How Inflation Dies: Demand Destruction and the Empty Reservoir

Builds-on: demand-side-audit-may-2026, energy-and-stagflation-forecast-2026-2031, cyclical-20-and-the-ai-capex-mask Related: regime-cascade-architecture, regime-check-june-10-2026, japan-debt-trap-thesis-audit, the-warsh-fed-and-the-financial-repression-thesis, ai-circular-financing-and-banking-exposure-audit, ai-crash-portfolio-defense Informs: portfolio-rebalance-april-2026, polly-fidelity-403b-allocation Led-to: failure-cascade-index, two-economy-gauge, normalization-vs-destruction-test (the open questions operationalized as standing gauges) Prior conversations: economic-deflation-and-interest-rate-forecast (June 2025 — quarterly deflation probabilities), economic-cycles-stellar-evolution-parallels (June 2025 — the fuel-exhaustion frame), japans-economic-cycles-us-stagflation-modeling

The question

The intuition under examination: at some point inflation stops because people stop paying. Businesses fail because input costs outrun what customers will bear. The question is the mechanics — how that break happens, what triggers it, how it has happened before, and how the cycle runs.

The one-line version of the answer, which the rest of this doc unpacks: cost-push inflation is a flow that has to be paid for out of a stock. When the stock runs dry, the flow stops — but it stops discretely, not smoothly, because businesses are options that fail all-at-once and layoffs are synchronized. The old commodity-trader line is "the cure for high prices is high prices" (demand destruction). What that line hides is that the cure usually arrives as a recession.

Reconciling the dam metaphor

regime-cascade-architecture already runs a hydraulic model of this system: reservoirs of accumulated stress, dams that hold the stress back, channels it cascades through. This doc adds the piece that model doesn't name: the demand reservoir — the pool of household purchasing power (wages + savings + credit capacity + transfers) that sits downstream of everything and powers the whole exhibit. Inflation is a turbine. It only spins while there's water behind it. "The dam can't be empty to operate" is exactly right: sellers can post any price they want, but a price only becomes inflation when someone pays it. The stress reservoirs in the cascade doc fill as this one drains — that's the relationship between the two models.

The stellar-evolution version from economic-cycles-stellar-evolution-parallels is the same claim: the phase transition comes when the fuel runs out, and the transition is not gradual.

Why cost-push inflation is self-limiting (theory)

Three frameworks converge on the same conclusion from different directions:

1. The monetarist real-balance argument (Friedman). A cost shock without monetary accommodation is not inflation — it's a relative-price change plus a recession. Higher prices for energy/food drain real balances; spending on everything else falls; those prices soften; the aggregate level stabilizes. Cost shocks "fail to generate economy-wide inflation without central bank accommodation" (Grokipedia summary of the cost-push literature; AEA retrospective on Friedman and the "cruel dilemma"). The 1970s persisted because the Fed accommodated; Volcker's contribution was refusing to.

2. Weber's sellers'-inflation phase model. Weber & Wasner (2023) break an inflation episode into phases: impulse (shortage in a systemically important input) → propagation (firms with pricing power pass through costs and fatten margins under cover of the shock) → conflict (labor tries to catch up; if it succeeds, the spiral continues). The episode ends when phase three fails — when labor can't claw back purchasing power and firms discover volume breaks before the next price increase sticks. Corporate profits accounted for roughly two-thirds of the 2020-2022 US GDP-deflator increase versus wages' one-third, which is why that episode could disinflate without mass unemployment: margins had room to give before jobs did.

3. The reservoir accounting. Household spending is funded by exactly four pumps: real wages, savings drawdown, credit expansion, and fiscal transfers. Sustained inflation requires at least one pump running. Check all four and you know whether the dam is refilling or draining. This is the practical monitoring version of #1 and #2.

(Addendum 2026-08-12 — the Kalecki channel, via EPB Research.) The fourth pump has a downstream effect the original frame missed: by the Kalecki profits identity, with household savings thin and net private investment flat, the ~6-7%-of-GDP federal deficit flows through transfer-funded consumption directly into corporate profit margins. Record profits alongside a drained household reservoir aren't a paradox — they're the same flow observed at two points in the pipe. Consequence: the margins justifying equity valuations are substantially fiscal artifacts, which gives the profits leg of the market and the government budget a single shared point of failure (austerity, a debt-ceiling standoff, or a bond-market revolt against issuance drains margins directly — and July's -53K government payrolls shows the pump already throttling at the employment level). Watch: deficit trajectory into the midterm fiscal fights, alongside the four household pumps.

How the break actually happens (the sequence)

The crash mechanics matter because they're non-linear. A business facing input costs above its customers' willingness to pay doesn't shrink smoothly — it operates at a loss out of working capital and hope, then fails discretely. Firms are options: they run until cash breaks, then exercise into bankruptcy all at once. That's why the demand side "holds up better than expected" right up until it doesn't.

The canonical sequence:

1. Margin squeeze. Input costs rise faster than realized prices. Firms absorb, hoping it's temporary. (Japan right now: firms eat ¥60 of every ¥100 cost increase — see below.)
2. Pass-through attempt → volume response. Price hikes stick at first (consumers grumble and pay), then volumes roll. Trade-down begins: private label, smaller baskets, deferred replacement. Private-label share becomes a hard ceiling on branded pricing.
3. The discounting phase (bullwhip). Retailers who over-ordered into the inflation meet the volume decline and liquidate. Goods prices actually fall — the 2022-23 US inventory glut produced 30-50% markdowns and was a major mechanical driver of the "immaculate disinflation."
4. Failure cascade. The weakest cost-absorbers (small, labor- and energy-intensive, thin-margin: restaurants, independent retail, small manufacturers) fail in a wave. This is where "business going out of business because input costs are too high" shows up in the data — always late in the episode, because of the option structure.
5. Layoffs → income destruction → second round. The failures and the survivors' cost-cutting hit employment together. Now the wage pump slows just as the credit pump maxes out. Demand falls further than the initial price shock justified — overshoot.
6. Pricing power inverts. Sellers who spent two years raising prices start competing on price. Headline inflation rolls over fast because energy demand destruction is the most non-linear of all (gasoline demand is inelastic until the marginal household is choosing between gas and rent — energy-and-stagflation-forecast-2026-2031 flagged exactly this threshold).

The whole loop is why the answer to "what triggers the crash" is usually the labor market, not the consumer directly. Consumers degrade gradually (savings → credit → delinquency); employers break discretely. Watching for the crash means watching for synchronized firing, not for the shopper giving up.

The historical episodes

What the episodes teach

Trigger taxonomy, ranked by speed:

1. Credit/financial event (2008) — fastest. Days to weeks. The reservoir doesn't drain; the pipes burst.
2. Policy shock (Volcker) — quarters. Deliberate draining via the credit pump.
3. Price spike past the rationing threshold (1920, Europe 2022) — one season. The commodity does the Fed's job.
4. Inventory/bullwhip mechanics (2022-23) — quarters, and can produce disinflation without a crash if the reservoir holds.
5. Margin-exhaustion bankruptcy cascade (UK, Japan) — slowest, 1-3 years. This is the "businesses die because input costs are too high" path, and it's the default path when no faster trigger fires first.

Three structural lessons:

• Demand destruction at the top of the cycle is often permanent (US oil post-1979, EU industry post-2022). The reservoir doesn't refill to the old waterline; the economy reorganizes around the lower flow. That's why these episodes mark regime boundaries, not just cycles.
• The crash needs no policy trigger. 1920 and Europe 2022 broke on price alone. Volcker is the exception people treat as the rule. If the Fed is politically constrained (the the-warsh-fed-and-the-financial-repression-thesis world), the market runs the Volcker function through the failure cascade instead — slower and less evenly distributed, but the same destination.
• Inflation persists only where the reservoir refills. Turkey and Argentina ran multi-year inflations because wage indexation and money growth kept pumping water uphill. The US 1970s persisted on accommodation. The corollary: watch the four pumps, not the price prints.

The 2026 US scoreboard: reservoir gauges

Where the four pumps stand as of July 2026:

Wages — slowing. Average hourly earnings +3.5% YoY in June, down from ~4% a year ago; June payrolls +57K, gains averaging ~40K/month since January 2025; unemployment 4.2% only because participation fell to 61.5% (lowest since March 2021). Quits low — workers staying put, no bargaining power. The wage-price spiral fuel is absent. Weber's phase three (labor conflict) is not happening; labor is losing the conflict quietly.

Savings — drained for the bottom, full at the top. Personal savings rate 4.0% (from 6.2% in early 2024); pandemic excess savings gone for the bottom quintiles. The K-shape is the load-bearing fact: credit stress is concentrated in lower-income cohorts while top-decile households — doing roughly half of all consumption — spend from asset wealth.

Credit — pump running hot, near redline. Card balances $1.33T record; 90+ day delinquencies 13.1%, worst since the financial crisis; 6.43% delinquency at small banks where lower-income borrowers concentrate; "survival debt" — cards covering groceries and utilities. This is the last pump before the delinquency wall, and demand-side-audit-may-2026 already logged the auto-loan version (90+ DPD at levels that implied 9% unemployment in 2010).

Transfers — off. No fiscal impulse aimed at households; tariffs are a negative transfer.

Meanwhile the cost side is still pushing: May CPI 4.2% headline (energy >60% of the print, core 2.9% — regime-check-june-10-2026), tariff pass-through peaking through Q2 2026 with PIIE arguing 4%+ by year-end is the modal case, oil ~$90 on the Iran regime. And the demand response is now visible in the corporate tape: consumer spending posted its sharpest decline in four years in January, real retail spending fell 0.7% in a single recent month, staples are slashing forecasts, and the P&G/Walmart earnings language (value pivots, private-label surge, "defend price or defend volume") is textbook stage 2-3 of the sequence above: pass-through failing, discounting beginning.

So the configuration is: cost-push accelerating into a draining reservoir. That's the 1974 and 2008-H1 configuration, not the 2021 one. The 2021-22 inflation ran on full tanks (stimulus savings + hot labor); the 2026 inflation is running on fumes, which is why the vault's read (supply-shock, not demand-pull; "the Fed can't fix this with rates") and this doc's read agree: the binding constraint on this inflation is not policy. It's the water level.

Where 2026 sits in the sequence: stage 2 crossing into stage 3 in the aggregate — but tier-corrected, further along. Margin squeeze is on (staples guidance), volume response is visible (trade-down, private label ceiling), discounting has started in pockets. Correction 2026-07-11, from the first failure-cascade-index reading: stage 4 has already fired at the small-business tier — Subchapter V filings +67% YoY in Q1 2026, +50% for H1, commercial Chapter 11s +28-42% — masked by low post-COVID base levels and by aggregate filing counts blending the small-firm surge with a calmer large-firm tier. Stage 5 (synchronized layoffs) has not fired: the low-hire/low-fire freeze is the option-holding phase for the firms still alive, and small-business employment is flat. The cyclical-20-and-the-ai-capex-mask falsification list (U-3 above 4.6-4.8%, Sahm above 0.30, continuing claims above 1.95M) is the stage-5 tripwire, now supplemented by the failure-cascade index's fuse gauges (SLOOS small-firm tightening crossing ~25-30% with the PPI-CPI absorption wedge still open).

The K-shape caveat — why the crash keeps not arriving. The aggregate dam is half-full because it's tiered. The bottom reservoir is dry (hence delinquencies), but aggregate spending is increasingly funded by the top tier, whose reservoir refills automatically from asset prices. This means the true trigger for a full demand crash in 2026 America is not the marginal consumer — she already broke, quietly, without moving the aggregate — it's an asset-price event that drains the top reservoir. Which connects this doc to ai-circular-financing-and-banking-exposure-audit: the 50-60% equity-correction base case over there is the demand-destruction trigger over here. The AI trade isn't just masking the cyclical data (cyclical-20-and-the-ai-capex-mask); via the wealth effect it is actively holding up the last funded tier of consumption. One dam, two jobs.

The other side of the dam: asset-side demand destruction

If the top-tier reservoir is asset-fed, then the question "when does demand destruction reach the top" reduces to "which asset classes still hold their price." Run the same demand-destruction audit on assets that the rest of this doc runs on goods, and the answer as of July 2026 is stark: almost every asset class has already had its demand-destruction moment. The list of what's still holding the top reservoir full is down to the AI-equity complex and supply-constrained housing.

Three patterns in the table:

Every asset market runs its own internal K-shape. "Flight to quality" in art, "rotation to accessible classics" in cars, "ex-Rolex/Patek/AP" in watches, "Northeast holds while the Sun Belt falls" in housing. That is the same trade-down behavior as private label vs. branded groceries, one level up the wealth ladder. Demand destruction doesn't hit an asset class uniformly — it kills the speculative tail and concentrates the remaining bid in the flagship. A 40%-concentrated S&P is exactly that: the flagship tier of the last un-drained market.

The aggregate-masking pattern repeats at every layer. Case-Shiller +0.67% while 77 metros fall is the same statistical structure as payrolls holding while bottom-quintile delinquencies hit crisis levels, and the same as GDP holding with AI capex included (cyclical-20-and-the-ai-capex-mask). Consumer, housing, equities: a concentrated strong segment holds up an average over a broadly declining base.

The drain moved up the liquidity ladder in order. Crypto and watches (2022 onward) → speculative art → collectibles → Sun Belt housing → offices → PE at the cash price → and the line currently sits just below AI equities. Illiquid assets show destruction through volume first (inventory surges, distributions drying up, lots pulled from auction) and price second, so the price data understates how far along the process is.

The implication that feeds back into the scenario weights below: the "asset-price event that drains the top reservoir" isn't waiting on a catalyst across a diversified wealth base. The base is already mostly drained — the wealth effect funding top-tier consumption is now substantially one trade. The system looks more stable than it is (the index holds) while being more fragile than it looks (a single reservoir; correlation goes to one when it breaks).

Caveat: part of this is normalization, not destruction. Watches, crypto, and Sun Belt housing are deflating from genuinely speculative 2021-22 peaks, and a fall back to trend is not a break below it. The distress tell is forced selling, and right now the secondaries discount is the only place it is clearly visible. Worth adding to the regime-check rotation as a standing gauge: secondaries discounts, REIT NAV gap, breadth, and the ETF-flow direction on the AI complex.

(Personal note, recorded because it's a clean instance of the mechanism: the 2025 house trade — selling the old house at the regional peak to fund the upgrade — was executing against exactly this liquidity gradient. Sell out of the still-full regional reservoir, rotate into the upgraded asset, and the gains stay proportional regardless of where the aggregate goes next. The households that will feel the drain are the ones who marked their wealth to the 2022 peak and never transacted.)

Scenarios for how this episode ends

Consistent with the vault's existing scenario stack, four endings. (Weights revised 2026-07-11 after the asset-side audit: B up 25→30, C down 20→15 — "the supply shock fades before the dam breaks" now requires the AI complex specifically to hold through 2027, because it's the only reservoir left standing.)

A. Stagflation grind (vault base case, ~40%). No fast trigger fires. Supply-shock inflation persists at 3.5-4.5%, the reservoir drains slowly, the bankruptcy cascade runs UK/Japan-style over 2026-28 — failures and quiet layoffs instead of a crash. Policy capitulates late and anchors at 3-4%. Inflation dies of exhaustion, not execution, around 2028.

B. The break (~30%). A discrete trigger — AI-equity correction draining the last full reservoir, a credit event out of private credit/BNPL, or a Hormuz spike past the rationing threshold — fires stages 4-6 within a quarter or two. Breakevens collapse, headline CPI rolls toward 1-2% inside a year, the Fed gets its emergency-cut permission slip. This is the disinflation crash — the 1920/2008 shape. The asset-side audit is why this weight rose: the trigger asset is no longer one of many, it's the only one whose failure transmits to the whole top tier.

C. Immaculate II (~15%). Tariff pass-through completes and fades (the Fed's own timeline: petering out by early 2027), Iran de-escalates, oil retreats, and the bullwhip discounting phase does the disinflating while the labor freeze thaws without breaking. Glide to 2.5-3%. Requires the energy regime to cooperate and the AI complex to hold through 2027 — two independent survivals, which is why the weight came down.

D. The refill (~15%). Financial repression Warsh-style, plus fiscal easing into the midterms: negative real rates and transfers refill the reservoir nominally. Inflation doesn't die — it anchors at 4%+ and the episode extends into the 2027-28 window. Turkey-lite. This is the scenario where the whole premise of this doc ("it has to stop") fails on the timeline that matters, because the dam gets refilled with printed water.

Weight history

Standing tracker — each regime check that revises these weights appends a reading. Note regime-check-july-11-2026 moved C back up to 20 (the oil round-trip is C's mechanism firing) and cut D to 10 (hawkish dots are the opposite of repression).

What this means for the TIPS/STIP position

This section is earned — the position is the stated context.

The uncomfortable structural fact: TIPS are insurance against scenarios A and D, and they are the wrong asset in scenario B — and B is precisely the scenario this doc describes. In a demand-destruction crash, breakevens collapse and nominals beat TIPS. 2008 is the template: 5-year breakevens went to ~zero, and TIPS took a second, non-fundamental hit because the crash was a liquidity event — Lehman's TIPS repo collateral hit the market and real yields spiked ~100-260bp in days even as recession made real yields "should" fall. TIPS are less liquid than nominals exactly when it matters.

STIP vs SCHP within that: STIP (0-5yr) is mostly a realized-CPI accrual instrument with little real-duration risk — it's the better hold in scenario A/D (inflation prints keep accruing, rate rises don't bite much). In scenario B it stops earning (accrual goes to ~zero for a while) but drawdown is modest. SCHP's intermediate real duration gets partially rescued in B if real yields fall — unless the 2008 liquidity dynamic repeats. Neither is a disaster; both underperform nominal duration in B. Two practical implications, consistent with the standing "duration question" from demand-side-audit-may-2026:

1. The rotation signal is the stage transition, not the CPI print. The sequence gives an ordered watch list: pass-through failure in earnings (now) → broadening discounting → the stage-5 tripwire (U-3 >4.6-4.8%, Sahm >0.30, claims >1.95M, or an AI-equity break draining the top tier). When stage 5 fires while breakevens are still above ~2.2%, that's the window where rotating part of the TIPS sleeve into nominal duration is buying crash insurance before it's priced. Breakevens are the market's water gauge, and they're a lagging one. The asset-side audit adds the leading gauges: S&P breadth and AI-complex ETF flows (the last reservoir's water level), secondaries discounts and the REIT NAV gap (where forced selling shows first).
2. Par floors make short TIPS crash-tolerant for a hold-to-maturity buyer. TIPS principal can't redeem below par, so deflation risk on individual short-dated TIPS bought near par is asymmetric. Fund wrappers (STIP/SCHP) dilute but retain some of this. The position survives being wrong about B; it just doesn't profit from it.

The honest summary: the current book is positioned for inflation persisting (A/D, ~55% combined here, consistent with the vault's ~60%), and this research says the mechanism that ends that inflation — the one the whole doc maps — is the scenario the book is least prepared for. That's not a flaw to fix today; it's the reason the duration lever exists and the reason the stage-5 tripwire list is the thing to actually watch.

Open questions

• Where is the non-linear threshold? The energy doc's $4.50/$5.50/$6.50 gas ladder is one attempt; nobody has a good ex-ante model of when inelastic demand snaps. Every historical episode found the threshold only by hitting it.
• Can the AI complex specifically hold long enough for the supply shock to fade first? The asset-side audit narrowed this question: it's no longer "does the top-tier wealth effect hold" in general — every other reservoir is already drained or draining — it's whether one trade survives through 2027. If yes, scenario C steals probability back from B and the crash gets skipped, not survived. This is the live question the masking-layer doc can't yet answer either.
• How much of the asset-side drain is normalization vs. destruction? A fall from a 2021-22 speculative peak back to trend is not a break below trend. The episodes that reorganized economies (US oil post-1979, EU industry post-2022) broke below and stayed. No good ex-ante test for which this is — the forced-selling gauges are the closest proxy.
• Does the failure cascade even show up in US data the way it did in the UK? Answered 2026-07-11 — operationalized in failure-cascade-index, and the first reading says yes, it's already visible: Subchapter V +50-67% YoY. The open question narrowed to breadth (does it climb the firm-size ladder) and transmission (when does it become layoffs).
• Measurement problem: if the bottom-half consumer has already crashed but aggregate stats are carried by the top decile, "demand destruction" may already be further along than any aggregate series shows. The dam gauge reads the average water level of two very different pools. Operationalized in two-economy-gauge; the falsifiable prior (bottom-half real spending negative for 4+ quarters) is a first-run computation.

Sources

• Demand destruction — Wikipedia; Bloomberg QuickTake (2022)
• Weber & Wasner, "Sellers' Inflation, Profits and Conflict" (PERI WP 571)
• AEA JEP retrospective: Cost-Push and Demand-Pull Inflation — Friedman and the "Cruel Dilemma"
• Recession of 1920-21 — Wikipedia; Berkeley Economic Review retrospective
• Federal Reserve History: Recession of 1981-82; The Great Inflation
• Hamilton, "Causes and Consequences of the Oil Shock of 2007-08" (Brookings)
• Columbia CGEP: Anatomy of the European Industrial Gas Demand Drop; Light Metal Age on smelter closures
• Morning Advertiser: UK hospitality insolvencies 2023 +37%; UKHospitality cost survey
• Teikoku Databank bankruptcy reports; Business Standard: Japan corporate bankruptcies at 12-year high; Reuters via Investing.com: bankruptcies seen rising as Iran conflict lifts costs
• Bloomberg: China's Involution Problem; CNBC: China's deflationary slide and price wars
• Logistics Viewpoints: bullwhip effect and the 2022 retail inventory glut
• CNBC: June 2026 jobs report (+57K, 4.2%, LFPR 61.5%); Marketplace: slow wage growth, workers staying put; US Bank: low-hire low-fire labor market
• Equifax: The K-Shaped Economy and Rising Consumer Credit Stress; credit card 90+ delinquencies at 15-year high; Forbes: Forced to Spend More, Affording Less (May 2026)
• Fed Notes: Detecting Tariff Effects on Consumer Prices in Real Time, Part II (Apr 2026); PIIE: The risk of higher US inflation in 2026
• Simon-Kucher US consumer affordability pulse 2026; MarketScreener: consumer companies' pricing stress test from the oil shock
• SF Fed: Inflation Expectations and the Risk of Deflation (2009) — TIPS/breakeven dislocations

Asset-side audit (added 2026-07-11):

• ResiClub: 77 major housing markets with YoY price declines (June 2026); Fortune: decline "no longer just a Sun Belt story" — LA and Dallas tumbling; Cotality US home price insights June 2026
• CoStar: US commercial property prices slip, dragged down by offices; PwC 2026 real estate deals outlook (REIT NAV discount); McKinsey Global Private Markets: real estate (secondaries volume/discounts)
• ScrewDownCrown: the luxury watch correction 2025-26; Automotive History: classic car market 2026 (Hagerty rating at 15-year low); Motorious: where the collector car market is heading in 2026; BofA Private Bank: Art Market Update Spring 2026
• InvestmentNews: Bitcoin's 2026 pullback (-51% from peak); VanEck: what triggered Bitcoin's February 2026 selloff
• Northwestern Mutual: market concentration in the age of AI (top 10 ≈ 40% of S&P); Schwab 2026 mid-year outlook (equities 47% of household financial assets, wealth-effect transmission)
• Cox Automotive: Manheim Used Vehicle Value Index June 2026 (+2.1% YoY, tariff base effects)