Tell a ministry where to look, not where to go

The request that arrives from a ministry is almost always the same one, whatever language it is written in. We have money, or a mandate, or a minister with a timetable. Where should we put it. The question is reasonable, the people asking it are serious, and the honest answer is not the one they asked for.

There are two things an adviser can hand over, and they cost about the same to produce. One is a conclusion: this state, that sector, these terms. The other is a capability: here is the question, here is the file that answers it, here is what would have to be true for the answer to change. The first is what gets bought. The second is what keeps working.

I think the first is almost always the wrong sale, and the reason is not manners. It is that a conclusion carries no instructions for its own revision. When the ground moves, the ministry that holds a conclusion has no way of knowing that it has stopped being true, and must buy another one.

Six dates on one desk

Consider what sits behind any recommendation about the American economy written from public sources, using this publication's own files as the example. The national series were retrieved on 19 September 2026 and carry annual observations, four to six points each. The state metrics were fetched on 7 September 2026, with monthly series running to July 2026 and a single prior month behind them. The occupational wage file is 2025. The tax collections file is fiscal year 2025. The population estimates are Census Vintage 2024. The health file is 2024.

That is six different as-of dates on one desk, and a recommendation built across them inherits all six. The document that lands with the ministry will not say so. It will read as though it describes the present, and it will keep reading that way long after two or three of its inputs have been superseded, because nothing in its form records which number it is standing on.

The thinness underneath is easy to underestimate. Take Mississippi in the state file. Its civilian labour force is recorded as 1,280,219 for July 2026 against 1,284,361 in June. Its total nonfarm employment is recorded as 1,196.2 thousand for July against 1,192.3 thousand in June. One series fell and the other rose, in the same state, in the same month, from the same agency, because they count different things. An adviser who needs a direction of travel can have either one. That is not a defect in the data; it is what a single month of two differently constructed series looks like. It is a defect in any recommendation that rests on one of them without saying which.

Decay is not a metaphor here, it is a property of the files. The national series are annual, so the freshest point in a recommendation citing them describes a year, not a moment, and the next point will arrive whether or not anyone revisits the advice. The state monthly series carry a July 2026 value, and nothing in the file can tell a reader whether that value still stands as published. A conclusion does not come with a shelf life printed on it. The evidence behind it does, and the adviser is the only party in the room who can read the date.

What these files answer, and what they do not

The limits are not vague, and a ministry is entitled to have them stated plainly rather than discovered later. These files will tell you what an occupation pays in a given state: the median for Healthcare Practitioners and Technical Occupations in 2025 is 123,330 dollars in California and 66,630 in Mississippi. They will tell you which of five taxes a state levies and what it collected in fiscal 2025. They will tell you how a national series has moved across its handful of annual points.

• They cannot compare the United States with anywhere else. No file here holds a figure for another country, so any sentence containing the words cheaper than, or faster than, with a foreign subject, is not coming from the evidence.
• They cannot go below the state line. No county, no metropolitan area, no utility district, no plant.
• They cannot describe a state's own past. One observation and one prior period is not a history, and four to six annual points is not a business cycle.
• They cannot look forward at all. There is no projection, no pipeline and no announced-investment figure anywhere in them.

Converting a request is mostly mechanical once you accept the limits. A ministry asking where its component supplier should build is really asking four separable questions: which states hold enough people already doing the specific work, at a wage the project can carry; which of the five recorded taxes those states levy and what that implies for the operating model; which of the project's costs appear in no federal file at all, the local property tax and the utility connection being the usual two; and what evidence would move a state from the shortlist to the discard pile. Three of those four can be started today with public files and no consultant. The fourth is the one worth paying for, and it is the one that almost never appears in the deliverable.

There is a virtue hidden in one of those limits that is worth naming. The state metrics file records two absences by name, manufacturing employment for Alabama and for the District of Columbia, and carries them as gaps rather than filling them. A ministry can work around a hole it has been shown. It cannot work around a hole that somebody smoothed over to make a table look complete, and it will not know to try.

The asymmetry that makes bad advice profitable

The adviser knows the provenance of every number in the document. The client sees the conclusion. That gap is where the commercial value of the relationship sits, and there is a standing incentive to keep it open, which is met less by dishonesty than by omission: the test that produced the recommendation is unglamorous, hard to present and harder to bill, so it stays on the adviser's side of the table. It is worth noticing how rarely a recommendation names the file it came from, and how often it names the firm instead. The first is checkable by anyone and the second is checkable by no one, and the choice between them is made early, quietly, and in the adviser's favour.

The result is an arrangement in which the client's dependence is the product. That is a commercial success and an advisory failure, and it is worth saying plainly because the same failure, at a larger scale and with public money, is what has given technical assistance between states its uneven reputation. Capability that does not transfer is consumption presented as investment. A ministry buying advice should ask what it will be able to do by itself at the end, and treat a blank answer as a price it has not been quoted.

The objection, which is a good one

The objection deserves its full weight. Ministries do not run on epistemology. There is a cabinet date, a minister who needs a name to say out loud, and a rival department with a simpler answer. An adviser who declines to name a place is not automatically being careful. They may be being safe, keeping their record clean by transferring the risk of being wrong back to the client, who has less information and more exposure. That is worse than a wrong recommendation honestly made.

So the answer is not to withhold a conclusion. It is to attach the test. Name the shortlist, and with it the two or three conditions that produced it and the thresholds at which a place drops off. If a state is on the list because its collections file shows no corporation net income tax line, as it does for Nevada, Texas, Washington and Wyoming, say so, and say what would have to change for that to stop mattering to this project. A recommendation with a falsifier attached is still a recommendation. It is simply one the ministry can maintain without you.

This is also a test of whether the advice was ever about the client. An adviser who cannot state the conditions under which their recommendation fails either has not found them or would rather not hand them over. Both are disqualifying, and the second is the more common.

What the discipline looks like in practice

The version of this that a publication can practise is narrow but real: print the source line, print the retrieval date, and print the figure in the units the file uses, so a reader who disagrees can go to the same file and check. It costs nothing except the temptation to sound more certain than the evidence allows, and it is the same discipline a ministry needs, at a smaller scale.

The best outcome of any advisory relationship across borders is that the client eventually disagrees with the adviser, using the adviser's own sources, and turns out to be right. A ministry that has been taught where to look can do that. A ministry that has been handed a destination can only ask again.