People forecast their own spending with a range that is far too narrow

The estimate is built from a version where nothing goes wrong

Asked what a renovation, a move or a holiday will cost, people mentally walk through the sequence of steps involved and add up what each one costs. That method is reasonable and it has one systematic flaw: the imagined sequence contains only the steps that were foreseen, and real sequences contain steps that were not.

The overrun is not usually caused by any single large error. It is the accumulation of things that were not in the picture — the additional thing that turned out to be required, the delay that cost money, the item that was cheaper in the version imagined. Each is individually unpredictable and their collective appearance is entirely predictable.

This tendency in estimating, sometimes described as the planning fallacy, is robust enough that it shows up in household projects and enormous institutional ones alike. Being aware of it turns out to reduce it only slightly, which is itself informative.

A single figure hides the real problem

Most estimates are given as one number. The more honest output would be a range, and the useful diagnostic is how wide someone is willing to make it. Ask for a figure you are ninety per cent confident the total will not exceed, and most people produce something only slightly above their original estimate.

That narrowness is the overconfidence, and it is separate from the estimate being low. It means the household has no view of how bad the outcome could be, which is the input that would determine whether it can withstand the project going wrong.

A more useful practice is to produce three numbers rather than one: what it costs if everything goes to plan, what it costs if the usual amount goes wrong, and what it costs in a genuinely bad case. The gap between the first and third is the thing worth having.

The past is a better guide than the plan

The most reliable correction is to stop reasoning about the specific project and look at how similar things have actually gone. What did the last renovation end up costing against its estimate? What did the last three expensive months look like against the budget? Those figures contain the unforeseen steps, because they already happened.

This feels wrong to people, because the current project seems knowable in a way that a general average does not. Each project has its own particular reasons why it will be different, and those reasons are exactly what makes the estimate optimistic in the first place.

It also requires having recorded what happened, which most households have not. A rough note of what the last few large expenditures actually cost against expectations is a small piece of record-keeping with unusually high value.

Ordinary months are forecast optimistically too

The same mechanism operates at a smaller scale on routine budgeting. An expected monthly figure is built from the regular items, and the irregular ones — the repair, the gift, the thing that broke — are omitted from the picture even though some subset of them occurs almost every month.

This is why budgets that look perfectly reasonable produce a shortfall with tiresome consistency. The forecast was for a month with no incidents, and months with no incidents are less common than the forecasting suggests.

The correction is not more careful categorisation. It is a deliberate allowance for the unlisted, sized from what actually happened over the past year rather than from what can be foreseen for the next one.

Building in a margin rather than trying to be more accurate

Since the bias survives awareness, the practical response is structural. Adding a contingency to any substantial estimate is not pessimism; it is a correction for a known and directional error, and the appropriate size can be estimated from past overruns rather than invented.

The margin has to be genuinely reserved to work. A contingency that gets spent on improving the specification, which is what usually happens, has functioned as a larger budget rather than as protection, and the project arrives at the same place.

What size of margin is appropriate depends on the type of project, the household’s position and how much a bad outcome would hurt, and none of that generalises. This describes a well-observed pattern in estimating rather than a rule, and any project large enough to require borrowing is a decision worth discussing with a regulated adviser.