Reference Class Forecasting

Origin. Kahneman and Tversky; Flyvbjerg's application to project planning.

Mechanism. Replaces the inside view — reasoning from this project's specifics — with the outside view: the distribution of outcomes for the class of projects to which this one belongs. The inside view is systematically optimistic and the bias does not diminish with expertise.

Procedure. Identify the reference class. Obtain its outcome distribution. Place this project in the distribution. Adjust only for differences that are demonstrably predictive, and expect that adjustment to be small.

Applies to. Schedule and cost estimation, where the inside view's failure is best documented.

Limitations. Reference class selection determines the answer and is where the optimism relocates. Narrow the class enough and every project is unique.