From reference class forecasting to decomposed contingency in infrastructure projects

Reference class forecasting (RCF) has become a central tool in infrastructure investment governance, primarily supporting funding approval by comparing outcomes with those of completed projects. Cost overruns that arise between funding commitment and contract award, and subsequently during construction, are often tied to how risk and uncertainty are managed after project approval, particularly when RCF is used to set or control contingency allowances. These patterns point to a problem in post-approval uncertainty management rather than a limitation of cost-estimation methods alone. While RCF provides a useful assurance mechanism for testing the plausibility of proposed cost estimates, it is not designed to manage uncertainty during the pre-contract phase, when project-specific information remains incomplete and subject to revision. A decomposed contingency approach (DCA) is proposed as a conceptual and methodological framework for managing uncertainty during this period. The DCA distinguishes among risk and aleatory uncertainty, epistemic uncertainty arising from incomplete or provisional information, and residual uncertainty associated with system-level effects that cannot be fully anticipated. It links different forms of uncertainty to decision approaches suited to their informational characteristics and provides a structured basis for adjusting contingency as information becomes more complete. Within this framework, contingency is determined through a decomposed treatment of uncertainty, while RCF is retained solely as an independent assurance mechanism.