Regimes names the stage of the cycle each month has been in, from State of the Economy's three gauges, by rules written down before the engine was built. This page is those rules, in full, and the record of the one test that decided what the third gauge is allowed to do.
The rules are transcribed from the ratified decision record (REGIME-DR-1 v3 (ratified 2026-09-19)). The stage table, the stress episodes and the settling counts are read from the engine's own route as this page renders. Two sets of figures are quoted from the gates that measured them: the Forward Stress test's, and the shares of changed months near turns.
| law | what it binds |
|---|---|
| Fixed rules, set in advance | The rules below were written into the decision record and ratified before the engine was built. They are not fitted to the record they classify. |
| Every claim tested against the record | A claim reaches a surface only after it has been run over the whole record and its result recorded — including when the result is that the claim fails. |
| Counts, never probabilities | Every "what followed" on the Regimes page is a count of periods in this record. None of them is a forecast, and none is expressed as a likelihood. |
| Point-in-time honesty | The gauges are read as first published, from the vintage score table — not as later revisions made them look — as far as the archived first prints reach; where they do not, the older record is partly today’s data, read only once it had been published, and the methodology measures how much. |
| Independence | The regime reads the State of the Economy’s three gauges and nothing else. No feature depends on another: State of the Economy, the Capitulation Monitor, the Housing Cycle and Economic Breadth each stand on their own principles. A synthesis of what they mean together is a separate surface, written after the pillars exist. Amended October 3, 2026: a feature classifies only on its own gauges; its record may describe the world it happened in. The Capitulation Monitor describes washouts by the regime’s stage, and is permitted under that wording. |
Three words, and they do not trade places.
| word | what it names |
|---|---|
| period | a run of months in one stage |
| episode | a run of the stress signal |
| stage | one of the four |
The feature is Regimes. The words spell, era and band do not name any of these.
Context per month, point-in-time: M, FS, IR (0–100; 50 the median; FS higher = healthier; IR lower = more strained); dM, dFS (3-month change); tr6 (IR’s 6-month change; positive = easing); side (+1 core PCE above 2%, −1 below); FS’s own 12-month high.
M is Momentum, FS is Forward Stress, IR is the Inflation Regime gauge — the three State of the Economy composites, read as first published — at least 85% first-published from Sep 2016; earlier years partly use today's revised data, read only once it had been published (how much).
Four ordered stages — Recovery · Expansion · Slowdown · Contraction — classified on Momentum. Forward Stress enters only in the frozen Contraction rule's compound branch. First match wins.
| # | stage | the rule, verbatim |
|---|---|---|
| 1 | Contraction | M < 30, or (M < 40 and dM ≤ −5 and FS < 40) — the frozen v5 form. |
| 2 | Recovery | entered after a Contraction period, with two months’ confirmation (the Contraction rule off and dM > 0, two months running); held while M < 60 and M ≥ (post-contraction trough + 5) and dM ≥ −3; exits to Expansion at M ≥ 60 or after 24 months; to Contraction by rule 1; to Slowdown if M falls back below trough + 5. A Contraction may exit directly to Expansion when M ≥ 60 on the exit month (2020’s V). |
| 3 | Slowdown | (M < 50 and dM ≤ 0) or M < 45. |
| 4 | Expansion | everything else with M ≥ 45. |
Read from the engine as this page loaded — the same rows the Regimes page's stage cards carry, over the record from Aug 1981.
| stage | periods | median length | became a Contraction within 18 months |
|---|---|---|---|
| Contraction | 6 | 6.5 months | — (it is the Contraction) |
| Recovery | 5 | 4 months | 1 of 5 |
| Slowdown | 13 | 3 months | 5 of 13 |
| Expansion | 13 | 27 months | 3 of 13 |
Counts of what followed in this record, never probabilities.
Persistence. M < 30 → Contraction immediately; the compound Contraction rule, or a Slowdown with dM ≤ −8 → 2 months’ confirmation; everything else → 3.
Exit thresholds. Slowdown leaves at M ≥ 55 or (M ≥ 50 and dM > 0 two months running); Contraction leaves only via Recovery, or straight to Expansion at M ≥ 60. Order: forward skips allowed; Contraction → Recovery (or Expansion at M ≥ 60) only. The latest months carry the standing stage with a forming note ("Slowdown forming, 2 of 3").
The direction words. cooling (dM ≤ −8 in 2 of the last 3 months) · firming (dM ≥ +8 in 2 of 3) · else none. Every direction word carries two-month persistence: a word changes only after two months of the new condition.
level = the Inflation Regime gauge’s own word (house_ladder: unusually calm · calm · neutral · strained · severely strained on the 80/60/40/20 edges); direction = rising (tr6 ≤ −5) · stable · easing (tr6 ≥ 5).
The overlay never changes the stage; it is read beside it. The top two bands of that ladder once shared the word “calm”; the top one now reads “unusually calm”.
On State of the Economy's banner the stage comes first, and inflation is read on the line below it — today, “Inflation: Neutral, stable”. The stage is named on Momentum; inflation is read beside it, never into it.
The stress signal opens when Forward Stress is ≥ 20 below its own 12-month high with M ≥ 50 for 2 consecutive months; it closes when Forward Stress is within 10 of its 12-month high for 2 months, or when the stage leaves Expansion.
| opened | closed | months | a Contraction followed within 18 months | S&P 500 TR, 12 months from the open |
|---|---|---|---|---|
| Oct 1984 | Mar 1985 | 6 | no | +17.9% |
| Apr 1989 | Jul 1989 | 4 | no | +15.5% |
| Sep 2000 | Dec 2000 | 4 | yes | −27.9% |
| Dec 2006 | Jul 2007 | 8 | yes | +6.3% |
| Mar 2019 | Nov 2019 | 9 | yes | −3.6% |
| Jun 2022 | May 2023 | 12 | no | +13.3% |
The fourth column tests the stage — whether a Contraction period began within eighteen months. The adjudication below tested something different: whether an NBER recession began. The two can disagree on one episode, and below they do.
Momentum names the stage. Forward Stress enters one branch of one rule and otherwise sits beside the stage as a signal. That is not an assumption; it is the result of a test that was written down in full — question, tiers, candidates and the bar — before it was run.
The question. Does Forward Stress carry information about coming deterioration that the four-stage growth classification does not already carry — and if so, which kind of deterioration, how far ahead, at what false-alarm cost?
| tier | the events |
|---|---|
| T1 economic | the 5 NBER recessions since 1981 (primary) |
| T2 economic deterioration without recession | Economic Breadth net < 0 for ≥ 3 consecutive months, on the revised record |
| T3 market | QUAD-1’s ≥ 20% drawdowns and, weaker, ≥ 10% corrections, from the S&P total-return chain, peak to trough |
| # | reading | the condition |
|---|---|---|
| 1 | level | FS < 40 while M ≥ 45 |
| 2 | direction | dFS ≤ −8 in 2 of 3 months while M ≥ 50 |
| 3 | drawdown from high | FS ≥ 20 below its own 12-month high while M ≥ 50 |
| 4 | stress-led Slowdown | M < 55 and FS < 40 joins Slowdown’s rule |
| 5 | stress-led Contraction | M < 45 and FS < 40 joins Contraction’s rule |
| 6 | tie-breaker | within 3 points of the 50 cut, FS ≥ 45 keeps Expansion, FS < 45 sends Slowdown |
The bar, applied per tier and stated as descriptive, not statistical: an episode-level hit on ≥ 4 of 5 T1 events with each lead listed; false alarms ≤ hits; AND the incremental test shows FS earlier than the growth stage by ≥ 2 months on at least 3 of 5. A candidate meeting the bar on T3 but not T1 is a market-stress signal, not an economic classifier, and is reported as such.
The original run's own verdict: no candidate met the bar on T1 or T3. Read as a level (FS < 40 at any Momentum; the stress-led Slowdown and Contraction; the tie-breaker at the 50 cut) FS added nothing the growth stage did not carry — one hit in four recessions, ten months after the growth stage’s own warning, or no tallied episode at all. Read as a fall from its own recent level it did.
Re-run using only the data available at the time, the verdict held. The one hit for a level reading was the tie-breaker's, and it came five months after the growth stage's own warning.
So FS stays out of the classification, and its direction is served beside the stage as a signal with its record.
The reading that did carry something is candidate 3 — the fall from Forward Stress's own twelve-month high — and it is the rule the stress signal is built on. Against the recessions it fired early three times:
| NBER recession | the growth stage's first warning | the signal fired | earlier by | lead to the recession |
|---|---|---|---|---|
| Aug 1990 | Jul 1989 | Apr 1989 | 3 months | 16 months |
| Jan 2008 | Jul 2007 | Dec 2006 | 7 months | 13 months |
| Mar 2020 | Mar 2020 | Mar 2019 | 12 months | 12 months |
Candidate 3 was tallied over five episodes: three hits, each earlier than the growth stage’s own first warning, and two false alarms in 45 years — 1984-10 and 2022-06. A sixth opening, 2000-09, was excluded by the pre-registered rule that an episode opening inside a drawdown already under way is not tallied. Three of five is not four of five, so the bar was not met.
The rows above and in the tiers are quoted from the adjudication re-run on the as-known record. The run's month-by-month tables live in gate publag_3_fs_adjudication_tables.md (the original run is regime_1_fs_adjudication.md). The episode table higher up is served live by the engine, which is why its fourth column and this one count different things.
The stage is classified on data as it was first published — at least 85% first-published from Sep 2016; earlier years partly use today's revised data, read only once it had been published (how much). That data is then revised, so the same rules run over today's record can name a different stage for a month already on the chart. Over the whole record that happens in 68 of 542 months.
The changes are not spread evenly, and three patterns in them are worth a reader's attention. They concentrate in one direction — 32 of the 68 changes (under half) are a month first read as an Expansion that settles as a Slowdown. They concentrate near turns, at roughly a third of changed months against a fifth of all months. And they concentrate in recent months: a month whose data is still inside the three-year settling window is about 3 times more likely to be carrying a stage that will change than one whose data has settled. Part of that contrast is built in, because an older first read already carries some of today's data.
For the stage the site reads today, the counts are on the Regimes page itself, under "How firm is this reading?". The three-year window is the same convention the Revisions Ledger uses to call an observation settled, so the two features measure settling the same way.
The site shows the point-in-time stage regardless, and that is deliberate rather than a limitation. A chart that showed each month as today's data classifies it would read the whole record with knowledge no reader ever had, and would quietly improve every past call the engine made. Showing the first read keeps the record as honest about what was knowable when as the archives allow — and this section is how the difference is disclosed instead of hidden.
Where the site reads the gauges “as first published” or “point-in-time”, each month is scored from the data as it stood about five weeks after the month began — where an archived copy of that data exists. Where none exists, that part of the month is scored from the latest figure that had been published by then, at today's revised value — never from a figure that had not yet come out. So the older record is only partly first-published in its values, though right in its timing; the recent record is first-published almost throughout.
How much, by era. Of the weight each gauge actually used, the share that was first-published — every gauge stays at or above 85% from Sep 2016:
| Years | Momentum | Forward Stress | Inflation Regime |
|---|---|---|---|
| 1964–80 | 35% | — | 0% |
| 1981–89 | 35% | 52% | 8% |
| 1990–99 | 33% | 52% | 12% |
| 2000–09 | 38% | 65% | 74% |
| 2010–14 | 84% | 76% | 80% |
| 2015–now | 100% | 90% | 97% |
The daily record reads the same, era by era (within 3 points). The Time Machine's coverage figure is a different measure — the share of each gauge's full design weight. It counts inputs that are never revised as first-published even in years before they existed, and any other input that did not exist yet as not first-published. So for the oldest years it reads higher than this table for Forward Stress and Inflation Regime, and lower for Momentum.
Which inputs lack first prints, and until when.
What it means for the older record. A reading from the years the table shows as partly first-published is partly today's understanding of that time, not only what a reader could have seen then. Comparisons between a month's first-published reading and its revised reading are narrower there than they would be with complete first prints, because part of both is the same revised data.
The served timeline is point-in-time — the gauges as the vintage score table records them as first published, where an archived first print exists, and otherwise at the latest figure that had been published by each month’s decision date (the section above gives how much). The revised record is computed by the same function from today’s data and differs from it in 68 of 542 months. Both are served, and each surface names which one it is reading.
The Regimes page reads the point-in-time record (this render: point-in-time). The what-changed table in State of the Economy's Time Machine reads the same engine on whichever basis that surface declares.