Energy, fuel and food price rises and UK households, 2027-28

Microsimulation-based analysis of what energy, motor fuel and food price rises would cost UK households in 2027-28, with the ongoing Middle East conflict (active since late February 2026, with recurrent Strait of Hormuz disruption) as the context — modelling impacts through energy bills, fuel costs and food inflation across 31.6 million weighted households using policyengine.py 5.3.0.

Live Dashboard

Scenarios

These are stress tests, not forecasts. Each scenario applies a set of price assumptions as full-year 2027-28 amounts, including where the cited source describes a 2026 disruption lasting a few months. No time path, quarterly profile or shock duration is modelled, and the pass-through coefficients implied by the percentages are judgements anchored to the cited sources rather than equations derived from them. The results file records the definition, source, reference period, derivation and uncertainty range of every parameter under parameters.registry.

Calibrated to conditions as of August 2026 (Brent ~$85/bbl after a ~$109 peak). Anchors: observed path (low), Goldman Sachs extended Hormuz-closure Brent >$100 (central), Goldman extreme-adverse and the Oxford Economics escalation case (high).

Scenario	Energy bills	Fuel price	Food price	CPI
Summer 2026 prices	+15%	+22%	+2.0%	+1.3pp
Sustained escalation	+45%	+46%	+4.0%	+3.1pp
Severe escalation	+90%	+62%	+6.5%	+5.3pp

The CPI column is not a forecast of inflation. It is the addition each scenario's own energy, fuel and food prices imply on ONS 2026 basket weights, measured against a no-conflict path, and it is used only to size the benefit uprating shortfall and the accelerated-uprating option. It never enters the household cost.

The Ofgem cap (£1,663 for July–September 2026 on the new typical-consumption basis, £1,723 for October–December) is reported as context and does not enter the calculation: the energy percentage is applied to each household's own baseline gas and electricity expenditure in the microdata. The model does not represent unit rates, standing charges, the gas/electricity split, region, payment method, quarterly cap periods or fixed-tariff coverage.

How it works

The pipeline applies price increases through three transmission channels and computes household-level impacts for the 2027-28 tax year:

1. Energy — Each household's own baseline gas and electricity spend, raised by the scenario percentage. This is a household-energy-expenditure sensitivity, not a price-cap calculation: no cap level, unit rate, standing charge or tariff type enters it
2. Fuel — Petrol and diesel spending (ONS Family Spending Table A6, allocated to vehicle-owning households only), raised by the scenario percentage
3. Food — Food and non-alcoholic drink spending (ONS Family Spending Table A6), raised by the scenario percentage

Plus a separately reported uprating compensation shortfall — CPI-uprated benefit rates are uprated each April from the previous September's annual CPI rate. The conflict began in late February 2026, so the September 2026 rate already carries part of the shock (about 0.9pp from the conflict to the annual rate on ONS energy and fuel indices, a proxy on August 2026 data with food omitted, so a lower bound) and April 2027 uprating passes it on. The gain from uprating is modelled by raising every CPI-uprated benefit rate, and the reported ESA, contribution-based JSA, industrial injuries, incapacity benefit and armed forces compensation awards, and the CPI-uprated additional State Pension, in PolicyEngine UK at each scenario's residual and running the benefit rules, so tapers, the benefit cap and housing costs apply; UC work allowances, the childcare cap and the non-dependant deduction (which lowers awards) and Scotland's Pension Age Winter Heating Payment are uprated too. Left alone because they are not CPI-uprated: the triple-locked basic and new State Pension, the Pension Credit minimum guarantee, UC's frozen LCWRA element, frozen earnings disregards and tax credits. The savings credit maximum and the partly CPI-linked maternity allowance and statutory maternity, paternity and sick pay are not raised; r% of their total bounds that omission. Only the residual addition goes unindexed, for the whole of 2027-28. This is reported as the compensation an immediate uprating would deliver, and is not added to the cost channels: the household's loss is the price rise itself, and adding an uprating term on top would count the same shock twice. It is what the accelerated-uprating policy pays.

Poverty is measured as people below 60% of the person-weighted median of equivalised HBAI household net income (BHC); the post-shock figure holds that baseline line fixed and nets modelled costs off income, so it is an anchored, consumption-adjusted measure rather than official HBAI poverty.

No fuel poverty figure is reported. The study previously published an indicative 10%-of-income energy-spend ratio; its level was not comparable with any official statistic (England's LILEE measure adds an energy-efficiency test, and DESNZ's own 10% indicator uses after-housing-costs income and modelled required rather than actual energy costs), so it was withdrawn rather than published alongside figures it invited comparison with. See #22.

The results file breaks impacts down by income quintile, region, country, tenure type and household composition. The dashboard presents all of these except the regional breakdown.

Headline results

All three scenarios, 2027-28, across 31.6 million weighted households. Regenerate with iran-impact-build --sync-dashboard; these figures are pinned as a regression test in tests/test_committed_output.py.

If energy and fuel stayed near their summer 2026 levels through 2027-28 (energy +15%, anchored to the July cap; fuel +22%, the August pump-price rise; food +2%, a judgement; mid-September pump prices and the October cap are higher), the direct cost would be £538 a household (£17.0bn). If gas stayed at about twice its pre-conflict price through 2027-28, as in sustained escalation, it would be £1,334 a household (£42.1bn). Quote the escalation figures as conditionals, not as the estimate of the war's effect.

	Summer 2026 prices	Sustained escalation	Severe escalation
Mean cost per household	£538	£1,334	£2,304
Total cost	£17.0bn	£42.1bn	£72.7bn
People newly below the anchored poverty line	379,109	1,508,923	2,337,905
Mean of per-household cost/income ratios	1.8%	4.5%	7.9%
Total cost ÷ total net income	0.9%	2.3%	4.0%

The last two rows are different statistics, not two estimates of one thing. The mean of ratios weights every household equally and so is pulled up by households with very low incomes; it is the figure the dashboard reports as mean_net_impact_pct, and it excludes households with non-positive income, for whom the ratio is undefined. The aggregate ratio is total cost divided by total income.

Sustained-escalation cost by channel, per household: energy £713, fuel £474, food £147. The uprating compensation shortfall is a further £115, reported separately rather than added (see above).

The shock is regressive on either basis. Under sustained escalation the poorest fifth of households bear a cost equal to 4.2% of their mean net income against 1.5% for the richest fifth (10.3% against 1.7% as a mean of ratios), even though the cash amounts run the other way — £981 against £1,613.

Only the sustained-escalation column's cost, total and poverty count, plus the baseline aggregates and the combined package, are pinned as a regression test in tests/test_committed_output.py. The other figures here are read from the generated output and are not test-guarded.

Policy responses evaluated

Aligned to the live Autumn Budget 2026 decisions where applicable:

• Energy Price Guarantee (cap subsidy)
• Flat energy rebate (£400/household)
• Council tax rebate (bands A–D)
• Universal Credit uplift (£20/week, matching the 2020-21 uplift)
• Fuel duty cut extension (5p/litre on each household's own modelled petrol and diesel volume, so non-drivers receive nothing; the current cut expires 31 Dec 2026)
• Means-tested cost-of-living payment (£650, on the 2022 Cost of Living Payment's qualifying benefits: Universal Credit, income-based JSA, income-related ESA, Income Support, Working Tax Credit, Child Tax Credit and Pension Credit)
• Accelerated benefit uprating
• Electricity VAT cut extension (5% → 0%, enacted Oct 2026–Mar 2027; modelled full-year)
• Social tariff
• Combined package — the eight measures above except the social tariff, applied jointly so the Energy Price Guarantee caps the bill before the electricity VAT relief applies to it. The social tariff is evaluated standalone only.

Every cost is a gross modelled household transfer, not an Exchequer costing: it excludes tax and benefit interactions, take-up, behavioural responses, administration, non-household fuel use and financing. Each policy is assessed on gross outlay (unclipped government spending), household protection and residual impact, targeting efficiency (share of spending reaching the bottom two quintiles), and the number of people lifted back above the anchored poverty line.

Project structure

src/iran_impact/        # Python microsimulation pipeline
  config.py             # Scenario parameters, parameter registry, limitations
  pipeline.py           # Core engine (baseline, shocks, policy responses)
  inputs.py             # Loader for the committed spending input tables
  inputs/               # Versioned source data (ONS Family Spending A6 extract)
  cli.py                # CLI entry point
  provenance.py         # Run provenance: revision, versions, bundle, hashes
scripts/
  extract_ons_a6.py     # Regenerates the A6 extract from the ONS workbook
run_pipeline.py         # Runs pipeline, writes JSON output
data/                   # CLI output, and where the certified dataset materializes
tests/                  # Unit tests (synthetic) and integration tests (managed data)
requirements.lock       # Fully pinned Python environment
.github/workflows/      # CI: tests, lockfile check, dashboard lint and build
dashboard/              # Next.js interactive dashboard
  app/                  # App Router entry (layout, page, global styles)
  src/components/       # React components (scenarios, policy, methodology tabs)
  src/lib/              # Data helpers, formatters, chart utils
  public/data/          # Pipeline JSON output consumed by frontend

Quick start

# Install the Python package. The `uk` extra pulls policyengine[uk]==5.3.0,
# which brings the UK country package the pipeline needs. Installing plain
# `policyengine` is not enough — the managed-simulation import fails.
pip install -e ".[uk,dev]"

# Or, to pin every dependency exactly (see Reproducibility below). The lock
# holds third-party pins only, so the project itself is installed separately:
#   uv pip sync requirements.lock
#   uv pip install -e . --no-deps

# Authenticate for the managed microdata (see Data access below)
export HF_TOKEN=...            # or: hf auth login

# Run the analysis pipeline and refresh the dashboard's copy of the results
iran-impact-build --sync-dashboard

# Launch dashboard
cd dashboard && bun install && bun run dev

Run the unit tests with pytest. They use synthetic households and need no data access. The end-to-end checks in tests/test_integration.py require the managed dataset and skip themselves without it — run them locally before pushing any change to the calculation.

Data access

The pipeline runs against policyengine.py's managed UK microsimulation, so the model version and data build are certified by the release bundle rather than chosen by this repository. The current bundle resolves to:

Bundle	uk-5.3.0
Model	policyengine-uk 2.90.2 (certified by the bundle — do not pin separately)
Dataset	enhanced_frs_2024_25
Data build	policyengine-uk-data-1.56.16

The microdata lives in the Hugging Face repository policyengine/policyengine-uk-data-private, which is gated: access is granted per account rather than being public. Running the pipeline requires a token from an account that has been granted access, supplied either through HUGGING_FACE_TOKEN, HF_TOKEN or HUGGINGFACE_HUB_TOKEN in the environment, or by hf auth login. Without it the managed_microsimulation() call fails when it tries to fetch the dataset.

A populated Hugging Face hub cache is not sufficient on its own: policyengine.py reuses only a SHA-verified artifact at its own materialization target (./data/enhanced_frs_2024_25.h5 relative to the working directory), and otherwise downloads it.

Do not pin policyengine-uk or the dataset directly. policyengine.py's release bundle selects both, and pinning them here could contradict it — the 5.3.0 bundle selects policyengine-uk 2.90.2, which is not the newest release of that package.

Reproducibility

Every generated results file carries a provenance block recording the git revision, the versions of every package that can move a number, the certified release bundle (model version, dataset, data build ID and artifact hash), and SHA-256 hashes of the source files that define the calculation.

Two runs of the same revision against the same certified data build produce identical output apart from the run timestamp and git revision.

requirements.lock pins the complete Python environment, including the policyengine-uk version the release bundle certifies. It is a compiled requirements file, so it contains third-party pins only — installing it takes two steps, the second of which puts iran_impact and the iran-impact-build console script on the path:

uv pip sync requirements.lock     # exact third-party versions
uv pip install -e . --no-deps     # this project, without touching those pins

CI checks that the lock still matches pyproject.toml.

CI also compares the source hashes recorded in the committed output against the working tree. A change to config.py, pipeline.py, provenance.py, pyproject.toml or requirements.lock that lands without a regenerated results file fails the build — a dependency change can move a number without touching any calculation source. Regenerate with iran-impact-build --sync-dashboard.

Data sources

• policyengine.py 5.3.0 managed UK microsimulation (enhanced_frs_2024_25; see Data access) — household incomes, benefits, energy and road-fuel consumption, vehicle ownership, council tax band, region, tenure and family type
• ONS Family Spending in the UK, FYE 2024, Table A6 — transport fuel and food spending by gross-income decile, committed as a versioned extract under src/iran_impact/inputs/
• Ofgem price cap announcements — reported as context; not used in the calculation
• Scenario anchors: Goldman Sachs Hormuz-closure analysis, Oxford Economics escalation scenario, Commons Library CBP-10601 and CBP-10403, Bank of England and NIESR inflation projections — see parameters.registry in the output for which figure each supports

Built with PolicyEngine.

License

AGPL-3.0, matching the PolicyEngine model packages this project builds on.