The finding and the prescription

Posted on: 31 May 2026

There is a paragraph about American public debt that appears, very nearly word for word, in two research notes from the same bank published within a few days of each other. The same figure, the net cost of debt rising to nineteen per cent of revenues by 2031 against the fourteen projected by the International Monetary Fund. The same mechanism set out, the one whereby higher costs generate fresh issuance that pushes yields higher still. The first time, that paragraph is there to suggest a little gold as protection inside an equity portfolio. The second time it is there to justify gold as a thesis in its own right. The same exhibit, two different conclusions, and in that small fact sits almost everything worth knowing about how to read the research of one of the world's largest private banks, namely that the argument does not lead to the conclusion; the conclusion goes looking for arguments to support it.

I say this as a matter of reading experience, not ideological suspicion. I have been receiving these notes for years, I read them when my work was something else, and I read them now that my work is watching systems from the outside. They are well written, by people who know what they are writing. The problem is not the quality. It is precisely the reverse: the quality is the packaging that lowers the reader's critical guard. A note full of precise percentages, of correlations at minus nought point six between yields and gold, of central bank purchases counted to the decimal tonne, communicates rigour, and rigour communicates disinterest. It looks like a laboratory report. It is not one.

Take the three documents in front of me, all from the same week at the end of May. On the Monday, equities have room to climb, earnings support the rally. On the Wednesday, equities remain attractive, the target on the American index is revised, and the debt paragraph appears in support of a little gold in the portfolio. In between, the note on gold itself: the year-end target trimmed after the metal had fallen fifteen per cent, and once again the same debt paragraph, this time in support of gold as a freestanding thesis. Three assets, three notes, the same handful of days, and the operating recommendation is always identical, which is to say stay positioned, indeed buy. The word sell appears in none of the three.

This is the point one sees only by laying the notes side by side, and never by reading one at a time in the moment it arrives. When yields rise and contradict both the equity thesis and the gold thesis, the response is not to revise the direction but to suggest shorter duration and a touch more gold. At which point the recommendation is not a variable that depends on the data; it is a constant, and it is the data that become the variable arranged around it.

The subtler mechanism lives in two words that recur in the gold note: cyclical and structural. Soft investor demand, Indian import duties raised to fifteen per cent, outflows from exchange-traded funds, all of this is filed under cyclical, short-term noise. Central bank purchases, which by contrast confirm the bullish thesis, become structural support. The same category, cyclical against structural, is assigned not according to the nature of the data but according to whether the datum confirms or undermines the conclusion already written. The number one likes is an underlying trend. The number that disturbs is a passing fluctuation. It is the tail wagging the dog, and it is set down in black and white in a document produced by one of the most serious institutions in the business.

Then there is the move I call the horoscope. The equity note recommends, within a single paragraph, the artificial intelligence value chain, industrials, automation, space, mainland China, Singapore, Japan, Swiss equities, healthcare, consumer discretionary, information technology, Germany. Everything is in there. And when a recommendation contains everything it is not selecting anything, it is covering. It is built like a horoscope: broad enough to be always partly right, never precise enough to be falsified. An adviser who says he likes all of it has not expressed a position, he has merely granted the reader psychological permission to stay where he already is, in whatever he already holds.

And here comes the line that explains all the rest, at the foot of the equity note, where the reader arrives already persuaded. Investors can use rallies in the megacap technology names to rebalance into structured products and multifactor strategies. There it is. Eighteen lines of earnings growth at nineteen per cent, of Asia at sixty-two, of European electrification, all converging on a single line that moves the client towards products carrying a higher margin for whoever places them. The eighteen lines are not the analysis to which the closing line is a footnote. They are the intellectual legitimacy that makes the closing line acceptable. The laboratory report existed in order to get the prescription signed.

None of this requires bad faith, and that is the part I struggle to convey to anyone expecting an exposé. The people who write these notes do their job extremely well. The job, however, is not the one the reader believes he is reading, since a bank that manages wealth has a structural interest in keeping the client invested, because the client who sells and goes to cash stops paying management fees and stops buying products. The forecast is not the outcome of a process of inquiry that might conclude with a sell. It is the product, and a well-designed product almost never lies; it simply selects, frames, emphasises and omits in one direction only, the one compatible with whoever produces it.

The shift of the first rate cut to December, after months in which it had been expected sooner, travels tucked inside a subordinate clause, without a line conceding that until recently the bank had been saying otherwise. The gold target falling seven per cent after a fifteen per cent decline in the metal, and the recommendation that stays buy. The diagnosis is updated continuously, the prescription never, and when I see a diagnosis that changes and a prescription that does not move, I have learned to trust the prescription as the signal of what is really needed by the one writing, rather than by me.

To read these notes, then, is not to distrust them, which would be another kind of naivety. It is to perform the one operation the document is built to discourage: to separate the diagnosis from the placement. The diagnosis is often excellent and worth reading. The placement is the line at the foot, and it should be read in the knowledge that it is the reason for every line above. Anyone who holds two or three notes from the same week together, and looks for the paragraph that repeats itself unchanged in the service of opposite conclusions, has already done half the work. The other half is remembering that on that side of the table, the side where the note is written, no one is sitting on your behalf.