Optimism bias, sensitivity analysis and switching values
Optimism bias is the tendency to underestimate costs and duration and overestimate benefits. The Green Book corrects it with explicit adjustments at the start of an appraisal, then reduces them as specific risks are handled. Sensitivity analysis and switching values then test how far the preferred option's assumptions can move before it stops being value for money.
Correcting optimism bias
At the outset, raise estimated costs and timeframes and lower estimated benefits. Size the adjustment from the organization's historical forecast errors: the gap between what past appraisals expected and what actually happened on similar proposals. The Green Book says appraisals should draw on evaluation evidence from the same or similar interventions for this.
As the appraisal develops, estimate specific risks in more detail. Where it is cheaper to prevent a risk than to bear it, include the prevention cost in the core costs. Then reduce the optimism bias adjustment in proportion to the risk prevented. What remains, plus the value of risks not prevented, is the contingency.
Sensitivity analysis and switching values
Sensitivity analysis tests how changes in key assumptions, such as input costs or demand, change an option's results. Stress-test the assumptions that are uncertain or likely to change.
A switching value is how far a key variable would need to move for an option to stop being value for money, for example the demand level at which the preferred option's net present social value becomes negative. At a minimum, do both for the preferred option and include them in the shortlist results.
Where the exam catches people
Believing the optimism bias adjustment stays fixed: it is reduced in proportion to the risk that funded prevention measures remove.
Believing the adjustment is dropped once risks are handled: only the prevented share comes off; the rest stays in contingency.
Believing the team's bottom-up estimate is enough where past proposals overran: the Green Book uses evidence from similar interventions to set the adjustment.
Believing central estimates are enough for the shortlist: the preferred option needs sensitivity analysis and switching values at a minimum.