The boundary of a public investment appraisal is drawn where the measurement stops

A public investment appraisal turns a project into three numbers: a capital cost, a stream of benefits, and a rate at which future years are discounted back to the present. The three are not measured to the same standard. The cost has a national counterpart that a statistical agency publishes and this brief carries as annual observations. The discount rate is observable, in the sense that the price of long public money is quoted every day. The benefit stream has no counterpart of that kind: no statistical agency publishes one, for any project or in aggregate. What an appraisal presents as a single answer is therefore a measurement, an observation and an assumption, stacked and given one decimal place.

Where this is written from

I should say what this vantage is, because it fixes everything that follows. I am an analyst writing inside The Cefa Brief, working from published statistics and public documents and from nothing else: annual series redistributed by FRED and retrieved for this site on 19 September 2026, together with the agencies' own documentation of what each series counts. I hold no seat in a finance ministry, no brief from a sponsor and no sight of a single project's business case.

That position is good at one thing and blind to another. It can test whether a claim about national aggregates is consistent with the published record, and it can name the unit a figure is stated in, which is where the errors in this field tend to start. It cannot audit an appraisal. No national annual series records a discount rate chosen by a sponsor, an optimism uplift, a benefit-cost ratio, a variation order or a contract. Everything below is about the shape of the instrument rather than about any project that has passed through it. Where the argument turns on something the statistics do not carry, the absence is named instead of filled.

Only the cost line has a published counterpart

Begin with the number that is measured. Nonresidential construction spending runs 856, 845, 968, 1193, 1295 and 1309 billion dollars across 2020 to 2025, each point stated at a seasonally adjusted annual rate. The unit is part of the fact.The series already states each period at a yearly rate, so 1309 billion is a figure for a year and not for a month. Across the six observations the series adds 453 billion, an increase of 52.9 percent on the 2020 figure. This is the only line of an appraisal with a national aggregate standing behind it.

It is also a price multiplied by a quantity, published as one number. Producer prices for all commodities, on a 1982 base of 100, run 194.4, 227.3, 264.5, 255.7, 254.6 and 260.3 over the same years, a rise of 65.9 index points or 33.9 percent. Both series moved, and the construction figure cannot say how much of its own movement belongs to each. I have not divided one by the other. The all-commodities index is not a construction cost index, and using it as one would be an invention rather than an adjustment. The point survives without the division: a cost fixed in one year's dollars and delivered over several is exposed to a quantity the appraisal does not measure and the agency does not publish.

The discount rate does more work than the project

The third number decides the answer. The ten-year Treasury constant maturity yield runs 0.89, 1.45, 2.95, 3.96, 4.21 and 4.29 percent across 2020 to 2025. No series in this brief publishes a sponsor's appraisal rate, and that yield is not one; it is the observable price of long public money, and the arithmetic of discounting behaves the same way whatever rate is put into it. A benefit of one dollar arriving in year thirty is worth one divided by 1.0429 raised to the thirtieth power, which is 28.4 cents. At the 0.89 percent of 2020 the same dollar is worth one divided by 1.0089 to the thirtieth, or 76.7 cents. Move the arrival to year twenty and the two figures are 43.2 cents and 83.8.

Nothing about the project changed between those calculations. The rate changed, and a rate does not scale a project down evenly. It scales the distant years much harder than the near ones, which gives appraisal a standing preference for benefits that arrive early. The assets public investment exists to build return late by their nature: a grid, a water system, a port, a research base. Research and development inside GDP runs 754, 839, 942, 991, 1029 and 1081 billion dollars over the same six years, each at a seasonally adjusted annual rate and so a yearly figure rather than a month's, ending 327 billion or 43.4 percent above where the window began. That is the category whose payoff sits furthest out, and therefore the category a higher rate penalises hardest. This is not a judgement about any project. It is a property of the arithmetic, and it applies before anyone has looked at the engineering.

What sits outside the boundary

Two things are excluded by construction rather than by oversight. The first is the stream of costs that begins when the building is finished. Construction spending measures value put in place, and it stops at handover; no series in this brief separates the cost of running an asset from the cost of building one, because the statistical system counts capital formation and does not count the obligation that follows it. The second is the comparison itself. Every appraisal is a comparison against a case in which the project is not built, and no statistical agency publishes that case. It is assembled by the party seeking approval, it carries the most leverage over the result, and it receives the least scrutiny of any input.

A third exclusion is less obvious. An appraisal is denominated in money, and money is rarely the resource in short supply. Job openings across the whole economy run 6.3, 10, 11.2, 9.3, 7.7 and 7.1 million, stated as a monthly average, which means each annual point is the average of that year's monthly readings and not the total of them; Average hourly earnings in the total private sector run 29.36, 30.61, 32.26, 33.7, 35.06 and 36.44 dollars per hour, a rise of 7.08 dollars or 24.1 percent across the window. An appraisal written against a 2020 wage and delivered against a 2025 one carries that much movement on its labour line before anything else has gone wrong. That figure is the whole private economy and not a construction payroll, and neither series tells a sponsor whether the particular people a particular site needs exist within reach of it.

What follows

My position is that an appraisal should publish its three numbers separately and never as one. The capital cost, with the year whose dollars it is priced in. The discount rate, as a number, with the year it was chosen and the reason. The benefit stream, with the plain statement that nothing published corroborates it. Where a quantity has no series at all, and the maintenance, the workforce and the do-nothing case have none, the appraisal should record the absence rather than substitute a dollar figure for it; a dollar figure standing in that position is not a measurement but a decision wearing one. What this vantage cannot do is say that any particular appraisal failed. Six annual observations of national aggregates cannot evaluate a project, cannot date a decision inside a year and cannot reach a business case. The argument is about where the instrument stops reading, and it stops reading at the edge of what someone once chose to count.