, statistical checks for multicollinearity, and a consideration of the existence of outlying or influential data points. In order to verify the continued applicability of the developed model, rigorous validation testing was performed using multiple cross-validation metrics. Finally, new project data have been collected and used to test the model to confirm its merit for continued use by industry practitioners and researchers.