Paulo Macro — Low correlations, dispersion, and a bond-curve divergence into payrolls (Sep 2026)

One-line take: a set-up note, not a trade note. It opens with his own month-end rule, stated as a joke that is really a discipline: "with month-end out of the way (remember what we say about month end: whenever I feel the urge to trade it, etc etc)." The market observation is a market-structure one: "My mind keeps coming back to just how low correlations (and how large dispersion) has become as the tension builds toward a market vulnerability or state of criticality that results in Risk Off." The timing caveat is the Dornbusch paraphrase: "it takes far longer for conditions to decompress than you think, but then it always happens faster than you imagined." Two things are on his screen in the meantime — "positioning in bonds," and "a notable divergence between 2yr, 5yr, and 10yr notes/bonds making new lows in futures while the 30yr has not broken down. Maybe nothing... maybe something." The catalyst is dated: "With August payrolls coming up this Friday, recall I have discussed at length since June the prospect of a soft patch in US data between midyear and mid autumn, and the Citi Economic Surprise index would suggest this is happening." The question he leaves open is the whole note: "Will a miss on NFP be enough to shake Risk and push some rotation in bonds' direction? We'll see..." Sign-off: "Stay frosty..."

Key points

Row-scope note. This is a pure macro / market-structure note and Paulo names no company, fund or ETF at all — the objects are cross-sectional correlation and dispersion, Treasury note and bond futures (2yr, 5yr, 10yr and 30yr), the August non-farm payrolls print due Friday 4 September, and the Citi Economic Surprise index. Per this hub's convention, rates, FX and macro indicators are not tabled as tickers unless he names a vehicle (see the Apr-27 JGB / USD-JPY note, likewise untabled), so there is no stock table on this page — nothing is invented. The longer "note later today" that he promises is a separate publication and is not part of this chat post.

Month-end — the rule is to not trade it

"Tue 1 Sept 2026... good morning... with month-end out of the way (remember what we say about month end: whenever I feel the urge to trade it, etc etc)."
The "etc etc" is doing the work of a standing house rule the readership already knows: the urge to trade month-end is itself the signal to sit out. Month-end flows (index rebalances, pension rebalancing, fund marks) move price for reasons unrelated to any thesis, so signal quality collapses for a few sessions.
Framing the first day of September as "month-end out of the way" tells you what he thinks the preceding sessions were worth as information: nothing. His clock restarts here.

Low correlation, high dispersion — the pre-condition, not the event

"My mind keeps coming back to just how low correlations (and how large dispersion) has become as the tension builds toward a market vulnerability or state of criticality that results in Risk Off."
Two measurements, one read: names are moving independently of each other (low correlation) and the spread between winners and losers is wide (high dispersion). On the surface that reads as a healthy stock-picker's tape.
His inversion is the point. A low-correlation tape is one where risk has been distributed into many separate, crowded single-name bets rather than one index position — and hedges calibrated to a low-correlation world are exactly the ones that fail when correlation snaps to one. Dispersion is the fuel; the Risk Off is the ignition.
"State of criticality" is his own recurring phrase — the same language he used at mid-year in Jul-08 ("among the most dangerous I have ever seen") and again in Aug-28 "When Things Diverge." This note is the next instalment of one continuing argument, not a new call.

The Dornbusch caveat — slower than you think, then faster than you imagined

"As with everything in the market, it takes far longer for conditions to decompress than you think, but then it always happens faster than you imagined (to paraphrase Dornbusch)."
Rudi Dornbusch's line was about crises: things take longer to happen than you think they will, and then they happen faster than you thought they could. Paulo applies it to decompression — the unwinding of a stretched condition, not the crisis itself.
It is a deliberate hedge against his own conviction: the diagnosis can be right and the position still wrong for months. That is the argument for owning convexity (cheap, dated optionality) rather than carrying a directional short — the same conclusion he reached in Aug-28 ("near-dated equity (and credit) volatility is grossly mispriced and cheap").

Bond positioning, and the 30yr that has not broken

"In the meantime, positioning in bonds continues to catch my attention, along with a notable divergence between 2yr, 5yr, and 10yr notes/bonds making new lows in futures while the 30yr has not broken down."
Read it in futures terms: contracts making new lows means those yields pushed to new highs across 2s, 5s and 10s — but the long bond has not confirmed. A move that fails to carry the whole curve is a move without a single macro driver behind it.
That non-confirmation is the tell he is flagging. If the selling were about inflation or term premium, the 30yr would lead, not lag; a front-and-belly-led sell-off looks more like positioning — which is precisely the word he leads the sentence with.
He refuses to over-claim it: "Maybe nothing... maybe something." It is logged as an anomaly to watch, not evidence for a trade.

Friday's payrolls, and a soft patch he has been early on since June

"With August payrolls coming up this Friday, recall I have discussed at length since June the prospect of a soft patch in US data between midyear and mid autumn, and the Citi Economic Surprise index would suggest this is happening."
The forecast is dated and bounded — a soft patch, "between midyear and mid autumn" — and it was made in June, so the August NFP print falls inside a window he called in advance. That is what makes it a test rather than a narrative.
The Citi Economic Surprise index measures data versus expectations, not data versus its own history: a falling reading means forecasters are being caught out on the downside, whatever the absolute level of activity. He cites it as corroboration that the soft patch is arriving on schedule.
Note what he is not doing: no payroll forecast, no number, no recession call. The window and the confirming gauge are the whole claim.

The open question — would an NFP miss actually rotate money into bonds?

"Will a miss on NFP be enough to shake Risk and push some rotation in bonds' direction? We'll see..."
This is where the three threads join: a low-correlation tape that is vulnerable, a bond market whose front and belly have already sold off on positioning, and a dated catalyst that could flip both. A weak print would be the first live test of whether bonds still work as the risk-off hedge.
Left explicitly unanswered. The value of the note is its conditional structure — if the soft patch is real and the payroll print misses, watch whether money rotates into duration or whether the bid simply fails to show up. The second outcome would be far more informative than the print itself.
Sign-off: "Stay frosty..."

In plain English

What he is actually saying

Picture a crowd walking across a footbridge. While everyone walks out of step the bridge is calm — that is a market with low correlation: shares are going up and down for their own individual reasons rather than all together, and the gap between the best and worst performers is wide (that gap is dispersion). It looks like a healthy market, and it is usually described as one. Paulo's worry is that the calm is exactly what lets investors take bigger and bigger separate bets — so when something finally makes the crowd fall into step, they all lean the same way at once. He calls the build-up "a state of criticality," and the falling-into-step is "Risk Off": the day everything sells together.

His warning about timing is the important part, and it is borrowed from the economist Rudi Dornbusch: these things take much longer to arrive than you expect, and then arrive much faster than you thought possible. In practice that is an argument against betting on a date and in favour of holding cheap insurance, because being early costs money in exactly the same way as being wrong.

The bond observation is a plumbing detail with a simple meaning. US government debt comes in different maturities — roughly, how long until you get your money back: 2, 5, 10 and 30 years. In the futures market the 2-, 5- and 10-year contracts have just made new lows (i.e. those interest rates pushed to new highs), but the 30-year has not followed. When the longest bond refuses to join in, it suggests the move is being driven by traders' positioning — who is short, who is being squeezed — rather than by a genuine change in the outlook for inflation or growth. He notes it without claiming to know which: "maybe nothing... maybe something."

Finally, the date to watch. Since June he has argued the US economy would pass through a "soft patch" — a stretch where the data comes in weaker than economists expected — somewhere between the middle of the year and the middle of autumn. The Citi Economic Surprise index, which simply tracks whether data is beating or missing forecasts, says that is now happening. Friday's August jobs report falls squarely inside that window. His open question is whether a weak jobs number is enough to frighten investors out of shares and into bonds. He does not answer it — and the honest reading is that how bonds respond will tell him more than the jobs number itself.

Key points extracted from the PauloMacro Substack chat note of 2026-SEP-01 (saved in transcript.txt) for personal study. No securities are named in the note and none are inferred. Not investment advice. © PauloMacro for source material.