What contingency in a capital budget is for

Contingency is money held inside a capital budget against events that are known to be possible but not yet specified. The ground may be worse than the boreholes suggested. A subcontractor may fail. A design detail that works on paper may not build. None of these is a surprise in the sense of being unimaginable. All of them are unscheduled, and that is the distinction contingency exists to handle.

It is not padding, because padding is unaccounted and contingency is a named line with a release rule attached to it. It is not a reserve for new scope either. Three distinct pots are routinely collapsed into one word: contingency covers uncertainty inside the agreed scope, escalation covers the movement of prices over the time the work takes, and management reserve covers decisions to change what is being built. Collapsing them is how a project comes to report that it spent its contingency on a scope change without anyone objecting.

How contingency is sized

Well-run estimating does not pick a percentage. It builds a risk register — a list of identified events, each with a judged likelihood and a judged cost — and then asks what total the project should hold in order to be confident of finishing inside its budget. That confidence level is a choice, and it is usually stated as the probability that the final cost lands at or below the number. A budget set at the median expects to be exceeded about half the time. A budget set high enough to be exceeded only rarely ties up money that could have funded something else. Neither choice is wrong; the error is making it without saying so.

1. Physical uncertainty: ground conditions, existing structures, weather exposure, anything discovered by digging.
2. Design maturity: how much of the design was actually complete when the estimate was made.
3. Market conditions: whether enough contractors will bid, and what they will charge for carrying risk themselves.
4. Interface risk: the joints between separate contracts, where responsibility is ambiguous and delay accumulates.

Who is allowed to release it

The governance rule matters more than the size of the pot. If a project manager can draw on contingency without telling anyone, it functions as a budget increase available on demand, and the early warning that a project is in difficulty never reaches the sponsor. If release requires a named decision at a named level, with a reason recorded, then drawdown becomes the single most useful leading indicator a sponsor has: the rate at which contingency is being consumed, set against the share of work completed, says whether the estimate is holding long before the outturn does.

Contingency should also fall as a project matures, and it falls because design work is done. At concept stage the range of possible outcomes is wide and the held amount is correspondingly large. By the time the drawings are fixed, the site is investigated and the contracts are let, most of the unknowns have either happened or been priced by somebody. That is engineering labour, and it is neither free nor evenly available — the chart shows how unevenly the workforce that performs it is distributed. A project whose contingency percentage has not moved between concept and construction has not been re-estimated. It has been re-typed.

There is a real disagreement about disclosure. Cost engineers commonly argue that a published contingency figure is consumed by definition, because everyone downstream treats it as available money and spends towards it. Auditors and public-finance officials argue the opposite: that concealed contingency produces serial reforecasting, destroys the credibility of every later estimate, and makes it impossible for a legislature to compare one project with another. The argument is not really about numbers. It is about whether a budget is a control instrument for the people building the thing or a public commitment to the people paying for it, and it is not settled.

So the question to put to any capital budget is not how large its contingency is. It is: a percentage of what, and as at which design stage. The same percentage held against a mature design with a completed site investigation is a different instrument from one held against a concept sketch — and the second is not really contingency at all. It is an estimate that has not been made yet.