Field Notes

A record that read as ordinary: the top-1% share, and oil joins the board

On 19 June the Federal Reserve published the Q4 2025 Distributional Financial Accounts, and the share of net worth held by the top 1% of households (S7) came in at 31.6%, the second-highest reading in the entire 1989 to 2026 record. The board responded by moving S7 down a notch, from EXTREME to WATCH. In the same week the financial press marked the minting of the world’s first trillionaire. A gauge of concentration had eased at the moment concentration was setting records.

The cause was not the data. It was the gauge, and it was the same trap the 12 June sweep was built to catch. S7 scored each reading as a z-score against its own trailing decade. But this is a series that has done almost nothing but rise: the last ten years all sit between 30 and 31%, so the window’s own average had climbed to about 30.5, with a standard deviation near half a point. A fresh near-record clears that recently-elevated average by only about 1.9 standard deviations, just under the line. A series that only ever rises will read each new high as ordinary, because its own recent past is the only thing it is compared against. S7 was the one concentration tile the June sweep did not reach.

We moved S7 to absolute thresholds anchored on the full record rather than on the recent decade: below 28.5% calm, 28.5 to 30.5% watch, above 30.5% extreme. The low cut sits at the long-run median, 28.1%, the pre-financialization norm; the high cut sits at the top fifth of the entire record, the floor of the post-2010 regime. At 31.6% the tile reads EXTREME. The late-1990s and 2000s broadening, when the share fell to 27 and 28%, reads CALM, because concentration genuinely was moderate then. It is not moderate now, and the tile now says so.

A second change, different in kind. WTI crude (C41) had been carried as context, an input price shown on the board but never scored. The week made the cost of that visible. Oil swung hard, the national average price of gasoline fell below four dollars for the first time in months, and the board’s only energy-strain signal was energy CPI (C35), which is monthly and trails the pump by a month or more. So the relief at the pump was real and the board could not yet see it. We promoted C41 to a scored fast-tell on absolute bands, 90 and 110 dollars a barrel, scored one way only: a price spike is household strain, while cheap oil is relief, not a recession tell, which the labor tiles carry. At 84.65 dollars it reads CALM, which is the relief the week delivered.

What this is, and what it is not. No underlying number changed. S7 is a recalibration of how a real reading is read, and C41 is an existing series finally given a state. Every threshold is pre-committed and written into the Methodology changelog, dated today. The instinct that a record share of the nation’s wealth should not read calm was the right one, and the gauge has been made to agree.

When the gauge cannot read red: recalibrating inflation and strain

For several months the board carried a quiet contradiction. Headline CPI (C2) climbed to 4.27% in May 2026, the highest since 2023 and the third straight monthly acceleration, with energy (C35) up 22% on the conflict with Iran. Yet C2, and every realized-inflation tile beside it, read CALM.

The cause was not the data. It was the gauge. These tiles scored each reading as a z-score against their own recent history, and that history includes the 2021 to 2023 inflation spike, when CPI touched 9%. A window that holds a 9% peak has a wide standard deviation, so 4.27% lands at only half of one, “below typical”, even while sitting at the 79th percentile of its own decade. The tile’s own plain-English twin said “higher than 79%”; its state said calm. The reference window had been poisoned by the very episode it was meant to detect.

We moved the realized-inflation levels to absolute thresholds, anchored on the 2% target rather than on their own worst year. C2 (CPI) and C39 (the month-over-month print) now read EXTREME; energy (C35) reads EXTREME; core PCE (C3), headline PCE (C40), shelter (C36), and core services (C38) read WATCH. Core goods (C37), genuinely below target at 1.0%, correctly stays CALM, which a rank-based rule would have wrongly flagged.

A full audit of the rest of the board found the same trap had caught a cluster more, and always in one direction: toward false calm, never false alarm. The clearest case sat in the open. Real wage growth for all workers (S2) read CALM at negative 0.82%, while the production-worker version (S21) read WATCH at negative 0.71%, the only difference being that S2’s window includes the 2022 real-wage collapse. Real wages are shrinking under the price surge; a negative print cannot honestly read calm.

Five more tiles moved to absolute thresholds and now read WATCH: S2 (real wages), F10 (Treasury auction demand, whose weakness the R6 relationship had already flagged as STRAINED), S1 (the Misery Index), F2 (the VIX), and C24 (credit-card delinquency). One, C16 (unemployment duration), was re-anchored but stays CALM: at 11.6 weeks it is genuinely low, its high percentile an artifact of an unusually calm decade rather than real strain.

What this is, and what it is not. None of the underlying numbers changed; this is a recalibration of how they are read, and every threshold is pre-committed and written into the Methodology changelog, dated today. The board’s census moved from 49 calm, 13 watch, 3 extreme to 38, 21, 6. The instrument now reads the inflation it was always seeing. A board that cannot read red when red is real fails the same way one that can only read green does.