Optimism Bias

Optimism bias is a systematic tendency for estimates to understate costs or delivery time, or overstate benefits. The UK Treasury’s Green Book requires explicit corrections in its public-appraisal context. This describes a pattern across estimates; it does not mean every individual estimate will be optimistic or every overrun was foreseeable.

Choose comparable evidence

Compare forecasts with outturns at a defined decision stage, using a common scope, price basis, and treatment of risk. Explain which projects belong in the reference class and why. Selecting comparable evidence involves judgment. A changed scope or an incomplete baseline can produce a difference that should not be treated as a pure forecast error. Track cost, duration, and benefit errors separately rather than applying one percentage to every variable.

The Treasury’s supplement, based on Mott MacDonald’s 2002 public-project study, gives equipment/development capital-cost starting values of 10–200% and outsourcing operating-cost values of 0–41%. These are not possible minimum and maximum outcomes. Within that method, start from the appropriate upper value at outline business case stage, then reduce it only with evidence that contributory risks have been mitigated. Earlier stages can require larger adjustments. Choose the applicable project components and cost categories, and prefer suitable primary evidence when available.

Connect the adjustment to the funding allowance

In a fictional university’s finance-system proposal, four comparable implementation packages overran matched-scope estimates by 20%, 30%, 50%, and 60%. Their unweighted mean is 40%. The separately described student-records project’s incomplete whole-program baseline is excluded pending reconstruction. The four observations are limited evidence, so sensitivity analysis remains necessary.

An implementation estimate of 4.0 million initially receives 1.6 million of optimism adjustment. Assume reviewed prevention work costs 0.2 and supports a 0.6 reduction in that adjustment. Core costs become 4.2; the remaining adjustment is 1.0. A further 0.3 of separately assessed residual risk is included only after confirming it is outside the reference-class allowance. The authority holds a reserve of 1.3, giving a total implementation funding envelope of 5.5. This invented cash example assumes zero inflation and uses undiscounted totals; financial schedules must otherwise convert appraisal values consistently.

Mitigation is not free, and an identified risk must not be counted both inside a historical allowance and again as an independent addition. Document the boundary and evidence before reducing an adjustment or adding a risk amount. A reserve is an allowance for uncertainty, not an expected invoice or a guarantee against overrun. If implementation spending is 5.0, it uses 0.8 of the 1.3 reserve and leaves 0.5; this says nothing by itself about the investment’s benefit.

The budget and cost reconciliation are developed in The Case Said Four Million and the Cost Was Nine. Switching value explains how a chosen comparison changes as an input moves, including tests around a corrected estimate.

References:

• HM Treasury, The Green Book (2026), chapter 6 (Open Government Licence v3.0).
• HM Treasury, Green Book supplementary guidance: optimism bias (PDF), table 1 and application steps; based on Mott MacDonald’s 2002 study (Open Government Licence v3.0).