Reference class forecasting (RCF) corrects chronically optimistic megaproject cost estimates by placing a planned project in the empirical distribution of outcomes from comparable completed projects and reading the required budget (an uplift on the base estimate) from that distribution. Awojobi and Jenkins (2016) applied this outside-view technique to hydroelectric dams and were explicit about what remains unresolved: the method presumes accessible, consistent historical records from major financiers and weakens when key variables such as site geology are missing or when the planned dam lies in a region dominated by outliers; it assumes the reference class stays relevant despite changes in macroeconomic conditions, policy regimes, inflation, and financing environments; it removes estimation bias but not residual technical and geological uncertainty; and it provides no settled rule linking institutional risk tolerance to the required contingency uplift. Later work has sharpened all four issues without settling any: a 2022 sequel shows RCF-informed rejections forfeit large net benefits in World Bank hydropower portfolios, a 2019 data update shows the overrun distribution drifts between eras, and a 2025 review of RCF still lists reference-class identification and uplift determination among its open research problems. The open problem is to establish testable criteria for when a historical dam reference class remains a valid basis for the outside view, together with a principled rule linking residual geological and technical uncertainty and institutional risk tolerance to the contingency uplift.

Background

Forecasting the construction cost of a hydroelectric dam before it is built is notoriously unreliable. Two systematic mechanisms are usually held responsible: optimism bias, the manifestation of the planning fallacy (the documented tendency of planners to underestimate durations and costs by imagining a best-case execution), and strategic misrepresentation (deliberately low-balled estimates advanced to win project approval). Reference class forecasting (RCF), brought into infrastructure planning by Bent Flyvbjerg building on Kahneman and Lovallo's 'outside view', counters both by setting aside the project's own details and instead locating it in the empirical distribution of cost overruns — final cost relative to the approved budget — from a reference class of completed comparable projects. The forecaster then reads the budget from that distribution: to be, say, 80 percent confident of covering actual costs, the base estimate is uplifted by the amount that 80 percent of past comparable projects would have needed. The uplift plays the role of contingency capital, the reserve held against cost risk during construction.

The outside view on dams is stark. Ansar, Flyvbjerg, Budzier, and Lunn (2014) assembled a reference class of 245 large dams built between 1934 and 2007 in 65 countries and found that three out of four dams overran their budgets, with actual costs on average 96 percent above estimates (median 27 percent) and implementation schedules on average 44 percent longer than estimated; geological surprises were among the identified but poorly hedgeable risk drivers. Awojobi and Jenkins (2015, 2016) brought the technique to bear on dams financed by the World Bank, demonstrating with the Bujagali dam in Uganda how an RCF-derived contingency can be embedded in investment appraisal, and concluded that access to the historical records held by the major financiers of dams is what makes learning from the past possible.

Awojobi and Jenkins (2016) were explicit, however, that the application leaves the method's foundations unresolved. First, the approach depends on consistent historical records from major financiers, and its predictive strength may be weak when such data are incomplete, when important variables such as site geology are missing, or when the planned dam falls into a region where outliers dominate the historical sample. Second, the method assumes the reference class remains relevant over time, but changes in macroeconomic conditions, policy regimes, inflation, currency devaluation, and financing environments can make older projects an imperfect guide for current appraisal. Third, the outside-view adjustment reduces bias from optimism and strategic misrepresentation but does not eliminate uncertainty from technical and geological factors, leaving the trade-off between better front-end information and residual construction risk unresolved. Fourth, no universal risk-tolerance rule exists: the appropriate uplift depends on the institution, the contractor, the project context, and the consequences of over- or under-providing contingency capital. These four issues jointly ask when the outside view is a legitimate basis for appraisal at all.

Problem Statement

The problem, as left open by Awojobi and Jenkins (2016) and still open in the literature, is to determine when a historical reference class of dam projects remains a valid basis for the outside view in reference class forecasting of a planned hydroelectric dam's construction cost, and how the resulting contingency uplift should be set. Concretely, the question has three parts, the first two of which jointly establish when the class remains a valid basis and the third of which sets the uplift: (i) Testable criteria for reference-class validity that address each of the following four aspects: the completeness and consistency of financier-held historical records; the presence or absence of key physical variables such as site geology; the trade-off between class granularity and sample size; and the treatment of regions whose historical sample is dominated by outliers. (ii) A method to assess or adjust for the temporal stability of the class, that is, whether projects built under different macroeconomic conditions, policy regimes, inflation, currency, and financing environments remain valid comparators for current appraisal. (iii) A principled rule for the contingency uplift, relating the residual technical and geological uncertainty that the outside view cannot remove, and the institution's risk tolerance, to the required uplift. The objective uniting these parts is to replace the currently informal judgment that a chosen reference class is 'similar enough' with validated methodology for dam cost appraisal.

References

  1. Awojobi, O., & Jenkins, G. P. (2016). Managing the cost overrun risks of hydroelectric dams: An application of reference class forecasting techniques. Renewable and Sustainable Energy Reviews, 63, 19-32. DOI: 10.1016/j.rser.2016.05.006. https://doi.org/10.1016/j.rser.2016.05.006
  2. Awojobi, O., & Jenkins, G. P. (2015). Were the hydro dams financed by the World Bank from 1976 to 2005 worthwhile? Energy Policy, 86, 222-232. DOI: 10.1016/j.enpol.2015.06.040. https://doi.org/10.1016/j.enpol.2015.06.040
  3. Ansar, A., Flyvbjerg, B., Budzier, A., & Lunn, D. (2014). Should we build more large dams? The actual costs of hydropower megaproject development. Energy Policy, 69, 43-56. DOI: 10.1016/j.enpol.2013.10.069; also arXiv:1409.0002. https://arxiv.org/abs/1409.0002
  4. Flyvbjerg, B. (2008). Curbing optimism bias and strategic misrepresentation in planning: Reference class forecasting in practice. European Planning Studies, 16(1), 3-21. DOI: 10.1080/09654310701747936. https://doi.org/10.1080/09654310701747936
  5. Kahneman, D., & Lovallo, D. (1993). Timid choices and bold forecasts: A cognitive perspective on risk taking. Management Science, 39(1), 17-31. DOI: 10.1287/mnsc.39.1.17. https://doi.org/10.1287/mnsc.39.1.17
  6. Jenkins, G. P., Olasehinde-Williams, G., & Baurzhan, S. (2022). Is there a net economic loss from employing reference class forecasting in the appraisal of hydropower projects? Renewable and Sustainable Energy Reviews, 159, 112218. DOI: 10.1016/j.rser.2022.112218. https://doi.org/10.1016/j.rser.2022.112218
  7. Plummer Braeckman, J., Disselhoff, T., & Kirchherr, J. (2019). Cost and schedule overruns in large hydropower dams: an assessment of projects completed since 2000. International Journal of Water Resources Development, 36(5), 839-854 (online first 2019; print issue 2020). DOI: 10.1080/07900627.2019.1568232. https://doi.org/10.1080/07900627.2019.1568232
  8. Cantarelli, C. C., Davis, K., Pinto, J. K., & Turner, N. (2025). Reference class forecasting: promises, problems, and a research agenda moving forward. Production Planning & Control, 37(7), 691-709 (online first 2025; print issue 2026). DOI: 10.1080/09537287.2025.2578708. https://doi.org/10.1080/09537287.2025.2578708
  9. Themsen, T. N. (2019). The processes of public megaproject cost estimation: The inaccuracy of reference class forecasting. Financial Accountability & Management, 35(4), 337-352. DOI: 10.1111/faam.12210. https://doi.org/10.1111/faam.12210
  10. Servranckx, T., Vanhoucke, M., & Aouam, T. (2021). Practical application of reference class forecasting for cost and time estimations: Identifying the properties of similarity. European Journal of Operational Research, 295(3), 1161-1179. DOI: 10.1016/j.ejor.2021.03.063. https://doi.org/10.1016/j.ejor.2021.03.063
  11. Theising, E., Wied, D., & Ziggel, D. (2023). Reference class selection in similarity-based forecasting of corporate sales growth. Journal of Forecasting, 42(5), 1069-1085 (online first 2022; print issue 2023). DOI: 10.1002/for.2927. https://doi.org/10.1002/for.2927
  12. Chadee, A., Martin, H., Gallage, S., & Rathnayake, U. (2023). Reducing cost overrun in public housing projects: a simplified reference class forecast for small island developing states. Buildings, 13(4), 998. DOI: 10.3390/buildings13040998. https://doi.org/10.3390/buildings13040998
  13. Bärenbold, R. (2023). Reducing risks in megaprojects: The potential of reference class forecasting. Project Leadership and Society, 4, 100103. DOI: 10.1016/j.plas.2023.100103. https://doi.org/10.1016/j.plas.2023.100103