Published June 2026Reading time ~25 minSeat convened Economics & Statistics RefereeStatus Refereed · accepted
Update 0322 September 2026 — The road around the Strait — Saudi Arabia’s pipeline to the Red Sea — was attacked and shut on 11 September, and the buffers that held this shock down are now visibly finite. The central finding is unaffected.Read the update ↓
Viewing: energy-realist stress case (Berman/Hagens adaptation)
— every figure recomputed at basket exposure +0.12, ≈2–3× the consensus burden. Not published figures.
What this means
Assumption setNot a theme switch: the energy-realist lens recomputes every figure on the page with
fuller energy-cost pass-through (basket exposure +0.12) — a flagged stress case, roughly 2–3×
the consensus burden. An adaptation of Berman/Hagens, not their published figures.
Household Economics · Forecast · June 2026
The Uneven Month
A war-driven oil shock is pushing U.S. inflation higher through late 2026. The national total is one number. What it does to a single mother, a young renter, or a paid-off retiree is a different number entirely.
Scenario-weighted modelSources: BLS · Census · IRS · Federal ReserveInteractive — adjust the scenario below
The short version
A 2026 war in the Persian Gulf has pushed up oil — and with it fuel, food, and shipping — through the year. This report estimates what that costs U.S. households by December, and who feels it most.
The most likely outcome is a status-quo rise of about $23 billion a month by December; the average across scenarios is $25 billion, and a full Monte Carlo puts the 90% range at roughly $4–49 billion — with most of that spread driven by how the war plays out, not by measurement (you set the odds with the sliders). A tougher “coupled” scenario, where a Gulf shock cascades across energy, shipping, food and credit at once, keeps that most-likely figure but fattens the bad tail — a severe-crisis December could top $70 billion, at low but non-trivial odds. The biggest dollar totals come from higher-income households, because they spend more. But the real squeeze lands at the bottom: the lowest-income households lose about 9% of their take-home pay, versus under 2% at the top. Most dollars at the top, most pain at the bottom — and that holds however you set the scenarios.
How to read this. (1) The “expected” figures are a weighted blend of three scenarios — ceasefire, stalemate, escalation — that you control with the sliders; they are not a forecast. (2) A figure “per month” is a December run-rate (the monthly cost once the price rise has fully built up), not the year’s total — the cumulative June–December total is about $110B. (3) “Burden” means the extra money needed to buy the same things as before: a real budget hit, and an upper bound, since people cut back.
Plain-language guide to the terms used here
Regressive — hits people with less money harder, as a share of income. Suits / Kakwani index — a single score (about −0.26) for how regressive the hit is; negative = lands hardest on those with least, roughly like a gasoline tax. Basket factor — how concentrated a household’s spending is in what spiked (food, fuel, utilities); higher for lower incomes. Run-rate vs. cumulative — run-rate = one month’s cost; cumulative = the total added across June–December. Marimekko (Figure 1) — width = number of households, height = cost each, so area = total dollars. Gini — a 0–1 score for how unequal incomes are. “Not-QE, QE” / Treasury liquidity — government borrowing choices that quietly pump money into markets, acting like stimulus. Liquidity-sensitive (crypto) — assets that move with how much money is flowing through markets, more than with oil itself. Floor vs. coupled tail (Figure 9) — the “floor” is the calmer estimate that treats each month as independent; the “coupled tail” allows a crisis where energy, shipping, food and credit go wrong together, which fattens the worst-case end. p95 / p99 — the 95th / 99th percentile: a bad outcome exceeded only about 1 month in 20, or 1 in 100. Copula / tail dependence — the statistical dial for how strongly different shocks move together specifically in the bad tail.
+2.9%
Expected cumulative price rise by Dec (probability-weighted)
$24B
Estimated added household cost per month, nationwide
Median AGI per tax return2 — pre-tax; ~$48.5k after tax
Income is not money in hand. Median household income ($83,730) counts a couple's combined earnings; median AGI per return ($50,400) is the pre-tax tax base. The budget views below use after-tax "available money" as the baseline.
The shock's size hinges on the Strait of Hormuz. Even a ceasefire is slow relief: the Pentagon estimates up to six months to clear mines, and clearing can't begin until the war ends.8 Drag the probabilities — the model reweights and updates every chart, trajectory, and market estimate on the page.
Probability of each path (Jun–Dec 2026)
Subjective weights, not market-implied. Each path's cumulative price impact is fixed from its monthly inflation track.
slow relief · +1.1% by Dec
grinding on · +2.8% by Dec
oil re-spikes · +5.2% by Dec
+2.9%expected rise by Dec
$25Badded cost / month (Dec)
$110BJun–Dec cumulative
$296BDec run-rate ×12
Probabilities normalize to 100%. Default weights reflect early-June 2026 — active ceasefire talks alongside ongoing strikes and regional spillover.
How this is measured — run-rate vs. cumulative
A run-rate is the cost in a single month once the price rise has fully built up; ×12 gives a steady-state annual rate, not the realized 2026 total. Because the shock builds month over month, the actual June–December cumulative (~$110B in the consensus lens) is about 4.45× the December run-rate, not 7×.
Figure 1 · Where the dollars come from
The biggest bills, the biggest backs
In plain termsmost of the national bill comes from the top, because they spend the most — even though each high-income household pays a smaller slice of its own income.
Each block is an income bracket. Width is the number of households; height is the monthly hit per household; so the area is the total dollars that bracket adds to the national bill. The top dominates — biggest budgets, ~21 million households — even at the lowest rate.
Aggregate monthly burden by income bracket
$ billions / month · national
Area = households × per-household monthly hit. Hover any block.
Reading itthe $200k+ block is ~21M households × ~$291 each ≈ $6.1B a month — wide and tall, so it dominates the national total even at the lowest rate.
A national total leans toward the top of the distribution simply because that's where spending — and much of the population — sits. That is the opposite of who feels it most.356
Figure 2 · What the month actually looks like
After the essentials, what's left to give?
In plain termsafter essentials, housing, and debt, this is what’s left — and where the shock bites. For some groups it bites into nothing.
The honest view. Take after-tax money, subtract what doesn't flex — food and utilities, housing, debt — and what remains is the monthly breathing room. The dashed line is the shock. When it lands in thin or negative space, something gives.
Monthly take-home split into where it goes — each block shows its dollar amount. The dashed line is where fixed commitments end; the hatched red bite is the expected December shock eating into what's left. Hover any block.
Tenure totals are exact CE 2024 figures (owner-with-mortgage $104,329/yr, owner-free $70,906, renter $57,108);129 age and household-type budgets use Census median incomes7 with CE category shares,13 and the new under-25 bracket draws on the CE youth cross-tab;26 the young-adult living-situation view uses Pew/Census shares of those living with family (its budget split is estimated).27 Dollar splits within each bar are derived from those sources; treat the shape as solid, the cents as estimates.
How the basket factor works
Lower-income households spend a larger share of their budget on the things that spiked — food, fuel, utilities — so the same headline inflation rate hits them harder. The basket factor scales the price rise by that share (about 1.32 at the bottom vs 0.80 at the top), per the New York Fed’s analysis of the Consumer Expenditure Survey.
Figure 3 · Month by month to December
The slow build of the squeeze
In plain termsthe cost is not a one-time jump; it builds month by month to December.
The shock doesn't arrive all at once — it compounds from June to December. These trajectories track the cumulative monthly hit as it accrues. Both charts are wired to the sliders above: reweight the scenario and the curves bend in real time. The second chart follows whichever lens you pick in Figure 2.
Cumulative monthly hit by income bracket
$ / household / month
$/household/month of added cost, accruing Jun→Dec under the current scenario weighting.
Cumulative monthly hit by demographic group
$ / household / month
Follows the Figure 2 lens. Switch the tabs above to change the groups shown.
Both charts share one vertical scale, so the income and demographic views are directly comparable. Each line's December value is that group's per-household monthly hit at the expected rate — the same number drawn as bar height in Figure 1. The color dots in Figure 2 match these lines.
Figure 4 · The flip
Most dollars at the top, most pain at the bottom
In plain termshigher-income households lose more dollars, but lower-income households lose a bigger slice of what they have — that flip is the unfairness.
Rank brackets by raw dollars lost and the affluent lead. Rank by share of after-tax income and the order inverts — lower-income households spend a far larger slice on the exact things that spiked.
The crossing lines are the regressivity signature of an energy-and-food shock. Collapsed to one signed scalar, the Suits index is −0.26 on pre-tax income (0 = proportional, negative = regressive) — squarely in the band of a gasoline excise tax (−0.25 to −0.35) — and −0.22 on after-tax income; we lead with pre-tax for benchmark comparability and report both, since the difference is purely the income base.31 The burden concentration (0.19) and the ~5.5× bottom/top share ratio need no income-base choice. These indices are invariant to the scenario weights — they measure the shape of the shock, not its size — and because they are computed on grouped brackets they are conservative (the true values are modestly more negative). A parametric bootstrap over the modeled inputs (basket factors, incomes, tax rates, counts) puts the pre-tax Suits at −0.26 (90% CI −0.31 to −0.21) and finds the shock regressive in 100% of 100,000 replicates on both income bases.33
How the index is computed
The Suits index plots each bracket’s cumulative share of income against its cumulative share of the burden: 0 = the burden falls proportionally to income, negative = it falls harder on lower incomes, positive = on higher incomes. Kakwani is the burden’s concentration coefficient minus the income Gini. Both use pre-tax income for comparability with the public-finance benchmark (after-tax shown alongside), computed from the seven income brackets in the Incidence & Sensitivity sheet.
Figure 5 · The debt underneath
The bills were heavy before the shock
In plain termsmany households were already stretched before the shock, especially the lowest earners, whose card balances equal about 85% of a month’s income.
From the Federal Reserve's 2022 Survey of Consumer Finances: who holds which debt, and how heavy credit-card balances are relative to monthly income across the distribution.
Credit-card debt as a share of monthly income, by income decile
% of monthly income
2022 SCF, lowest-income decile to highest. Carry the broad gradient — ~85% of a month's income at the bottom, ~8% at the top — alongside the upper-middle bump noted below.
card balance ÷ monthly income5th–7th deciles — highest share carrying a balance (up to 61%)
The ratio generally falls as income rises — ~85% of a month's income at the bottom to ~8% at the top — but it is not monotonic: the 5th–7th deciles bump up (66 / 49 / 48%) because those upper-middle deciles have the highest share of households carrying a balance (57–61%, versus 28% at the bottom and 26% at the top). Verified against the corrected 2022 SCF analysis.4
What U.S. families owe — 2022 Survey of Consumer Finances
Share holding each debt type, and the median balance among those who hold it.110
Debt type
% holding
Median (holders)
Mean (holders)
Any debt
77.4%
$80,200
$163,800
Credit card
~45%
$2,700
$6,100
Student loans
22%
~$25,000
~$47,000
Mortgage (primary home)
42%
$156,400
—
Auto loan
~35%
—
—
Other installment (incl. BNPL)
18.5%
—
—
Student debt skews toward higher earners in dollars,5 but delinquency skews low: 27% of borrowers under $50k are behind versus 10% of those over $100k.11
Figure 6 · What gives first
The order of sacrifice
In plain termswhen money runs short, households pay bills in a fairly predictable order, and the lowest-priority ones slip first.
When the shock eats the breathing room, households rarely cut at random. Counseling guidance describes a consistent payment hierarchy, and cross-sectional delinquency data are consistent with it — lower-priority unsecured bills tend to fall behind first. This is a description of typical prioritization, not a prediction that this shock will cause a particular order; establishing that would need an event study around prior energy spikes.
1
Food & essential utilities
Survival, and utilities can be shut off within weeks. Paid first, almost always.
2
Housing — rent or mortgage
Eviction (renters, faster) or foreclosure (owners, slower) is the hardest setback to recover from.
3
Transportation — auto loan, insurance, fuel
Usually required to keep earning. Repossession threatens income itself.
4
Insurance & required taxes
Legal exposure and loss of coverage; real but less immediate.
5
Credit-card minimums
Protect the score and revolving access, but the first unsecured bill to slip when cash is short.
6
Student loans (especially federal)
Most deferrable — forbearance and income-driven plans — so the slowest severe consequences. Typically the first to go.
Renter-heavy and single-earner households reach this ladder fastest, because their breathing room was thinnest to begin with.
In plain termshow the shock might ripple into stocks, bonds, and crypto — illustrative directions, not predictions. Don’t trade on them.
An oil-driven inflation shock ripples beyond the kitchen table. The mainstream macro chain: higher inflation keeps the Fed higher-for-longer, which lifts bond yields (pressuring bond prices), squeezes growth equities, and leaves crypto ambiguous. These are illustrative directional sensitivities, scaled to the scenario you set above — not forecasts or advice.
Scenario-linked market sensitivity
% · bps
Weighted column updates live with the sliders. Values are illustrative scenario impacts, not price targets.
Metric
Ceasefire
Status quo
Escalation
Weighted
For market watchers — Treasury-driven liquidity overlay (illustrative)
Treasury-driven liquidity overlay
Howell’s “not-QE, QE”: the Treasury’s financing mix — T-bill share, its cash balance at the Fed (the TGA), and reverse-repo / bank reserves — now injects or drains money-market liquidity like stealth QE, independent of the oil shock.2829 Pick the impulse; the liquidity-sensitive lines re-price.
Treasury / Fed lever
Injecting
Draining
2026 reading
How the selected impulse re-prices the liquidity-sensitive lines from the table above (weighted column):
Liquidity-sensitive line
Oil-shock read
Treasury overlay
Net read
Not investment advice. These directional figures illustrate how an oil-inflation shock tends to transmit to markets; the relationships are contested and regime-dependent. The inflation and yield links are the firmest — oil shocks reliably lift headline CPI16 and tend to push Treasury yields up.17 The equity and crypto responses are regime-dependent, not reliable directional bets: historically the S&P 500 rose in six of seven oil-spike episodes since 1986 (averaging +24% over the following year), with supply-driven shocks like this one skewing more negative and recession the swing factor.18 A stagflationary oil shock can weigh on both stocks and bonds at once.14 Crypto is best read as a liquidity gauge rather than a simple risk asset: this conflict has been characterized as tightening global liquidity through the capital account,24 and with Bitcoin the most liquidity-sensitive major asset, a deepening shock reads as a near-term headwind — while the same refinancing wall and currency-debasement pressures could later flip liquidity, and crypto, to a tailwind.25 The U.S. is a net exporter of petroleum products but a net importer of crude oil, and pump and diesel prices are set globally — so households are not meaningfully shielded by “energy independence.”2023 Yields sat near 4.1–4.3% with the funds rate at 3.5–3.75% in mid-2026.15 Do not trade on this.
In plain termsthe Fed meets five more times before December; what it decides is the biggest wildcard the model cannot price.
The shock builds across the exact window in which the FOMC meets five more times in 2026. Each is a decision point; June, September, and December also publish projections (the “dot plot”). What the Fed does at these dates is the largest swing factor the model itself cannot price — and the two lenses imply different Fed reactions.
FOMC 2026 — June to December
★ meeting includes the Summary of Economic Projections. Schedule: Federal Reserve.
Figure 9 · The number is a distribution
From an envelope to a forecast
In plain termsthe most likely outcome is a status-quo ~$23B a month, and the wide range around it is about how the war goes, not measurement. The second line asks a harder question: if a Gulf shock sets off a chain reaction — energy, shipping, food and credit going wrong at once — the bad end gets fatter. A truly awful month (worse than 99 out of 100) rises from about $55B to $69B, and a month above $60B goes from a 1-in-1,000 rarity to roughly 1-in-50.
A 200,000-draw Monte Carlo replaces the corner range with a predictive distribution — reported as two bands. The floor calibrates volatility on history (independent month-to-month noise); the coupled tail adds what a war shock actually does — a crisis regime, fat tails, and a tail-dependent copula so energy, shipping, food and credit go wrong together, with the realist basket-broadening switched on harder in the tail. The gap between the two bands is the quantified “novel territory.”
Predictive distribution of the monthly aggregate burden
density · $B / month (Dec, national)
Median ~$23B in both bands — coupling does not move the center. The story is the upper tail: the 99th percentile rises from $55B (floor) to $69B (t-copula) to $75B (rotated-Clayton).
Two things the floor distribution already showed. The spread is structural, not statistical — 82% of it is which scenario realizes (geopolitics), only 18% is path noise, so a tighter forecast needs better information on the war, not finer measurement — and it is multimodal, with the mean sitting in a low-density valley between a status-quo mode (~$23B) and an escalation tail.32
What the coupling layer adds is all in the upper tail. The center holds, but the chance of a burden above $60B/month rises from 0.1% to 2.1% (~18× once a cascade is admitted) and the p99 gap is ~$14B.34 One finding worth your scrutiny: the copula’s tail-dependence parameter turns out second-order — a discrete crisis regime already supplies most of the joint-tail co-movement — so the honest levers to argue over are how likely and how severe a crisis is, and how far the basket broadens, not the copula family.
How the two bands are built — and the one nuance for the thesis
Floor: independent Gaussian monthly noise (σ=0.08 / 0.15pp/mo), no crisis regime, consensus basket — the future as a draw from the past, hence a lower bound. Coupled: a Student-t copula across the seven months (ρ=0.35, ν=4; tail-dependence λ=0.18) with fat-tailed Student-t marginals (ν=5), a Bernoulli crisis regime (p=0.12) that multiplies variance ×2.5, and a state-dependent basket-broadening (the realist +0.12 becomes +0.18 in a crisis) so a bad draw is hit on both the rate and the breadth; a rotated-Clayton variant (λU=0.71) gives a directional upper bound. Independently reproduced from the raw draw vectors and on a different seed. The nuance for the thesis: that state-dependent broadening is the only channel that touches incidence, and it attenuates the pre-tax Suits only slightly (−0.26 → −0.25 at +0.12, −0.24 at +0.30) — regressive throughout. The coupling parameters are deliberately assumptions with reported sensitivity, not fitted estimates.
Methods & sources
How this was built
A probability-weighted blend of three scenarios produces an expected cumulative price rise by December (default +2.91%), anchored to realized monthly CPI (March +0.9%, April +0.6%). That rate is adjusted per group by a basket factor derived from food-and-energy budget shares by income — about 18.5% for the lowest quintile versus a ~13.75% aggregate, per the New York Fed's analysis of the Consumer Expenditure Survey19 — then applied to monthly spending and scaled by household counts for the national total. The monthly trajectories compound each scenario's inflation track and reweight live. Measured (sourced) inputs: incomes, AGI, CE spending by quintile / age / tenure, debt holdings, household counts, current rate and yield levels. Modeled (estimated): basket factors, within-bar dollar splits, market sensitivities, and the scenario probabilities themselves.
Reading the numbers like an analyst. The headline monthly figure is a December run-rate; ×12 gives a steady-state annualized rate, not the realized 2026 total. Because the shock builds month over month, the honest Jun–Dec cumulative (in the panel) is far smaller — about $110B in the consensus lens — roughly 4.45× the December run-rate, because the burden builds over the window rather than sitting at its December level all seven months. Figures are nominal and undiscounted (immaterial over seven months). The burden is the cost to hold consumption constant — a real-income hit, and therefore an upper bound, since households substitute and economize. In the energy-realist lens the basket factors carry an extra +0.12 (embedded-energy broadening) on top of higher scenario inflation; that double lever makes it the most aggressive assumption in the model and should be read as a stress case, not a forecast.
Not financial, legal, or tax advice. The triage ordering describes how obligations are typically prioritized in counseling; the right action depends on individual circumstances. Aggregate totals are a central estimate (read as ~$20–30B/month), not a precise figure.
Sources
Sources are grouped by which dataset they power. The shared foundation underlies both lenses; the lens you are not viewing is dimmed — flip the toggle at the top to compare.
Shared foundation — both lenses
U.S. Census Bureau, Income in the United States: 2024 (median $83,730; mean $121,000). census.gov
IRS Statistics of Income / Tax Foundation, Latest Federal Income Tax Data, TY2022 (median AGI ~$50,400/return). taxfoundation.org
U.S. Bureau of Labor Statistics, Consumer Expenditures — 2024 (spending by quintile; housing share). bls.gov/cex
Yu-Ting Chiang & Mick Dueholm, Which U.S. Households Have Credit Card Debt?, Federal Reserve Bank of St. Louis (May 2024; figures corrected July–Sept 2024). Card-balance-to-monthly-income by income decile (1st–10th): 85, 62, 58, 42, 66, 49, 48, 36, 27, 8 percent — verified against the corrected source. The non-monotonic 5th–7th-decile bump is a real feature: those deciles hold the highest share of balance-carriers (28→61% peak at the 7th decile). stlouisfed.org
Federal Reserve Bank of Cleveland, The Evolution of Student Debt 2019–2022. clevelandfed.org
U.S. Census Bureau, household income distribution & counts, 2024. census.gov
U.S. Census Bureau, median household income by age of householder, 2024 (peak $116,800 at 45–54; $56,680 at 65+). census.gov (PDF)
Washington Post / Associated Press, Strait of Hormuz mine-clearing could take six months (Apr 2026). washingtonpost.com
USAFacts / U.S. Census ACS, renter housing cost burden (median rent $1,487; 32.8% of renter income). usafacts.org
Federal Reserve, Changes in U.S. Family Finances from 2019 to 2022 (2022 SCF). federalreserve.gov
Federal Reserve, Survey of Household Economics and Decisionmaking (student-loan delinquency by income). federalreserve.gov/shed
BLS Consumer Expenditure Survey, age of reference person & composition tables, 2023–24 (category shares; single-parent housing 37.3%). bls.gov/cex/tables
CNBC, Federal Reserve rate & Treasury market coverage (May 2026) — funds rate 3.5–3.75%, markets pricing no further 2026 cuts. cnbc.com
Federal Reserve Bank of New York, Liberty Street Economics — food-and-energy budget shares by income quintile (lowest ~18.5% vs ~13.75% aggregate; the 4th quintile sits near the aggregate); cross-checked with BLS experimental CPI by income. newyorkfed.org
U.S. Energy Information Administration, U.S. energy facts — imports and exports — the U.S. remained a net crude oil importer in 2024; crude oil was about 67% of total U.S. energy imports. eia.gov
Darius Dale / 42 Macro, A Global Liquidity Crisis Is Underway… What’s Next? (Mar 2026) — the U.S.–Israel–Iran oil shock has moved beyond a supply disruption into a global liquidity crisis, tightening financial conditions via the capital account and pressuring risk assets and safe havens alike; 42 Macro’s inflation regimes feature elevated cross-asset correlation. 42macro.com
Michael Howell, CrossBorder Capital / Capital Wars — the ~65-month global liquidity cycle peaked in Q3 2025 with a downswing likely into 2027; “Bitcoin is the most liquidity-sensitive asset on the planet”; a ~$40T debt-refinancing wall (2026–2028) absorbs liquidity, though fiscal dominance and currency debasement support crypto over the long run. capitalwars.substack.com
BLS Consumer Expenditure Survey, consumer units with reference person under age 25 (Table 3203, 2021–22): income before taxes ~$47,100, after taxes ~$45,400; the youngest, most renter-heavy cohort, with 1.4 earners and the thinnest cushion. bls.gov/cex/tables
Pew Research Center (2024–25) & U.S. Census ACS 2024 — 57% of 18–24-year-olds (and ~32.5% of 18–34) live with a parent, driven by housing affordability. The personal-budget split for this non-householder group is estimated from young-adult earnings and the housing-buffer / transport-exposure structure. pewresearch.org
Michael Howell, CrossBorder Capital / Capital Wars, “Prospects for a Puzzling 2025” — a “not-QE, QE” (and “not-YCC, YCC”) regime: heavy short-dated T-bill issuance plus reverse-repo (RRP) and Treasury General Account (TGA) drawdowns add net money-market liquidity even during Fed QT, while bill issuance creates coupon scarcity that suppresses long yields. capitalwars.substack.com
Hudson Bay Capital (S. Miran & N. Nordvig), Activist Treasury Issuance and the Tug-of-War over Monetary Policy — tilting issuance toward bills / cutting coupons is “stealth QE” with many of the same liquidity consequences as Fed QE (a quarter’s ~$245B of “missing coupons” scaled to a ~$800B-equivalent in their example). hudsonbaycapital.com
Federal Reserve Bank of New York (Perli, Mar 2026) & BNY, Treasury Road Trip — a TGA drawdown lifts system liquidity (reserves + ON RRP); after the July 2025 debt-limit increase the Treasury rebuilt the TGA to ~$900B via >$600B of net bill issuance, dropping system liquidity to its lowest since QT began and pressuring repo rates — after which the Fed halted QT on Dec 1, 2025. newyorkfed.org · bny.com
Incidence indices computed from the seven-bracket burden and representative incomes (see the Incidence & Sensitivity sheet). Suits = 1 − 2∫(burden-vs-income concentration curve); Kakwani = burden concentration − income Gini, by trapezoidal integration. The pre-tax base is comparable to the public-finance benchmark literature; the after-tax base measures burden as a share of disposable income. Grouped point-masses understate within-population inequality, so both are conservative.
Monte Carlo predictive distribution — companion methodological supplement (research_analysis.py, seed 20260605, N=200,000), independently reproduced here with a different seed. Realized December cumulative rate 90% interval 0.4–5.8% (median 2.8%); monthly aggregate burden 90% interval $3.7–49.3B (median $23.3B). A law-of-total-variance decomposition attributes ~82% of the spread to which scenario realizes and ~18% to within-scenario path noise; weight uncertainty is negligible. About 28% of mass falls outside the deterministic $9.4–44.2B corners (~15% below, ~13% above). Within-scenario noise is Gaussian (persistent σ=0.08pp/mo, transitory σ=0.15pp/mo); this is the historical-volatility floor — a coupled-tail extension is reported alongside it (note 34).
Parametric bootstrap of the incidence indices over modeled inputs (basket factors 6% CV, representative incomes 10% CV, effective tax ±2pp, counts/spend 2.5%/3%; 100,000 replicates), independently reproduced. Pre-tax Suits −0.26 (90% CI −0.31, −0.21), Kakwani −0.25 (−0.29, −0.21); after-tax −0.22 / −0.21; regressive in 100% of replicates on both bases. Captures modeled-input, not survey-sampling, uncertainty; a microdata resample (SCF/CE public-use files) would add survey confidence intervals and is specified but unestimated.
Coupled-tail Monte Carlo — companion specification (coupled_tail_mc.py, seed 20260605, N=200,000), independently reproduced from the raw draw vectors and on a different seed. Retains the floor and adds a tunable coupling layer: a Student-t copula across months (ρ=0.35, νcop=4; tail-dependence λ=0.18), fat-tailed Student-t marginals (ν=5), a Bernoulli crisis regime (p=0.12, variance ×2.5), and a state-dependent basket-broadening (BFADD 0→0.18 in crisis); a rotated-Clayton variant (λU=0.71) gives a directional upper bound. The median is preserved (~$23–24B in every run); coupling reweights the upper tail — aggregate p99 $54.8B (floor) → $68.7B (t) → $75.3B (Clayton); P(burden>$60B) 0.1%→2.1%; p99 “novel-territory” gap ≈$14–20B. Knobs set to the agreed central values (p_crisis=0.12, crisis_mult=2.5, bfadd_crisis=0.18). The tail is driven by crisis probability/severity and breadth; the copula degrees-of-freedom are second-order once a regime is present. The broadening attenuates pre-tax Suits only mildly (−0.26→−0.25 at +0.12, −0.24 at +0.30), never flipping sign.
Consensus / macro lens only
Vanguard, Oil shock complicates central bank outlooks (Apr 2026) — a stagflationary oil shock can weigh on both stocks and bonds, with policy rates likely higher-for-longer. vanguard.com
IMF, Oil Prices and Inflation Dynamics (Choi et al., 2017) & Federal Reserve DSGE model (2024) — a 10% oil rise adds ~0.15–0.4 percentage point to headline CPI on impact. imf.org · federalreserve.gov
European Central Bank working paper, Global financial markets and oil price shocks in real time — 10-yr government bond yields tend to rise on oil price shocks. ecb.europa.eu
Real Investment Advice / Kedia Advisory (2026) — the S&P 500 rose in 6 of 7 oil-spike episodes since 1986 (avg +24% over the following year); the equity response is regime-dependent. realinvestmentadvice.com
Energy-realist lens only — Berman & Hagens
Art Berman, America Has Plenty of Oil—Just Not the Right Kind (artberman.com, May 2026) — the U.S. imported ~6.3 Mb/d of crude and exported ~4.0 Mb/d in 2024–25 (a net crude importer); refineries need imported medium/heavy grades, losing them implies a diesel deficit and margin dislocation; “energy independence” is an accounting artifact. artberman.com
Arthur Berman & Nate Hagens, The Great Simplification — “Shale Oil and the Slurping Sound” and “Peak Oil: The Hedonic Adjustment”: U.S. shale has been high-graded and cannibalized, implying steep declines; roughly 40% of what the U.S. labels “oil” is NGPLs, ethanol, biofuels, and refinery gain. thegreatsimplification.com
Nate Hagens, The Great Simplification — oil underpins transport, agriculture, mining, and trade (“oil is the economy”), so energy shocks transmit more broadly than a narrow cost-share implies; “Wide Boundary News” (May 2026) flags the misleading blurring of crude vs. petroleum-product statistics. thegreatsimplification.com
Updates since publication
What we have learned since this went out
The study above is preserved exactly as published. Nothing in it has
been rewritten, re-run, or quietly corrected — it is a reading of the world taken at a moment,
and its value depends on it staying that way. Everything learned since lives down here, dated,
so you can see what held, what didn’t, and when we knew.
Update 01Scope: June FOMC · oil path · § Five Fed meetings
What was validated
The study’s consensus lens projected that the Fed would hold at 3.50–3.75% and signal
higher-for-longer, with markets pricing no further 2026 cuts. On 17 June the FOMC held at
3.50–3.75% on a unanimous 12–0 vote — a fourth consecutive hold — struck the
easing-bias language from its statement, and removed the 2026 cut from its dot plot. The median
year-end projection rose to 3.8% (from 3.4% in March), and 9 of the 18 submitting
participants penciled in at least one hike; markets began pricing one as early as October. The Committee also raised its 2026 headline PCE inflation
projection to 3.6% (core 3.3%), up from 2.7% in March, attributing part of the rise to energy
supply shocks — and the June minutes name the closure of the Strait of Hormuz explicitly.
The direction of the call was right. If anything the Committee moved more hawkish than the
consensus lens described.
What has not resolved
Which scenario the world is actually on. Within days of publication a U.S.–Iran framework
agreement pulled Brent from a wartime peak above $120 down to roughly $72 by late June
— briefly back to pre-war levels, tracking the study’s ceasefire path (+1.10% cumulative
by December). It has since reversed: renewed strikes, an attack on a Qatari LNG tanker on 7 July,
and a reinstated U.S. blockade of Iranian vessels with a proposed 20% transit fee pushed Brent back
above $80 on 13 July, its highest in about a month.
The scenario has not collapsed onto a single path. It is oscillating across the range the study
modeled — which is the case for reporting a distribution and a set of sliders rather than a
forecast, not against it.
What is untouched
The central finding. The incidence indices are invariant to the inflation rate — Suits and
Kakwani describe the shock’s shape, not its size. Whichever path realizes, the bill
still concentrates in dollars at the top and in share of income at the bottom. A smaller shock is a
smaller version of the same regressive shape, not a different shape. Nothing in this update moves
the thesis; it moves the magnitude’s odds.
What we are watching
The 28–29 July FOMC meeting, and whether the Hormuz transit regime stabilizes. If the
December run-rate resolves materially away from the ~$24.7B/month central estimate, we will
say so here — with the same specificity we are asking of you.
Update 02Scope: oil-path buffers · China · strategic reserves · diesel · § The scenario · § Portfolios
What was validated
The study said pump and diesel prices are set globally, and that “energy
independence” would not shield U.S. households. That is how it has played out — through
the products, not the crude. U.S. distillate inventories stood at 107.1 million barrels on
7 August, the lowest for late summer in roughly three decades (late-July stocks ran
109–127 million in 2022–25), and the IEA’s August report has diesel cracks at
record margins, with diesel exports from the major exporting regions down 1.3 mb/d
year-over-year. Russia — the world’s second-largest diesel exporter — has reportedly
banned gasoline exports into January 2027 and restricted diesel exports after strikes on its
refineries, and China reportedly directed its refiners to halt product exports in March, relenting
only partially since. Part of the mechanism is likely the barrel itself: record diesel cracks
against a softening crude headline are the pattern the study’s realist-lens sources argued a
Hormuz loss would produce20 — the Strait’s barrels are
mostly medium and heavy sour, the grades complex refineries are built to turn into diesel, so the
replacement slate makes less distillate per barrel. Strike-hit Russian refineries and China’s
run cuts then took capacity out on top, and export restrictions keep part of what is refined at
home. The crude headline softened; the diesel channel — the one this study runs household
incidence through — tightened.
What has not resolved
Which path — and the reason is now as interesting as the fact. The world’s biggest
buffer turned out to be stock drawdowns, not new supply. China, which had stockpiled through
the cheap window, cut its crude imports 32% in the second quarter — to 8.1 mb/d,
its first months below 8 since 2016 — and ran its refineries off inventory instead (imports
fell 3.9 mb/d; refinery runs only 2.2 mb/d). IEA members released a record 400 million
barrels. The U.S. Strategic Petroleum Reserve fell from 415 million barrels before the
war to 298.7 million on 7 August — its lowest since 1983 — en route to roughly
243 million when the reported release order completes. The draws muted the price without closing
the gap: the IEA now expects a 1.8 mb/d deficit in the third quarter — double its
July estimate — with observed global inventories down 410 million barrels since late
February. Pre-war Hormuz carried ~20 mb/d, about a fifth of global supply; the realized average
loss ran ~13–14% of global flows, with bypass pipelines and residual transit covering part.
Analyst estimates put China’s remaining cover near 100 days at the current burn.
The can has been kicked — competently, and at scale — but it was kicked with
inventory. The time it bought is measured in months, not years, and every month of buffer
spent is a month of buffer the next disruption will not have.
What is untouched
The central finding. The incidence indices are invariant to the inflation rate — they
describe the shock’s shape, not its size. A buffered shock is a smaller
version of the same regressive shape, not a different shape. If anything, a distillate-led squeeze
leans harder on the bottom of the distribution — diesel is trucking, farming, and heating
fuel before it is anyone’s discretionary mile — though the study modeled the household
basket, not product-level splits, and we hold that boundary honestly.
What we are watching
China’s restocking decision — resuming imports supports prices; continuing to draw
thins the world’s largest discretionary buffer. Whether the SPR release runs to its
~243 million barrel floor. Whether Russia’s export restrictions — and its
strike-reduced refinery output — persist into the heating season, and whether China’s
export restraint returns with the next escalation. Whether U.S. and Atlantic-basin distillate
stocks rebuild before the heating season — the same weekly EIA series this update is verified
against — because entering winter at a three-decade seasonal low is how a products squeeze
reaches home heating directly. And the standing
commitment stands: if the December run-rate resolves materially away from ~$24.7B/month,
we will say so here.
Update 03Scope: the pipeline closure · crude · the pump · CPI · reserves · the September Fed · § The scenario · § Five Fed meetings
What was validated
The study said the size of this shock hinges on the Strait of Hormuz. September tested
that from an unexpected direction: not the Strait, but the road around it. An official source at
the Saudi Ministry of Energy stated on 11 September that the East–West Pipeline, in the
Riyadh and Madinah regions, had been attacked on the morning of 10 September and was “shut down
as a precautionary measure,” with injuries, and further developments to be announced in due
course. The line carries crude from the eastern fields to Yanbu on the Red Sea; with Hormuz
constrained it had become the principal way Saudi barrels reached water that Iran does not sit
astride.
The market priced the route, not the barrel. Brent spot rose 10.5% in the session after the
attack, from $109.51 on 9 September to $120.98 on the 10th, reaching $130.80 on
the 15th. But the more telling number compares two crudes rather than two days: the gap between
Brent, which must travel by sea, and West Texas Intermediate, which does not, widened from
$4.54 a barrel on 1 September to $23.78 on the 15th. Both crudes rose — Brent
from $96.02 to $130.80, WTI from $91.48 to $107.02 — so this is not one benchmark falling
away from another; it is the waterborne barrel repricing 2.2 times harder than the
landlocked one. That is not a shortage of oil being priced. It is a shortage of ways to move it — but it is not a shortage that began on
10 September. The spread had already reached $12.25 by the 9th, the last session before the
attack: $7.71 of the $19.24 widening, two fifths of it, was in the price before a single drone
arrived. The same caution applies to the level — Brent had already climbed above $100 on
3 September. The attack steepened a climb that was under way; and our pinned daily series ends on
15 September, so this update says nothing about the week that followed, including reported progress
on restoring the line.
The products-not-crude claim is now in the record in dollars. Retail diesel reached
$6.285 a gallon in the week ending 14 September — the highest nominal price in the EIA
series, which begins in 1994, passing June 2022’s $5.810. Deflated, the record does not hold:
that 2022 peak is about $6.57 in current dollars, so the real price is about 4% below
its 2022 peak, and we would rather say so than keep the cleaner sentence. That is a statement about
price, not about burden: what a household actually pays turns on how much it buys, and on how much
of the diesel it never buys directly reaches it through freight. Against their
February averages diesel is up 69% and gasoline 47%; the gap between them, $0.75 a
gallon in early March, is now $1.83. The IEA’s September report gives the supply-side
mechanism: net diesel and gasoil exports from the Gulf averaged 390 kb/d in August, just
over a quarter of pre-war levels. And the Fed: on 16 September the Committee raised the target range
a quarter point to 3¾–4%, its first increase since 2023, and did it 12–0
— where July’s hold had drawn three dissents the other way. Update 01 graded a
higher-for-longer call; it has now been answered in the strongest form available.
What has not resolved
The size of the bill, and the reading has turned. August CPI-U rose 0.40% on the month,
enough to carry the index back above its May base: cumulative May-to-August is now +0.05%,
where the July print had it 0.35% below. Realized inflation is still beneath every track the
console draws — 1.49 points under the probability-weighted path, 0.86 under the all-ceasefire
corner — but the two gaps moved in opposite directions in August: the distance to the
mildest path narrowed, the distance to the weighted path widened slightly. On two months we were
prepared to say the December run-rate was tracking materially below the ~$24.7 billion central
estimate. On three, with the turn in hand, we are not saying it, and that restraint is the finding
rather than a hedge. August’s reference period closed before the pipeline was hit and before
diesel crossed $6. The September print, due 13 October, is the first that can contain either.
One more piece of arithmetic, and it cuts against everything above. The model’s scenarios are
not oil-price paths with an inflation layer on top; they are inflation paths, and the bill is
the cumulative December figure multiplied by a fixed constant. That makes the realized prints
load-bearing in a way the pipeline and the pump are not. Run the remaining four months at each
track’s own rate, starting from where the index actually is, and December lands at
$2.1 billion a month on the ceasefire path, $12.3 billion on status quo
— and $22.6 billion even on the all-escalation path, below the
~$24.7 billion central estimate and barely half the $44 billion escalation corner the
console draws. To reach the central estimate from here, the remaining four months would each have
to run 0.71%, hotter than the escalation track’s own 0.65%. That is arithmetic, not a
forecast. It is also why we lean the way we do about December — though not to the point of
calling it closed. Taking the realized months as given, the model as published still puts about a
7.5% chance on December reaching the central estimate; widen its monthly variance to what we
now measure and that becomes 9.2%, and on the escalation branch alone 29.7%. The shortfall from June
and July is banked, and that is a lean, not a verdict.
Why did it miss? Not, it turns out, because the mechanism is wrong. The monthly gap against the
weighted track has closed from 0.97 points in June to 0.47 in July to 0.04 in
August — two thirds of the entire shortfall is June alone, and August very nearly landed on
the track. The model was built on 5 June, with Brent averaging $107 in May; Brent then fell to $85
in June and $84 in July before returning to $110 in September. It was overtaken within days of
being seeded, by a de-escalation window that has since closed. The deeper limitation is ours to
own. The model assumes you are on one path for seven months and then blends whole paths by fixed
weights; it has no way to say ceasefire until August, escalation after, which is, as far as
we can tell, what happened. When the world switches branch mid-period, a fixed blend of whole
branches runs low, and stays low, because the early months are already spent.
And the shape of the error is worth naming precisely, because the obvious version of it is wrong.
The simulation lets a month come in off its path by about a sixth of a percentage point; the
realized month-to-month volatility of headline CPI, measured from the same government series the
model is scored against, is 1.6 times that — a gap wide enough that the model’s
own figure sits outside the range the data supports — and June came in off even the
mildest path by five times it. Conditional on any single
branch, the June-to-August sequence is something this model essentially cannot produce. The
study’s floor band for December is wide — about $4 billion to $49 billion
at ninety percent, and the coupled bands wider still — and it is the June-to-August
total that is roughly a one-in-four-hundred draw from the model that produced it: unlikely,
not impossible. December itself is a far less surprising question — the model gives a December
as low as the weighted landing about a one-in-five chance, not one in four hundred. So the verdict is narrower than our uncertainty was too tight. The headline
interval was wide enough. What the machinery underneath it could not do was represent a mild summer
followed by a sharp September — one branch is held for all seven months, and the monthly
wobble is set to about half what inflation actually does. A successor needs month-to-month variance
matched to the regime being modelled and the ability to change branch part-way through. We have
begun building one, and it contains this model as the special case where the branch never changes.
(These are seasonally adjusted prints and they revise; the conclusion does not turn on the third
decimal.)
Behind the prints, the buffers — and this is the update’s substance. Update 02 said the
blow was being buffered by stockpile drawdowns rather than new supply. Name them now and the picture
sharpens, because two of them are stocks and two are flows, and the two do different work:
stocks buy time, flows close the gap. The largest was a route — flow
capacity — and it was cut in an afternoon. The second is inventory: the IEA puts global observed stocks down
507 million barrels since the war began, an average draw of 2.8 mb/d — a stock being
spent, not a cushion being held. The third is the Strategic Petroleum Reserve, and its
behaviour has changed: weekly draws have collapsed from 6.1 million barrels to 1.2 million to
403 thousand in the week ending 11 September, with the reserve at 285.0 million barrels
and roughly 41.5 million still above what we have been calling the floor — and that number
deserves naming properly. The reserve held 415.4 million barrels on 6 March, the last reading
before the release order; 415.4 less the 172 million ordered is 243.4. The
“floor” is where the announced release finishes, not where the reserve runs out.
At 243 million barrels the SPR still holds 243 million barrels. The programme is about
three-quarters delivered, and how far off completion is depends entirely on which stretch you
average: about twenty weeks at last month’s pace, about eight at the pace of the
last six months, about two years at the pace of the last week alone. That spread is the point
— arithmetic, not a forecast. And the slowdown is not a wall being hit, and the cleanest way
to show that uses nothing but the reserve’s own record. In the week to 11 September it delivered 58
thousand barrels a day. Five weeks earlier it delivered 874 thousand — fifteen times as much
— and did it while holding about 5% more oil than it holds now. Whatever is slowing the
release, it is not the reserve’s ability to deliver. This is a delivery schedule finishing, not a reserve failing. What the
Energy Department does when it finishes is a different question, and one we are not going to guess
at. The fourth buffer was never ours at all: Chinese
restraint. China’s August crude imports were 37.93 million tonnes, up 6.2% on July —
the second monthly rise. A buffer that consists of someone else choosing not to buy lasts exactly as
long as they keep choosing. Whether that choosing has changed we cannot yet say: August cargoes were
bought against June and July prices — the year’s trough — with a month or more of
shipping behind them, so two rises are as consistent with bargain-hunting as with restraint ending,
and barrels going into Chinese storage rather than through Chinese refineries would have
added to the world’s buffer instead of spending it. September and October arrivals,
bought against $110 crude, are the ones that will tell. So: a route, a stock being spent, a stock
with a valve on its outlet, and another country’s restraint. Three of the four
can stop without warning. One did, in an afternoon.
What is untouched
The central finding. The incidence indices are invariant to the size of the bill: Suits and Kakwani
describe the shock’s shape, not its magnitude. A larger bill is a larger version of
the same regressive shape, and a distillate-led one — diesel up 69% on February against
gasoline’s 47% — leans on trucking, farming and heating fuel before it touches
anyone’s discretionary mile. Distillate stocks have begun the rebuild Update 02 told readers to
watch for, gaining 4.5 million barrels from the 21 August low to 107.9 million on
11 September. But the rebuild has not left record territory: 21 August, 28 August, 4 September and
11 September were each the lowest print for their own week of the year in the entire
1982–2026 weekly record, the last of them 4.4 million barrels below 1996, the next-lowest year
— and the heating season begins from there. The boundary Update 02 held stands: the study modeled the household basket, not
product-level splits, and we hold it.
What we are watching
The pipeline, and what Saudi Arabia says about it — the ministry’s statement promised
further announcements, and our pinned prices stop on 15 September. The September CPI on
13 October, the first print that can carry any of this. Distillate through the restocking window,
from the lowest base the record holds. The SPR, and whether the near-halt in deliveries is choice or
constraint. Russia’s restrictions, reported extended to 31 October. And the standing commitment,
which we are tightening rather than restating: Updates 01 and 02 promised to speak if the
December run-rate resolved materially away from the ~$24.7B/month central estimate. We now drop
the condition and the materiality threshold both — when December resolves against that
estimate we will say by how much and in which direction, including if the answer is that it landed on
the number.
Every study here has a surface a non-specialist can read, and a floor directly beneath it with
the numbers, code, and assumptions a specialist needs to trust it. This is the floor. The engines
run from a fixed seed (20260605, N = 200,000), and every headline statistic
in the report above is recomputable from what follows — including the raw draw vectors, so you
can re-derive any percentile yourself rather than take ours.
coupled_tail_mc.py (Economics & Statistics Referee seat) generated
coupled_mc_results.json, the two-band figures, and the raw draws — knobs:
p_crisis = 0.12, crisis_mult = 2.5, bfadd_crisis = 0.18,
Student-t copula ν = 4.
coupled_mc.py is a deliberately independent reimplementation by the Author, written to
check the first without reusing its code: it parameterizes the crisis regime differently
(scenario-conditional probabilities 0.03 / 0.08 / 0.30, crisis_mult scaling probability
rather than variance, m_crisis as the variance multiplier). It will not reproduce
coupled_mc_results.json line-for-line, and it is not meant to — it is the
cross-implementation check that confirmed the first engine’s tail. Both are published so the check
itself is auditable.
If a number in the report and a number in these files disagree, the files
win — and we want to know. That is what the next box is for.
Found a discrepancy?
Hold this study to the standard it holds others to.
This study exists because someone caught an annual inflation rate being applied as a monthly one
— an error that overstated a real problem by roughly fifteen- to twentyfold. A number that
exaggerates a genuine crisis doesn’t help the people it describes; it hands ammunition to
anyone who wants to dismiss them. The same standard applies here. Be specific, cite something,
and we’ll do the work — and if you’re right, we amend the study and log the change in the open.