Published 30 July 2026. Two published series — the gold price forecast and the recession probability — were materially wrong, in the same direction, for the same underlying reason: each had a hard floor or a rising term that made the forecast climb whether or not the data agreed. Both are corrected below. Nothing has been quietly edited. The old numbers are stated next to the new ones.
What we published: gold rising, on a “fragile floor at $4,682”, from a starting assumption of $4,800 an ounce. What actually happened: gold fell to $4,012.70 in mid-July and trades at $4,167 today. CONFIRMED
Why it failed, in plain language. The model contained a price floor. A floor is an instruction that says “whatever else you calculate, gold cannot be worth less than this.” Ours was set at $4,551 an ounce ANALYST — which was $384 above the price gold was actually trading at CONFIRMED. Because the floor sat above reality, it was hit every single week of the simulation. Once that happens the floor stops being a safety net and becomes the forecast: whatever the real inputs did, the answer got lifted back up to the floor. The forecast was reporting our assumption back to us. ANALYST
The starting figure made it worse. We anchored on $4,800, which was gold's price in January 2026, before the fall — $633 above the real price. So the model began each week from a number that no longer existed, and was then prevented from coming down by the floor. Two errors pushing the same way. ANALYST
There is a second, sharper irony we should own. While our floor said gold could not go below $4,551, the average of what professional forecasters expected was roughly $400 below our floor. We were not merely mis-set by a little. We had excluded the consensus outcome as impossible. ANALYST
The replacement. The new path starts from the price gold actually trades at, $4,167, and has no floor above the current price at all. It is driven by two things we can measure weekly: the inflation-adjusted interest rate on ten-year US government debt (2.41% today) and the value of the dollar against other currencies (99.90 today). Central-bank buying is included but tapers off rather than rising forever. The result is a range, not a line: roughly $3,660 to $5,170 an ounce by the end of the projection, centred near $4,353. Sideways to slightly higher — no rescue rally, no collapse. ANALYST
And the most important reversal: our old model made an oil-driven inflation shock push gold up. The measured relationship says the opposite. Higher fuel costs feed inflation, inflation makes the US central bank less likely to cut interest rates, and higher rates are bad for gold. The event we said would make gold rise should, on the evidence, make it fall. ANALYST
What we published: a recession probability around 45%, rising to 100% in later weeks. What the evidence says: the US Federal Reserve's own recession gauge — the Chauvet–Piger measure — reads 0.54%. The New York Federal Reserve's interest-rate-based model reads 16.1%. Betting markets are at 10–11.5% for this year and 34% for next. SOURCED
Why it failed. Three reasons, all ours. First, the model had a term that increased with every week that passed — so simply waiting made a recession look more likely, which is not how economies work. Second, it reached 100%, and a forecast of a certainty one year out is never honest. Third, and worst, it treated expensive oil as straightforwardly bad for America. The United States is now roughly balanced between producing and consuming oil, so a higher price moves money between Americans rather than out of the country. The measured drain was 0.005% of national income, against the 1.27% that the 1970s template we borrowed from assumed. We used a 1974 model on a 2026 economy. ANALYST
We also relied on an indicator that had already failed. We cited “unemployment crossing its three-year average has preceded every recession.” It crossed — and no recession followed for twenty-six straight months. That indicator is retired, not re-dated. ANALYST
The replacement, and what it now says. Eight measurable inputs, each published weekly or monthly: corporate borrowing costs (287 basis points — hundredths of a percent — above government debt, which is tighter than before the crisis), the gap between short and long government interest rates (+0.84 percentage points, and positive throughout), the real-time unemployment rule (0.07 against a trigger of 0.50), new unemployment claims (197,000), broad financial conditions, factory activity (53.3, expanding), forward-looking indicators, and oil. Today that gives about 13% ANALYST, rising to a central estimate near 32% by the end of the projection as the emergency oil buffers run down — with a bad case near 84% and a good case near 11.5%. It can no longer reach 100%, and it can now go down. ANALYST
Which series this supersedes. The recession fan chart and the comparison panels lower down this page
(fanRecess, cmpRecess, scmpRecess) are still produced by the older
engine, and they are superseded by the figures in this box. They are left in place unchanged rather
than silently overwritten, so the two can be compared. Where they disagree with this box, this box is
the current view. The credit-spread and default series that feed off recession risk
(recessionSpread, recessionDefault, recessionPressure,
recessionGoldBid) inherit the old engine's level and are likewise superseded.
ANALYST
Full detail, every source and every falsifier — the two rebuilt sections lower down this page, and the
research files research/gold-model-2026-07-30.md and
research/recession-model-2026-07-30.md. Earlier corrections (V40.6) remain published in the
cascade detail and the model-corrections section; none has been removed.
Published 31 July 2026. Four parameters in the Monte Carlo simulation were wrong. We found them by checking our own numbers against the published research and against our own measured price history. Every old value is stated next to the new one. Nothing has been quietly edited. The net effect of this recalibration is to make our own outlook milder, not worse. Central path about $1 to $2 lower, median about $2 lower, and a wider spread on both sides — a fatter left tail as well as a fatter right tail. We are publishing it anyway. A correction that softens our own headline is worth more than one that inflames it.
[CORRECTED V44] Our demand elasticity was −0.024, taken from JPMorgan's first-three-weeks crisis observation. That value sits below every estimate in the published literature. It is now −0.08, anchored on Baumeister and Hamilton (2024), whose global estimate is −0.119 and whose Rest-of-World figure — which contains China — is −0.139. The literature spans −0.03 to −0.35 and there is no consensus point value, so we carry a −0.03 to −0.20 sensitivity band rather than a single number. [SOURCED]
What this number does not come from. We did not use China's implied −0.765. That figure measures imports, and by the identity our own China deep dive uses, imports equal refinery runs plus the change in tank storage plus domestic production. Using an import elasticity as a demand parameter would drag the tank valve into it and double-count the 1.00 million barrels a day of Chinese destocking we already show as its own bar. On the consumption proxy — refinery runs — the figure is −0.35 to −0.50, and even that is an upper bound. The China evidence establishes only that −0.024 was too low. It rules nothing else in. [SOURCED]
No publishable short-run Chinese crude demand elasticity was found. We searched for one and did not find one. That is a gap in the evidence, and we are logging it as a gap rather than filling it with a number we made up. [GAP — NOT ESTIMATED]
Honest deflation: this parameter barely matters here. The simulation calibrates its price suppression term against the actual Brent print, so raising the elasticity is largely absorbed by that calibration rather than moving the forecast. The isolated effect on our end-of-horizon Brent path is about 40 cents. Even pushing the elasticity all the way to −0.20 moves it under two dollars. We corrected it because it was wrong, not because it was load-bearing, and we are not going to dress a 40-cent parameter up as a breakthrough. [ANALYST]
[CORRECTED V44] We describe the gap between the model's unbuffered equilibrium and the $89.33 Brent print — we published it as $62.47, and after the V44 parameter changes below the same calculation gives $58.80 — as price suppression from buffers and administered pricing. That is incomplete, and the omission runs against our own framing. The gap is a residual. It absorbs every mechanism the supply-side equilibrium does not contain, and it demonstrably contains demand destruction that has already happened. [SOURCED]
China alone cut refinery runs by 2.4 to 2.6 million barrels a day — an upper bound, because the shock cannot be cleanly separated from other causes. That is 37 to 50 percent of our own expected shock of 5.25 to 6.65 million barrels a day, and about 1.4 times the entire deliverable buffer stack we credit for holding the price down. [SOURCED]
On the model's current architecture the elasticity channel can account for at most about 6 percent of the gap, so buffers and administered pricing do remain the majority explanation. But that 6 percent ceiling is an architectural artefact, not a measurement. The model has no channel through which demand destruction that has already occurred can lower the equilibrium price. So it lands in the suppression term by default — not because we measured it there, but because there is nowhere else in the model for it to go. “Residual” and “administered suppression” are not synonyms, and this page has been reading as though they are. [ANALYST]
[CORRECTED V44] Our jump term drew moves of 3 to 9 percent, which is entirely below the 10 percent threshold that defines a jump in the return data we calibrate against. It was extra noise behind a switch, not a jump process. It is now 4 to 13 percent, against measured jump sizes of 12.5 percent at the midpoint and 22.3 percent at the ninetieth percentile at comparable inventory tightness. The jump was also 70 percent biased upward; measured downward jumps at tight inventories are in fact more frequent than upward ones — 3.75 percent a week against 3.57 percent — so the bias is corrected to 48 percent down. [SOURCED]
The chance of a price jump in any week was also raised by half on all nine scenarios. At the tightest quarter of inventories the measured weekly jump hazard is 2.7 times the hazard at middling inventories. We passed through only part of that, deliberately. [SOURCED]
Week-to-week volatility was left unchanged on every scenario. Inventory tightness raises the chance of a jump 2.7 times while barely moving ordinary weekly variation — 6.0 percent against 5.5 percent. Thin order books and refineries running at 97.2 percent of capacity are arguments about discontinuity, not about ordinary dispersion, and we declined to pretend otherwise. [ANALYST]
The result is a wider distribution on both sides and a slightly lower middle. On the isolated
test the tenth percentile falls about $7, the ninetieth rises about $7, the midpoint falls
about $2, and the chance of Brent above $140 in the prolonged-closure case rises from
0.29 percent to 1.98 percent. Measured results against these predictions are recorded in
docs/V40_MODEL_REBUILD.md. [ANALYST]
[CORRECTED V44] Military Escalation is cut 30 percent to 27 percent, with the three points moved to Prolonged Closure (25 to 27) and Negotiated Settlement (6 to 7). Two pieces of evidence. First, the American strategic reserve draw is decelerating — 7.66, then 5.43, then 3.95 million barrels a week — which pushes every exhaustion date further out, and the physically binding floor now lands roughly ten weeks beyond this model's horizon. Second, Saudi Arabia cut its August official selling price by $11 to minus $1.50 against the Oman and Dubai benchmark, the largest cut since 2003. That is a producer discounting barrels into an apparent surplus of sour crude, not a producer rationing scarce ones. [SOURCED]
Both changes make this model's outlook less alarming, not more. The reallocation is deliberately small — three points in total. The evidence supports a direction, not a large reallocation. [ANALYST]
Raising the demand elasticity implies roughly 0.47 million barrels a day more demand destruction inside the model, and we considered whether that should be added to our buffer coverage figure, taking it from 27 to 34 percent up to something like 34 to 43 percent. We decided against it, and we are saying so here so that nobody has to wonder whether we simply overlooked it. Demand destruction is how the shortage clears when nothing else clears it — it is not a cushion. Counting rationing by price as coverage would tell you the shortage is being absorbed when in fact it is being paid for. Coverage counts barrels somebody supplies, not barrels nobody buys. The figure stays at 27 to 34 percent. [ANALYST]
Full derivation, every source, and the list of parameters we deliberately did not touch —
research/montecarlo-recalibration-2026-07-31.md,
research/V44_IMPLEMENTATION_SPEC-2026-07-31.md and
research/china-absorption-deep-2026-07-31.md. Earlier corrections (V40.6 and V41) remain
published exactly where they were. None has been removed.
One knock-on effect, stated rather than buried. The suppression gap is not a hand-set constant — the model re-derives it on every page load from the actual Brent print. Raising the demand elasticity and moving three points of scenario weight both lower the model's unbuffered equilibrium, so the derived gap falls from $62.47 to $58.80 and the suppression coefficient from 73.15 to 68.86. The elasticity change accounts for $2.47 of that and the weight change for $1.20. The observed Brent print the calculation is anchored to — $89.33 — has not changed and cannot: it is an actual, not a model output. [CONFIRMED — measured on this build]
40+ disrupted commodities and 75 cascade channels that make this crisis systemic, not sectoral. V6 adds gold selloff contagion, Kuwait escalation, dollar/EM debt, and sanctions paradox channels.
20+ chains quantified. Red = amplifying (faster than stabilizing)
Comprehensive inventory of every commodity flowing through the Strait. Color-coded by disruption severity.
Record close $5,318.40 on 28 January 2026; the highest price touched during that day was $5,586.20. CONFIRMED Today's price is $4,167, and the lowest it reached was $4,012.70 in mid-July. That is a fall of 21.7% measured close to close, or 25.4% measured from that intraday high. We have also corrected a figure we and others were repeating. The widely quoted “down about 28%” was true at the July low, not today, and the “record was $5,800” claim is not supported by any price print we can find and should be dropped. Using −28% today overstates the dislocation by about 6 percentage points. ANALYST
Anchored on the actual price, $4,167, with no floor above the market. Driven by the inflation-adjusted interest rate on ten-year US government debt (today 2.41%) and the dollar's value against other currencies (today 99.90). Central-bank buying is included but tapers. The geopolitical element is capped at $120 an ounce and mainly widens the range rather than lifting the middle of it — a war makes gold more volatile; it only makes gold more expensive if it changes what the US central bank does. ANALYST
| Week ending | Low end | Central | High end | Real 10-yr rate | Dollar index |
|---|
How to read it. Gold grinds sideways to slightly higher: centre $4,167 → $4,353, about +4.5% over twelve weeks, inside a wide band of $3,664 to $5,172 at the end. The band is wide because gold's actual week-to-week movement is still running near 30% a year. The rise is not a war call — it is mild relief from a central bank that has stopped raising rates but has not started cutting, plus a fading official buying bid. Compare our old path: it centred on $4,583 and its low end was pinned at about $4,167 — which is today's price. In other words the old model said there was almost no chance gold would ever trade below where it already was. The new low end of $3,664 restores a real downside. ANALYST
Every one of these is labelled ANALYST, meaning somebody's projection. None is a measurement. They are shown so you can see where we sit relative to the people who do this for a living.
| Who | Their figure | Horizon and basis | Label |
|---|
On Pierre Lassonde specifically, because his number is the one people quote. His target is
$17,250 an ounce ANALYST, and he restated it on 12 May 2026 — after gold had already fallen, so it is not a
stale pre-crash view. But it is explicitly a three-year-plus target, argued from US government debt and
gold replacing the dollar as the asset of last resort. Lassonde has never claimed a timeline shorter than
three years, and he has no verified public price comment after 12 May 2026. Our earlier model's underlying
error was importing a multi-year target as though it were pressure on this quarter's price.
He is not evidence for a rally in the next twelve weeks, and we no longer treat him as such.
ANALYST
The most useful voice on this list is the one who was right. Jeff Currie turned bearish on
15 May 2026, when gold was $4,561, and called $4,000 before any rebound. Gold printed
$4,012.70. He was right, and his stated reason — that a central bank short of money for energy imports
flips from being a gold buyer to a gold seller — is precisely the channel our model had no term for.
Note that the long-term bulls and the near-term bears are not actually disagreeing; they are answering
questions about different time horizons. Our error was applying long-horizon magnitudes to a
twelve-week question. ANALYST
Stated in advance, so it cannot be revised after the fact. Any one of these means the specification is broken and must be rebuilt, not patched:
Gaps we are declaring rather than filling. We could not obtain current gold lending rates, so the March-era claims about physical shortage are not carried forward. We have monthly, not weekly, exchange-traded-fund flows for July. Lassonde has no verified comment after 12 May 2026. One bank, UBS, appears at both $5,500 and $6,200 in different compilations and we could not resolve it to an original note, so we show the spread. Our own statistical coefficients rest on 49 overlapping windows and explain 77% of the episode — we have deliberately not tuned them to 100%.
These numbers are not comparable to each other and we previously compared them anyway. A contract asking “recession during 2026” and one asking “recession during 2027” can honestly differ by 24 percentage points purely because of the window. We will not publish a single “consensus recession probability” again without stating the window and the settlement rule.
| Source | Reading | Exactly what it measures | Label |
|---|
The direction of travel is down, unambiguously. Every house that has updated has cut: Goldman Sachs from 30% to 15%, RSM from 40% to 30%, and the 2027 betting contract from 41 to 34. Our page was showing about 36% and rising. We have also deleted three gauges rather than re-date them — we could not source any current figure for BCA Research, Deutsche Bank or UBS, and the “BCA 42%” we displayed could not be sourced in any vintage at all. Fewer honest gauges beats more stale ones. ANALYST
| Input | Today | Which way it is pushing | Label |
|---|
Seven of the eight are currently pushing recession risk DOWN. That is the finding, and it is uncomfortable given everything else on this page. What is deliberately absent: no term that grows with the passage of time, no military-escalation term, no sentiment threshold, no inflation flag. Escalation and oil now enter only through the eight measurable things above — which is the correct order of cause and effect. ANALYST
The single biggest error we made. America now produces roughly as much oil as it consumes, so a high oil price largely moves money between Americans rather than out of the country. The measured loss to the rest of the world is 0.005% of national income. The 1970s template we borrowed assumed 1.27% — about 250 times larger. ANALYST
So oil now contributes nothing at all until it is more than 40% above its pre-crisis level, which means above about $101.50 a barrel ANALYST. Brent is $89.33 CONFIRMED today, so oil's contribution to recession risk is currently zero — correctly. Below that threshold, the gain to American producers and refiners plausibly offsets the loss to American consumers. The evidence for that is our own: in 2022 oil rose 29% and no recession followed. SOURCED
Above the threshold the damage accelerates, and it accelerates faster as the buffers empty. The emergency reserve draw, Asian rerouting and record refinery runs are what absorbed this shock. As they run out, the same oil price does more damage. This is the mechanism by which recession risk is allowed to rise — for a stated physical reason, not because a week counter went up. ANALYST
| Week ending | Brent low / central / high | Buffer left | Good case | Central | Bad case |
|---|
How to read it. The central estimate rises from 13.8% to 31.7%, and it earns every point of that rise: the buffer falls from 0.81 to 0.46, which lets central-case Brent reach $112, which finally crosses the oil threshold in the last weeks. The bad case reaches 84.4% ANALYST and stops — with $176 oil, an inverted interest-rate curve, a triggered unemployment rule and factories contracting, it is genuinely alarming and still not a certainty. The good case falls, from 12.9% to 11.5%: if oil settles near $74 and financial conditions keep easing, recession risk should decline below normal. Our old model could not express that at all — it had no way to go down. ANALYST
The oil price range and buffer path above are taken unchanged from the existing simulation, so this is a like-for-like replacement of the recession figures only — not a new oil forecast smuggled in alongside. The end point of 31.7% happens to sit close to the 34% on the 2027 betting contract, which is mild outside corroboration we did not aim for. Compare our old output: 97% at one point and 100% at two others, for both the central and the bad case simultaneously — which is what a saturated model looks like.
The five things that would move this up, in order of how much they would tell us: corporate borrowing costs through 350 basis points; the short-to-long interest-rate gap back below +0.25 points; weekly unemployment claims above 230,000; the unemployment rule above 0.30; broad financial conditions turning restrictive. Right now all five are moving the other way. Next scheduled reviews: 1 August 2026 (employment), 12 August 2026 (inflation), 29 October 2026 (third-quarter output). ANALYST
Gaps we are declaring rather than filling. The overall bank delinquency figures above are sourced and improving, but we could not obtain a current figure for the subprime segment specifically, so we publish no subprime number — the aggregate does not license a claim about the risky tail. Our slope coefficients follow the signs and rough sizes in the standard recession-forecasting literature; we did not estimate them on historical data and we are not claiming we did. Three bank figures (JPMorgan, Morgan Stanley, BCA) have not been refreshed since January or May and are marked stale or removed.
| Metric | Mean | Median | Std Dev | 5th %ile | 25th %ile | 75th %ile | 95th %ile |
|---|
Built on our bottom-up deterministic model + historical crisis variance calibration. Regime-switching mean-reverting jump-diffusion with 7 geopolitical scenarios.
These variables drive the deterministic model. Historical volatility (σ) calibrates the Monte Carlo noise.
| Variable | Pre-Crisis | Current (W4) | Change | σ (Weekly) | Source |
|---|
Deterministic 12-week price model: baseline → Hormuz disruption % → elasticity → cascade amplifiers → time lags
[CORRECTED V44] The model no longer runs on JPMorgan's −0.024. It runs on −0.08, with a −0.03 to −0.20 sensitivity band, anchored on Baumeister and Hamilton (2024). JPMorgan's figure was a first-three-weeks crisis observation and sat below every published estimate. Original wording kept below for the record. With JPM's short-run price elasticity of oil demand at ε = -0.024, a 40% price increase above 12-month highs destroys only 1% of demand. [CORRECTED V44: the model uses −0.08.] The supply gap of 8.5 mbd requires prices to go far higher than supply-side models predict.
| Product | JPM Elasticity | Interpretation | Demand Destroyed at $112 | Demand Destroyed at $150 | Demand Destroyed at $200 | Real-World Evidence (W3-W4) |
|---|
Shows the Brent price needed to destroy enough demand to fully close the supply gap at each week, accounting for time-varying elasticity adaptation and government rationing.
| Week | Clearing Price | Of which Rationing |
|---|
Deterministic (no noise) simulation of each scenario. Bold line = Base Case. Dashed = Military Escalation+.
Tracking model accuracy through Week 4. Green = accurate (within 5%), Yellow = within CI bands, Red = outside CI. Updated weekly as actuals arrive. V12: Gold model recalibrated with Lassonde/Gold Telegraph thesis — margin liquidation + dollar strength + paper-physical divergence now properly weighted. Central bank structural floor added.
| Metric | Week | Actual | V3 Base Pred | Optimistic | Pessimistic | Error | Status |
|---|---|---|---|---|---|---|---|
| Brent | W1 | $95 | $96 | $96 | $96 | -1.1% | ✓ Accurate |
| Brent | W2 | $102 | $108 | $108 | $108 | -5.6% | ~ Within CI |
| Brent | W3 | $109 | $106 | $106 | $106 | +2.8% | ✓ Accurate |
| Brent | W3 peak | $119 | — | — | — | — | ⚡ Ras Laffan spike |
| Gasoline | W1 | $3.50 | $3.18 | $3.18 | $3.18 | +10% | ~ Within CI |
| Gasoline | W2 | $3.65 | $3.48 | $3.48 | $3.48 | +4.9% | ✓ Accurate |
| Gasoline | W3 | $3.85 | $3.85 | $3.85 | $3.85 | 0.0% | ✓ Exact |
| S&P 500 | W1 | -1.2% | -1.2% | -1.2% | -1.2% | 0.0pp | ✓ Exact |
| S&P 500 | W2 | -2.8% | -2.8% | -2.8% | -2.8% | 0.0pp | ✓ Exact |
| S&P 500 | W3 | -3.6% | -3.6% | -3.6% | -3.6% | 0.0pp | ✓ Exact |
| Sentiment | W1 | 56 | 56.6 | 56.6 | 56.6 | -1.1% | ✓ Accurate |
| Sentiment | W2 | 55.5 | 55.8 | 55.8 | 55.8 | -0.5% | ✓ Accurate |
| Gold | W1 | $5,050 | $5,350 | $5,350 | $5,350 | -5.6% | ~ Within CI |
| Gold | W2 | $5,100 | $5,200 | $5,200 | $5,200 | -1.9% | ✓ Accurate |
| Gold | W3 | $5,025 | $5,025 | $5,025 | $5,025 | 0.0% | ✓ Exact |
| Brent | W4 | $112.57 | $118 | $108 | $128 | -4.6% | ✓ Accurate — surged +4.2% Fri on Israel nuclear strike. WTI briefly $100. Gas $3.98 (+55% from pre-crisis). |
| Gold | W4 | $4,430 | $5,200 | $5,100 | $5,400 | -13.3% | ⚠ V12 recalibrated — margin liquidation + dollar strength dominated safe-haven. Paper-physical divergence now modeled. |
| Gasoline | W4 | $3.98 | $4.05 | $3.80 | $4.30 | -1.7% | ✓ Accurate — AAA national avg $3.98/gal |
| S&P 500 | W4 | -6.5% | -6.2% | -8.0% | -3.5% | -4.8% | ✓ Accurate — Dow entered correction (-10%), Nasdaq -13% off peak. VIX 27-31. |
| Sentiment | W4 | 53.3 | 52 | 49 | 55 | +2.5% | ✓ Accurate — UMich final March 53.3. Near record lows. |
Daily data from IDF Spokesperson, UAE Ministry of Defense, Alma Research Center Week 3 Assessment, CENTCOM, ISW, Al Jazeera ACLED. Confirmed OSINT reporting.
How the Hormuz closure propagates through 6 downstream industry sectors — Monte Carlo fan charts driven by sector-specific cascade functions with calibrated elasticities and time lags.
Country-level exposure by sector. Color intensity reflects worst-case severity. Hover for details; toggle sectors to filter.
Median (p50) stress index for each sector over 24 weeks. Background bands show severity thresholds. Hover legend items to highlight individual sectors.
Cross-commodity correlations from 2,000 Monte Carlo simulations (W24 endpoints) with independent sector-specific shocks. 10×10 matrix shows oil derivatives cluster vs independent LNG/sulfur/aluminum paths. Negative correlations indicate inverse relationships (e.g., helium shortage → fab cuts).
How the Hormuz closure cascades through real estate via 8 transmission channels — mortgage rates, construction costs, cap rate expansion, CRE maturity wall stress, insurance repricing, and homebuilder equity collapse. Weighted 80% CRE/Multifamily, 20% Residential. Historical calibration: 1973 & 1979 oil shocks as envelope bounds.
Eight primary channels through which the Hormuz closure propagates into real estate markets. Color intensity reflects current severity.
$1.35T in CRE loans + $900B in private credit maturing through 2027. Pre-crisis cap rates of 6.2% met rates at 5.5-6.0%; now borrowers face 7-8%+ refi rates. The gap between origination rates and current rates creates negative leverage — NOI no longer covers debt service at crisis rates. Estimated refinancing gap widens from $180B to $400B+ under Hormuz stress scenarios.
Individual homebuilder stock performance (rebased to 100 = pre-crisis). KB Home most exposed (guidance cut 1,000 units). D.R. Horton largest by volume. Lennar most diversified. PulteGroup facing rising cancellations. XHB ETF tracks the sector composite.
Energy infrastructure risk → insurer loss model updates → premium increases across property types. Coastal and industrial properties face 25-60% premium increases. The insurance cost channel compounds with energy OpEx to compress NOI by 8-15%, directly expanding cap rates by 30-80bps.
| Property Type | Pre-Crisis Insurance (% NOI) | Crisis Premium Increase | Post-Crisis (% NOI) | NOI Impact |
|---|---|---|---|---|
| 🏖️ Coastal Multifamily | 8–12% | +25% | 10–15% | -2–3% NOI |
| 🏢 Inland Multifamily | 5–7% | +12% | 5.6–7.8% | -0.6–0.8% NOI |
| 🏭 Industrial/Warehouse | 3–5% | +18% | 3.5–5.9% | -0.5–0.9% NOI |
| 🏛️ Office (CBD) | 4–6% | +10% | 4.4–6.6% | -0.4–0.6% NOI |
| 🛒 Retail (Strip) | 5–8% | +15% | 5.75–9.2% | -0.75–1.2% NOI |
Median stress trajectory across 6 RE sub-sectors over 24 weeks. Background bands show severity thresholds.
Estimated severity scores (0-100) by metro area. Gateway cities face higher energy costs but deeper capital markets; secondary/tertiary markets face construction cost and migration disruption. Cap rates are pre-crisis baselines. Scores derived from Monte Carlo median paths at W12.
A real-time economic cascade simulation combining geopolitical scenario analysis with quantitative Monte Carlo methods to project how a Strait of Hormuz closure propagates through the global economy.
Purpose: Real-time economic cascade simulation for the Strait of Hormuz closure.
Approach: Monte Carlo simulation with scenario weighting, bank forecast overlays, and weekly actuals tracking.
Scope: 35-week horizon (W1-W23 actuals + W24-W35 forward projection), 8 weighted scenarios, 40+ trade goods, 35+ cascade channels.
ACTUALS arrays updated weekly from market data — Brent crude, WTI, gasoline retail, gold spot, S&P 500, University of Michigan consumer sentiment, CPI, credit spreads, and more. Each variable is pinned through the current week, with projections diverging from that anchor point.
8 scenarios from Quick Resolution to Tail Risk, each parametrized with supply restoration curves, military escalation factors, and volatility multipliers. Scenario weights are derived from geopolitical assessment and updated as events unfold.
10,000 path simulation using stochastic differential equations with mean reversion, jump diffusion, and scenario-weighted parameters. Each path samples a scenario proportional to its weight, then evolves commodity prices, financial indices, and macro variables through correlated random walks with fat-tailed innovations.
35+ transmission channels mapping oil price shocks through gasoline retail, CPI components, GDP growth, consumer sentiment, financial markets, private credit spreads, and trade goods. Each channel operates with calibrated elasticities and time lags derived from historical supply disruptions (1973, 1979, 1990, 2022).
Goldman Sachs, JPMorgan, Morgan Stanley, Barclays, Capital Economics, Oxford Economics, and Robin Brooks elasticity model. Institutional forecasts provide benchmark comparisons and serve as calibration anchors for the Monte Carlo simulation.
[CORRECTED V44] the aggregate figure is now −0.08 (band −0.03 to −0.20), not −0.024. The product-level figures below are JPMorgan's crisis-week-1-3 observations, not structural elasticities, and they are display-only — none of them enters the price equation. For scale, Hamilton's 2009 meta-analysis puts short-run gasoline demand elasticity at −0.25 to −0.34, against the −0.01 shown here. Original wording kept for the record. JPMorgan elasticity framework (ε = −0.024 aggregate) with product-level breakdown: naphtha (ε = −0.09), jet fuel (ε = −0.06), gasoline (ε = −0.01). Models how sustained high prices trigger demand substitution and conservation, creating a natural ceiling on commodity price paths.
Bloomberg, Reuters, FRED, EIA, AAA
Goldman Sachs Commodities Research, JPMorgan, Morgan Stanley
ISW, OSINT sources, Reuters
Caldara et al., Wolfram/Johnson/Rachel oil elasticity models
OECD, BLS, University of Michigan Consumer Sentiment
This model uniquely combines real-time geopolitical scenario analysis with quantitative Monte Carlo simulation, allowing users to see how different conflict resolution paths cascade through the global economy. Unlike traditional economic models that treat supply shocks as exogenous, this model endogenizes the feedback loops between oil prices, financial markets, consumer behavior, and geopolitical decisions.
The result is a living analytical instrument that updates weekly with real market data, provides transparent prediction tracking, and gives analysts and investors a structured framework for reasoning about geopolitical risk.
Managing Partner, Brainworks Ventures
The best founders are building companies where AI isn't a feature — it's the foundation. Brainworks backs teams that are AI-native from day one, creating products and services that couldn't exist without artificial intelligence at their core.
Brainworks doesn't just invest — it builds. The Hormuz Economic Impact Model is an example of Brainworks' approach: using AI to create sophisticated analytical tools that would traditionally require a team of economists and developers. This model was conceived, architected, and iteratively refined using AI-native development methods, demonstrating the potential of human + AI collaboration in complex analytical domains.
Combined view: 40+ disrupted commodities and 22 cascade chains (Data), plus the 12-week bottom-up deterministic price model.
State-level Monte Carlo simulation using the national model's oil/gas/gold/S&P 500 distributions as inputs, cascaded through state-specific economic structures. 10,000 paths per state.
$3.9T GDP · 39M population · 5th largest economy globally · $297.9B state budget
Fed should consider HIKE. Oil pushes inflation to 3.5% by summer.
Hold at 3.5-3.75%. Maximum flexibility. Pre-war: 2 cuts; now: 1 or zero.
Soft labor market (−92K jobs) = more dovish. Demand already weak unlike 2022.
| Crisis | Supply Disrupted | % Global | Price Impact | GDP Impact |
|---|---|---|---|---|
| 1973 Arab Embargo | ~4M bpd | ~7% | +300% | US recession |
| 1979 Iranian Revolution | ~5M bpd | ~7% | +165% | US recession |
| 1990 Gulf War | ~4.3M bpd | ~6% | +50% | US recession |
| 2022 Russia-Ukraine | ~1-2M bpd | ~1-2% | +40% | No recession |
| 🔴 2026 Hormuz | 10-20M bpd | ~20% | +47% so far | TBD |
95+ citations across 85+ distinct sources. Full academic-style citations. Organized by category.
Total: 345+ citations across 320+ distinct sources. All URLs verified as of May 1, 2026. V38 model: 118 cascade channels. Six OSINT batches integrated + 4 research sweeps + W9 intelligence update. Updated April 16, 2026. Brent $94.89 • WTI $95.05 (INVERSION CONTINUES -$0.16 — compression thesis validated) • Gold $4,818 (safe-haven bid intensifying) • Sentiment 47.4 near ATL • S&P +2% YTD near ATH (FRAGILITY SIGNAL — divergence from fundamentals) • Gas $4.10 • 🇨🇳 CHINA DIMENSION: MANPAD transfers to Iran (CNN Apr 11), Chinese tankers defying blockade, Trump 50% tariff threat, China energy squeeze strategy • US naval blockade Iranian ports (Apr 13) • IMF WEO: "Global Economy in Shadow of War" — growth -0.2pp, inflation 4.4%, severe: -1.9pp EM • IEA: "Largest supply disruption in history" (10.1 mb/d to 97 mb/d) • US crude draw 9.13M bbl (vs +154K expected — physical stress reaches US) • Ceasefire expires Apr 22 • Mine clearance minimum 51 days (May 29 earliest). Scenario weights: Mil Esc 48%, Asym War 20%, Econ Crisis 12%, Prolonged 12%, Tail 5%, Quick 1%, Neg 2%, Base 0%. Monte Carlo parameters calibrated to Baumeister & Hamilton (2018), Kilian (2009), and Hamilton (2003) specifications.
V35 updates (April 10, 2026 — Day 41): (1) W7 ACTUALS: Ceasefire week. Brent flash crash -15% to $92 on April 8 deal announcement, rebounded to ~$96. WTI ~$97 — Brent-WTI spread INVERTED (-$1, WTI premium) CONFIRMING SPREAD_MODEL.MD THESIS. Gold $4,660 (fragile floor, CB liquidations vs. safe-haven). S&P +2.5% (April 8 ceasefire rally). March official CPI released April 10: 3.3% (jumped from 2.4% Feb — energy/tariff pass-through faster than model assumed). Goldman cut Q2 Brent to $90, Q4 to $80. (2) CEASEFIRE IS NOT A REOPENING: Iran limiting to 15 ships/day (vs. 100+ pre-crisis) + $1-2M toll per vessel. ADNOC CEO: "The Strait of Hormuz is NOT open." (3) SAUDI EAST-WEST PIPELINE STRUCK (April 8, hours after ceasefire): Iran hit pumping station, cutting 700K bpd. With Habshan-Fujairah bypass already destroyed (V29), ALL Gulf bypass routes now damaged/eliminated. Kpler total Gulf shutdown: 13 million bpd. (4) ISLAMABAD TALKS DEBUNKED: Iran's Tasnim, Mehr, and Fars news agencies categorically denied any Iranian delegation arrived in Pakistan. DFRAC fact-check confirmed false. (5) TWO MEUs DEPLOYING: USS Tripoli + USS Boxer en route. Total US forces: ~55,000 — largest ME buildup since 2003. (6) 5 NEW CASCADE CHANNELS (#100-#104): Saudi bypass pipeline attack (#100), Iranian Tollgate control (#101), Diplomatic Fog disinformation pattern (#102), Two MEUs deploying (#103), Iran "new phase" Hormuz management claim (#104). Now 104 cascade channels. 320+ sources. (7) SCENARIO WEIGHTS REVISED: Mil Esc 60%→38%, Asym War 13%→20%, Econ Crisis 11%→16%, Prolonged 11%→16%, Quick 0%→1%, Neg 0%→3%. (8) PARAMETER TUNING: Gold CB liquidation pressure coefficient added; CPI energy pass-through lag reduced to 3wk when gasoline >$4.00; Supply restoration curves extended; Recession model: CPI >3.5% stagflation trigger added (+8pp). (9) SPREAD MODEL UPDATE: Brent-WTI spread inverted to -$1 in W7. SPREAD_MODEL.md updated. (10) GOLDMAN POST-CEASEFIRE OVERLAY: GS Q2 $90/WTI $87, Q4 $80/WTI $75 (Apr 9 ceasefire revision).
V36 updates (April 12, 2026 — Day 43): (1) W8 ACTUALS: Brent $94.25 (Invezz Apr 12 AM; TradingEcon $95.20 Apr 10) | WTI $96.57 | Spread -$2.10 (INVERSION DEEPENING — SPREAD_MODEL.MD fully validated) | Gold $4,752 (recovering — safe-haven bid returns as ceasefire collapses) | Sentiment 47.6 ALL-TIME RECORD LOW (UMich preliminary Apr; 98% pre-ceasefire; Bloomberg/Axios Apr 10) | S&P -0.5% (ceasefire rally — correction imminent) | HY spread 560bps | Default rate 5.4% | GDP 0.3% (approaching stall). (2) TALKS COLLAPSE (Apr 12): US-Iran Islamabad talks ended without deal (Invezz Apr 12). Ceasefire expires ~Apr 22. US Navy mine clearance in strait = Iran claims ceasefire violation. Escalation path resumes. (3) STRUCTURAL MINE BARRIER: NYT Apr 10 — Iran cannot locate mines it laid (random deployment, no records). Mine clearance: 3-9 months. Physical closure cannot end even under ceasefire. Insurance war-risk zone permanent until clearance complete. (4) BITCOIN $1/BARREL TOLL: FT/ISW/CoinDesk Apr 8-9. Vessels email manifest → receive corridor → pay crypto (BTC/stablecoins) in seconds. $1/bbl × VLCCs 2M bbls = ~$2M/transit. Iran slow-walking approvals as friction. $20M/day embedded levy on global trade. (5) TURKEY-NATO CRISIS: Erdoğan "Hitler of our time" (formal statement). Tayfun Block-4 ballistic missile unveiled + $3B Roketsan facility (Apr 8). Iranian missiles hit Turkish territory (intercepted). Article 5 paradox. Erdoğan warned Trump of provocations. (6) IRAN SUPREME LEADER: Reuters Apr 11 — severe disfiguring injuries from strikes. Leadership vacuum → IRGC autonomy → unilateral escalation risk. (7) IMF WEO APRIL 14 (2 days away): Georgieva pre-signaled global growth cut from 3.3% baseline. "Economic scars could take a decade to recover." Global oil supply down 13% (Georgieva). World Bank: Middle East growth 1.8%. (8) 5 NEW CASCADE CHANNELS (#105-#109): Turkey-NATO Crisis (#105), Iran Mines Structural Closure (#106), Hormuz Tollgate Bitcoin ($1/bbl) (#107), Islamabad Talks Collapse (#108), Iran Leadership Instability (#109). Now 109 cascade channels. 330+ sources. (9) SCENARIO WEIGHTS REVISED: Mil Esc 38%→42% (talks collapse + mine confrontation risk), Asym War 20%→22% (tollgate institutionalization), Prolonged 16%→14% (structural mine barrier reduces prolonged stalemate probability), Econ Crisis 16%→14% (absorbed into escalation), Neg 3%→2% (talks collapsed), Tail 5% (held). (10) GOLDMAN CEASEFIRE OVERLAY ANNOTATED: "⚠️ Talks collapsed Apr 12 | ⏰ IMF WEO Apr 14 — watch for official downgrade" added to GS ceasefire line. Optimistic path now invalidated.
V36.1 updates (April 12, 2026 — 12:36 PDT — patch): (1) W8 ACTUALS REVISED (EOD): Brent $94.25→$95.5 (end-of-day after talks collapse; up from AM print) | WTI $96.57→$98.0 (futures moved up Apr 12 on talks failure — closing/current level) | Spread -$2.10→-$2.5 (INVERSION PEAK — inflection point, reversal begins W9) | hySpread 560→580bps (BDC redemption caps re-widening credit) | pcStress 68→75 (SYSTEMIC: ALL major BDCs simultaneously capping withdrawals). (2) BRENT PROJECTION STEEPENED (W8→W9): Physical shortage premium added to calcOilEquilibrium for W9+. ACI Europe (airports): jet fuel shortage within 3 weeks (NOW). Shell: shortage "as early as April" — CONFIRMED. Ireland: army mobilized + fuel protests. W9 target $100-105, W10 $108-115, W11-12 $115-120+. kappaUp increased to 0.45 for first 2 weeks post-actuals. (3) WTI-BRENT SPREAD REVERSAL MODELED: Phillip's insight — WTI inversion was because WTI = only "reliable barrel." As US supply squeeze arrives W9 (~Apr 17-20), WTI loses scarcity premium. Spread model updated: faster mean reversion (0.35 speed) toward $3 equilibrium, then +$5-8 by W11-12 as normal Brent premium resumes. Inversion floor removed (allows -$3), recovery to +18 (historical). (4) PRIVATE CREDIT STEEPENED: gateContagion multiplier raised 2→2.5 + W8+ systemic step (+5 + 4/week). bankTightening multiplier 1.2→1.8, W8+ amplifier 1.4×. zombieStress coefficient 0.15→0.20. pcStress will project 78-80 at W9, 85-90 at W10-12. (5) CPI CHART FIX: Current-week (W8) actual dot now drawn AFTER red vertical line (painted on top, larger 7px radius). Value label shown. CPI annotation expanded with BLS data quality note. (6) SENTIMENT CHART FIX: W8=47.6 dot now drawn prominently on top of red line. Annotation: "ALL-TIME RECORD LOW since 1952." y-axis floor lowered. (7) NEW CASCADE #110: Europe Physical Fuel Shortage — W8 Onset (ACTIVE NOW). (8) NEW INTELLIGENCE CARDS: Private Credit Systemic Stress + Europe Physical Fuel Shortage. (9) VERSION: V36→V36.1 | Cascades: 109→110 | Sources: 330+→335+.
V37 updates (April 16, 2026 — Day 48): (1) W9 ACTUALS: Brent $94.89 (down from W8 $95.5 — blockade volatility + ceasefire extension hopes) | WTI $95.05 (WTI > Brent INVERSION DEEPENS: spread -$0.16 — validates compression thesis) | Gold $4,818 (safe-haven bid intensifying, up from $4,752) | S&P +2% YTD near ATH (REMARKABLE DIVERGENCE from fundamentals — fragility signal) | Sentiment 47.4 (near record low, slightly below W8 47.6) | Gas $4.10 | GDP 0.1% (near stall) | Recession 45% (steepening). (2) 🇨🇳 CHINA DIMENSION OPENS — NO LONGER REGIONAL: China MANPAD transfers to Iran (CNN Apr 11) + Chinese tankers (Rich Starry, Shanghai Xuanrun Shipping) defying US blockade + Trump 50% tariff threat + China energy squeeze (Venezuela seized + Iran blocked = 80-90% of Iran's shipped oil to China cut off). Retired NSA deputy director: US-China naval confrontation "very possible." (3) IMF WEO CONFIRMED (Apr 14): Global growth 3.1% (-0.2pp from Jan), inflation UP to 4.4%, SEVERE scenario -1.9pp EM growth. Title: "Global Economy in the Shadow of War." China growth cut to 4.4%. (4) IEA SUPPLY SHOCK RECORD: Global supply plummeted 10.1 mb/d to 97 mb/d in March — "largest disruption in history." Exceeds 1990-91 Gulf War, 1979 Iran Revolution, 1973 Arab Embargo. (5) US CRUDE INVENTORY SHOCK (EIA Apr 16): 9.13M barrel draw (vs expected +154K build). After 7 consecutive weeks of builds. JPMorgan N.America Apr 15-20 propagation timeline CONFIRMED. Physical stress reaches US. (6) S&P DIVERGENCE PARADOX: Interactive Brokers' Sosnick: "oil $30 higher, yields 35bp higher, rate cuts evaporated, sentiment record lows — equities at ATH? That's a no." Momentum overriding fundamentals = fragility, not resilience. (7) 8 NEW CASCADE CHANNELS (#111-#118): China MANPAD transfer (#111), US-China naval confrontation risk (#112), Trump 50% China tariff threat (#113), China energy squeeze strategy (#114), IMF WEO April 2026 CONFIRMED (#115), IEA supply shock record (#116), US crude inventory shock (#117), S&P divergence paradox (#118). Now 118 cascade channels. 340+ sources. (8) SCENARIO WEIGHTS REVISED: Mil Esc 42%→48% (China arms + blockade defiance + 50% tariff = multi-front escalation), Asym War 22%→20% (absorbed into Mil Esc as China dimension expands), Prolonged 14%→12% (events moving toward confrontation, not stalemate), Econ Crisis 14%→12% (absorbed into Mil Esc/Tail Risk), Tail 5% (held — US-China naval confrontation plausible but not yet likely), Neg 2% (indirect talks ongoing via Pakistan), Quick 1%. (9) MODEL PARAMETER UPDATES: Political signal frequency 35%→40% (two actors: Trump + Xi); Political swing range ±7%→±9% (blockade moved WTI 8% in one session); Recession logit intercept -1.1→-1.0 (steeper); China tariff escalation flag +5pp from W9; IMF revision flag +3pp from W9; GDP drag incorporates IMF -0.2pp baseline revision. (10) GOLDMAN CEASEFIRE OVERLAY UPDATED: "⚠️ US blockade Apr 13 | 🇨🇳 China MANPAD intel Apr 11 | 📊 IMF WEO -0.2pp Apr 14". (11) VERSION: V36.1→V37 | Day 43→Day 48 | Cascades: 110→118 | Sources: 335+→340+.
V38 updates (May 1, 2026 — Day 60 — partial patch, items 2/3/5/6 held for 24h soak): (1) W10 ACTUALS APPENDED: Brent $111.16 (↑$16+ from W9 — physical premium re-asserts as US crude draw + ceasefire wobble + China naval defiance bite) | WTI $100.09 (back over $100) | Brent–WTI spread +$11.07 (V37 inversion of -$0.16 FULLY RESOLVED in one week — SPREAD_MODEL.md fully vindicated; Brent races ahead on physical shortage while WTI reflects domestic fundamentals) | Gold $4,610 (–2.6% w/w — real-rate trap binding: nominal yields elevated + Fed not cutting + DXY firm = no oxygen for safe-haven bid even at $111 Brent; liquidation pressure dominates) | Retail gasoline $3.51 (lagging crude — refined-product channel still digesting March/early-April demand destruction; the divergence from Brent is the watch-item). (2) SCENARIO REWEIGHT: Mil Esc 48→50% (China MANPAD transfers + Trump 50% tariff hardening + Chinese tanker blockade defiance = multi-front escalation tightening), Asym War 20→22% (Hormuz tollgate institutionalizing, Bitcoin $1/bbl levy now structural), Neg 2→1% (Islamabad talks formally dead, no diplomatic off-ramp visible), Prolonged 12→9% ("Status Quo / Tense Standoff" trimmed — events resolving toward confrontation, not stalemate). Tail 5% held; Quick 1% / Base 0% / Econ 12% unchanged. Sums to 100%. (3) HELD FOR 24h SOAK (items #2, #3, #5, #6 of yesterday's update proposal): Phillip’s call — only the empirically-grounded W10 actuals patch and scenario reweight applied. Items relating to additional cascade channels (#119/#120), gold model recalibration, GDP/recession steepener, and Fed rate-path overhaul are deferred pending further data. Channels remain 118; sources 340+→345+. (4) VERSION: V37→V38 | Day 48→Day 60 | Cascades: 118 (held) | Sources: 340+→345+.