The US/Israel–Iran war: the economic costs of a continued conflict
The 2026 US/Israel–Iran war disrupted Gulf supply in several ways at once. Iran retaliated against economic targets in Gulf countries that host US bases, striking oil and gas facilities and an aluminium smelter; 17% of Qatar’s LNG export capacity was damaged. Oil fields, refineries, and aluminium and fertiliser plants across the Gulf cut output, many because their exports could no longer get out. Iran's own exports were blockaded, and shipping through the Strait of Hormuz became restricted and far more expensive.
We take these measured disruptions and trace how they spread through the world economy, using two models. A dynamic (year-by-year) CGE model, GTAP-RD (Aguiar, Corong and van der Mensbrugghe 2019), links every economy’s industries, households, governments and trade, so it follows the knock-on effects: dearer oil and gas raise costs for firms and households everywhere, Gulf economies earn and spend less, and other suppliers step in. This is how a supply shock in the Gulf ends up changing GDP, trade, industry output and incomes in India, Japan, Europe and beyond.
A short-run price impact model estimates what the lost supply does to oil, gas and consumer prices. It sets world energy prices from how strongly prices actually responded to the lost supply in 2026 (following Kilian and Murphy 2014), then passes them into consumer prices using econometric estimates of how much past oil price rises reached consumers within two years (Jordà 2005; Känzig 2021).
Using this combination of models, we ask what the war did to economies around the world in 2026, and what six more months of fighting in 2027 would add.
How to read the numbers Unless stated otherwise, every CGE model result is the gap from a world with no war, in that year.
Zero is a world where the war never happened
Unless stated otherwise, all results from the dynamic CGE model (GTAP-RD) are measured as the difference from a no-war baseline: the path each economy would have followed with no war at all, projected forward from 2022 using the Shared Socioeconomic Pathways (SSP2, the “middle of the road” pathway), the standard long-run projections behind IPCC scenarios: GDP from the OECD and population from IIASA. On every chart, the zero line is that baseline.
For example, a figure of −5% in 2026 means that country's real GDP in 2026 is 5% lower than it would have been without the war. It does not mean the economy shrank by 5% from one year to the next.
The energy and consumer price results come from a separate short-run price impact model. Brent oil is shown as the price itself, in dollars a barrel, next to its no-war level. Gas and consumer prices are shown as the gap from a world with no war, in that year.
What is a CGE model?
A computable general equilibrium (CGE) model is a map of the whole world economy in numbers. It covers every industry, household and government in each region, and the trade between them, all linked through prices (Hertel 1997; Corong et al. 2017).
Because everything is linked, a shock in one place ripples through the rest of the economy. When Gulf oil supply falls, oil gets dearer; refiners, airlines and power plants pay more; households have less to spend on other things; other countries produce more oil; and trade shifts around the world. The model follows all of these knock-on effects at once.
It is built on standard economic theory: firms keep their costs down, households make the most of their income, and prices move until what is supplied matches what is bought. It is matched to real data on how each economy produces, spends and trades. Our dynamic version also tracks how investment builds up each economy’s capital year by year, so a shock today can still be felt years later.
A CGE model answers “what if” questions: it compares a world with the war to the same world without it. It is not a forecast of what GDP will be.
Like any model, it is a stylised picture of the economy and does not capture every link. For example, it leaves out money and the financial sector, so effects through banks, financial markets, interest rates and business and consumer confidence are not included. The results give useful insight into the size and direction of the war’s effects and who is hit hardest, but they should not be read as precise forecasts.
Headline results
Central case: basic GTAP-RD. Each GDP figure is the gap from the no-war baseline in a given year. Gulf economies: Saudi Arabia, the UAE, Qatar, Iraq, Kuwait and Bahrain (Iran excluded), weighted by baseline GDP in that year. Oil price card: short-run price model. Consumer price card: range between the two methods of measuring pass-through (estimated from past oil shocks, and measured in 2026).
Two paths for the war
The shocks below are the only things we change; every other result, in the Gulf and elsewhere, is the model working out how they ripple through the world economy. Both paths start from the same measured 2026. They differ only in whether the fighting stops at the end of 2026 or goes on into 2027. The difference between scenario B and scenario A is the extra cost of the war continuing, compared with a scenario where peace is reached at the end of 2026, measured year by year.
The war ends in 2026
- 2026 shocks as measured*: Gulf oil, LNG, refined fuel and fertiliser supply cut country by country, aluminium output down, and shipping through the Strait of Hormuz several times more expensive.
- Recovery: oil and fuel recover through 2027, based on the US EIA's Short-Term Energy Outlook (Oct 2026), which expects most Gulf production back to pre-war levels by mid-2027.
- Damaged LNG plants (17% of Qatar's capacity) are back by 2029, based on QatarEnergy's estimate of 3 to 5 years to repair; we use the short end, in line with Qatar's energy minister's figure of about 3 years.
The war goes on into 2027
- 2026 the same as A.
- Six more months of fighting in 2027, with Gulf supply losses near their worst measured level (spring 2026); higher shipping costs; Iran's exports down 80%.
- Recovery: the war ends after 2027, leaving further damage, with recovery delayed until the fighting stops.
- Oil and fuel are repaired over three years**, back to normal in 2030.
- LNG is repaired over five years, the long end of QatarEnergy's 3 to 5 year estimate, assumed because a longer war would likely cause greater overall damage; Qatar's LNG exports are still 14% down in 2030.
*Measured: the 2026 shocks (lost oil, gas and other supply) come from real-world data. GDP results are the model's estimates, not measurements.
**Three years is the rule of thumb for rebuilding oil production capacity in war-hit exporters (Alhajji 2003). Past cases ran from about 2 years (Iraq after 1988) to 4 years (Iran after 1988).
How we put the war into the model
We cut what Gulf oil fields and refineries produce, by the amount the 2026 data show was lost: US EIA monthly figures for crude oil and Kpler refinery-run figures for refining. This follows the observed data most closely, because in 2026 fields shut in and refineries cut runs, and some facilities were damaged. The oil was not produced and kept at home. What we capture is idle fields and refineries, plus damaged plants that take years to repair. Every other shock follows its own observed measure: Qatar’s LNG, Gulf urea and Iran’s oil as lost exports, and aluminium as lost output.
Sensitivity. As a check, we also applied the crude oil and refined fuel losses to what the Gulf ships abroad rather than what it produces, since the closure first showed up as cargoes that could not leave. This gives smaller losses for the Gulf and somewhat larger ones for oil importers. The results are in the full technical paper.
See the full shocks, sources and assumptions
Every figure is a yearly average. The model works in whole years, so a shock that lasts only part of a year is averaged across the twelve months. For example, six months at −60% and six months at −10% go in as −35% for that year. This is how B's 2027 (six months of fighting, then recovery) and 2026 (built from monthly data) become single yearly figures.
Data sources and a worked example
- 2026 data sources: crude oil from the US EIA (Short-Term Energy Outlook, Oct 2026); refined fuel from IEA, Kpler and JODI; LNG from cargo data; aluminium and urea from company reports and importers’ trade data; shipping costs from war-risk insurance quotes.
- Other assumption: LNG cargoes for November and December 2026 (no cargo data yet).
- Worked example, UAE crude oil in 2027 (scenario B): half the year at −39% (worst month of 2026, measured), half at −10% (war damage), plus −9% for a three-month pipeline outage, gives −34%.
Peace in 2027 or a longer war: how the two scenarios differ
Compare economies and models Click to open the full comparison: pick several economies and any of the model versions.
Compare economies and models
A second round of fighting would push energy costs much higher again
Annual average price levels compared with no war in that year. 2026 is measured; 2027 is projected. If fighting continues into 2027, Brent would average about $140 a barrel if oil stocks (commercial and government) can be drawn as in 2026, and more if they cannot, against $84 if the war ends this year and a hypothetical $53 if no war had occurred. Brent averaged $69 in 2025, before the war.
if oil stocks (commercial and government) can be drawn as in 2026; low end about $125 to $130; could pass $150 if stocks run short
EIA forecast; a hypothetical $53 if no war had occurred
against $69 in 2025, and a hypothetical $58 if no war had occurred in 2026
| 2026 (both scenarios) | 2027, war ends in 2026 | 2027, fighting continues | |
|---|---|---|---|
| European gas, vs. no war | about +85% range +65% to +110% | about +25% range +11% to +52% | about +75% range +35% to +210% |
| Asian LNG, vs. no war | about +110% range +90% to +130% | about +25% range +11% to +52% | about +75% range +35% to +210% |
Short-run price model (world energy market calibrated to measured 2026 prices, plus input-output cost pass-through). 2025 and 2026 from the EIA Short-Term Energy Outlook (October against February 2026; 2025 is the same in both) and IMF commodity prices. 2027 if fighting continues: the extra lost Gulf supply priced with the market response measured in 2026, including stock draws (commercial and government) (central), or the published oil price response of Caldara, Cavallo and Iacoviello (2019) (low end). Gas gaps are probably on the low side, because the no-war level is taken from the pre-war winter.
The war is already raising living costs, and a longer war would raise them much further
The difference in the consumer price index compared with if there were no war that year. +3% means the same shopping costs 3% more than it would without the war. It is a gap in the price level, not the yearly inflation rate.
against +0.9% to +2.0% if the war ends in 2026
against +1.0% to +1.3% if the war ends in 2026
US +1.0% to +1.8%; euro area +0.9% to +1.1%
Each range runs between the two methods of measuring pass-through (see the chart below). The main method is the total effect within two years of a typical supply-driven oil price rise, estimated from past oil shocks (1992 to 2019) and applied to each scenario's oil prices. It is not a Gulf-specific estimate. Prices keep rising into 2027 even if the war ends, because pass-through builds over one to two years. Independently checked.
Ranges presented as two different methodologies of measuring the pass-through of oil prices into inflation. For more information on the methodologies, please view the annex.
**UK dashed line: direct effect on household energy bills and motor fuel only, because the energy price cap delays bills. *Japan: less certain (see notes).
Why these five? They have long, clean price histories and clear estimates from past oil shocks. For others, such as Korea, China, India and Brazil, the estimates are too uncertain to show. Japan* is less certain (see the annex).
The consumer price figures shown already include the level of government support that was typical in past oil shocks, because they are estimated from how prices actually moved then. If governments do more this time (price caps, fuel tax cuts, subsidies), consumer prices would rise less than shown, and more of the cost would fall on public budgets.
Brent at about $140, if oil stocks (commercial and government) can be drawn as in 2026. The 2027 supply loss (10.7 million barrels a day) is larger than any past oil shock, so the price response is pushed beyond the range it was measured on, and nothing in the model caps the price where demand would collapse into recession. Treat it as indicative.
The annex shows two other measures: the price rises already visible in 2026 data, and the full effect if higher energy costs were passed on through every industry in full. Each answers a different question from the main estimate, so their figures should not be compared directly with it.
Short-run price model. Oil-to-consumer-price pass-through from local projections with an oil supply shock instrument (Jordà 2005; Känzig 2021), estimated on 1992 to 2019 monthly data (OECD Prices database; IMF consumer price data; Brent from IMF via FRED). A sample running to 2025 gives higher figures (US 2.4% / 3.7% / 6.5%; euro area 1.4% / 3.1% / 4.7%), because energy passed through faster in the 2021 to 2023 inflation surge.
Which industries gain and which lose
Each cell is one industry in one economy: the difference from the no-war baseline in that year. Coral means lower than without the war, navy means higher. Hover over or tap a cell for the detail. Output is the volume an industry produces; exports and imports are the volumes it trades, all valued at no-war prices, so price changes are taken out. Choose the model in each panel; the default is our central case (basic GTAP-RD). A choice in one panel applies to all three sections.
US refinery output is +3.9% in 2026, while Japan and Korea’s is −2.3%, and −6.3% in 2027 if the war goes on. Japan and Korea’s refiners rely on Gulf crude (82% of their crude imports).
UAE aluminium and other metals output is −32% in 2026. India (+2.7%) and Japan and Korea (+2.8%) produce more. A small story: the Gulf makes about 2% of world output in this industry.
Qatar’s gas output is −63% in 2026. If the war goes on into 2027, it is still −13% in 2030, as the damaged LNG plants are repaired over five years (assumed, the long end of QatarEnergy’s 3 to 5 year estimate). If the war ends in 2026, it is back to about −3% by 2029.
Grey hatched cells are changes of 1% (or $0.5 billion) or more that the model likely overstates, so read only their direction: oil and gas gains outside the Gulf, and gains in other Gulf industries (see below). Smaller changes are shown as they are and should be read as about zero. A dot (·) marks an industry too small to show: under $200 million a year without the war. Output is gross output; exports and imports are bilateral trade volumes valued at no-war export prices (before freight and insurance). $ billion at 2022 prices; % is the change in volume.
Greyed-out values: why the model likely overstates some industry changes
- Oil and gas output outside the Gulf. In the model, producers elsewhere replace the lost Gulf supply within the same year: crude output in the US, the EU and China rises by about 18 to 20% in 2027 if the war goes on. In reality little spare capacity can come on that fast. In 2026 most of the gap was covered by oil stocks and lower consumption instead, which the model does not include. So these gains are far larger than could happen within a year.
- Other Gulf industries. Some Gulf industries outside energy grow strongly in the model: in 2027, if the war goes on, Qatar’s other manufacturing output is up 81%. Two features of the model drive this. First, Qatar’s LNG is cut at exports, so the workers and capital in its idle LNG plants are released (both down about 60%), and in the model they move straight into other industries within the year. Second, the model lets wages and other costs fall sharply, which acts like a large fall in the currency and makes Gulf goods much cheaper abroad: the cost of workers and capital falls 52% against world prices in Qatar, 30% in Iraq, Kuwait and Bahrain and 15% in Saudi Arabia. In practice Gulf currencies are pegged to the dollar, wages (largely of migrant workers) adjust slowly, and new factories take years to build. The sticky-wages version damps this only a little (Qatar’s other manufacturing +68% instead of +81%).
How the war bends the energy trade map
Who buys from the Gulf, who supplies instead, and how quickly trade returns to normal. Default: central case (basic GTAP-RD). Every figure compares a year with the same year without the war.
Gulf crude exports, including Iran’s, are $194 billion lower in 2026 than without the war (at 2022 prices), led by Iraq, Kuwait and Bahrain (−$85 billion) and Saudi Arabia (−$58 billion). In the model, producers in the rest of the world (+$148 billion) and the United States (+$35 billion) export more to fill the gap. If the war goes on, the Gulf loses $338 billion in 2027, and the rest of the world (+$235 billion) and the US (+$59 billion) fill most of it. The model refills the gap too quickly, since in 2026 oil stocks did much of that job, so read this as where supply shifts rather than how much.
The Gulf’s share of China’s crude imports falls from 52% to 37% in 2026, and to 27% in 2027 if the war goes on. By 2030 it is back to about 52%.
The Gulf supplies 71% of India’s gas imports without the war. That falls to 30% in 2026, and is still 63% in 2030 if the war goes on.
Gulf incomes fall, so Gulf imports fall. Egypt’s exports to the Gulf are −19% in 2026, a loss equal to about 2.4% of all Egypt’s exports.
Share of each importer’s imports of the good that comes from the Gulf (Saudi Arabia, the UAE, Qatar, Iraq, Kuwait, Bahrain and Iran), valued at no-war prices.
Change in each economy’s exports of the good chosen above, $ million at 2022 prices, against the no-war baseline in that year. *Oil and gas gains outside the Gulf are too large: the model replaces all the lost Gulf supply within the year, while in 2026 oil stocks did most of that job. Read them as the direction trade moves.
Trade between any two economies
Pick a good, an economy and a direction to see how its trade with each partner changes against the no-war baseline in that year. Volumes valued at no-war prices. Default: central case (basic GTAP-RD).
Change in trade with each partner, $ million at 2022 prices, valued at no-war export prices (before freight and insurance). "Within the region" is trade between countries inside a group, such as within the EU. Hover over a bar for the % change and the no-war level.
In-depth comparison: industries and trade Click to open. Build up to six cases, each with its own economy, industry, scenario and model, and compare them over time or in one year.
Each value is the gap from the no-war baseline in that year. Output is gross output; exports, imports and bilateral trade are volumes valued at no-war export prices (before freight and insurance). $ million at 2022 prices. Oil and gas gains outside the Gulf, and gains in Gulf industries, show the direction only (see the heatmap note).
Consumer prices: higher or lower than without the war?
Cost of the usual household basket, % above or below its no-war level in the chosen year. One scale for all economies, years, versions and models. This is a relative price level, not inflation.
Exploratory. The model has no money or exchange rates, so prices are measured against a fixed world anchor. Large falls in the Gulf mean consumer prices there fall relative to world prices as incomes collapse; in practice this shows up as lower real wages, not cheaper shops. Both indexes come straight from the model. The fixed basket prices the no-war household basket at the model's household prices, as official CPIs do; the model's own index (ppriv) uses what households actually buy. They agree within about 0.3 points everywhere except Qatar, where petrol prices nearly triple and households cut petrol purchases by 85% (2026: −12% fixed basket, −22% model index). The world figure weights regions by baseline household spending (our choice).
Who loses, and by how much
Choose a measure. Click a row to follow that economy over time. In 2026 the two scenarios give the same results, because they only diverge from 2027.
Show all regions as a table
What six more months of fighting adds
B minus A: how much further below its no-war level an economy is in each year if the war runs into 2027.
Exports and output by industry
Difference from the no-war baseline in the same year, $ million at 2022 prices, so small industries do not look dramatic. Uses the economy and year chosen above.
How much the answer depends on the model
The same shocks run through each model set-up. Choose an economy, a measure and a scenario; B minus A shows the extra cost of the war going on into 2027. Each value is the gap from the no-war baseline in that year.
Lines on the right show 2025 to 2030. Industry output and exports are volumes.
How the 2026 results compare with the IMF
Our modelled real GDP for 2026 against the IMF: its October 2026 forecasts, set against its pre-war October 2025 forecasts, give a rough read of the war's hit to real GDP in 2026.
How the GTAP-RD model works
The model
A global computable general equilibrium model: every economy’s industries, households, governments and trade, linked by prices. We use the standard GTAP model, version 7 (Corong et al. 2017), with recursive dynamics, GTAP-RD (Aguiar, Corong and van der Mensbrugghe 2019), so capital builds up year by year and investment follows returns. We coded GTAP-RD into Julia ourselves, on top of Maroš Ivanič’s Julia code for the standard GTAP model (Ivanič 2025). The data cover 12 regions with 2022 as the base year, on an SSP2 baseline (Riahi et al. 2017). We change only the measured shocks (Gulf supply, Iran’s exports and shipping costs); the model then works out how prices, output, trade, jobs and incomes adjust in every economy.
Model extensions coded by us
We coded these on top of the Julia model and tested each one independently.
- Sticky wages: real wages adjust to unemployment with a lag, so jobs can fall in the slump (wage rule from Dixon and Rimmer 2002).
- GTAP-E: energy and capital substitute, and fuels substitute for each other (Burniaux and Truong 2002; McDougall and Golub 2007).
- Gulf risk premium: investors demand a higher return in the Gulf during and after the war, sized from sovereign spread moves using Damodaran’s country risk premium (Damodaran 2026) and the IMF’s conflict estimate (IMF 2025a); central, high and low paths.
What to keep in mind
Working paper
Built on a beta release of our dataset. Results may be revised as the dataset and model are developed further.
Yearly averages
The model works in years, so price spikes within a year are smoothed out. Peak prices were far higher.
How recovery is modelled
Every shocked flow is held at its sourced level in every year, recovery years included. The recovery is modelled directly as changes to Gulf oil and refinery production, based on forecasts.
Gas substitution
Buyers switch gas suppliers at a fifth of the standard GTAP rate, since most gas moves on fixed routes. A departure from GTAP, chosen by us.
Supply elsewhere
Other producers raise output at the model's supply response, with no explicit spare-capacity limit, so importer losses are a lower bound: with sticky wages, the world’s 2026 loss doubles (−0.66% against −0.32%). Oil stock releases and demand restraint, which covered much of the 2026 gap, are not modelled.
Assumed paths for the longer war
Six more months of fighting, a three-month pipeline outage (the one in September 2026 lasted about 10 days), the size of the damage, three-year repairs for oil and five-year repairs for LNG, and shipping costs at their peak through 2027 are all our assumptions: there is no forecast for a longer war. A small shipping-cost add-on after the war, fading to zero over three years, is also assumed.
Jobs
In the basic model, employment stays at its baseline level and wages adjust. Only the sticky-wage version lets jobs fall.
Idle plants
Under the shock to production, idle fields and refineries keep most of their workers and capital while producing less. This is how the model applies an output cut. It makes Gulf losses larger, Qatar’s most (−30% in 2026 against about −20% implied by the IMF).
War damage
Damage enters as lost output only: no capital is destroyed and repairs cost nothing. A test run that also destroys capital adds about 1 to 1.5 points to Qatar’s 2030 loss.
No money or finance
The CGE model has no money, exchange rates, interest rates or confidence effects. Energy and consumer prices come from the separate short-run price model, which is not linked to it.
Small results
Small GDP gains should be read as zero or near-zero effects, not as gains from the war. Small results for industries outside energy are not robust.
Non-oil losses
The UAE, and probably Iran, fare worse in IMF forecasts than in the model, which points to losses outside oil and gas (tourism, finance, trade hubs) that the shocks do not cover. Iran’s results should not be read as an estimate of the war’s impact on Iran: they do not include the physical destruction of the war there.
The GTAP-RD model versions
Every version is the same dynamic CGE model, GTAP-RD (Aguiar, Corong and van der Mensbrugghe 2019), which we coded into Julia ourselves on top of Ivanič’s Julia code for the standard GTAP model (Ivanič 2025), with the same shocks; each changes one part of how the economy responds. The extensions were coded by us.
- Basic GTAP-RD
- The standard GTAP model (version 7) with recursive dynamics: capital builds up year by year and investment follows rates of return (Corong et al. 2017; Aguiar, Corong and van der Mensbrugghe 2019). Employment stays at its baseline level and wages adjust fully each year.
- GTAP-RD + sticky wages
- Real wages adjust to unemployment only with a lag, so jobs can be lost in the slump and output falls further while wages catch up (wage rule from Dixon and Rimmer 2002).
- GTAP-RD + GTAP-E (energy substitution)
- Firms can substitute between energy and capital, and between fuels such as oil, gas and coal, when energy prices change (Burniaux and Truong 2002; McDougall and Golub 2007).
- GTAP-RD + sticky wages + GTAP-E
- Both of the above together.
- GTAP-RD + Gulf risk premium (central, high, low)
- Investors demand a higher return on capital in the Gulf while the war lasts, which slows investment there. The premium is 1.52 times the rise in sovereign credit spreads, following Damodaran’s country risk premium (Damodaran 2026). Central: about 0.15 percentage points a year for Saudi Arabia, 0.25 for the UAE and Qatar, 0.9 for Iraq, Kuwait and Bahrain, and 2.7 for Iran, from the spread moves in this war and the IMF's conflict estimate for emerging economies (IMF 2025a). It holds in each war year, halves in the first year after the war and then ends; for Iran it fades to zero over eight years. High: the Gulf states at the size of Qatar's 2017 crisis move (0.85; 2.7 for Iraq, Kuwait and Bahrain), with Iran's premium not fading before 2030. Low: no premium for Saudi Arabia, the UAE and Qatar, 0.25 for Iraq, Kuwait and Bahrain, 0.6 for Iran. Fade paths are our assumption.
- GTAP-RD + sticky wages + Gulf risk premium, central
- Sticky wages and the central risk premium together.
How the short-run price impact model works
World energy prices
Prices move until buyers and other suppliers make up the lost Gulf supply. How strongly they respond is measured from 2026 itself: the actual price rise against the actual supply loss, counting stock draws (commercial and government) as part of the response (Kilian and Murphy 2014; Caldara, Cavallo and Iacoviello 2019). 2025 and 2026 prices come from the US EIA’s Short-Term Energy Outlook and IMF commodity prices; the no-war path is the EIA’s pre-war February 2026 forecast.
Consumer prices
Oil prices are passed into consumer prices using how much of a typical supply-driven oil price rise has reached consumer prices within two years in the past, estimated by local projections on 1992 to 2019 data (Jordà 2005; Känzig 2021). The second method uses the speed measured in 2026 so far. A third, the full effect if energy costs were passed through every industry in full, uses input-output tables (Miller and Blair 2009). The annex compares the methods.
Consumer prices: two methods of measuring pass-through
The shaded ranges in the main consumer price charts show two different methods of measuring how energy prices pass through to consumer prices. The solid line is the central estimate of the main method, estimated econometrically from past oil shocks; its 90% statistical band is given in the table below. The dashed line, set out here, is the pass-through actually measured from March to September 2026, carried forward at the same speed.
Why the two methods differ. The main estimate learns from past oil shocks how prices respond over one to two years, so it keeps prices rising into 2027 even if the war ends. The measured method only sees the first seven months of this shock and assumes that speed simply continues, so it misses the slower knock-on effects still to come. It also picks up this year's support policies (price caps, fuel tax cuts, subsidies), which hold prices down for now. That is why it is lower for the US and Japan. In the euro area it is slightly higher, because European prices have risen faster this year than past shocks would suggest. Where the two are close, as for Canada and the UK if fighting continues, the figure is more solid.
Method 1 (main): econometrically estimated from past oil shocks
We look at how consumer prices responded to oil supply shocks from 1992 to 2019, using local projections with an oil supply shock instrument that separates oil supply news from other events (Jordà 2005; Känzig 2021). We then feed each scenario's monthly oil prices through that response. It gives the total effect within two years of a typical supply-driven oil price rise, not a Gulf-specific estimate. These are the solid lines in the main chart.
The 90% band under each figure is the statistical uncertainty of the estimate: given how varied past oil shocks were, we are 90% confident the true effect lies in that range. Every band except Japan's in 2026 stays above zero, so the effect is real, but its size is uncertain; the bands widen in 2027 if fighting continues because that oil shock is far larger than any in the data. Japan* is less certain: its 2026 band reaches zero, and its history includes consumption tax rises.
| 2026 | 2027, war ends in 2026 | 2027, fighting continues | |
|---|---|---|---|
| United States | about +1.8% 90% band +1.3% to +2.5% | about +2.0% 90% band +1.0% to +3.9% | about +3.9% 90% band +2.5% to +6.7% |
| Euro area | about +0.9% 90% band +0.7% to +1.2% | about +1.3% 90% band +0.7% to +1.9% | about +2.3% 90% band +1.4% to +3.2% |
| United Kingdom | about +0.7% 90% band +0.2% to +1.3% | about +1.4% 90% band +0.4% to +3.7% | about +2.2% 90% band +0.7% to +5.2% |
| Canada | about +1.0% 90% band +0.7% to +1.5% | about +1.0% 90% band +0.4% to +1.6% | about +2.1% 90% band +1.3% to +3.2% |
| Japan* | about +0.6% 90% band +0.0% to +1.1% | about +1.5% 90% band +0.2% to +3.4% | about +2.2% 90% band +0.2% to +4.7% |
Method 2: measured in 2026 so far
How much consumer prices have actually risen above their pre-war trend from March to September 2026, per 1% rise in oil prices, carried forward at the same speed. These are the dashed lines in the main chart.
The main weakness: many other things also moved prices this year. Unlike the main method, this one does not separate the oil shock from everything else, so any other price change since March is counted as the war. Examples:
- Government support (price caps, fuel tax cuts, subsidies) has held prices down for now.
- Gas and electricity prices have moved differently from oil, especially in Europe.
- Exchange rates have shifted, changing what imports cost in local currency.
- Central banks have responded, and weaker confidence has cooled spending.
- Other price changes, such as food, tariffs and wages, would have happened without the war.
- The no-war trend itself is a guess: if it is drawn too high or too low, the measured effect moves with it.
It also covers only seven months, so it misses the slower knock-on effects that come later. We therefore use it as a check on the main method, not as the central estimate.
| 2026 | 2027, war ends in 2026 | 2027, fighting continues | |
|---|---|---|---|
| United States | about +1% | about +1% | about +2.5% |
| Euro area | about +1% | about +1% | about +3% |
| Canada* | about +0.8% | about +0.7% | about +2% |
| Korea* | about +0.7% | about +0.6% | about +1.8% |
| United Kingdom (direct energy only, different method) | about +0.8% | about +0.7% | about +2% |
| ** Held down by policy so far: what has reached shoppers, not the full cost | |||
| India** (March to August) | about +0.7% | about +0.6% | about +1.7% |
| Japan** | about +0.3% | about +0.3% | about +0.8% |
| Brazil** | about +0.3% | about +0.2% | about +0.7% |
| Australia** | about +0.2% | about +0.2% | about +0.5% |
| China** | about +0.1% | about +0.1% | about +0.4% |
Which economies? Every figure except the UK uses the same method: the 2026 pass-through measured from monthly consumer price data. The US and euro area are the most reliable. Starred (*) figures carry a caveat. Double-starred (**) figures are measured too, but governments there probably hold fuel and power prices down, so they show what has reached shoppers so far, not the full cost.
Scrapping the consumer carbon charge in April 2025 may make the measured figure a little too high.
Fuel tax cuts and regulated electricity prices may dampen what shoppers see.
State fuel pricing, subsidies and bill rebates probably hold prices down (not yet traced in the data). The cost may land on budgets and state firms instead. Full cost: see the next annex section.
The energy price cap delays bills by about a quarter, so the standard method fails. Instead we count the direct effect on household energy bills and motor fuel only, so this is probably a lower bound (2026 range +0.6% to +1.0%).
Subsidy changes distort the measurement.
Short-run price model. Pass-through measured as the March to September 2026 rise in consumer prices above their pre-war trend, per 1% rise in oil prices, applied to each scenario's oil price (data: FRED, OECD Prices database, IMF consumer price data for Japan; UK: Ofgem price cap timing). Other economies, if higher energy costs are passed through in full: see the next annex section.
Consumer prices by economy, if higher energy costs are passed through in full
| If passed through in full | 2026 | 2027, war ends in 2026 | 2027, fighting continues |
|---|---|---|---|
| World | about +3% | about +2% | about +6% |
| United States | about +1% | about +1% | about +3% |
| EU | about +4% | about +2% | about +7% |
| Japan & Korea | about +5% | about +3% | about +8% |
| China | about +2% | about +1% | about +4% |
| India | about +5% | about +3% | about +10% |
Short-run price model, input-output cost pass-through on the Matra dataset (beta v5.5) (Miller and Blair 2009). Each year is a separate calculation: the same input-output tables, fed with that year's energy price gaps from no war. Nothing carries over between years.
References and data
References
- Aguiar, A., Corong, E. and van der Mensbrugghe, D. (2019). The GTAP Recursive Dynamic (GTAP-RD) Model: Version 1.0. West Lafayette: Center for Global Trade Analysis, Purdue University, February. Available at: ageconsearch.umn.edu/record/333133.
- Alhajji, A.F. (2003). Oil production capacity rebuilding experience has implications for Iraq. Oil & Gas Journal, 3 November.
- Burniaux, J.-M. and Truong, T. (2002). GTAP-E: An Energy-Environmental Version of the GTAP Model. GTAP Technical Paper 16. West Lafayette: Center for Global Trade Analysis, Purdue University.
- Caldara, D., Cavallo, M. and Iacoviello, M. (2019). Oil price elasticities and oil price fluctuations. Journal of Monetary Economics, 103, 1–20.
- Corong, E., Hertel, T., McDougall, R., Tsigas, M. and van der Mensbrugghe, D. (2017). The Standard GTAP Model, Version 7. Journal of Global Economic Analysis, 2(1), 1–119.
- Damodaran, A. (2026). Country Default Spreads and Risk Premiums, January 2026 update. New York: NYU Stern School of Business. Available at: pages.stern.nyu.edu/~adamodar/New_Home_Page/datafile/ctryprem.html.
- Dixon, P. and Rimmer, M. (2002). Dynamic General Equilibrium Modelling for Forecasting and Policy: A Practical Guide and Documentation of MONASH. Contributions to Economic Analysis 256. Amsterdam: North-Holland.
- Hertel, T.W. (ed.) (1997). Global Trade Analysis: Modeling and Applications. Cambridge: Cambridge University Press.
- IIASA (2024). SSP Scenario Database, version 3.0. Laxenburg: International Institute for Applied Systems Analysis. Available at: zenodo.org/records/14718294.
- IMF (2025a). Global Financial Stability Report: Enhancing Resilience amid Uncertainty, April 2025, chapter 2. Washington, DC: International Monetary Fund.
- IMF (2025b). World Economic Outlook database, October 2025. Washington, DC: International Monetary Fund.
- IMF (2026). World Economic Outlook database, October 2026. Washington, DC: International Monetary Fund.
- Ivanič, M. (2025). GlobalTradeAnalysisProjectModelV7.jl: the standard GTAP model, version 7, in Julia. Julia package, version 0.1.11. GitHub: github.com/mivanic/GlobalTradeAnalysisProjectModelV7.jl (MIT licence).
- Jordà, Ò. (2005). Estimation and inference of impulse responses by local projections. American Economic Review, 95(1), 161–182.
- Känzig, D.R. (2021). The macroeconomic effects of oil supply news: evidence from OPEC announcements. American Economic Review, 111(4), 1092–1125.
- Kilian, L. and Murphy, D.P. (2014). The role of inventories and speculative trading in the global market for crude oil. Journal of Applied Econometrics, 29(3), 454–478.
- McDougall, R. and Golub, A. (2007). GTAP-E: A Revised Energy-Environmental Version of the GTAP Model. GTAP Research Memorandum 15. West Lafayette: Center for Global Trade Analysis, Purdue University.
- Miller, R.E. and Blair, P.D. (2009). Input-Output Analysis: Foundations and Extensions, 2nd edn. Cambridge: Cambridge University Press.
- OECD (2026). Prices database: consumer price indices, all items. Paris: OECD. Accessed 31 October 2026.
- Riahi, K. et al. (2017). The Shared Socioeconomic Pathways and their energy, land use, and greenhouse gas emissions implications: an overview. Global Environmental Change, 42, 153–168.
- US Energy Information Administration (2026). Short-Term Energy Outlook, February and October 2026. Washington, DC: EIA.
- Wittgenstein Centre for Demography and Global Human Capital (2023). Wittgenstein Centre Human Capital Data Explorer: population projections (WIC2023). Vienna: Wittgenstein Centre. Available at: zenodo.org/records/7896627.
Data
- International Energy Agency, Oil Market Reports, March to October 2026; JODI oil database; Kpler; AXSMarine.
- IMF Primary Commodity Prices (via FRED, Federal Reserve Bank of St Louis).
- Company reports: QatarEnergy, Qatalum, Alba, EGA, SABIC, Industries Qatar.
- QatarEnergy on LNG damage: 17% of export capacity (trains 4 and 6) damaged, 3 to 5 years to repair. Reported by Qatar Commodities Intelligence, S&P Global and gCaptain, 2026.
- Matra global dataset, beta (built from the OECD Inter-Country Input-Output tables).
- Baseline: Shared Socioeconomic Pathways, SSP2 (Riahi et al. 2017); SSP Scenario Database v3.0 (IIASA 2024), with population from the Wittgenstein Centre projections (Wittgenstein Centre 2023); history from the World Bank.
Our own choices, marked as assumptions: putting the crude oil and refining losses on production; Dolphin held at capacity; the rule for B's 2027; pipeline outages in B; repair paths (except the QatarEnergy LNG range); gas trade elasticities at a fifth of GTAP's; risk premium fade paths.