Regime-Balanced Allocation with a Debasement Tilt
Derived from the Systemic Money Model — credit-money, r < g erosion, the debasement/Fisher channel.
Spread capital so something carries in every macro regime — with a deliberate but bounded tilt toward scarce real assets.
Headline finding
The framework’s stagflation case understates a 2022-style regime by ~10–14pp. The document modeled −4.3% (conservative); the real 2022 cost −14.9% to −18.3% — because the long-bond hedge broke (TLT −29%), exactly the vulnerability the strategy names but under-sizes.
The framework
The economic thesis the portfolio is built on — stated as falsifiable claims with evidence grades.
A balance-sheet-consistent reformalization of the credit-money economy. Most money is created endogenously by bank lending — but under capital, liquidity and default constraints, not without limit. A debt ratio stays bounded through any of four valves, not by "flight forward" alone. The operative erosion condition is r < g, not a permanently negative real rate. Inflation redistributes along net-nominal positions (the Fisher channel), not as a one-way transfer upward.
Pillars
Falsifiers
- P1 — If the real refinancing rate rises ≥ 2pp over ≥ 8 quarters and corporate insolvencies do NOT rise (with intact bank buffers), the rate→insolvency mechanism is falsified. (2022–25 direction: confirmed.)
- P2 — If measured household inflation incidence is orthogonal to net-nominal position, the Fisher-channel claim fails.
- P4 — If a state with debt > 120% and a primary deficit > 3% sustains i − gₙ > +2pp for ≥ 5 years without consolidation or crisis, the fiscal-dominance risk thesis is falsified.
- P5 — An unremunerated retail CBDC with a low holding cap should cause deposit outflow < 5%; without a cap, measurable credit tightening. Testable only on introduction.
Read the full framework summary9 sections ▾
Balance-sheet frame
Five sectors (households, firms — solvent and zombie —, banks, the central bank, the state) and three accounting identities: the bank balance sheet; money created by lending and destroyed by repayment; and inside money, where every nominal claim is another sector’s liability — so the sectors’ net financial wealth sums to zero and consolidated public debt equals the private sector’s net financial wealth. Pure accounting, no behavioural assumptions yet.
Constrained endogenous money
Most money is created endogenously when banks lend (balance-sheet extension), but realized credit is the minimum of demand and supply — bounded by capital regulation, refinancing cost and expected loss. "Created from nothing at a keystroke" is a correct booking description and a misleading behavioural one: endogenous does not mean unconstrained.
Debt dynamics — four valves, no growth imperative
A non-exploding debt ratio needs only one of four valves — nominal growth (real or inflationary), net repayment, write-offs, or continued credit creation — not a perpetual "flight forward". The old exponential-debt law is the degenerate case of fully capitalised interest; it describes the zombie segment, not total debt. And a stationary economy with positive rates stays stock-flow-consistent as long as bank net interest is spent rather than hoarded — the "missing interest money" worry confuses a stock with a flow.
Zombie dynamics
A bank evergreens a bad loan when its capital buffer is thin, refinancing is cheap and forbearance is tolerated, so the zombie share falls as rates rise. That yields a testable prediction: a sharp rate rise makes evergreening expensive and insolvencies climb — as seen in 2022–25. "Cleansing" is state-dependently damped, not switched off.
Distribution — decomposition, not zero-sum
An unexpected price-level jump revalues nominal positions; that revaluation core sums to zero by identity (the Cantillon advantage is a subset of it). But the total welfare effect adds a non-zero aggregate output term, so the whole is not zero-sum. Incidence runs along net-nominal position and real-asset share — no blanket upward transfer: the asset-price channel favours owners, the employment channel the low-income, the Fisher channel nominal debtors.
Fiscal regime — r < g, not r < 0
The operative erosion condition for the public debt ratio is r below g, not a negative real rate. Three regimes are told apart — monetary dominance, fiscal dominance, financial repression — by the persistence of the nominal-rate-minus-nominal-growth gap, the fiscal reaction function, and creditor structure. For the USA in mid-2026 the real short rate is mildly positive while r is barely below g: weak erosion arithmetic under monetary dominance, with fiscal dominance a risk scenario rather than a finding.
US institutional state — conditional
Only the GENIUS Act (payment stablecoins) is law. The Strategic Bitcoin Reserve rests on a revocable executive order, and the Bitcoin-reserve bill was only introduced. A retail CBDC would substitute bank deposits (disintermediation), not nationalise credit creation. So institutional claims must be stated as conditionals — dated, and revisited when the legal status changes.
Why this model is preferred
Against the original version (internally inconsistent, and refuted on a "permanently negative real rate") and a New-Keynesian representative-agent benchmark (consistent but thin on balance-sheet and distribution channels), this revision earns its keep: wider explanatory reach with fewer assumptions and higher falsifiability. The price is deliberate — weaker but true and testable claims.
Limits
It is a closed-economy model — the dollar’s reserve status widens US fiscal space asymmetrically and does not transfer one-to-one to other states. The evergreening calculus is plausibilised, not micro-estimated; incidence is household-heterogeneous; expectations are exogenous. The earlier normative framing of "money as a domination medium" was dropped because it imported its conclusion into its premise.
From framework to portfolio
If erosion is real but not guaranteed-dovish — the new Fed chair (Warsh) is a balance-sheet hawk, and the Bitcoin reserve rests on a revocable order — then betting everything on permanent debasement risks ruin if the regime flips. So the model’s own falsifiers become portfolio construction: give every macro regime a defender, tilt toward scarce real assets deliberately but boundedly, and replace permanence claims with pre-defined de-risking rules.
The portfolio
Instead of a single-thesis bet on permanently-dovish fiscal dominance (whose risk is ruin if the thesis flips), capital is spread so something carries in every macro regime, with a bounded overweight of real/scarce assets. Three risk tiers, no leverage. Pre-defined rebalancing and de-risking rules replace permanence claims.
| Building block | Role / regime defended | Conservative | Balanced | Offensive |
|---|---|---|---|---|
| World equities (core) | Growth — broad; defuses the US concentration | 20% | 28% | 30% |
| US / Tech tilt | "System toll" as a tilt, not the main bet | 8% | 12% | 20% |
| Gold (physically backed) | Crisis + inflation hedge — the real "sponge" | 15% | 15% | 15% |
| Bitcoin | Asymmetric option — drawdown-bounded | 3% | 7% | 15% |
| Broad commodities | Stagflation diversifier | 5% | 5% | 5% |
| Long Treasuries | The missing deflation / QT hedge | 22% | 18% | 8% |
| Inflation linkers (TIPS) | Protection against unexpected inflation | 12% | 8% | 5% |
| Cash / money market (USD) | Dry powder + rebalancing reservoir | 15% | 7% | 2% |
How the critical sizes were chosen
- Bitcoin is sized by drawdown survivability, not conviction — it has lost 75–85% several times, hence 3–15% rather than 35%. Even an isolated 80% crash stays bearable (−2.4 / −5.6 / −12.0 portfolio pp).
- Long Treasuries insure the strategy’s OWN falsifier (deflation / QT). Caveat: in the 2022 rate shock long bonds failed — so linkers + cash are added, which carry precisely then. The combination, not any single instrument, is the hedge.
- Gold takes the role the original thesis assigned to Bitcoin: it currently stays stable while BTC trades like a risk asset.
The stress evidence
Regime-switching SV factor model — 50 synthetic paths × 6 profiles, 252-day horizon, plus 6 real historical episodes in both rebalancing conventions. Descriptive: assumptions stated, not a forecast.
Tiers — real backtest
| Tier | Return p.a. | Vol p.a. | Sharpe | Max DD | Worst regime |
|---|---|---|---|---|---|
| Conservative Survives every regime at the lowest drawdown | +8.8% | 7.9% | 1.11 | −20.9% | Rate shock −15% (resilient) |
| Balanced More growth, still bond-anchored | +13.1% | 10.7% | 1.20 | −25.4% | Rate shock −19% (moderate) |
| Offensive Real-asset tilt to the fore | +20.3% | 15.9% | 1.24 | −31.9% | Rate shock −24% (moderate) |
Real backtest over the common window (2014–2026). Worst regime = synthetic profile with the deepest median drawdown.
Synthetic regimes — drawdown depth
Cells: median max-drawdown across n=50 synthetic paths. p5 = 5th-percentile 1-year return. Colour ∝ drawdown depth.
The hedge that breaks
Long bonds rise in a flight-to-safety deflation crash (COVID +14%, GFC +25%) but fall WITH equities in a rate shock (2022 −29%). That break is the documented limit of the hedge — which is why linkers + cash are added to carry precisely then. The combination, not any single instrument, is the hedge.
Real historical episodes
| Episode | Window | Conservative | Balanced | Offensive | Single-thesis |
|---|---|---|---|---|---|
| COVID crash Deflation shock; bonds insured (TLT +14%). | Feb–Mar 2020 | −8.7% | −14.5% | −21.4% | −26.0% |
| 2022 rate shock The hedge broke — TLT −29%. The headline finding. | full-year 2022 | −14.9% | −18.7% | −23.7% | −39.0% |
| 2018 Q4 selloff | Sep–Dec 2018 | −5.6% | −9.6% | −15.2% | −23.6% |
| Crypto winter Concentration test — single-thesis −42% vs −25.6%. | Nov 2021–Nov 2022 | −16.3% | −20.3% | −25.6% | −42.0% |
| Current regime Diversified tiers carry the recent BTC drawdown without loss. | Oct 2025–Jun 2026 | +5.3% | +4.6% | +2.0% | −8.4% |
| GFC (proxy) Proxy: no BTC, VT→SPY. Bonds insured strongly (TLT +25%). | Oct 2007–Mar 2009 | −6.6% | −14.3% | −22.0% | −21.2% |
Buy-and-hold end return per episode. Single-thesis = Single-thesis benchmark (50% Tech / 35% BTC / 15% Gold).
Document-claim verdicts
| # | Document claim | Verdict | Conf. | Note |
|---|---|---|---|---|
| B1 | Volatility per annum (per tier) | Supported | High | Δ 0.0 / 0.7 / 2.1pp vs the document. |
| B2 | Risk-off / deflation scenario | Supported as tail↳ Overstated | High | Matches the synthetic 5% tail; real single crashes were milder — a pessimistic tail, not a point forecast. |
| B3 | Stagflation scenario | Not supported↳ Understated | High | Real 2022 was ~10–14pp worse than the document. Mechanism correct, magnitude under-stated. |
| B4 | Debasement-boom scenario | Not testable | Low | No boom profile in the engine; a post-hoc window confirms direction/ranking only. |
| B5 | Single-thesis concentration loss | Supported | High | +16.3pp deeper median drawdown; real crypto-winter −42% vs −25.6%. |
| B6 | Bonds insure deflation / break in a rate shock | Supported | High | TLT solo +14% / +25% in deflation, −29% in the rate shock. |
| B7* | Isolated BTC −80% crash cost | Descriptively plausible | High | Arithmetic check exact; real crypto-winter −77% anchor. |
* added post-hoc — not part of the pre-registered claim set.
Caveats
- Single-factor model (mean R² 0.53) — sub-clusters captured only via the mean loading.
- Tail bands indicative: ~11 independent 252-day blocks (effective sample size ≈ 11).
- GFC is a proxy only (no BTC; VT→SPY, PDBC→DBC).
- Daily constant-mix ≠ the document’s threshold rebalancing — the true strategy lies between constant-mix and buy-and-hold.
- No EUR/USD dimension — all figures in USD.
- Monte-Carlo spread ≤ 3pp for the three tiers, but 5.8pp for the BTC-heavy single-thesis benchmark.
Found this useful?
I run bespoke stress analyses and build portfolio frameworks like this one on request.
Commission your own →Descriptive stress-test case study — not investment advice, and not a suitability or recommendation statement. The framework is a falsifiable hypothesis; the portfolio is an illustrative, rule-based framework for independent decisions. Stress results are model-dependent and not a forecast of future market behaviour.