Cost overruns in Norwegian projects – An econometric study

Abstract

Cost overruns are a global phenomenon. By assuming that companies are profit maximizing, we imply that inaccurate estimates of project costs are unwanted as the basis for investment analysis is weakened. The aim of this thesis is therefore to identify which factors affect the ability to set accurate budgets and meet the estimated costs. I highlight this topic by analyzing the differences in cost overruns between projects from the Norwegian oil industry and the public sector, by introducing macroeconomic variables for analysis and by looking into whether the cost overruns from one sector affects the other.

Firstly, descriptive statistics and univariate regressions were run in order to obtain a better overview of the topic. Both public and oil projects are statistically more prone to cost overruns than underruns, however, oil projects experience overruns of larger magnitude. Public projects show a trend where increasing project size reduces cost overruns, while cost overruns in oil projects tend to increase with the duration of the project.

To further analyze the dynamics of cost overruns, multivariate regressions were performed. This includes using forward selection by iterative processes in order to arrive at the final models. For oil projects, I find the variables Duration, Pension fund surprise, GDP growth and NCS investment surprise to significantly affect the magnitude of cost overruns, explaining about 25% of the variability in cost overruns for oil projects. As for public projects, the corresponding model includes the variables Duration, Employment level, GDP from marine activities and Export, explaining about 13% of the variability.

The models above indicate that cost overruns in both sectors depend on the macroeconomic environment at the time of project execution. The explanatory power for the oil model is rather acceptable, however, I fail to find a good general model for cost overruns in public projects. When comparing the two sectors by running the models on the opposite dataset, no causal relationship in cost overruns between the two was found, and I therefore fail to confirm whether the different sectors affect the cost overruns of each other.