Bias and Error in Mine Project Capital Cost Estimation

Abstract

Previous analyses of small samples of mining projects have found that feasibility studies tend to underestimate the as-built capital costs of the project. Our review of 63 international mining and smelting projects confirms that as-built capital costs are, on average, 14% higher than as estimated in the bankable feasibility study. There is little attenuation over time of this bias in capital cost estimation, appearing to reflect an absence of learning on the part of the project sponsor or the consulting engineering firm. We argue that this persistence of bias is instead intentional and rational, driven by a scarcity of project financing and the need by project sponsors to inflate the project economics in a bid to secure financing. We find some empirical support for our contention. A second phase of the analysis examines estimation error. Roughly half of all projects' as-built capital costs fall outside of the expected ± 15% of the feasibility study capital cost estimate, even after allowing for intentional estimation bias. Cost overruns of 100% or more happen in roughly 1 out of 13 projects. Smaller projects have less estimation accuracy than large projects. Finally, our analysis of the cost overrun data reveals that a shifted lognormal probability distribution should be used when modeling mining project capital costs in a Monte Carlo analysis.